NAME

libPARI - Functions and Operations Available in PARI and GP

The functions and operators available in PARI and in the GP/PARI calculator are numerous and ever-expanding. Here is a description of the ones available in version 2.7.4. It should be noted that many of these functions accept quite different types as arguments, but others are more restricted. The list of acceptable types will be given for each function or class of functions. Except when stated otherwise, it is understood that a function or operation which should make natural sense is legal. In this chapter, we will describe the functions according to a rough classification. The general entry looks something like:

foo(x,{flag = 0}): short description.

The library syntax is GEN foo(GEN x, long fl = 0).

This means that the GP function foo has one mandatory argument x, and an optional one, flag, whose default value is 0. (The {} should not be typed, it is just a convenient notation we will use throughout to denote optional arguments.) That is, you can type foo(x,2), or foo(x), which is then understood to mean foo(x,0). As well, a comma or closing parenthesis, where an optional argument should have been, signals to GP it should use the default. Thus, the syntax foo(x,) is also accepted as a synonym for our last expression. When a function has more than one optional argument, the argument list is filled with user supplied values, in order. When none are left, the defaults are used instead. Thus, assuming that foo's prototype had been

   foo({x = 1},{y = 2},{z = 3}),

typing in foo(6,4) would give you foo(6,4,3). In the rare case when you want to set some far away argument, and leave the defaults in between as they stand, you can use the ``empty arg'' trick alluded to above: foo(6,,1) would yield foo(6,2,1). By the way, foo() by itself yields foo(1,2,3) as was to be expected.

In this rather special case of a function having no mandatory argument, you can even omit the (): a standalone foo would be enough (though we do not recommend it for your scripts, for the sake of clarity). In defining GP syntax, we strove to put optional arguments at the end of the argument list (of course, since they would not make sense otherwise), and in order of decreasing usefulness so that, most of the time, you will be able to ignore them.

Finally, an optional argument (between braces) followed by a star, like {x}*, means that any number of such arguments (possibly none) can be given. This is in particular used by the various print routines.

@3Flags. A flag is an argument which, rather than conveying actual information to the routine, instructs it to change its default behavior, e.g. return more or less information. All such flags are optional, and will be called flag in the function descriptions to follow. There are two different kind of flags

@3* generic: all valid values for the flag are individually described (``If flag is equal to 1, then...'').

@3* binary: use customary binary notation as a compact way to represent many toggles with just one integer. Let (p_0,...,p_n) be a list of switches (i.e. of properties which take either the value 0 or 1), the number 2^3 + 2^5 = 40 means that p_3 and p_5 are set (that is, set to 1), and none of the others are (that is, they are set to 0). This is announced as ``The binary digits of flag mean 1: p_0, 2: p_1, 4: p_2'', and so on, using the available consecutive powers of 2.

@3Mnemonics for flags. Numeric flags as mentioned above are obscure, error-prone, and quite rigid: should the authors want to adopt a new flag numbering scheme (for instance when noticing flags with the same meaning but different numeric values across a set of routines), it would break backward compatibility. The only advantage of explicit numeric values is that they are fast to type, so their use is only advised when using the calculator gp.

As an alternative, one can replace a numeric flag by a character string containing symbolic identifiers. For a generic flag, the mnemonic corresponding to the numeric identifier is given after it as in

  fun(x, {flag = 0} ):
    If flag is equal to 1 = AGM, use an agm formula ...

@3 which means that one can use indifferently fun(x, 1) or fun(x, "AGM").

For a binary flag, mnemonics corresponding to the various toggles are given after each of them. They can be negated by prepending no_ to the mnemonic, or by removing such a prefix. These toggles are grouped together using any punctuation character (such as ',' or ';'). For instance (taken from description of ploth(X = a,b,expr,{flag = 0},{n = 0}))

    Binary digits of flags mean: C<1 = Parametric>,
    C<2 = Recursive>,...

@3so that, instead of 1, one could use the mnemonic "Parametric; no_Recursive", or simply "Parametric" since Recursive is unset by default (default value of flag is 0, i.e. everything unset). People used to the bit-or notation in languages like C may also use the form "Parametric | no_Recursive".

@3Pointers. \varsidx{pointer} If a parameter in the function prototype is prefixed with a & sign, as in

foo(x,&e)

@3it means that, besides the normal return value, the function may assign a value to e as a side effect. When passing the argument, the & sign has to be typed in explicitly. As of version 2.7.4, this pointer argument is optional for all documented functions, hence the & will always appear between brackets as in Z_issquare(x,{&e}).

@3About library programming. The library function foo, as defined at the beginning of this section, is seen to have two mandatory arguments, x and flag: no function seen in the present chapter has been implemented so as to accept a variable number of arguments, so all arguments are mandatory when programming with the library (usually, variants are provided corresponding to the various flag values). We include an = default value token in the prototype to signal how a missing argument should be encoded. Most of the time, it will be a NULL pointer, or -1 for a variable number. Refer to the User's Guide to the PARI library for general background and details.


Standard monadic or dyadic operators

\subseckbd{+/-} The expressions +x and -x refer to monadic operators (the first does nothing, the second negates x).

The library syntax is GEN gneg(GEN x) for -x.

\subseckbd{+} The expression x + y is the sum of x and y. Addition between a scalar type x and a t_COL or t_MAT y returns respectively [y[1] + x, y[2],...] and y + x {Id}. Other additions between a scalar type and a vector or a matrix, or between vector/matrices of incompatible sizes are forbidden.

The library syntax is GEN gadd(GEN x, GEN y).

\subseckbd{-} The expression x - y is the difference of x and y. Subtraction between a scalar type x and a t_COL or t_MAT y returns respectively [y[1] - x, y[2],...] and y - x {Id}. Other subtractions between a scalar type and a vector or a matrix, or between vector/matrices of incompatible sizes are forbidden.

The library syntax is GEN gsub(GEN x, GEN y) for x - y.

\subseckbd{*} The expression x * y is the product of x and y. Among the prominent impossibilities are multiplication between vector/matrices of incompatible sizes, between a t_INTMOD or t_PADIC Restricted to scalars, * is commutative; because of vector and matrix operations, it is not commutative in general.

Multiplication between two t_VECs or two t_COLs is not allowed; to take the scalar product of two vectors of the same length, transpose one of the vectors (using the operator ~ or the function mattranspose, see Label se:linear_algebra) and multiply a line vector by a column vector:

  ? a = [1,2,3];
  ? a * a
    ***   at top-level: a*a
    ***                  ^--
    *** _*_: forbidden multiplication t_VEC * t_VEC.
  ? a * a~
  %2 = 14

If x,y are binary quadratic forms, compose them; see also qfbnucomp and qfbnupow. If x,y are t_VECSMALL of the same length, understand them as permutations and compose them.

The library syntax is GEN gmul(GEN x, GEN y) for x * y. Also available is GEN gsqr(GEN x) for x * x.

/ The expression x / y is the quotient of x and y. In addition to the impossibilities for multiplication, note that if the divisor is a matrix, it must be an invertible square matrix, and in that case the result is x*y^{-1}. Furthermore note that the result is as exact as possible

in particular, division of two integers always gives a rational number (which may be an integer if the quotient is exact) and not the Euclidean quotient (see x \ y for that), and similarly the quotient of two polynomials is a rational function in general. To obtain the approximate real value of the quotient of two integers, add 0. to the result; to obtain the approximate p-adic value of the quotient of two integers, add O(p^k) to the result; finally, to obtain the Taylor series expansion of the quotient of two polynomials, add O(X^k) to the result or use the taylor function (see Label se:taylor).

The library syntax is GEN gdiv(GEN x, GEN y) for x / y.

\subseckbd{} The expression x \y is the Euclidean quotient of x and y. If y is a real scalar, this is defined as floor(x/y) if y > 0, and ceil(x/y) if y < 0 and the division is not exact. Hence the remainder x - (x\y)*y is in [0, |y|[.

Note that when y is an integer and x a polynomial, y is first promoted to a polynomial of degree 0. When x is a vector or matrix, the operator is applied componentwise.

The library syntax is GEN gdivent(GEN x, GEN y) for x \ y.

\/ The expression x \/ y evaluates to the rounded Euclidean quotient of x and y. This is the same as x \y except for scalar division

the quotient is such that the corresponding remainder is smallest in absolute value and in case of a tie the quotient closest to + oo is chosen (hence the remainder would belong to ]{-}|y|/2, |y|/2]).

When x is a vector or matrix, the operator is applied componentwise.

The library syntax is GEN gdivround(GEN x, GEN y) for x \/ y.

% The expression x % y evaluates to the modular Euclidean remainder of x and y, which we now define. When x or y is a non-integral real number, x%y is defined as x - (x\y)*y. Otherwise, if y is an integer, this is the smallest non-negative integer congruent to x modulo y. (This actually coincides with the previous definition if and only if x is an integer.) If y is a polynomial, this is the polynomial of smallest degree congruent to x modulo y. For instance

  ? (1/2) % 3
  %1 = 2
  ? 0.5 % 3
  %2 = 0.5000000000000000000000000000
  ? (1/2) % 3.0
  %3 = 1/2

Note that when y is an integer and x a polynomial, y is first promoted to a polynomial of degree 0. When x is a vector or matrix, the operator is applied componentwise.

The library syntax is GEN gmod(GEN x, GEN y) for x % y.

\subseckbd{^} The expression x^n is powering. If the exponent is an integer, then exact operations are performed using binary (left-shift) powering techniques. In particular, in this case x cannot be a vector or matrix unless it is a square matrix (invertible if the exponent is negative). If x is a p-adic number, its precision will increase if v_p(n) > 0. Powering a binary quadratic form (types t_QFI and t_QFR) returns a reduced representative of the class, provided the input is reduced. In particular, x^1 is identical to x.

PARI is able to rewrite the multiplication x * x of two identical objects as x^2, or sqr(x). Here, identical means the operands are two different labels referencing the same chunk of memory; no equality test is performed. This is no longer true when more than two arguments are involved.

If the exponent is not of type integer, this is treated as a transcendental function (see Label se:trans), and in particular has the effect of componentwise powering on vector or matrices.

As an exception, if the exponent is a rational number p/q and x an integer modulo a prime or a p-adic number, return a solution y of y^q = x^p if it exists. Currently, q must not have large prime factors. Beware that

  ? Mod(7,19)^(1/2)
  %1 = Mod(11, 19) /* is any square root */
  ? sqrt(Mod(7,19))
  %2 = Mod(8, 19)  /* is the smallest square root */
  ? Mod(7,19)^(3/5)
  %3 = Mod(1, 19)
  ? %3^(5/3)
  %4 = Mod(1, 19)  /* Mod(7,19) is just another cubic root */

If the exponent is a negative integer, an inverse must be computed. For non-invertible t_INTMOD, this will fail and implicitly exhibit a non trivial factor of the modulus:

  ? Mod(4,6)^(-1)
    ***   at top-level: Mod(4,6)^(-1)
    ***                         ^-----
    *** _^_: impossible inverse modulo: Mod(2, 6).

(Here, a factor 2 is obtained directly. In general, take the gcd of the representative and the modulus.) This is most useful when performing complicated operations modulo an integer N whose factorization is unknown. Either the computation succeeds and all is well, or a factor d is discovered and the computation may be restarted modulo d or N/d.

For non-invertible t_POLMOD, this will fail without exhibiting a factor.

  ? Mod(x^2, x^3-x)^(-1)
    ***   at top-level: Mod(x^2,x^3-x)^(-1)
    ***                               ^-----
    *** _^_: non-invertible polynomial in RgXQ_inv.
  ? a = Mod(3,4)*y^3 + Mod(1,4); b = y^6+y^5+y^4+y^3+y^2+y+1;
  ? Mod(a, b)^(-1);
    ***   at top-level: Mod(a,b)^(-1)
    ***                         ^-----
    *** _^_: impossible inverse modulo: Mod(0, 4).

In fact the latter polynomial is invertible, but the algorithm used (subresultant) assumes the base ring is a domain. If it is not the case, as here for Z/4Z, a result will be correct but chances are an error will occur first. In this specific case, one should work with 2-adics. In general, one can try the following approach

  ? inversemod(a, b) =
  { my(m);
    m = polsylvestermatrix(polrecip(a), polrecip(b));
    m = matinverseimage(m, matid(#m)[,1]);
    Polrev( vecextract(m, Str("..", poldegree(b))), variable(b) )
  }
  ? inversemod(a,b)
  %2 = Mod(2,4)*y^5 + Mod(3,4)*y^3 + Mod(1,4)*y^2 + Mod(3,4)*y + Mod(2,4)

This is not guaranteed to work either since it must invert pivots. See Label se:linear_algebra.

The library syntax is GEN gpow(GEN x, GEN n, long prec) for x^n.

cmp(x,y)

Gives the result of a comparison between arbitrary objects x and y (as -1, 0 or 1). The underlying order relation is transitive, the function returns 0 if and only if x === y, and its restriction to integers coincides with the customary one. Besides that, it has no useful mathematical meaning.

In case all components are equal up to the smallest length of the operands, the more complex is considered to be larger. More precisely, the longest is the largest; when lengths are equal, we have matrix > vector > scalar. For example:

  ? cmp(1, 2)
  %1 = -1
  ? cmp(2, 1)
  %2 = 1
  ? cmp(1, 1.0)   \\ note that 1 == 1.0, but (1===1.0) is false.
  %3 = -1
  ? cmp(x + Pi, [])
  %4 = -1

@3This function is mostly useful to handle sorted lists or vectors of arbitrary objects. For instance, if v is a vector, the construction vecsort(v, cmp) is equivalent to Set(v).

The library syntax is GEN cmp_universal(GEN x, GEN y).

divrem(x,y,{v})

Creates a column vector with two components, the first being the Euclidean quotient (x \y), the second the Euclidean remainder (x - (x\y)*y), of the division of x by y. This avoids the need to do two divisions if one needs both the quotient and the remainder. If v is present, and x, y are multivariate polynomials, divide with respect to the variable v.

Beware that divrem(x,y)[2] is in general not the same as x % y; no GP operator corresponds to it:

  ? divrem(1/2, 3)[2]
  %1 = 1/2
  ? (1/2) % 3
  %2 = 2
  ? divrem(Mod(2,9), 3)[2]
   ***   at top-level: divrem(Mod(2,9),3)[2
   ***                 ^--------------------
   ***   forbidden division t_INTMOD \ t_INT.
  ? Mod(2,9) % 6
  %3 = Mod(2,3)

The library syntax is GEN divrem(GEN x, GEN y, long v = -1), where v is a variable number. Also available is GEN gdiventres(GEN x, GEN y) when v is not needed.

lex(x,y)

Gives the result of a lexicographic comparison between x and y (as -1, 0 or 1). This is to be interpreted in quite a wide sense: It is admissible to compare objects of different types (scalars, vectors, matrices), provided the scalars can be compared, as well as vectors/matrices of different lengths. The comparison is recursive.

In case all components are equal up to the smallest length of the operands, the more complex is considered to be larger. More precisely, the longest is the largest; when lengths are equal, we have matrix > vector > scalar. For example:

  ? lex([1,3], [1,2,5])
  %1 = 1
  ? lex([1,3], [1,3,-1])
  %2 = -1
  ? lex([1], [[1]])
  %3 = -1
  ? lex([1], [1]~)
  %4 = 0

The library syntax is GEN lexcmp(GEN x, GEN y).

max(x,y)

Creates the maximum of x and y when they can be compared.

The library syntax is GEN gmax(GEN x, GEN y).

min(x,y)

Creates the minimum of x and y when they can be compared.

The library syntax is GEN gmin(GEN x, GEN y).

shift(x,n)

Shifts x componentwise left by n bits if n >= 0 and right by |n| bits if n < 0. May be abbreviated as x << n or x >> (-n). A left shift by n corresponds to multiplication by 2^n. A right shift of an integer x by |n| corresponds to a Euclidean division of x by 2^{|n|} with a remainder of the same sign as x, hence is not the same (in general) as x \ 2^n.

The library syntax is GEN gshift(GEN x, long n).

shiftmul(x,n)

Multiplies x by 2^n. The difference with shift is that when n < 0, ordinary division takes place, hence for example if x is an integer the result may be a fraction, while for shifts Euclidean division takes place when n < 0 hence if x is an integer the result is still an integer.

The library syntax is GEN gmul2n(GEN x, long n).

sign(x)

sign (0, 1 or -1) of x, which must be of type integer, real or fraction.

The library syntax is GEN gsigne(GEN x).

vecmax(x,{&v})

If x is a vector or a matrix, returns the largest entry of x, otherwise returns a copy of x. Error if x is empty.

If v is given, set it to the index of a largest entry (indirect maximum), when x is a vector. If x is a matrix, set v to coordinates [i,j] such that x[i,j] is a largest entry. This flag is ignored if x is not a vector or matrix.

  ? vecmax([10, 20, -30, 40])
  %1 = 40
  ? vecmax([10, 20, -30, 40], &v); v
  %2 = 4
  ? vecmax([10, 20; -30, 40], &v); v
  %3 = [2, 2]

The library syntax is GEN vecmax0(GEN x, GEN *v = NULL). Also available is GEN vecmax(GEN x).

vecmin(x,{&v})

If x is a vector or a matrix, returns the smallest entry of x, otherwise returns a copy of x. Error if x is empty.

If v is given, set it to the index of a smallest entry (indirect minimum), when x is a vector. If x is a matrix, set v to coordinates [i,j] such that x[i,j] is a smallest entry. This is ignored if x is not a vector or matrix.

  ? vecmin([10, 20, -30, 40])
  %1 = -30
  ? vecmin([10, 20, -30, 40], &v); v
  %2 = 3
  ? vecmin([10, 20; -30, 40], &v); v
  %3 = [2, 1]

The library syntax is GEN vecmin0(GEN x, GEN *v = NULL). Also available is GEN vecmin(GEN x).

Comparison and Boolean operators

The six standard comparison operators <= , < , >= , > , == , != are available in GP. The result is 1 if the comparison is true, 0 if it is false. The operator == is quite liberal : for instance, the integer 0, a 0 polynomial, and a vector with 0 entries are all tested equal.

The extra operator === tests whether two objects are identical and is much stricter than == : objects of different type or length are never identical.

For the purpose of comparison, t_STR objects are strictly larger than any other non-string type; two t_STR objects are compared using the standard lexicographic order.

GP accepts < > as a synonym for != . On the other hand, = is definitely not a synonym for == : it is the assignment statement.

The standard boolean operators || (inclusive or), && (and) and ! (not) are also available.


Conversions and similar elementary functions or commands

Many of the conversion functions are rounding or truncating operations. In this case, if the argument is a rational function, the result is the Euclidean quotient of the numerator by the denominator, and if the argument is a vector or a matrix, the operation is done componentwise. This will not be restated for every function.

Col(x, {n})

Transforms the object x into a column vector. The dimension of the resulting vector can be optionally specified via the extra parameter n.

If n is omitted or 0, the dimension depends on the type of x; the vector has a single component, except when x is

@3* a vector or a quadratic form (in which case the resulting vector is simply the initial object considered as a row vector),

@3* a polynomial or a power series. In the case of a polynomial, the coefficients of the vector start with the leading coefficient of the polynomial, while for power series only the significant coefficients are taken into account, but this time by increasing order of degree. In this last case, Vec is the reciprocal function of Pol and Ser respectively,

@3* a matrix (the column of row vector comprising the matrix is returned),

@3* a character string (a vector of individual characters is returned).

In the last two cases (matrix and character string), n is meaningless and must be omitted or an error is raised. Otherwise, if n is given, 0 entries are appended at the end of the vector if n > 0, and prepended at the beginning if n < 0. The dimension of the resulting vector is |n|.

Note that the function Colrev does not exist, use Vecrev.

The library syntax is GEN gtocol0(GEN x, long n). GEN gtocol(GEN x) is also available.

Colrev(x, {n})

As Col(x, n), then reverse the result. In particular

The library syntax is GEN gtocolrev0(GEN x, long n). GEN gtocolrev(GEN x) is also available.

List({x = []})

Transforms a (row or column) vector x into a list, whose components are the entries of x. Similarly for a list, but rather useless in this case. For other types, creates a list with the single element x. Note that, except when x is omitted, this function creates a small memory leak; so, either initialize all lists to the empty list, or use them sparingly.

The library syntax is GEN gtolist(GEN x = NULL). The variant GEN listcreate(void) creates an empty list.

Mat({x = []})

Transforms the object x into a matrix. If x is already a matrix, a copy of x is created. If x is a row (resp. column) vector, this creates a 1-row (resp. 1-column) matrix, unless all elements are column (resp. row) vectors of the same length, in which case the vectors are concatenated sideways and the associated big matrix is returned. If x is a binary quadratic form, creates the associated 2 x 2 matrix. Otherwise, this creates a 1 x 1 matrix containing x.

  ? Mat(x + 1)
  %1 =
  [x + 1]
  ? Vec( matid(3) )
  %2 = [[1, 0, 0]~, [0, 1, 0]~, [0, 0, 1]~]
  ? Mat(%)
  %3 =
  [1 0 0]
  [0 1 0]
  [0 0 1]
  ? Col( [1,2; 3,4] )
  %4 = [[1, 2], [3, 4]]~
  ? Mat(%)
  %5 =
  [1 2]
  [3 4]
  ? Mat(Qfb(1,2,3))
  %6 =
  [1 1]
  [1 3]

The library syntax is GEN gtomat(GEN x = NULL).

Mod(a,b)

In its basic form, creates an intmod or a polmod (a mod b); b must be an integer or a polynomial. We then obtain a t_INTMOD and a t_POLMOD respectively:

  ? t = Mod(2,17); t^8
  %1 = Mod(1, 17)
  ? t = Mod(x,x^2+1); t^2
  %2 = Mod(-1, x^2+1)

@3If a % b makes sense and yields a result of the appropriate type (t_INT or scalar/t_POL), the operation succeeds as well:

  ? Mod(1/2, 5)
  %3 = Mod(3, 5)
  ? Mod(7 + O(3^6), 3)
  %4 = Mod(1, 3)
  ? Mod(Mod(1,12), 9)
  %5 = Mod(1, 3)
  ? Mod(1/x, x^2+1)
  %6 = Mod(-1, x^2+1)
  ? Mod(exp(x), x^4)
  %7 = Mod(1/6*x^3 + 1/2*x^2 + x + 1, x^4)

If a is a complex object, ``base change'' it to Z/bZ or K[x]/(b), which is equivalent to, but faster than, multiplying it by Mod(1,b):

  ? Mod([1,2;3,4], 2)
  %8 =
  [Mod(1, 2) Mod(0, 2)]
  [Mod(1, 2) Mod(0, 2)]
  ? Mod(3*x+5, 2)
  %9 = Mod(1, 2)*x + Mod(1, 2)
  ? Mod(x^2 + y*x + y^3, y^2+1)
  %10 = Mod(1, y^2 + 1)*x^2 + Mod(y, y^2 + 1)*x + Mod(-y, y^2 + 1)

This function is not the same as x % y, the result of which has no knowledge of the intended modulus y. Compare

  ? x = 4 % 5; x + 1
  %1 = 5
  ? x = Mod(4,5); x + 1
  %2 = Mod(0,5)

The library syntax is GEN gmodulo(GEN a, GEN b).

Pol(t,{v = 'x})

Transforms the object t into a polynomial with main variable v. If t is a scalar, this gives a constant polynomial. If t is a power series with non-negative valuation or a rational function, the effect is similar to truncate, i.e. we chop off the O(X^k) or compute the Euclidean quotient of the numerator by the denominator, then change the main variable of the result to v.

The main use of this function is when t is a vector: it creates the polynomial whose coefficients are given by t, with t[1] being the leading coefficient (which can be zero). It is much faster to evaluate Pol on a vector of coefficients in this way, than the corresponding formal expression a_n X^n +...+ a_0, which is evaluated naively exactly as written (linear versus quadratic time in n). Polrev can be used if one wants x[1] to be the constant coefficient:

  ? Pol([1,2,3])
  %1 = x^2 + 2*x + 3
  ? Polrev([1,2,3])
  %2 = 3*x^2 + 2*x + 1

The reciprocal function of Pol (resp. Polrev) is Vec (resp.  Vecrev).

  ? Vec(Pol([1,2,3]))
  %1 = [1, 2, 3]
  ? Vecrev( Polrev([1,2,3]) )
  %2 = [1, 2, 3]

@3Warning. This is not a substitution function. It will not transform an object containing variables of higher priority than v.

  ? Pol(x + y, y)
    ***   at top-level: Pol(x+y,y)
    ***                 ^----------
    *** Pol: variable must have higher priority in gtopoly.

The library syntax is GEN gtopoly(GEN t, long v = -1), where v is a variable number.

Polrev(t,{v = 'x})

Transform the object t into a polynomial with main variable v. If t is a scalar, this gives a constant polynomial. If t is a power series, the effect is identical to truncate, i.e. it chops off the O(X^k).

The main use of this function is when t is a vector: it creates the polynomial whose coefficients are given by t, with t[1] being the constant term. Pol can be used if one wants t[1] to be the leading coefficient:

  ? Polrev([1,2,3])
  %1 = 3*x^2 + 2*x + 1
  ? Pol([1,2,3])
  %2 = x^2 + 2*x + 3

The reciprocal function of Pol (resp. Polrev) is Vec (resp.  Vecrev).

The library syntax is GEN gtopolyrev(GEN t, long v = -1), where v is a variable number.

Qfb(a,b,c,{D = 0.})

Creates the binary quadratic form ax^2+bxy+cy^2. If b^2-4ac > 0, initialize Shanks' distance function to D. Negative definite forms are not implemented, use their positive definite counterpart instead.

The library syntax is GEN Qfb0(GEN a, GEN b, GEN c, GEN D = NULL, long prec). Also available are GEN qfi(GEN a, GEN b, GEN c) (assumes b^2-4ac < 0) and GEN qfr(GEN a, GEN b, GEN c, GEN D) (assumes b^2-4ac > 0).

Ser(s,{v = 'x},{d = seriesprecision})

Transforms the object s into a power series with main variable v (x by default) and precision (number of significant terms) equal to d ( = the default seriesprecision by default). If s is a scalar, this gives a constant power series in v with precision d. If s is a polynomial, the polynomial is truncated to d terms if needed

  ? Ser(1, 'y, 5)
  %1 = 1 + O(y^5)
  ? Ser(x^2,, 5)
  %2 = x^2 + O(x^7)
  ? T = polcyclo(100)
  %3 = x^40 - x^30 + x^20 - x^10 + 1
  ? Ser(T, 'x, 11)
  %4 = 1 - x^10 + O(x^11)

@3The function is more or less equivalent with multiplication by 1 + O(v^d) in theses cases, only faster.

If s is a vector, on the other hand, the coefficients of the vector are understood to be the coefficients of the power series starting from the constant term (as in Polrev(x)), and the precision d is ignored: in other words, in this case, we convert t_VEC / t_COL to the power series whose significant terms are exactly given by the vector entries. Finally, if s is already a power series in v, we return it verbatim, ignoring d again. If d significant terms are desired in the last two cases, convert/truncate to t_POL first.

  ? v = [1,2,3]; Ser(v, t, 7)
  %5 = 1 + 2*t + 3*t^2 + O(t^3)  \\ 3 terms: 7 is ignored!
  ? Ser(Polrev(v,t), t, 7)
  %6 = 1 + 2*t + 3*t^2 + O(t^7)
  ? s = 1+x+O(x^2); Ser(s, x, 7)
  %7 = 1 + x + O(x^2)  \\ 2 terms: 7 ignored
  ? Ser(truncate(s), x, 7)
  %8 = 1 + x + O(x^7)

The warning given for Pol also applies here: this is not a substitution function.

The library syntax is GEN gtoser(GEN s, long v = -1, long precdl), where v is a variable number.

Set({x = []})

Converts x into a set, i.e. into a row vector, with strictly increasing entries with respect to the (somewhat arbitrary) universal comparison function cmp. Standard container types t_VEC, t_COL, t_LIST and t_VECSMALL are converted to the set with corresponding elements. All others are converted to a set with one element.

  ? Set([1,2,4,2,1,3])
  %1 = [1, 2, 3, 4]
  ? Set(x)
  %2 = [x]
  ? Set(Vecsmall([1,3,2,1,3]))
  %3 = [1, 2, 3]

The library syntax is GEN gtoset(GEN x = NULL).

Str({x}*)

Converts its argument list into a single character string (type t_STR, the empty string if x is omitted). To recover an ordinary GEN from a string, apply eval to it. The arguments of Str are evaluated in string context, see Label se:strings.

  ? x2 = 0; i = 2; Str(x, i)
  %1 = "x2"
  ? eval(%)
  %2 = 0

This function is mostly useless in library mode. Use the pair strtoGEN/GENtostr to convert between GEN and char*. The latter returns a malloced string, which should be freed after usage.

Strchr(x)

Converts x to a string, translating each integer into a character.

  ? Strchr(97)
  %1 = "a"
  ? Vecsmall("hello world")
  %2 = Vecsmall([104, 101, 108, 108, 111, 32, 119, 111, 114, 108, 100])
  ? Strchr(%)
  %3 = "hello world"

The library syntax is GEN Strchr(GEN x).

Strexpand({x}*)

Converts its argument list into a single character string (type t_STR, the empty string if x is omitted). Then perform environment expansion, see Label se:envir. This feature can be used to read environment variable values.

  ? Strexpand("$HOME/doc")
  %1 = "/home/pari/doc"

The individual arguments are read in string context, see Label se:strings.

Strtex({x}*)

Translates its arguments to TeX format, and concatenates the results into a single character string (type t_STR, the empty string if x is omitted).

The individual arguments are read in string context, see Label se:strings.

Vec(x, {n})

Transforms the object x into a row vector. The dimension of the resulting vector can be optionally specified via the extra parameter n.

If n is omitted or 0, the dimension depends on the type of x; the vector has a single component, except when x is

@3* a vector or a quadratic form (in which case the resulting vector is simply the initial object considered as a row vector),

@3* a polynomial or a power series. In the case of a polynomial, the coefficients of the vector start with the leading coefficient of the polynomial, while for power series only the significant coefficients are taken into account, but this time by increasing order of degree. In this last case, Vec is the reciprocal function of Pol and Ser respectively,

@3* a matrix: return the vector of columns comprising the matrix.

@3* a character string: return the vector of individual characters.

@3* an error context (t_ERROR): return the error components, see iferr.

In the last three cases (matrix, character string, error), n is meaningless and must be omitted or an error is raised. Otherwise, if n is given, 0 entries are appended at the end of the vector if n > 0, and prepended at the beginning if n < 0. The dimension of the resulting vector is |n|. Variant: GEN gtovec(GEN x) is also available.

The library syntax is GEN gtovec0(GEN x, long n).

Vecrev(x, {n})

As Vec(x, n), then reverse the result. In particular In this case, Vecrev is the reciprocal function of Polrev: the coefficients of the vector start with the constant coefficient of the polynomial and the others follow by increasing degree.

The library syntax is GEN gtovecrev0(GEN x, long n). GEN gtovecrev(GEN x) is also available.

Vecsmall(x, {n})

Transforms the object x into a row vector of type t_VECSMALL. The dimension of the resulting vector can be optionally specified via the extra parameter n.

This acts as Vec(x,n), but only on a limited set of objects: the result must be representable as a vector of small integers. If x is a character string, a vector of individual characters in ASCII encoding is returned (Strchr yields back the character string).

The library syntax is GEN gtovecsmall0(GEN x, long n). GEN gtovecsmall(GEN x) is also available.

binary(x)

Outputs the vector of the binary digits of |x|. Here x can be an integer, a real number (in which case the result has two components, one for the integer part, one for the fractional part) or a vector/matrix.

The library syntax is GEN binaire(GEN x).

bitand(x,y)

Bitwise and of two integers x and y, that is the integer

  sum_i (x_i and y_i) 2^i

Negative numbers behave 2-adically, i.e. the result is the 2-adic limit of bitand(x_n,y_n), where x_n and y_n are non-negative integers tending to x and y respectively. (The result is an ordinary integer, possibly negative.)

  ? bitand(5, 3)
  %1 = 1
  ? bitand(-5, 3)
  %2 = 3
  ? bitand(-5, -3)
  %3 = -7

The library syntax is GEN gbitand(GEN x, GEN y). Also available is GEN ibitand(GEN x, GEN y), which returns the bitwise and of |x| and |y|, two integers.

bitneg(x,{n = -1})

bitwise negation of an integer x, truncated to n bits, n >= 0, that is the integer

  sum_{i = 0}^{n-1} not(x_i) 2^i.

The special case n = -1 means no truncation: an infinite sequence of leading 1 is then represented as a negative number.

See Label se:bitand for the behavior for negative arguments.

The library syntax is GEN gbitneg(GEN x, long n).

bitnegimply(x,y)

Bitwise negated imply of two integers x and y (or not (x ==> y)), that is the integer

  sum (x_i and not(y_i)) 2^i

See Label se:bitand for the behavior for negative arguments.

The library syntax is GEN gbitnegimply(GEN x, GEN y). Also available is GEN ibitnegimply(GEN x, GEN y), which returns the bitwise negated imply of |x| and |y|, two integers.

bitor(x,y)

bitwise (inclusive) or of two integers x and y, that is the integer

  sum (x_i or y_i) 2^i

See Label se:bitand for the behavior for negative arguments.

The library syntax is GEN gbitor(GEN x, GEN y). Also available is GEN ibitor(GEN x, GEN y), which returns the bitwise ir of |x| and |y|, two integers.

bittest(x,n)

Outputs the n-th bit of x starting from the right (i.e. the coefficient of 2^n in the binary expansion of x). The result is 0 or 1.

  ? bittest(7, 3)
  %1 = 1 \\ the 3rd bit is 1
  ? bittest(7, 4)
  %2 = 0 \\ the 4th bit is 0

See Label se:bitand for the behavior at negative arguments.

The library syntax is GEN gbittest(GEN x, long n). For a t_INT x, the variant long bittest(GEN x, long n) is generally easier to use, and if furthermore n >= 0 the low-level function ulong int_bit(GEN x, long n) returns bittest(abs(x),n).

bitxor(x,y)

Bitwise (exclusive) or of two integers x and y, that is the integer

  sum (x_i xor y_i) 2^i

See Label se:bitand for the behavior for negative arguments.

The library syntax is GEN gbitxor(GEN x, GEN y). Also available is GEN ibitxor(GEN x, GEN y), which returns the bitwise xor of |x| and |y|, two integers.

ceil(x)

Ceiling of x. When x is in R, the result is the smallest integer greater than or equal to x. Applied to a rational function, ceil(x) returns the Euclidean quotient of the numerator by the denominator.

The library syntax is GEN gceil(GEN x).

centerlift(x,{v})

Same as lift, except that t_INTMOD and t_PADIC components are lifted using centered residues:

@3* for a t_INTMOD x belongs to Z/nZ, the lift y is such that -n/2 < y <= n/2.

@3* a t_PADIC x is lifted in the same way as above (modulo p^padicprec(x)) if its valuation v is non-negative; if not, returns the fraction p^v centerlift(x p^{-v}); in particular, rational reconstruction is not attempted. Use bestappr for this.

For backward compatibility, centerlift(x,'v) is allowed as an alias for lift(x,'v).

The library syntax is centerlift(GEN x).

characteristic(x)

Returns the characteristic of the base ring over which x is defined (as defined by t_INTMOD and t_FFELT components). The function raises an exception if incompatible primes arise from t_FFELT and t_PADIC components.

  ? characteristic(Mod(1,24)*x + Mod(1,18)*y)
  %1 = 6

The library syntax is GEN characteristic(GEN x).

component(x,n)

Extracts the n-th-component of x. This is to be understood as follows: every PARI type has one or two initial code words. The components are counted, starting at 1, after these code words. In particular if x is a vector, this is indeed the n-th-component of x, if x is a matrix, the n-th column, if x is a polynomial, the n-th coefficient (i.e. of degree n-1), and for power series, the n-th significant coefficient.

For polynomials and power series, one should rather use polcoeff, and for vectors and matrices, the [$] operator. Namely, if x is a vector, then x[n] represents the n-th component of x. If x is a matrix, x[m,n] represents the coefficient of row m and column n of the matrix, x[m,] represents the m-th row of x, and x[,n] represents the n-th column of x$.

Using of this function requires detailed knowledge of the structure of the different PARI types, and thus it should almost never be used directly. Some useful exceptions:

      ? x = 3 + O(3^5);
      ? component(x, 2)
      %2 = 81   \\ p^(p-adic accuracy)
      ? component(x, 1)
      %3 = 3    \\ p
      ? q = Qfb(1,2,3);
      ? component(q, 1)
      %5 = 1

The library syntax is GEN compo(GEN x, long n).

conj(x)

Conjugate of x. The meaning of this is clear, except that for real quadratic numbers, it means conjugation in the real quadratic field. This function has no effect on integers, reals, intmods, fractions or p-adics. The only forbidden type is polmod (see conjvec for this).

The library syntax is GEN gconj(GEN x).

conjvec(z)

Conjugate vector representation of z. If z is a polmod, equal to Mod(a,T), this gives a vector of length {degree}(T) containing:

@3* the complex embeddings of z if T has rational coefficients, i.e. the a(r[i]) where r = polroots(T);

@3* the conjugates of z if T has some intmod coefficients;

@3if z is a finite field element, the result is the vector of conjugates [z,z^p,z^{p^2},...,z^{p^{n-1}}] where n = {degree}(T).

@3If z is an integer or a rational number, the result is z. If z is a (row or column) vector, the result is a matrix whose columns are the conjugate vectors of the individual elements of z.

The library syntax is GEN conjvec(GEN z, long prec).

denominator(x)

Denominator of x. The meaning of this is clear when x is a rational number or function. If x is an integer or a polynomial, it is treated as a rational number or function, respectively, and the result is equal to 1. For polynomials, you probably want to use

  denominator( content(x) )

instead. As for modular objects, t_INTMOD and t_PADIC have denominator 1, and the denominator of a t_POLMOD is the denominator of its (minimal degree) polynomial representative.

If x is a recursive structure, for instance a vector or matrix, the lcm of the denominators of its components (a common denominator) is computed. This also applies for t_COMPLEXs and t_QUADs.

@3Warning. Multivariate objects are created according to variable priorities, with possibly surprising side effects (x/y is a polynomial, but y/x is a rational function). See Label se:priority.

The library syntax is GEN denom(GEN x).

digits(x,{b = 10})

Outputs the vector of the digits of |x| in base b, where x and b are integers.

The library syntax is GEN digits(GEN x, GEN b = NULL).

floor(x)

Floor of x. When x is in R, the result is the largest integer smaller than or equal to x. Applied to a rational function, floor(x) returns the Euclidean quotient of the numerator by the denominator.

The library syntax is GEN gfloor(GEN x).

frac(x)

Fractional part of x. Identical to x-{floor}(x). If x is real, the result is in [0,1[.

The library syntax is GEN gfrac(GEN x).

hammingweight(x)

If x is a t_INT, return the binary Hamming weight of |x|. Otherwise x must be of type t_POL, t_VEC, t_COL, t_VECSMALL, or t_MAT and the function returns the number of non-zero coefficients of x.

  ? hammingweight(15)
  %1 = 4
  ? hammingweight(x^100 + 2*x + 1)
  %2 = 3
  ? hammingweight([Mod(1,2), 2, Mod(0,3)])
  %3 = 2
  ? hammingweight(matid(100))
  %4 = 100

The library syntax is long hammingweight(GEN x).

imag(x)

Imaginary part of x. When x is a quadratic number, this is the coefficient of omega in the ``canonical'' integral basis (1,omega).

The library syntax is GEN gimag(GEN x).

length(x)

Length of x; #x is a shortcut for length(x). This is mostly useful for

@3* vectors: dimension (0 for empty vectors),

@3* lists: number of entries (0 for empty lists),

@3* matrices: number of columns,

@3* character strings: number of actual characters (without trailing \0, should you expect it from C char*).

   ? #"a string"
   %1 = 8
   ? #[3,2,1]
   %2 = 3
   ? #[]
   %3 = 0
   ? #matrix(2,5)
   %4 = 5
   ? L = List([1,2,3,4]); #L
   %5 = 4

The routine is in fact defined for arbitrary GP types, but is awkward and useless in other cases: it returns the number of non-code words in x, e.g. the effective length minus 2 for integers since the t_INT type has two code words.

The library syntax is long glength(GEN x).

lift(x,{v})

If v is omitted, lifts intmods from Z/nZ in Z, p-adics from Q_p to Q (as truncate), and polmods to polynomials. Otherwise, lifts only polmods whose modulus has main variable v. t_FFELT are not lifted, nor are List elements: you may convert the latter to vectors first, or use apply(lift,L). More generally, components for which such lifts are meaningless (e.g. character strings) are copied verbatim.

  ? lift(Mod(5,3))
  %1 = 2
  ? lift(3 + O(3^9))
  %2 = 3
  ? lift(Mod(x,x^2+1))
  %3 = x
  ? lift(Mod(x,x^2+1))
  %4 = x

Lifts are performed recursively on an object components, but only by one level: once a t_POLMOD is lifted, the components of the result are not lifted further.

  ? lift(x * Mod(1,3) + Mod(2,3))
  %4 = x + 2
  ? lift(x * Mod(y,y^2+1) + Mod(2,3))
  %5 = y*x + Mod(2, 3)   \\ do you understand this one?
  ? lift(x * Mod(y,y^2+1) + Mod(2,3), 'x)
  %6 = Mod(y, y^2 + 1)*x + Mod(Mod(2, 3), y^2 + 1)
  ? lift(%, y)
  %7 = y*x + Mod(2, 3)

@3To recursively lift all components not only by one level, but as long as possible, use liftall. To lift only t_INTMODs and t_PADICs components, use liftint. To lift only t_POLMODs components, use liftpol. Finally, centerlift allows to lift t_INTMODs and t_PADICs using centered residues (lift of smallest absolute value).

The library syntax is GEN lift0(GEN x, long v = -1), where v is a variable number. Also available is GEN lift(GEN x) corresponding to lift0(x,-1).

liftall(x)

Recursively lift all components of x from Z/nZ to Z, from Q_p to Q (as truncate), and polmods to polynomials. t_FFELT are not lifted, nor are List elements: you may convert the latter to vectors first, or use apply(liftall,L). More generally, components for which such lifts are meaningless (e.g. character strings) are copied verbatim.

  ? liftall(x * (1 + O(3)) + Mod(2,3))
  %1 = x + 2
  ? liftall(x * Mod(y,y^2+1) + Mod(2,3)*Mod(z,z^2))
  %2 = y*x + 2*z

The library syntax is GEN liftall(GEN x).

liftint(x)

Recursively lift all components of x from Z/nZ to Z and from Q_p to Q (as truncate). t_FFELT are not lifted, nor are List elements: you may convert the latter to vectors first, or use apply(liftint,L). More generally, components for which such lifts are meaningless (e.g. character strings) are copied verbatim.

  ? liftint(x * (1 + O(3)) + Mod(2,3))
  %1 = x + 2
  ? liftint(x * Mod(y,y^2+1) + Mod(2,3)*Mod(z,z^2))
  %2 = Mod(y, y^2 + 1)*x + Mod(Mod(2*z, z^2), y^2 + 1)

The library syntax is GEN liftint(GEN x).

liftpol(x)

Recursively lift all components of x which are polmods to polynomials. t_FFELT are not lifted, nor are List elements: you may convert the latter to vectors first, or use apply(liftpol,L). More generally, components for which such lifts are meaningless (e.g. character strings) are copied verbatim.

  ? liftpol(x * (1 + O(3)) + Mod(2,3))
  %1 = (1 + O(3))*x + Mod(2, 3)
  ? liftpol(x * Mod(y,y^2+1) + Mod(2,3)*Mod(z,z^2))
  %2 = y*x + Mod(2, 3)*z

The library syntax is GEN liftpol(GEN x).

norm(x)

Algebraic norm of x, i.e. the product of x with its conjugate (no square roots are taken), or conjugates for polmods. For vectors and matrices, the norm is taken componentwise and hence is not the L^2-norm (see norml2). Note that the norm of an element of R is its square, so as to be compatible with the complex norm.

The library syntax is GEN gnorm(GEN x).

numerator(x)

Numerator of x. The meaning of this is clear when x is a rational number or function. If x is an integer or a polynomial, it is treated as a rational number or function, respectively, and the result is x itself. For polynomials, you probably want to use

  numerator( content(x) )

instead.

In other cases, numerator(x) is defined to be denominator(x)*x. This is the case when x is a vector or a matrix, but also for t_COMPLEX or t_QUAD. In particular since a t_PADIC or t_INTMOD has denominator 1, its numerator is itself.

@3Warning. Multivariate objects are created according to variable priorities, with possibly surprising side effects (x/y is a polynomial, but y/x is a rational function). See Label se:priority.

The library syntax is GEN numer(GEN x).

numtoperm(n,k)

Generates the k-th permutation (as a row vector of length n) of the numbers 1 to n. The number k is taken modulo n!, i.e. inverse function of permtonum. The numbering used is the standard lexicographic ordering, starting at 0.

The library syntax is GEN numtoperm(long n, GEN k).

padicprec(x,p)

Absolute p-adic precision of the object x. This is the minimum precision of the components of x. The result is LONG_MAX (2^{31}-1 for 32-bit machines or 2^{63}-1 for 64-bit machines) if x is an exact object.

The library syntax is long padicprec(GEN x, GEN p).

permtonum(x)

Given a permutation x on n elements, gives the number k such that x = numtoperm(n,k), i.e. inverse function of numtoperm. The numbering used is the standard lexicographic ordering, starting at 0.

The library syntax is GEN permtonum(GEN x).

precision(x,{n})

The function has two different behaviors according to whether n is present or not.

If n is missing, the function returns the precision in decimal digits of the PARI object x. If x is an exact object, the largest single precision integer is returned.

  ? precision(exp(1e-100))
  %1 = 134                \\ 134 significant decimal digits
  ? precision(2 + x)
  %2 = 2147483647         \\ exact object
  ? precision(0.5 + O(x))
  %3 = 28                 \\ floating point accuracy, NOT series precision
  ? precision( [ exp(1e-100), 0.5 ] )
  %4 = 28                 \\ minimal accuracy among components

The return value for exact objects is meaningless since it is not even the same on 32 and 64-bit machines. The proper way to test whether an object is exact is

  ? isexact(x) = precision(x) == precision(0)

If n is present, the function creates a new object equal to x with a new ``precision'' n. (This never changes the type of the result. In particular it is not possible to use it to obtain a polynomial from a power series; for that, see truncate.) Now the meaning of precision is different from the above (floating point accuracy), and depends on the type of x:

For exact types, no change. For x a vector or a matrix, the operation is done componentwise.

For real x, n is the number of desired significant decimal digits. If n is smaller than the precision of x, x is truncated, otherwise x is extended with zeros.

For x a p-adic or a power series, n is the desired number of significant p-adic or X-adic digits, where X is the main variable of x. (Note: yes, this is inconsistent.) Note that the precision is a priori distinct from the exponent k appearing in O(*^k); it is indeed equal to k if and only if x is a p-adic or X-adic unit.

  ? precision(1 + O(x), 10)
  %1 = 1 + O(x^10)
  ? precision(x^2 + O(x^10), 3)
  %2 = x^2 + O(x^5)
  ? precision(7^2 + O(7^10), 3)
  %3 = 7^2 + O(7^5)

For the last two examples, note that x^2 + O(x^5) = x^2(1 + O(x^3)) indeed has 3 significant coefficients

The library syntax is GEN precision0(GEN x, long n). Also available are GEN gprec(GEN x, long n) and long precision(GEN x). In both, the accuracy is expressed in words (32-bit or 64-bit depending on the architecture).

\subsec{random({N = 2^{{31}}})} Returns a random element in various natural sets depending on the argument N.

@3* t_INT: returns an integer uniformly distributed between 0 and N-1. Omitting the argument is equivalent to random(2^31).

@3* t_REAL: returns a real number in [0,1[ with the same accuracy as N (whose mantissa has the same number of significant words).

@3* t_INTMOD: returns a random intmod for the same modulus.

@3* t_FFELT: returns a random element in the same finite field.

@3* t_VEC of length 2, N = [a,b]: returns an integer uniformly distributed between a and b.

@3* t_VEC generated by ellinit over a finite field k (coefficients are t_INTMODs modulo a prime or t_FFELTs): returns a ``random'' k-rational affine point on the curve. More precisely if the curve has a single point (at infinity!) we return it; otherwise we return an affine point by drawing an abscissa uniformly at random until ellordinate succeeds. Note that this is definitely not a uniform distribution over E(k), but it should be good enough for applications.

@3* t_POL return a random polynomial of degree at most the degree of N. The coefficients are drawn by applying random to the leading coefficient of N.

  ? random(10)
  %1 = 9
  ? random(Mod(0,7))
  %2 = Mod(1, 7)
  ? a = ffgen(ffinit(3,7), 'a); random(a)
  %3 = a^6 + 2*a^5 + a^4 + a^3 + a^2 + 2*a
  ? E = ellinit([3,7]*Mod(1,109)); random(E)
  %4 = [Mod(103, 109), Mod(10, 109)]
  ? E = ellinit([1,7]*a^0); random(E)
  %5 = [a^6 + a^5 + 2*a^4 + 2*a^2, 2*a^6 + 2*a^4 + 2*a^3 + a^2 + 2*a]
  ? random(Mod(1,7)*x^4)
  %6 = Mod(5, 7)*x^4 + Mod(6, 7)*x^3 + Mod(2, 7)*x^2 + Mod(2, 7)*x + Mod(5, 7)

These variants all depend on a single internal generator, and are independent from your operating system's random number generators. A random seed may be obtained via getrand, and reset using setrand: from a given seed, and given sequence of randoms, the exact same values will be generated. The same seed is used at each startup, reseed the generator yourself if this is a problem. Note that internal functions also call the random number generator; adding such a function call in the middle of your code will change the numbers produced.

@3Technical note. Up to version 2.4 included, the internal generator produced pseudo-random numbers by means of linear congruences, which were not well distributed in arithmetic progressions. We now use Brent's XORGEN algorithm, based on Feedback Shift Registers, see http://wwwmaths.anu.edu.au/~brent/random.html. The generator has period 2^{4096}-1, passes the Crush battery of statistical tests of L'Ecuyer and Simard, but is not suitable for cryptographic purposes: one can reconstruct the state vector from a small sample of consecutive values, thus predicting the entire sequence.

The library syntax is GEN genrand(GEN N = NULL).

Also available: GEN ellrandom(GEN E) and GEN ffrandom(GEN a).

real(x)

Real part of x. In the case where x is a quadratic number, this is the coefficient of 1 in the ``canonical'' integral basis (1,omega).

The library syntax is GEN greal(GEN x).

round(x,{&e})

If x is in R, rounds x to the nearest integer (rounding to + oo in case of ties), then and sets e to the number of error bits, that is the binary exponent of the difference between the original and the rounded value (the ``fractional part''). If the exponent of x is too large compared to its precision (i.e. e > 0), the result is undefined and an error occurs if e was not given.

@3Important remark. Contrary to the other truncation functions, this function operates on every coefficient at every level of a PARI object. For example

  {truncate}((2.4*X^2-1.7)/(X)) = 2.4*X,

whereas

  {round}((2.4*X^2-1.7)/(X)) = (2*X^2-2)/(X).

An important use of round is to get exact results after an approximate computation, when theory tells you that the coefficients must be integers.

The library syntax is GEN round0(GEN x, GEN *e = NULL). Also available are GEN grndtoi(GEN x, long *e) and GEN ground(GEN x).

simplify(x)

This function simplifies x as much as it can. Specifically, a complex or quadratic number whose imaginary part is the integer 0 (i.e. not Mod(0,2) or 0.E-28) is converted to its real part, and a polynomial of degree 0 is converted to its constant term. Simplifications occur recursively.

This function is especially useful before using arithmetic functions, which expect integer arguments:

  ? x = 2 + y - y
  %1 = 2
  ? isprime(x)
    ***   at top-level: isprime(x)
    ***                 ^----------
    *** isprime: not an integer argument in an arithmetic function
  ? type(x)
  %2 = "t_POL"
  ? type(simplify(x))
  %3 = "t_INT"

Note that GP results are simplified as above before they are stored in the history. (Unless you disable automatic simplification with \y, that is.) In particular

  ? type(%1)
  %4 = "t_INT"

The library syntax is GEN simplify(GEN x).

sizebyte(x)

Outputs the total number of bytes occupied by the tree representing the PARI object x.

The library syntax is long gsizebyte(GEN x). Also available is long gsizeword(GEN x) returning a number of words.

sizedigit(x)

Outputs a quick bound for the number of decimal digits of (the components of) x, off by at most 1. If you want the exact value, you can use #Str(x), which is slower.

The library syntax is long sizedigit(GEN x).

truncate(x,{&e})

Truncates x and sets e to the number of error bits. When x is in R, this means that the part after the decimal point is chopped away, e is the binary exponent of the difference between the original and the truncated value (the ``fractional part''). If the exponent of x is too large compared to its precision (i.e. e > 0), the result is undefined and an error occurs if e was not given. The function applies componentwise on vector / matrices; e is then the maximal number of error bits. If x is a rational function, the result is the ``integer part'' (Euclidean quotient of numerator by denominator) and e is not set.

Note a very special use of truncate: when applied to a power series, it transforms it into a polynomial or a rational function with denominator a power of X, by chopping away the O(X^k). Similarly, when applied to a p-adic number, it transforms it into an integer or a rational number by chopping away the O(p^k).

The library syntax is GEN trunc0(GEN x, GEN *e = NULL). The following functions are also available: GEN gtrunc(GEN x) and GEN gcvtoi(GEN x, long *e).

valuation(x,p)

Computes the highest exponent of p dividing x. If p is of type integer, x must be an integer, an intmod whose modulus is divisible by p, a fraction, a q-adic number with q = p, or a polynomial or power series in which case the valuation is the minimum of the valuation of the coefficients.

If p is of type polynomial, x must be of type polynomial or rational function, and also a power series if x is a monomial. Finally, the valuation of a vector, complex or quadratic number is the minimum of the component valuations.

If x = 0, the result is LONG_MAX (2^{31}-1 for 32-bit machines or 2^{63}-1 for 64-bit machines) if x is an exact object. If x is a p-adic numbers or power series, the result is the exponent of the zero. Any other type combinations gives an error.

The library syntax is long gvaluation(GEN x, GEN p).

variable({x})

Gives the main variable of the object x (the variable with the highest priority used in x), and p if x is a p-adic number. Return 0 if x has no variable associated to it.

  ? variable(x^2 + y)
  %1 = x
  ? variable(1 + O(5^2))
  %2 = 5
  ? variable([x,y,z,t])
  %3 = x
  ? variable(1)
  %4 = 0

@3The construction

     if (!variable(x),...)

@3can be used to test whether a variable is attached to x.

If x is omitted, returns the list of user variables known to the interpreter, by order of decreasing priority. (Highest priority is x, which always come first.)

The library syntax is GEN gpolvar(GEN x = NULL). However, in library mode, this function should not be used for x non-NULL, since gvar is more appropriate. Instead, for x a p-adic (type t_PADIC), p is gel(x,2); otherwise, use long gvar(GEN x) which returns the variable number of x if it exists, NO_VARIABLE otherwise, which satisfies the property varncmp(NO_VARIABLE, v) > 0 for all valid variable number v, i.e. it has lower priority than any variable.


Transcendental functions

Since the values of transcendental functions cannot be exactly represented, these functions will always return an inexact object: a real number, a complex number, a p-adic number or a power series. All these objects have a certain finite precision.

As a general rule, which of course in some cases may have exceptions, transcendental functions operate in the following way:

@3* If the argument is either a real number or an inexact complex number (like 1.0 + I or Pi*I but not 2 - 3*I), then the computation is done with the precision of the argument. In the example below, we see that changing the precision to 50 digits does not matter, because x only had a precision of 19 digits.

  ? \p 15
     realprecision = 19 significant digits (15 digits displayed)
  ? x = Pi/4
  %1 = 0.785398163397448
  ? \p 50
     realprecision = 57 significant digits (50 digits displayed)
  ? sin(x)
  %2 = 0.7071067811865475244

Note that even if the argument is real, the result may be complex (e.g. {acos}(2.0) or {acosh}(0.0)). See each individual function help for the definition of the branch cuts and choice of principal value.

@3* If the argument is either an integer, a rational, an exact complex number or a quadratic number, it is first converted to a real or complex number using the current precision held in the default realprecision. This precision (the number of decimal digits) can be changed using \p or default(realprecision,...)). After this conversion, the computation proceeds as above for real or complex arguments.

In library mode, the realprecision does not matter; instead the precision is taken from the prec parameter which every transcendental function has. As in gp, this prec is not used when the argument to a function is already inexact. Note that the argument prec stands for the length in words of a real number, including codewords. Hence we must have prec >= 3.

Some accuracies attainable on 32-bit machines cannot be attained on 64-bit machines for parity reasons. For example the default gp accuracy is 28 decimal digits on 32-bit machines, corresponding to prec having the value 5, but this cannot be attained on 64-bit machines.

@3* If the argument is a polmod (representing an algebraic number), then the function is evaluated for every possible complex embedding of that algebraic number. A column vector of results is returned, with one component for each complex embedding. Therefore, the number of components equals the degree of the t_POLMOD modulus.

@3* If the argument is an intmod or a p-adic, at present only a few functions like sqrt (square root), sqr (square), log, exp, powering, teichmuller (Teichmüller character) and agm (arithmetic-geometric mean) are implemented.

Note that in the case of a 2-adic number, sqr(x) may not be identical to x*x: for example if x = 1+O(2^5) and y = 1+O(2^5) then x*y = 1+O(2^5) while sqr(x) = 1+O(2^6). Here, x * x yields the same result as sqr(x) since the two operands are known to be identical. The same statement holds true for p-adics raised to the power n, where v_p(n) > 0.

@3Remark. If we wanted to be strictly consistent with the PARI philosophy, we should have x*y = (4 mod 8) and sqr(x) = (4 mod 32) when both x and y are congruent to 2 modulo 4. However, since intmod is an exact object, PARI assumes that the modulus must not change, and the result is hence (0 mod 4) in both cases. On the other hand, p-adics are not exact objects, hence are treated differently.

@3* If the argument is a polynomial, a power series or a rational function, it is, if necessary, first converted to a power series using the current series precision, held in the default seriesprecision. This precision (the number of significant terms) can be changed using \ps or default(seriesprecision,...). Then the Taylor series expansion of the function around X = 0 (where X is the main variable) is computed to a number of terms depending on the number of terms of the argument and the function being computed.

Under gp this again is transparent to the user. When programming in library mode, however, it is strongly advised to perform an explicit conversion to a power series first, as in x = gtoser(x, seriesprec), where the number of significant terms seriesprec can be specified explicitly. If you do not do this, a global variable precdl is used instead, to convert polynomials and rational functions to a power series with a reasonable number of terms; tampering with the value of this global variable is deprecated and strongly discouraged.

@3* If the argument is a vector or a matrix, the result is the componentwise evaluation of the function. In particular, transcendental functions on square matrices, which are not implemented in the present version 2.7.4, will have a different name if they are implemented some day.

\subseckbd{^} If y is not of type integer, x^y has the same effect as exp(y*log(x)). It can be applied to p-adic numbers as well as to the more usual types.

The library syntax is GEN gpow(GEN x, GEN n, long prec) for x^n.

Catalan

Catalan's constant G = sum_{n >= 0}((-1)^n)/((2n+1)^2) = 0.91596.... Note that Catalan is one of the few reserved names which cannot be used for user variables.

The library syntax is GEN mpcatalan(long prec).

Euler

Euler's constant gamma = 0.57721.... Note that Euler is one of the few reserved names which cannot be used for user variables.

The library syntax is GEN mpeuler(long prec).

I

The complex number sqrt {-1}.

The library syntax is GEN gen_I().

Pi

The constant Pi (3.14159...). Note that Pi is one of the few reserved names which cannot be used for user variables.

The library syntax is GEN mppi(long prec).

abs(x)

Absolute value of x (modulus if x is complex). Rational functions are not allowed. Contrary to most transcendental functions, an exact argument is not converted to a real number before applying abs and an exact result is returned if possible.

  ? abs(-1)
  %1 = 1
  ? abs(3/7 + 4/7*I)
  %2 = 5/7
  ? abs(1 + I)
  %3 = 1.414213562373095048801688724

If x is a polynomial, returns -x if the leading coefficient is real and negative else returns x. For a power series, the constant coefficient is considered instead.

The library syntax is GEN gabs(GEN x, long prec).

acos(x)

Principal branch of {cos}^{-1}(x) = -i log (x + i sqrt {1-x^2}). In particular, {Re(acos}(x)) belongs to [0,Pi] and if x belongs to R and |x| > 1, then {acos}(x) is complex. The branch cut is in two pieces: ]- oo ,-1] , continuous with quadrant II, and [1,+ oo [, continuous with quadrant IV. We have {acos}(x) = Pi/2 - {asin}(x) for all x.

The library syntax is GEN gacos(GEN x, long prec).

acosh(x)

Principal branch of {cosh}^{-1}(x) = 2 log ( sqrt {(x+1)/2} + sqrt {(x-1)/2}). In particular, {Re}({acosh}(x)) >= 0 and {In}({acosh}(x)) belongs to ]-Pi,Pi]0; if x belongs to R and x < 1, then {acosh}(x) is complex.

The library syntax is GEN gacosh(GEN x, long prec).

agm(x,y)

Arithmetic-geometric mean of x and y. In the case of complex or negative numbers, the optimal AGM is returned (the largest in absolute value over all choices of the signs of the square roots). p-adic or power series arguments are also allowed. Note that a p-adic agm exists only if x/y is congruent to 1 modulo p (modulo 16 for p = 2). x and y cannot both be vectors or matrices.

The library syntax is GEN agm(GEN x, GEN y, long prec).

arg(x)

Argument of the complex number x, such that -Pi < {arg}(x) <= Pi.

The library syntax is GEN garg(GEN x, long prec).

asin(x)

Principal branch of {sin}^{-1}(x) = -i log (ix + sqrt {1 - x^2}). In particular, {Re(asin}(x)) belongs to [-Pi/2,Pi/2] and if x belongs to R and |x| > 1 then {asin}(x) is complex. The branch cut is in two pieces: ]- oo ,-1], continuous with quadrant II, and [1,+ oo [ continuous with quadrant IV. The function satisfies i {asin}(x) = {asinh}(ix).

The library syntax is GEN gasin(GEN x, long prec).

asinh(x)

Principal branch of {sinh}^{-1}(x) = log (x + sqrt {1+x^2}). In particular {Im(asinh}(x)) belongs to [-Pi/2,Pi/2]. The branch cut is in two pieces: [-i oo ,-i], continuous with quadrant III and [i,+i oo [ continuous with quadrant I.

The library syntax is GEN gasinh(GEN x, long prec).

atan(x)

Principal branch of {tan}^{-1}(x) = log ((1+ix)/(1-ix)) / 2i. In particular the real part of {atan}(x)) belongs to ]-Pi/2,Pi/2[. The branch cut is in two pieces: ]-i oo ,-i[, continuous with quadrant IV, and ]i,+i oo [ continuous with quadrant II. The function satisfies i {atan}(x) = -i{atanh}(ix) for all x != +- i.

The library syntax is GEN gatan(GEN x, long prec).

atanh(x)

Principal branch of {tanh}^{-1}(x) = log ((1+x)/(1-x)) / 2. In particular the imaginary part of {atanh}(x) belongs to [-Pi/2,Pi/2]; if x belongs to R and |x| > 1 then {atanh}(x) is complex.

The library syntax is GEN gatanh(GEN x, long prec).

bernfrac(x)

Bernoulli number B_x, where B_0 = 1, B_1 = -1/2, B_2 = 1/6,..., expressed as a rational number. The argument x should be of type integer.

The library syntax is GEN bernfrac(long x).

bernpol(n, {v = 'x})

Bernoulli polynomial B_n in variable v.

  ? bernpol(1)
  %1 = x - 1/2
  ? bernpol(3)
  %2 = x^3 - 3/2*x^2 + 1/2*x

The library syntax is GEN bernpol(long n, long v = -1), where v is a variable number.

bernreal(x)

Bernoulli number B_x, as bernfrac, but B_x is returned as a real number (with the current precision).

The library syntax is GEN bernreal(long x, long prec).

bernvec(x)

Creates a vector containing, as rational numbers, the Bernoulli numbers B_0, B_2,..., B_{2x}. This routine is obsolete. Use bernfrac instead each time you need a Bernoulli number in exact form.

@3Note. This routine is implemented using repeated independent calls to bernfrac, which is faster than the standard recursion in exact arithmetic. It is only kept for backward compatibility: it is not faster than individual calls to bernfrac, its output uses a lot of memory space, and coping with the index shift is awkward.

The library syntax is GEN bernvec(long x).

besselh1(nu,x)

H^1-Bessel function of index nu and argument x.

The library syntax is GEN hbessel1(GEN nu, GEN x, long prec).

besselh2(nu,x)

H^2-Bessel function of index nu and argument x.

The library syntax is GEN hbessel2(GEN nu, GEN x, long prec).

besseli(nu,x)

I-Bessel function of index nu and argument x. If x converts to a power series, the initial factor (x/2)^nu/Gamma(nu+1) is omitted (since it cannot be represented in PARI when nu is not integral).

The library syntax is GEN ibessel(GEN nu, GEN x, long prec).

besselj(nu,x)

J-Bessel function of index nu and argument x. If x converts to a power series, the initial factor (x/2)^nu/Gamma(nu+1) is omitted (since it cannot be represented in PARI when nu is not integral).

The library syntax is GEN jbessel(GEN nu, GEN x, long prec).

besseljh(n,x)

J-Bessel function of half integral index. More precisely, besseljh(n,x) computes J_{n+1/2}(x) where n must be of type integer, and x is any element of C. In the present version 2.7.4, this function is not very accurate when x is small.

The library syntax is GEN jbesselh(GEN n, GEN x, long prec).

besselk(nu,x)

K-Bessel function of index nu and argument x.

The library syntax is GEN kbessel(GEN nu, GEN x, long prec).

besseln(nu,x)

N-Bessel function of index nu and argument x.

The library syntax is GEN nbessel(GEN nu, GEN x, long prec).

cos(x)

Cosine of x.

The library syntax is GEN gcos(GEN x, long prec).

cosh(x)

Hyperbolic cosine of x.

The library syntax is GEN gcosh(GEN x, long prec).

cotan(x)

Cotangent of x.

The library syntax is GEN gcotan(GEN x, long prec).

dilog(x)

Principal branch of the dilogarithm of x, i.e. analytic continuation of the power series log _2(x) = sum_{n >= 1}x^n/n^2.

The library syntax is GEN dilog(GEN x, long prec).

eint1(x,{n})

Exponential integral int_x^ oo (e^{-t})/(t)dt = incgam(0, x), where the latter expression extends the function definition from real x > 0 to all complex x != 0.

If n is present, we must have x > 0; the function returns the n-dimensional vector [eint1(x),...,eint1(nx)]. Contrary to other transcendental functions, and to the default case (n omitted), the values are correct up to a bounded absolute, rather than relative, error 10^-n, where n is precision(x) if x is a t_REAL and defaults to realprecision otherwise. (In the most important application, to the computation of L-functions via approximate functional equations, those values appear as weights in long sums and small individual relative errors are less useful than controlling the absolute error.) This is faster than repeatedly calling eint1(i * x), but less precise.

The library syntax is GEN veceint1(GEN x, GEN n = NULL, long prec). Also available is GEN eint1(GEN x, long prec).

erfc(x)

Complementary error function, analytic continuation of (2/ sqrt Pi)int_x^ oo e^{-t^2}dt = incgam(1/2,x^2)/ sqrt Pi, where the latter expression extends the function definition from real x to all complex x != 0.

The library syntax is GEN gerfc(GEN x, long prec).

eta(z,{flag = 0})

Variants of Dedekind's eta function. If flag = 0, return prod_{n = 1}^ oo (1-q^n), where q depends on x in the following way:

@3* q = e^{2iPi x} if x is a complex number (which must then have positive imaginary part); notice that the factor q^{1/24} is missing!

@3* q = x if x is a t_PADIC, or can be converted to a power series (which must then have positive valuation).

If flag is non-zero, x is converted to a complex number and we return the true eta function, q^{1/24}prod_{n = 1}^ oo (1-q^n), where q = e^{2iPi x}.

The library syntax is GEN eta0(GEN z, long flag, long prec).

Also available is GEN trueeta(GEN x, long prec) (flag = 1).

exp(x)

Exponential of x. p-adic arguments with positive valuation are accepted.

The library syntax is GEN gexp(GEN x, long prec). For a t_PADIC x, the function GEN Qp_exp(GEN x) is also available.

expm1(x)

Return exp (x)-1, computed in a way that is also accurate when the real part of x is near 0. Only accept real or complex arguments. A naive direct computation would suffer from catastrophic cancellation; PARI's direct computation of exp (x) alleviates this well known problem at the expense of computing exp (x) to a higher accuracy when x is small. Using expm1 is recommended instead:

  ? default(realprecision, 10000); x = 1e-100;
  ? a = expm1(x);
  time = 4 ms.
  ? b = exp(x)-1;
  time = 28 ms.
  ? default(realprecision, 10040); x = 1e-100;
  ? c = expm1(x);  \\ reference point
  ? abs(a-c)/c  \\ relative error in expm1(x)
  %7 = 0.E-10017
  ? abs(b-c)/c  \\ relative error in exp(x)-1
  %8 = 1.7907031188259675794 E-9919

@3As the example above shows, when x is near 0, expm1 is both faster and more accurate than exp(x)-1.

The library syntax is GEN gexpm1(GEN x, long prec).

gamma(s)

For s a complex number, evaluates Euler's gamma function

  Gamma(s) = int_0^ oo t^{s-1} exp (-t)dt.

Error if s is a non-positive integer, where Gamma has a pole.

For s a t_PADIC, evaluates the Morita gamma function at s, that is the unique continuous p-adic function on the p-adic integers extending Gamma_p(k) = (-1)^k prod_{j < k}'j, where the prime means that p does not divide j.

  ? gamma(1/4 + O(5^10))
  %1= 1 + 4*5 + 3*5^4 + 5^6 + 5^7 + 4*5^9 + O(5^10)
  ? algdep(%,4)
  %2 = x^4 + 4*x^2 + 5

The library syntax is GEN ggamma(GEN s, long prec). For a t_PADIC x, the function GEN Qp_gamma(GEN x) is also available.

gammah(x)

Gamma function evaluated at the argument x+1/2.

The library syntax is GEN ggammah(GEN x, long prec).

hyperu(a,b,x)

U-confluent hypergeometric function with parameters a and b. The parameters a and b can be complex but the present implementation requires x to be positive.

The library syntax is GEN hyperu(GEN a, GEN b, GEN x, long prec).

incgam(s,x,{g})

Incomplete gamma function int_x^ oo e^{-t}t^{s-1}dt, extended by analytic continuation to all complex x, s not both 0. The relative error is bounded in terms of the precision of s (the accuracy of x is ignored when determining the output precision). When g is given, assume that g = Gamma(s). For small |x|, this will speed up the computation.

The library syntax is GEN incgam0(GEN s, GEN x, GEN g = NULL, long prec). Also available is GEN incgam(GEN s, GEN x, long prec).

incgamc(s,x)

Complementary incomplete gamma function. The arguments x and s are complex numbers such that s is not a pole of Gamma and |x|/(|s|+1) is not much larger than 1 (otherwise the convergence is very slow). The result returned is int_0^x e^{-t}t^{s-1}dt.

The library syntax is GEN incgamc(GEN s, GEN x, long prec).

lambertw(y)

Lambert W function, solution of the implicit equation xe^x = y, for y > 0.

The library syntax is GEN glambertW(GEN y, long prec).

lngamma(x)

Principal branch of the logarithm of the gamma function of x. This function is analytic on the complex plane with non-positive integers removed, and can have much larger arguments than gamma itself.

For x a power series such that x(0) is not a pole of gamma, compute the Taylor expansion. (PARI only knows about regular power series and can't include logarithmic terms.)

  ? lngamma(1+x+O(x^2))
  %1 = -0.57721566490153286060651209008240243104*x + O(x^2)
  ? lngamma(x+O(x^2))
   ***   at top-level: lngamma(x+O(x^2))
   ***                 ^-----------------
   *** lngamma: domain error in lngamma: valuation != 0
  ? lngamma(-1+x+O(x^2))
   *** lngamma: Warning: normalizing a series with 0 leading term.
   ***   at top-level: lngamma(-1+x+O(x^2))
   ***                 ^--------------------
   *** lngamma: domain error in intformal: residue(series, pole) != 0

The library syntax is GEN glngamma(GEN x, long prec).

log(x)

Principal branch of the natural logarithm of x belongs to C^*, i.e. such that {Im(log}(x)) belongs to ]-Pi,Pi]. The branch cut lies along the negative real axis, continuous with quadrant 2, i.e. such that lim _{b\to 0^+} log (a+bi) = log a for a belongs to R^*. The result is complex (with imaginary part equal to Pi) if x belongs to R and x < 0. In general, the algorithm uses the formula

   log (x) ~ (Pi)/(2{agm}(1, 4/s)) - m log 2,

if s = x 2^m is large enough. (The result is exact to B bits provided s > 2^{B/2}.) At low accuracies, the series expansion near 1 is used.

p-adic arguments are also accepted for x, with the convention that log (p) = 0. Hence in particular exp ( log (x))/x is not in general equal to 1 but to a (p-1)-th root of unity (or +-1 if p = 2) times a power of p.

The library syntax is GEN glog(GEN x, long prec). For a t_PADIC x, the function GEN Qp_log(GEN x) is also available.

polylog(m,x,{flag = 0})

One of the different polylogarithms, depending on flag:

If flag = 0 or is omitted: m-th polylogarithm of x, i.e. analytic continuation of the power series {Li}_m(x) = sum_{n >= 1}x^n/n^m (x < 1). Uses the functional equation linking the values at x and 1/x to restrict to the case |x| <= 1, then the power series when |x|^2 <= 1/2, and the power series expansion in log (x) otherwise.

Using flag, computes a modified m-th polylogarithm of x. We use Zagier's notations; let Re _m denote Re or Im depending on whether m is odd or even:

If flag = 1: compute ~ D_m(x), defined for |x| <= 1 by

   Re _m(sum_{k = 0}^{m-1} ((- log |x|)^k)/(k!){Li}_{m-k}(x) +((- log |x|)^{m-1})/(m!) log |1-x|).

If flag = 2: compute D_m(x), defined for |x| <= 1 by

   Re _m(sum_{k = 0}^{m-1}((- log |x|)^k)/(k!){Li}_{m-k}(x) -(1)/(2)((- log |x|)^m)/(m!)).

If flag = 3: compute P_m(x), defined for |x| <= 1 by

   Re _m(sum_{k = 0}^{m-1}(2^kB_k)/(k!)( log |x|)^k{Li}_{m-k}(x) -(2^{m-1}B_m)/(m!)( log |x|)^m).

These three functions satisfy the functional equation f_m(1/x) = (-1)^{m-1}f_m(x).

The library syntax is GEN polylog0(long m, GEN x, long flag, long prec). Also available is GEN gpolylog(long m, GEN x, long prec) (flag = 0).

psi(x)

The psi-function of x, i.e. the logarithmic derivative Gamma'(x)/Gamma(x).

The library syntax is GEN gpsi(GEN x, long prec).

sin(x)

Sine of x.

The library syntax is GEN gsin(GEN x, long prec).

sinh(x)

Hyperbolic sine of x.

The library syntax is GEN gsinh(GEN x, long prec).

sqr(x)

Square of x. This operation is not completely straightforward, i.e. identical to x * x, since it can usually be computed more efficiently (roughly one-half of the elementary multiplications can be saved). Also, squaring a 2-adic number increases its precision. For example,

  ? (1 + O(2^4))^2
  %1 = 1 + O(2^5)
  ? (1 + O(2^4)) * (1 + O(2^4))
  %2 = 1 + O(2^4)

Note that this function is also called whenever one multiplies two objects which are known to be identical, e.g. they are the value of the same variable, or we are computing a power.

  ? x = (1 + O(2^4)); x * x
  %3 = 1 + O(2^5)
  ? (1 + O(2^4))^4
  %4 = 1 + O(2^6)

(note the difference between %2 and %3 above).

The library syntax is GEN gsqr(GEN x).

sqrt(x)

Principal branch of the square root of x, defined as sqrt {x} = exp ( log x / 2). In particular, we have {Arg}({sqrt}(x)) belongs to ]-Pi/2, Pi/2], and if x belongs to R and x < 0, then the result is complex with positive imaginary part.

Intmod a prime p, t_PADIC and t_FFELT are allowed as arguments. In the first 2 cases (t_INTMOD, t_PADIC), the square root (if it exists) which is returned is the one whose first p-adic digit is in the interval [0,p/2]. For other arguments, the result is undefined.

The library syntax is GEN gsqrt(GEN x, long prec). For a t_PADIC x, the function GEN Qp_sqrt(GEN x) is also available.

sqrtn(x,n,{&z})

Principal branch of the nth root of x, i.e. such that {Arg}({sqrt}(x)) belongs to ]-Pi/n, Pi/n]. Intmod a prime and p-adics are allowed as arguments.

If z is present, it is set to a suitable root of unity allowing to recover all the other roots. If it was not possible, z is set to zero. In the case this argument is present and no square root exist, 0 is returned instead or raising an error.

  ? sqrtn(Mod(2,7), 2)
  %1 = Mod(4, 7)
  ? sqrtn(Mod(2,7), 2, &z); z
  %2 = Mod(6, 7)
  ? sqrtn(Mod(2,7), 3)
    ***   at top-level: sqrtn(Mod(2,7),3)
    ***                 ^-----------------
    *** sqrtn: nth-root does not exist in gsqrtn.
  ? sqrtn(Mod(2,7), 3,  &z)
  %2 = 0
  ? z
  %3 = 0

The following script computes all roots in all possible cases:

  sqrtnall(x,n)=
  { my(V,r,z,r2);
    r = sqrtn(x,n, &z);
    if (!z, error("Impossible case in sqrtn"));
    if (type(x) == "t_INTMOD" || type(x)=="t_PADIC",
      r2 = r*z; n = 1;
      while (r2!=r, r2*=z;n++));
    V = vector(n); V[1] = r;
    for(i=2, n, V[i] = V[i-1]*z);
    V
  }
  addhelp(sqrtnall,"sqrtnall(x,n):compute the vector of nth-roots of x");

The library syntax is GEN gsqrtn(GEN x, GEN n, GEN *z = NULL, long prec). If x is a t_PADIC, the function GEN Qp_sqrt(GEN x, GEN n, GEN *z) is also available.

tan(x)

Tangent of x.

The library syntax is GEN gtan(GEN x, long prec).

tanh(x)

Hyperbolic tangent of x.

The library syntax is GEN gtanh(GEN x, long prec).

teichmuller(x)

Teichmüller character of the p-adic number x, i.e. the unique (p-1)-th root of unity congruent to x / p^{v_p(x)} modulo p.

The library syntax is GEN teich(GEN x).

theta(q,z)

Jacobi sine theta-function

   theta_1(z, q) = 2q^{1/4} sum_{n >= 0} (-1)^n q^{n(n+1)} sin ((2n+1)z).

The library syntax is GEN theta(GEN q, GEN z, long prec).

thetanullk(q,k)

k-th derivative at z = 0 of theta(q,z).

The library syntax is GEN thetanullk(GEN q, long k, long prec).

GEN vecthetanullk(GEN q, long k, long prec) returns the vector of all (d^itheta)/(dz^i)(q,0) for all odd i = 1, 3,..., 2k-1. GEN vecthetanullk_tau(GEN tau, long k, long prec) returns vecthetanullk_tau at q = exp (2iPi tau).

weber(x,{flag = 0})

One of Weber's three f functions. If flag = 0, returns

  f(x) = exp (-iPi/24).eta((x+1)/2)/eta(x) such that j = (f^{24}-16)^3/f^{24},

where j is the elliptic j-invariant (see the function ellj). If flag = 1, returns

  f_1(x) = eta(x/2)/eta(x) such that j = (f_1^{24}+16)^3/f_1^{24}.

Finally, if flag = 2, returns

  f_2(x) = sqrt {2}eta(2x)/eta(x) such that j = (f_2^{24}+16)^3/f_2^{24}.

Note the identities f^8 = f_1^8+f_2^8 and ff_1f_2 = sqrt 2.

The library syntax is GEN weber0(GEN x, long flag, long prec). Also available are GEN weberf(GEN x, long prec), GEN weberf1(GEN x, long prec) and GEN weberf2(GEN x, long prec).

zeta(s)

For s a complex number, Riemann's zeta function zeta(s) = sum_{n >= 1}n^{-s}, computed using the Euler-Maclaurin summation formula, except when s is of type integer, in which case it is computed using Bernoulli numbers for s <= 0 or s > 0 and even, and using modular forms for s > 0 and odd.

For s a p-adic number, Kubota-Leopoldt zeta function at s, that is the unique continuous p-adic function on the p-adic integers that interpolates the values of (1 - p^{-k}) zeta(k) at negative integers k such that k = 1 (mod p-1) (resp. k is odd) if p is odd (resp. p = 2).

The library syntax is GEN gzeta(GEN s, long prec).


Arithmetic functions

These functions are by definition functions whose natural domain of definition is either Z (or Z_{ > 0}). The way these functions are used is completely different from transcendental functions in that there are no automatic type conversions: in general only integers are accepted as arguments. An integer argument N can be given in the following alternate formats:

@3* t_MAT: its factorization fa = factor(N),

@3* t_VEC: a pair [N, fa] giving both the integer and its factorization.

This allows to compute different arithmetic functions at a given N while factoring the latter only once.

    ? N = 10!; faN = factor(N);
    ? eulerphi(N)
    %2 = 829440
    ? eulerphi(faN)
    %3 = 829440
    ? eulerphi(S = [N, faN])
    %4 = 829440
    ? sigma(S)
    %5 = 15334088

Arithmetic functions and the factoring engine

All arithmetic functions in the narrow sense of the word --- Euler's totient function, the Moebius function, the sums over divisors or powers of divisors etc.--- call, after trial division by small primes, the same versatile factoring machinery described under factorint. It includes Shanks SQUFOF, Pollard Rho, ECM and MPQS stages, and has an early exit option for the functions moebius and (the integer function underlying) issquarefree. This machinery relies on a fairly strong probabilistic primality test, see ispseudoprime, but you may also set

    default(factor_proven, 1)

@3to ensure that all tentative factorizations are fully proven. This should not slow down PARI too much, unless prime numbers with hundreds of decimal digits occur frequently in your application.

Orders in finite groups and Discrete Logarithm functions

The following functions compute the order of an element in a finite group: ellorder (the rational points on an elliptic curve defined over a finite field), fforder (the multiplicative group of a finite field), znorder (the invertible elements in Z/nZ). The following functions compute discrete logarithms in the same groups (whenever this is meaningful) elllog, fflog, znlog.

All such functions allow an optional argument specifying an integer N, representing the order of the group. (The order functions also allows any non-zero multiple of the order, with a minor loss of efficiency.) That optional argument follows the same format as given above:

@3* t_INT: the integer N,

@3* t_MAT: the factorization fa = factor(N),

@3* t_VEC: this is the preferred format and provides both the integer N and its factorization in a two-component vector [N, fa].

When the group is fixed and many orders or discrete logarithms will be computed, it is much more efficient to initialize this data once and for all and pass it to the relevant functions, as in

  ? p = nextprime(10^40);
  ? v = [p-1, factor(p-1)]; \\ data for discrete log & order computations
  ? znorder(Mod(2,p), v)
  %3 = 500000000000000000000000000028
  ? g = znprimroot(p);
  ? znlog(2, g, v)
  %5 = 543038070904014908801878611374

addprimes({x = []})

Adds the integers contained in the vector x (or the single integer x) to a special table of ``user-defined primes'', and returns that table. Whenever factor is subsequently called, it will trial divide by the elements in this table. If x is empty or omitted, just returns the current list of extra primes.

The entries in x must be primes: there is no internal check, even if the factor_proven default is set. To remove primes from the list use removeprimes.

The library syntax is GEN addprimes(GEN x = NULL).

bestappr(x, {B})

Using variants of the extended Euclidean algorithm, returns a rational approximation a/b to x, whose denominator is limited by B, if present. If B is omitted, return the best approximation affordable given the input accuracy; if you are looking for true rational numbers, presumably approximated to sufficient accuracy, you should first try that option. Otherwise, B must be a positive real scalar (impose 0 < b <= B).

@3* If x is a t_REAL or a t_FRAC, this function uses continued fractions.

  ? bestappr(Pi, 100)
  %1 = 22/7
  ? bestappr(0.1428571428571428571428571429)
  %2 = 1/7
  ? bestappr([Pi, sqrt(2) + 'x], 10^3)
  %3 = [355/113, x + 1393/985]

By definition, a/b is the best rational approximation to x if |b x - a| < |v x - u| for all integers (u,v) with 0 < v <= B. (Which implies that n/d is a convergent of the continued fraction of x.)

@3* If x is a t_INTMOD modulo N or a t_PADIC of precision N = p^k, this function performs rational modular reconstruction modulo N. The routine then returns the unique rational number a/b in coprime integers |a| < N/2B and b <= B which is congruent to x modulo N. Omitting B amounts to choosing it of the order of sqrt {N/2}. If rational reconstruction is not possible (no suitable a/b exists), returns [].

  ? bestappr(Mod(18526731858, 11^10))
  %1 = 1/7
  ? bestappr(Mod(18526731858, 11^20))
  %2 = []
  ? bestappr(3 + 5 + 3*5^2 + 5^3 + 3*5^4 + 5^5 + 3*5^6 + O(5^7))
  %2 = -1/3

@3In most concrete uses, B is a prime power and we performed Hensel lifting to obtain x.

The function applies recursively to components of complex objects (polynomials, vectors,...). If rational reconstruction fails for even a single entry, return [].

The library syntax is GEN bestappr(GEN x, GEN B = NULL).

bestapprPade(x, {B})

Using variants of the extended Euclidean algorithm, returns a rational function approximation a/b to x, whose denominator is limited by B, if present. If B is omitted, return the best approximation affordable given the input accuracy; if you are looking for true rational functions, presumably approximated to sufficient accuracy, you should first try that option. Otherwise, B must be a non-negative real (impose 0 <= {degree}(b) <= B).

@3* If x is a t_RFRAC or t_SER, this function uses continued fractions.

  ? bestapprPade((1-x^11)/(1-x)+O(x^11))
  %1 = 1/(-x + 1)
  ? bestapprPade([1/(1+x+O(x^10)), (x^3-2)/(x^3+1)], 1)
  %2 =  [1/(x + 1), -2]

@3* If x is a t_POLMOD modulo N or a t_SER of precision N = t^k, this function performs rational modular reconstruction modulo N. The routine then returns the unique rational function a/b in coprime polynomials, with {degree}(b) <= B which is congruent to x modulo N. Omitting B amounts to choosing it of the order of N/2. If rational reconstruction is not possible (no suitable a/b exists), returns [].

  ? bestapprPade(Mod(1+x+x^2+x^3+x^4, x^4-2))
  %1 = (2*x - 1)/(x - 1)
  ? % * Mod(1,x^4-2)
  %2 = Mod(x^3 + x^2 + x + 3, x^4 - 2)
  ? bestapprPade(Mod(1+x+x^2+x^3+x^5, x^9))
  %2 = []
  ? bestapprPade(Mod(1+x+x^2+x^3+x^5, x^10))
  %3 = (2*x^4 + x^3 - x - 1)/(-x^5 + x^3 + x^2 - 1)

The function applies recursively to components of complex objects (polynomials, vectors,...). If rational reconstruction fails for even a single entry, return [].

The library syntax is GEN bestapprPade(GEN x, long B).

bezout(x,y)

Deprecated alias for gcdext

The library syntax is GEN gcdext0(GEN x, GEN y).

bigomega(x)

Number of prime divisors of the integer |x| counted with multiplicity:

  ? factor(392)
  %1 =
  [2 3]
  [7 2]
  ? bigomega(392)
  %2 = 5;  \\ = 3+2
  ? omega(392)
  %3 = 2;  \\ without multiplicity

The library syntax is long bigomega(GEN x).

binomial(x,y)

binomial coefficient binom{x}{y}. Here y must be an integer, but x can be any PARI object.

The library syntax is GEN binomial(GEN x, long y). The function GEN binomialuu(ulong n, ulong k) is also available, and so is GEN vecbinome(long n), which returns a vector v with n+1 components such that v[k+1] = binomial(n,k) for k from 0 up to n.

chinese(x,{y})

If x and y are both intmods or both polmods, creates (with the same type) a z in the same residue class as x and in the same residue class as y, if it is possible.

  ? chinese(Mod(1,2), Mod(2,3))
  %1 = Mod(5, 6)
  ? chinese(Mod(x,x^2-1), Mod(x+1,x^2+1))
  %2 = Mod(-1/2*x^2 + x + 1/2, x^4 - 1)

This function also allows vector and matrix arguments, in which case the operation is recursively applied to each component of the vector or matrix.

  ? chinese([Mod(1,2),Mod(1,3)], [Mod(1,5),Mod(2,7)])
  %3 = [Mod(1, 10), Mod(16, 21)]

For polynomial arguments in the same variable, the function is applied to each coefficient; if the polynomials have different degrees, the high degree terms are copied verbatim in the result, as if the missing high degree terms in the polynomial of lowest degree had been Mod(0,1). Since the latter behavior is usually not the desired one, we propose to convert the polynomials to vectors of the same length first:

   ? P = x+1; Q = x^2+2*x+1;
   ? chinese(P*Mod(1,2), Q*Mod(1,3))
   %4 = Mod(1, 3)*x^2 + Mod(5, 6)*x + Mod(3, 6)
   ? chinese(Vec(P,3)*Mod(1,2), Vec(Q,3)*Mod(1,3))
   %5 = [Mod(1, 6), Mod(5, 6), Mod(4, 6)]
   ? Pol(%)
   %6 = Mod(1, 6)*x^2 + Mod(5, 6)*x + Mod(4, 6)

If y is omitted, and x is a vector, chinese is applied recursively to the components of x, yielding a residue belonging to the same class as all components of x.

Finally chinese(x,x) = x regardless of the type of x; this allows vector arguments to contain other data, so long as they are identical in both vectors.

The library syntax is GEN chinese(GEN x, GEN y = NULL). GEN chinese1(GEN x) is also available.

content(x)

Computes the gcd of all the coefficients of x, when this gcd makes sense. This is the natural definition if x is a polynomial (and by extension a power series) or a vector/matrix. This is in general a weaker notion than the ideal generated by the coefficients:

  ? content(2*x+y)
  %1 = 1            \\ = gcd(2,y) over Q[y]

If x is a scalar, this simply returns the absolute value of x if x is rational (t_INT or t_FRAC), and either 1 (inexact input) or x (exact input) otherwise; the result should be identical to gcd(x, 0).

The content of a rational function is the ratio of the contents of the numerator and the denominator. In recursive structures, if a matrix or vector coefficient x appears, the gcd is taken not with x, but with its content:

  ? content([ [2], 4*matid(3) ])
  %1 = 2

The library syntax is GEN content(GEN x).

contfrac(x,{b},{nmax})

Returns the row vector whose components are the partial quotients of the continued fraction expansion of x. In other words, a result [a_0,...,a_n] means that x ~ a_0+1/(a_1+...+1/a_n). The output is normalized so that a_n != 1 (unless we also have n = 0).

The number of partial quotients n+1 is limited by nmax. If nmax is omitted, the expansion stops at the last significant partial quotient.

  ? \p19
    realprecision = 19 significant digits
  ? contfrac(Pi)
  %1 = [3, 7, 15, 1, 292, 1, 1, 1, 2, 1, 3, 1, 14, 2, 1, 1, 2, 2]
  ? contfrac(Pi,, 3)  \\ n = 2
  %2 = [3, 7, 15]

x can also be a rational function or a power series.

If a vector b is supplied, the numerators are equal to the coefficients of b, instead of all equal to 1 as above; more precisely, x ~ (1/b_0)(a_0+b_1/(a_1+...+b_n/a_n)); for a numerical continued fraction (x real), the a_i are integers, as large as possible; if x is a rational function, they are polynomials with deg a_i = deg b_i + 1. The length of the result is then equal to the length of b, unless the next partial quotient cannot be reliably computed, in which case the expansion stops. This happens when a partial remainder is equal to zero (or too small compared to the available significant digits for x a t_REAL).

A direct implementation of the numerical continued fraction contfrac(x,b) described above would be

  \\ "greedy" generalized continued fraction
  cf(x, b) =
  { my( a= vector(#b), t );
    x *= b[1];
    for (i = 1, #b,
      a[i] = floor(x);
      t = x - a[i]; if (!t || i == #b, break);
      x = b[i+1] / t;
    ); a;
  }

@3There is some degree of freedom when choosing the a_i; the program above can easily be modified to derive variants of the standard algorithm. In the same vein, although no builtin function implements the related Engel expansion (a special kind of Egyptian fraction decomposition: x = 1/a_1 + 1/(a_1a_2) +... ), it can be obtained as follows:

  \\ n terms of the Engel expansion of x
  engel(x, n = 10) =
  { my( u = x, a = vector(n) );
    for (k = 1, n,
      a[k] = ceil(1/u);
      u = u*a[k] - 1;
      if (!u, break);
    ); a
  }

@3Obsolete hack. (don't use this): If b is an integer, nmax is ignored and the command is understood as contfrac(x,, b).

The library syntax is GEN contfrac0(GEN x, GEN b = NULL, long nmax). Also available are GEN gboundcf(GEN x, long nmax), GEN gcf(GEN x) and GEN gcf2(GEN b, GEN x).

contfracpnqn(x, {n = -1})

When x is a vector or a one-row matrix, x is considered as the list of partial quotients [a_0,a_1,...,a_n] of a rational number, and the result is the 2 by 2 matrix [p_n,p_{n-1};q_n,q_{n-1}] in the standard notation of continued fractions, so p_n/q_n = a_0+1/(a_1+...+1/a_n). If x is a matrix with two rows [b_0,b_1,...,b_n] and [a_0,a_1,...,a_n], this is then considered as a generalized continued fraction and we have similarly p_n/q_n = (1/b_0)(a_0+b_1/(a_1+...+b_n/a_n)). Note that in this case one usually has b_0 = 1.

If n >= 0 is present, returns all convergents from p_0/q_0 up to p_n/q_n. (All convergents if x is too small to compute the n+1 requested convergents.)

  ? a=contfrac(Pi,20)
  %1 = [3, 7, 15, 1, 292, 1, 1, 1, 2, 1, 3, 1, 14, 2, 1, 1, 2, 2, 2, 2]
  ? contfracpnqn(a,3)
  %2 =
  [3 22 333 355]
  [1  7 106 113]
  ? contfracpnqn(a,7)
  %3 =
  [3 22 333 355 103993 104348 208341 312689]
  [1  7 106 113  33102  33215  66317  99532]

The library syntax is GEN contfracpnqn(GEN x, long n). also available is GEN pnqn(GEN x) for n = -1.

core(n,{flag = 0})

If n is an integer written as n = df^2 with d squarefree, returns d. If flag is non-zero, returns the two-element row vector [d,f]. By convention, we write 0 = 0 x 1^2, so core(0, 1) returns [0,1].

The library syntax is GEN core0(GEN n, long flag). Also available are GEN core(GEN n) (flag = 0) and GEN core2(GEN n) (flag = 1)

coredisc(n,{flag = 0})

A fundamental discriminant is an integer of the form t = 1 mod 4 or 4t = 8,12 mod 16, with t squarefree (i.e. 1 or the discriminant of a quadratic number field). Given a non-zero integer n, this routine returns the (unique) fundamental discriminant d such that n = df^2, f a positive rational number. If flag is non-zero, returns the two-element row vector [d,f]. If n is congruent to 0 or 1 modulo 4, f is an integer, and a half-integer otherwise.

By convention, coredisc(0, 1)) returns [0,1].

Note that quaddisc(n) returns the same value as coredisc(n), and also works with rational inputs n belongs to Q^*.

The library syntax is GEN coredisc0(GEN n, long flag). Also available are GEN coredisc(GEN n) (flag = 0) and GEN coredisc2(GEN n) (flag = 1)

dirdiv(x,y)

x and y being vectors of perhaps different lengths but with y[1] != 0 considered as Dirichlet series, computes the quotient of x by y, again as a vector.

The library syntax is GEN dirdiv(GEN x, GEN y).

direuler(p = a,b,expr,{c})

Computes the Dirichlet series associated to the Euler product of expression expr as p ranges through the primes from a to b. expr must be a polynomial or rational function in another variable than p (say X) and expr(X) is understood as the local factor expr(p^{-s}).

The series is output as a vector of coefficients. If c is present, output only the first c coefficients in the series. The following command computes the sigma function, associated to zeta(s)zeta(s-1):

  ? direuler(p=2, 10, 1/((1-X)*(1-p*X)))
  %1 = [1, 3, 4, 7, 6, 12, 8, 15, 13, 18]

The library syntax is direuler(void *E, GEN (*eval)(void*,GEN), GEN a, GEN b)

dirmul(x,y)

x and y being vectors of perhaps different lengths representing the Dirichlet series sum_n x_n n^{-s} and sum_n y_n n^{-s}, computes the product of x by y, again as a vector.

  ? dirmul(vector(10,n,1), vector(10,n,moebius(n)))
  %1 = [1, 0, 0, 0, 0, 0, 0, 0, 0, 0]

The product length is the minimum of #x*v(y) and #y*v(x), where v(x) is the index of the first non-zero coefficient.

  ? dirmul([0,1], [0,1]);
  %2 = [0, 0, 0, 1]

The library syntax is GEN dirmul(GEN x, GEN y).

divisors(x)

Creates a row vector whose components are the divisors of x. The factorization of x (as output by factor) can be used instead.

By definition, these divisors are the products of the irreducible factors of n, as produced by factor(n), raised to appropriate powers (no negative exponent may occur in the factorization). If n is an integer, they are the positive divisors, in increasing order.

The library syntax is GEN divisors(GEN x).

eulerphi(x)

Euler's phi (totient) function of the integer |x|, in other words |(Z/xZ)^*|.

  ? eulerphi(40)
  %1 = 16

According to this definition we let phi(0) := 2, since Z^ *= {-1,1}; this is consistent with znstar(0): we have znstar(n).no = eulerphi(n) for all n belongs to Z.

The library syntax is GEN eulerphi(GEN x).

factor(x,{lim})

General factorization function, where x is a rational (including integers), a complex number with rational real and imaginary parts, or a rational function (including polynomials). The result is a two-column matrix: the first contains the irreducibles dividing x (rational or Gaussian primes, irreducible polynomials), and the second the exponents. By convention, 0 is factored as 0^1.

@3Q and Q(i). See factorint for more information about the algorithms used. The rational or Gaussian primes are in fact pseudoprimes (see ispseudoprime), a priori not rigorously proven primes. In fact, any factor which is <= 2^{64} (whose norm is <= 2^{64} for an irrational Gaussian prime) is a genuine prime. Use isprime to prove primality of other factors, as in

  ? fa = factor(2^2^7 + 1)
  %1 =
  [59649589127497217 1]
  [5704689200685129054721 1]
  ? isprime( fa[,1] )
  %2 = [1, 1]~   \\ both entries are proven primes

@3 Another possibility is to set the global default factor_proven, which will perform a rigorous primality proof for each pseudoprime factor.

A t_INT argument lim can be added, meaning that we look only for prime factors p < lim. The limit lim must be non-negative. In this case, all but the last factor are proven primes, but the remaining factor may actually be a proven composite! If the remaining factor is less than lim^2, then it is prime.

  ? factor(2^2^7 +1, 10^5)
  %3 =
  [340282366920938463463374607431768211457 1]

@3Deprecated feature. Setting lim = 0 is the same as setting it to primelimit + 1. Don't use this: it is unwise to rely on global variables when you can specify an explicit argument.

This routine uses trial division and perfect power tests, and should not be used for huge values of lim (at most 10^9, say): factorint(, 1 + 8) will in general be faster. The latter does not guarantee that all small prime factors are found, but it also finds larger factors, and in a much more efficient way.

  ? F = (2^2^7 + 1) * 1009 * 100003; factor(F, 10^5)  \\ fast, incomplete
  time = 0 ms.
  %4 =
  [1009 1]
  [34029257539194609161727850866999116450334371 1]
  ? factor(F, 10^9)    \\ very slow
  time = 6,892 ms.
  %6 =
  [1009 1]
  [100003 1]
  [340282366920938463463374607431768211457 1]
  ? factorint(F, 1+8)  \\ much faster, all small primes were found
  time = 12 ms.
  %7 =
  [1009 1]
  [100003 1]
  [340282366920938463463374607431768211457 1]
  ? factor(F)   \\ complete factorisation
  time = 112 ms.
  %8 =
  [1009 1]
  [100003 1]
  [59649589127497217 1]
  [5704689200685129054721 1]

@3Over Q, the prime factors are sorted in increasing order.

@3Rational functions. The polynomials or rational functions to be factored must have scalar coefficients. In particular PARI does not know how to factor multivariate polynomials. The following domains are currently supported: Q, R, C, Q_p, finite fields and number fields. See factormod and factorff for the algorithms used over finite fields, factornf for the algorithms over number fields. Over Q, van Hoeij's method is used, which is able to cope with hundreds of modular factors.

The routine guesses a sensible ring over which to factor: the smallest ring containing all coefficients, taking into account quotient structures induced by t_INTMODs and t_POLMODs (e.g. if a coefficient in Z/nZ is known, all rational numbers encountered are first mapped to Z/nZ; different moduli will produce an error). Factoring modulo a non-prime number is not supported; to factor in Q_p, use t_PADIC coefficients not t_INTMOD modulo p^n.

  ? T = x^2+1;
  ? factor(T);                         \\ over Q
  ? factor(T*Mod(1,3))                 \\ over F_3
  ? factor(T*ffgen(ffinit(3,2,'t))^0)  \\ over F_{3^2}
  ? factor(T*Mod(Mod(1,3), t^2+t+2))   \\ over F_{3^2}, again
  ? factor(T*(1 + O(3^6))              \\ over Q_3, precision 6
  ? factor(T*1.)                       \\ over R, current precision
  ? factor(T*(1.+0.*I))                \\ over C
  ? factor(T*Mod(1, y^3-2))            \\ over Q(2^{1/3})

@3In most cases, it is clearer and simpler to call an explicit variant than to rely on the generic factor function and the above detection mechanism:

  ? factormod(T, 3)           \\ over F_3
  ? factorff(T, 3, t^2+t+2))  \\ over F_{3^2}
  ? factorpadic(T, 3,6)       \\ over Q_3, precision 6
  ? nffactor(y^3-2, T)        \\ over Q(2^{1/3})
  ? polroots(T)               \\ over C

Note that factorization of polynomials is done up to multiplication by a constant. In particular, the factors of rational polynomials will have integer coefficients, and the content of a polynomial or rational function is discarded and not included in the factorization. If needed, you can always ask for the content explicitly:

  ? factor(t^2 + 5/2*t + 1)
  %1 =
  [2*t + 1 1]
  [t + 2 1]
  ? content(t^2 + 5/2*t + 1)
  %2 = 1/2

@3 The irreducible factors are sorted by increasing degree. See also nffactor.

The library syntax is GEN gp_factor0(GEN x, GEN lim = NULL). This function should only be used by the gp interface. Use directly GEN factor(GEN x) or GEN boundfact(GEN x, ulong lim). The obsolete function GEN factor0(GEN x, long lim) is kept for backward compatibility.

factorback(f,{e})

Gives back the factored object corresponding to a factorization. The integer 1 corresponds to the empty factorization.

If e is present, e and f must be vectors of the same length (e being integral), and the corresponding factorization is the product of the f[i]^{e[i]}.

If not, and f is vector, it is understood as in the preceding case with e a vector of 1s: we return the product of the f[i]. Finally, f can be a regular factorization, as produced with any factor command. A few examples:

  ? factor(12)
  %1 =
  [2 2]
  [3 1]
  ? factorback(%)
  %2 = 12
  ? factorback([2,3], [2,1])   \\ 2^3 * 3^1
  %3 = 12
  ? factorback([5,2,3])
  %4 = 30

The library syntax is GEN factorback2(GEN f, GEN e = NULL). Also available is GEN factorback(GEN f) (case e = NULL).

factorcantor(x,p)

Factors the polynomial x modulo the prime p, using distinct degree plus Cantor-Zassenhaus. The coefficients of x must be operation-compatible with Z/pZ. The result is a two-column matrix, the first column being the irreducible polynomials dividing x, and the second the exponents. If you want only the degrees of the irreducible polynomials (for example for computing an L-function), use factormod(x,p,1). Note that the factormod algorithm is usually faster than factorcantor.

The library syntax is GEN factcantor(GEN x, GEN p).

factorff(x,{p},{a})

Factors the polynomial x in the field F_q defined by the irreducible polynomial a over F_p. The coefficients of x must be operation-compatible with Z/pZ. The result is a two-column matrix: the first column contains the irreducible factors of x, and the second their exponents. If all the coefficients of x are in F_p, a much faster algorithm is applied, using the computation of isomorphisms between finite fields.

Either a or p can omitted (in which case both are ignored) if x has t_FFELT coefficients; the function then becomes identical to factor:

  ? factorff(x^2 + 1, 5, y^2+3)  \\ over F_5[y]/(y^2+3) ~ F_25
  %1 =
  [Mod(Mod(1, 5), Mod(1, 5)*y^2 + Mod(3, 5))*x
   + Mod(Mod(2, 5), Mod(1, 5)*y^2 + Mod(3, 5)) 1]
  [Mod(Mod(1, 5), Mod(1, 5)*y^2 + Mod(3, 5))*x
   + Mod(Mod(3, 5), Mod(1, 5)*y^2 + Mod(3, 5)) 1]
  ? t = ffgen(y^2 + Mod(3,5), 't); \\ a generator for F_25 as a t_FFELT
  ? factorff(x^2 + 1)   \\ not enough information to determine the base field
   ***   at top-level: factorff(x^2+1)
   ***                 ^---------------
   *** factorff: incorrect type in factorff.
  ? factorff(x^2 + t^0) \\ make sure a coeff. is a t_FFELT
  %3 =
  [x + 2 1]
  [x + 3 1]
  ? factorff(x^2 + t + 1)
  %11 =
  [x + (2*t + 1) 1]
  [x + (3*t + 4) 1]

@3 Notice that the second syntax is easier to use and much more readable.

The library syntax is GEN factorff(GEN x, GEN p = NULL, GEN a = NULL).

factorial(x)

Factorial of x. The expression x! gives a result which is an integer, while factorial(x) gives a real number.

The library syntax is GEN mpfactr(long x, long prec). GEN mpfact(long x) returns x! as a t_INT.

factorint(x,{flag = 0})

Factors the integer n into a product of pseudoprimes (see ispseudoprime), using a combination of the Shanks SQUFOF and Pollard Rho method (with modifications due to Brent), Lenstra's ECM (with modifications by Montgomery), and MPQS (the latter adapted from the LiDIA code with the kind permission of the LiDIA maintainers), as well as a search for pure powers. The output is a two-column matrix as for factor: the first column contains the ``prime'' divisors of n, the second one contains the (positive) exponents.

By convention 0 is factored as 0^1, and 1 as the empty factorization; also the divisors are by default not proven primes is they are larger than 2^{64}, they only failed the BPSW compositeness test (see ispseudoprime). Use isprime on the result if you want to guarantee primality or set the factor_proven default to 1. Entries of the private prime tables (see addprimes) are also included as is.

This gives direct access to the integer factoring engine called by most arithmetical functions. flag is optional; its binary digits mean 1: avoid MPQS, 2: skip first stage ECM (we may still fall back to it later), 4: avoid Rho and SQUFOF, 8: don't run final ECM (as a result, a huge composite may be declared to be prime). Note that a (strong) probabilistic primality test is used; thus composites might not be detected, although no example is known.

You are invited to play with the flag settings and watch the internals at work by using gp's debug default parameter (level 3 shows just the outline, 4 turns on time keeping, 5 and above show an increasing amount of internal details).

The library syntax is GEN factorint(GEN x, long flag).

factormod(x,p,{flag = 0})

Factors the polynomial x modulo the prime integer p, using Berlekamp. The coefficients of x must be operation-compatible with Z/pZ. The result is a two-column matrix, the first column being the irreducible polynomials dividing x, and the second the exponents. If flag is non-zero, outputs only the degrees of the irreducible polynomials (for example, for computing an L-function). A different algorithm for computing the mod p factorization is factorcantor which is sometimes faster.

The library syntax is GEN factormod0(GEN x, GEN p, long flag).

ffgen(q,{v})

Return a t_FFELT generator for the finite field with q elements; q = p^f must be a prime power. This functions computes an irreducible monic polynomial P belongs to F_p[X] of degree f (via ffinit) and returns g = X (mod P(X)). If v is given, the variable name is used to display g, else the variable x is used.

  ? g = ffgen(8, 't);
  ? g.mod
  %2 = t^3 + t^2 + 1
  ? g.p
  %3 = 2
  ? g.f
  %4 = 3
  ? ffgen(6)
   ***   at top-level: ffgen(6)
   ***                 ^--------
   *** ffgen: not a prime number in ffgen: 6.

@3Alternative syntax: instead of a prime power q, one may input directly the polynomial P (monic, irreducible, with t_INTMOD coefficients), and the function returns the generator g = X (mod P(X)), inferring p from the coefficients of P. If v is given, the variable name is used to display g, else the variable of the polynomial P is used. If P is not irreducible, we create an invalid object and behaviour of functions dealing with the resulting t_FFELT is undefined; in fact, it is much more costly to test P for irreducibility than it would be to produce it via ffinit.

The library syntax is GEN ffgen(GEN q, long v = -1), where v is a variable number.

To create a generator for a prime finite field, the function GEN p_to_GEN(GEN p, long v) returns 1+ffgen(x*Mod(1,p),v).

ffinit(p,n,{v = 'x})

Computes a monic polynomial of degree n which is irreducible over F_p, where p is assumed to be prime. This function uses a fast variant of Adleman and Lenstra's algorithm.

It is useful in conjunction with ffgen; for instance if P = ffinit(3,2), you can represent elements in F_{3^2} in term of g = ffgen(P,'t). This can be abbreviated as g = ffgen(3^2, 't), where the defining polynomial P can be later recovered as g.mod.

The library syntax is GEN ffinit(GEN p, long n, long v = -1), where v is a variable number.

fflog(x,g,{o})

Discrete logarithm of the finite field element x in base g, i.e.  an e in Z such that g^e = o. If present, o represents the multiplicative order of g, see Label se:DLfun; the preferred format for this parameter is [ord, factor(ord)], where ord is the order of g. It may be set as a side effect of calling ffprimroot.

If no o is given, assume that g is a primitive root. The result is undefined if e does not exist. This function uses

@3* a combination of generic discrete log algorithms (see znlog)

@3* a cubic sieve index calculus algorithm for large fields of degree at least 5.

@3* Coppersmith's algorithm for fields of characteristic at most 5.

  ? t = ffgen(ffinit(7,5));
  ? o = fforder(t)
  %2 = 5602   \\ I<not> a primitive root.
  ? fflog(t^10,t)
  %3 = 10
  ? fflog(t^10,t, o)
  %4 = 10
  ? g = ffprimroot(t, &o);
  ? o   \\ order is 16806, bundled with its factorization matrix
  %6 = [16806, [2, 1; 3, 1; 2801, 1]]
  ? fforder(g, o)
  %7 = 16806
  ? fflog(g^10000, g, o)
  %8 = 10000

The library syntax is GEN fflog(GEN x, GEN g, GEN o = NULL).

ffnbirred(q,n{,fl = 0})

Computes the number of monic irreducible polynomials over F_q of degree exactly n, (flag = 0 or omitted) or at most n (flag = 1).

The library syntax is GEN ffnbirred0(GEN q, long n, long fl). Also available are GEN ffnbirred(GEN q, long n) (for flag = 0) and GEN ffsumnbirred(GEN q, long n) (for flag = 1).

fforder(x,{o})

Multiplicative order of the finite field element x. If o is present, it represents a multiple of the order of the element, see Label se:DLfun; the preferred format for this parameter is [N, factor(N)], where N is the cardinality of the multiplicative group of the underlying finite field.

  ? t = ffgen(ffinit(nextprime(10^8), 5));
  ? g = ffprimroot(t, &o);  \\ o will be useful!
  ? fforder(g^1000000, o)
  time = 0 ms.
  %5 = 5000001750000245000017150000600250008403
  ? fforder(g^1000000)
  time = 16 ms. \\ noticeably slower, same result of course
  %6 = 5000001750000245000017150000600250008403

The library syntax is GEN fforder(GEN x, GEN o = NULL).

ffprimroot(x, {&o})

Return a primitive root of the multiplicative group of the definition field of the finite field element x (not necessarily the same as the field generated by x). If present, o is set to a vector [ord, fa], where ord is the order of the group and fa its factorisation factor(ord). This last parameter is useful in fflog and fforder, see Label se:DLfun.

  ? t = ffgen(ffinit(nextprime(10^7), 5));
  ? g = ffprimroot(t, &o);
  ? o[1]
  %3 = 100000950003610006859006516052476098
  ? o[2]
  %4 =
  [2 1]
  [7 2]
  [31 1]
  [41 1]
  [67 1]
  [1523 1]
  [10498781 1]
  [15992881 1]
  [46858913131 1]
  ? fflog(g^1000000, g, o)
  time = 1,312 ms.
  %5 = 1000000

The library syntax is GEN ffprimroot(GEN x, GEN *o = NULL).

fibonacci(x)

x-th Fibonacci number.

The library syntax is GEN fibo(long x).

gcd(x,{y})

Creates the greatest common divisor of x and y. If you also need the u and v such that x*u + y*v = gcd(x,y), use the bezout function. x and y can have rather quite general types, for instance both rational numbers. If y is omitted and x is a vector, returns the {gcd} of all components of x, i.e. this is equivalent to content(x).

When x and y are both given and one of them is a vector/matrix type, the GCD is again taken recursively on each component, but in a different way. If y is a vector, resp. matrix, then the result has the same type as y, and components equal to gcd(x, y[i]), resp. gcd(x, y[,i]). Else if x is a vector/matrix the result has the same type as x and an analogous definition. Note that for these types, gcd is not commutative.

The algorithm used is a naive Euclid except for the following inputs:

@3* integers: use modified right-shift binary (``plus-minus'' variant).

@3* univariate polynomials with coefficients in the same number field (in particular rational): use modular gcd algorithm.

@3* general polynomials: use the subresultant algorithm if coefficient explosion is likely (non modular coefficients).

If u and v are polynomials in the same variable with inexact coefficients, their gcd is defined to be scalar, so that

  ? a = x + 0.0; gcd(a,a)
  %1 = 1
  ? b = y*x + O(y); gcd(b,b)
  %2 = y
  ? c = 4*x + O(2^3); gcd(c,c)
  %3 = 4

@3A good quantitative check to decide whether such a gcd ``should be'' non-trivial, is to use polresultant: a value close to 0 means that a small deformation of the inputs has non-trivial gcd. You may also use gcdext, which does try to compute an approximate gcd d and provides u, v to check whether u x + v y is close to d.

The library syntax is GEN ggcd0(GEN x, GEN y = NULL). Also available are GEN ggcd(GEN x, GEN y), if y is not NULL, and GEN content(GEN x), if y = NULL.

gcdext(x,y)

Returns [u,v,d] such that d is the gcd of x,y, x*u+y*v = gcd(x,y), and u and v minimal in a natural sense. The arguments must be integers or polynomials.

  ? [u, v, d] = gcdext(32,102)
  %1 = [16, -5, 2]
  ? d
  %2 = 2
  ? gcdext(x^2-x, x^2+x-2)
  %3 = [-1/2, 1/2, x - 1]

If x,y are polynomials in the same variable and inexact coefficients, then compute u,v,d such that x*u+y*v = d, where d approximately divides both and x and y; in particular, we do not obtain gcd(x,y) which is defined to be a scalar in this case:

  ? a = x + 0.0; gcd(a,a)
  %1 = 1
  ? gcdext(a,a)
  %2 = [0, 1, x + 0.E-28]
  ? gcdext(x-Pi, 6*x^2-zeta(2))
  %3 = [-6*x - 18.8495559, 1, 57.5726923]

@3For inexact inputs, the output is thus not well defined mathematically, but you obtain explicit polynomials to check whether the approximation is close enough for your needs.

The library syntax is GEN gcdext0(GEN x, GEN y).

hilbert(x,y,{p})

Hilbert symbol of x and y modulo the prime p, p = 0 meaning the place at infinity (the result is undefined if p != 0 is not prime).

It is possible to omit p, in which case we take p = 0 if both x and y are rational, or one of them is a real number. And take p = q if one of x, y is a t_INTMOD modulo q or a q-adic. (Incompatible types will raise an error.)

The library syntax is long hilbert(GEN x, GEN y, GEN p = NULL).

isfundamental(x)

True (1) if x is equal to 1 or to the discriminant of a quadratic field, false (0) otherwise.

The library syntax is long isfundamental(GEN x).

ispolygonal(x,s,{&N})

True (1) if the integer x is an s-gonal number, false (0) if not. The parameter s > 2 must be a t_INT. If N is given, set it to n if x is the n-th s-gonal number.

  ? ispolygonal(36, 3, &N)
  %1 = 1
  ? N

The library syntax is long ispolygonal(GEN x, GEN s, GEN *N = NULL).

ispower(x,{k},{&n})

If k is given, returns true (1) if x is a k-th power, false (0) if not.

If k is omitted, only integers and fractions are allowed for x and the function returns the maximal k >= 2 such that x = n^k is a perfect power, or 0 if no such k exist; in particular ispower(-1), ispower(0), and ispower(1) all return 0.

If a third argument &n is given and x is indeed a k-th power, sets n to a k-th root of x.

@3For a t_FFELT x, instead of omitting k (which is not allowed for this type), it may be natural to set

  k = (x.p ^ poldegree(x.pol) - 1) / fforder(x)

The library syntax is long ispower(GEN x, GEN k = NULL, GEN *n = NULL). Also available is long gisanypower(GEN x, GEN *pty) (k omitted).

ispowerful(x)

True (1) if x is a powerful integer, false (0) if not; an integer is powerful if and only if its valuation at all primes is greater than 1.

  ? ispowerful(50)
  %1 = 0
  ? ispowerful(100)
  %2 = 1
  ? ispowerful(5^3*(10^1000+1)^2)
  %3 = 1

The library syntax is long ispowerful(GEN x).

isprime(x,{flag = 0})

True (1) if x is a prime number, false (0) otherwise. A prime number is a positive integer having exactly two distinct divisors among the natural numbers, namely 1 and itself.

This routine proves or disproves rigorously that a number is prime, which can be very slow when x is indeed prime and has more than 1000 digits, say. Use ispseudoprime to quickly check for compositeness. See also factor. It accepts vector/matrices arguments, and is then applied componentwise.

If flag = 0, use a combination of Baillie-PSW pseudo primality test (see ispseudoprime), Selfridge ``p-1'' test if x-1 is smooth enough, and Adleman-Pomerance-Rumely-Cohen-Lenstra (APRCL) for general x.

If flag = 1, use Selfridge-Pocklington-Lehmer ``p-1'' test and output a primality certificate as follows: return

@3* 0 if x is composite,

@3* 1 if x is small enough that passing Baillie-PSW test guarantees its primality (currently x < 2^{64}, as checked by Jan Feitsma),

@3* 2 if x is a large prime whose primality could only sensibly be proven (given the algorithms implemented in PARI) using the APRCL test.

@3* Otherwise (x is large and x-1 is smooth) output a three column matrix as a primality certificate. The first column contains prime divisors p of x-1 (such that prod p^{v_p(x-1)} > x^{1/3}), the second the corresponding elements a_p as in Proposition 8.3.1 in GTM 138 , and the third the output of isprime(p,1).

The algorithm fails if one of the pseudo-prime factors is not prime, which is exceedingly unlikely and well worth a bug report. Note that if you monitor isprime at a high enough debug level, you may see warnings about untested integers being declared primes. This is normal: we ask for partial factorisations (sufficient to prove primality if the unfactored part is not too large), and factor warns us that the cofactor hasn't been tested. It may or may not be tested later, and may or may not be prime. This does not affect the validity of the whole isprime procedure.

If flag = 2, use APRCL.

The library syntax is GEN gisprime(GEN x, long flag).

isprimepower(x,{&n})

If x = p^k is a prime power (p prime, k > 0), return k, else return 0. If a second argument &n is given and x is indeed the k-th power of a prime p, sets n to p.

The library syntax is long isprimepower(GEN x, GEN *n = NULL).

ispseudoprime(x,{flag})

True (1) if x is a strong pseudo prime (see below), false (0) otherwise. If this function returns false, x is not prime; if, on the other hand it returns true, it is only highly likely that x is a prime number. Use isprime (which is of course much slower) to prove that x is indeed prime. The function accepts vector/matrices arguments, and is then applied componentwise.

If flag = 0, checks whether x is a Baillie-Pomerance-Selfridge-Wagstaff pseudo prime (strong Rabin-Miller pseudo prime for base 2, followed by strong Lucas test for the sequence (P,-1), P smallest positive integer such that P^2 - 4 is not a square mod x).

There are no known composite numbers passing this test, although it is expected that infinitely many such numbers exist. In particular, all composites <= 2^{64} are correctly detected (checked using http://www.cecm.sfu.ca/Pseudoprimes/index-2-to-64.html).

If flag > 0, checks whether x is a strong Miller-Rabin pseudo prime for flag randomly chosen bases (with end-matching to catch square roots of -1).

The library syntax is GEN gispseudoprime(GEN x, long flag).

issquare(x,{&n})

True (1) if x is a square, false (0) if not. What ``being a square'' means depends on the type of x: all t_COMPLEX are squares, as well as all non-negative t_REAL; for exact types such as t_INT, t_FRAC and t_INTMOD, squares are numbers of the form s^2 with s in Z, Q and Z/NZ respectively.

  ? issquare(3)          \\ as an integer
  %1 = 0
  ? issquare(3.)         \\ as a real number
  %2 = 1
  ? issquare(Mod(7, 8))  \\ in Z/8Z
  %3 = 0
  ? issquare( 5 + O(13^4) )  \\ in Q_13
  %4 = 0

If n is given, a square root of x is put into n.

  ? issquare(4, &n)
  %1 = 1
  ? n
  %2 = 2

For polynomials, either we detect that the characteristic is 2 (and check directly odd and even-power monomials) or we assume that 2 is invertible and check whether squaring the truncated power series for the square root yields the original input.

The library syntax is long issquareall(GEN x, GEN *n = NULL). Also available is long issquare(GEN x). Deprecated GP-specific functions GEN gissquare(GEN x) and GEN gissquareall(GEN x, GEN *pt) return gen_0 and gen_1 instead of a boolean value.

issquarefree(x)

True (1) if x is squarefree, false (0) if not. Here x can be an integer or a polynomial.

The library syntax is long issquarefree(GEN x).

istotient(x,{&N})

True (1) if x = phi(n) for some integer n, false (0) if not.

  ? istotient(14)
  %1 = 0
  ? istotient(100)
  %2 = 0

If N is given, set N = n as well.

  ? istotient(4, &n)
  %1 = 1
  ? n
  %2 = 10

The library syntax is long istotient(GEN x, GEN *N = NULL).

kronecker(x,y)

Kronecker symbol (x|y), where x and y must be of type integer. By definition, this is the extension of Legendre symbol to Z x Z by total multiplicativity in both arguments with the following special rules for y = 0, -1 or 2:

@3* (x|0) = 1 if |x |= 1 and 0 otherwise.

@3* (x|-1) = 1 if x >= 0 and -1 otherwise.

@3* (x|2) = 0 if x is even and 1 if x = 1,-1 mod 8 and -1 if x = 3,-3 mod 8.

The library syntax is long kronecker(GEN x, GEN y).

lcm(x,{y})

Least common multiple of x and y, i.e. such that lcm(x,y)*gcd(x,y) = {abs}(x*y). If y is omitted and x is a vector, returns the {lcm} of all components of x.

When x and y are both given and one of them is a vector/matrix type, the LCM is again taken recursively on each component, but in a different way. If y is a vector, resp. matrix, then the result has the same type as y, and components equal to lcm(x, y[i]), resp. lcm(x, y[,i]). Else if x is a vector/matrix the result has the same type as x and an analogous definition. Note that for these types, lcm is not commutative.

Note that lcm(v) is quite different from

  l = v[1]; for (i = 1, #v, l = lcm(l, v[i]))

Indeed, lcm(v) is a scalar, but l may not be (if one of the v[i] is a vector/matrix). The computation uses a divide-conquer tree and should be much more efficient, especially when using the GMP multiprecision kernel (and more subquadratic algorithms become available):

  ? v = vector(10^4, i, random);
  ? lcm(v);
  time = 323 ms.
  ? l = v[1]; for (i = 1, #v, l = lcm(l, v[i]))
  time = 833 ms.

The library syntax is GEN glcm0(GEN x, GEN y = NULL).

logint(x,b,&z)

Return the largest integer e so that b^e <= x, where the parameters b > 1 and x > 0 are both integers. If the parameter z is present, set it to b^e.

  ? logint(1000, 2)
  %1 = 9
  ? 2^9
  %2 = 512
  ? logint(1000, 2, &z)
  %3 = 9
  ? z
  %4 = 512

@3The number of digits used to write b in base x is 1 + logint(x,b):

  ? #digits(1000!, 10)
  %5 = 2568
  ? logint(1000!, 10)
  %6 = 2567

@3This function may conveniently replace

    floor( log(x) / log(b) )

@3which may not give the correct answer since PARI does not guarantee exact rounding.

The library syntax is long logint0(GEN x, GEN b, GEN *z = NULL).

moebius(x)

Moebius mu-function of |x|. x must be of type integer.

The library syntax is long moebius(GEN x).

nextprime(x)

Finds the smallest pseudoprime (see ispseudoprime) greater than or equal to x. x can be of any real type. Note that if x is a pseudoprime, this function returns x and not the smallest pseudoprime strictly larger than x. To rigorously prove that the result is prime, use isprime.

The library syntax is GEN nextprime(GEN x).

numbpart(n)

Gives the number of unrestricted partitions of n, usually called p(n) in the literature; in other words the number of nonnegative integer solutions to a+2b+3c+.. .= n. n must be of type integer and n < 10^{15} (with trivial values p(n) = 0 for n < 0 and p(0) = 1). The algorithm uses the Hardy-Ramanujan-Rademacher formula. To explicitly enumerate them, see partitions.

The library syntax is GEN numbpart(GEN n).

numdiv(x)

Number of divisors of |x|. x must be of type integer.

The library syntax is GEN numdiv(GEN x).

omega(x)

Number of distinct prime divisors of |x|. x must be of type integer.

  ? factor(392)
  %1 =
  [2 3]
  [7 2]
  ? omega(392)
  %2 = 2;  \\ without multiplicity
  ? bigomega(392)
  %3 = 5;  \\ = 3+2, with multiplicity

The library syntax is long omega(GEN x).

partitions(k,{a = k},{n = k}))

Returns the vector of partitions of the integer k as a sum of positive integers (parts); for k < 0, it returns the empty set [], and for k = 0 the trivial partition (no parts). A partition is given by a t_VECSMALL, where parts are sorted in nondecreasing order:

  ? partitions(3)
  %1 = [Vecsmall([3]), Vecsmall([1, 2]), Vecsmall([1, 1, 1])]

@3correspond to 3, 1+2 and 1+1+1. The number of (unrestricted) partitions of k is given by numbpart:

  ? #partitions(50)
  %1 = 204226
  ? numbpart(50)
  %2 = 204226

@3Optional parameters n and a are as follows:

@3* n = nmax (resp. n = [nmin,nmax]) restricts partitions to length less than nmax (resp. length between nmin and nmax), where the length is the number of nonzero entries.

@3* a = amax (resp. a = [amin,amax]) restricts the parts to integers less than amax (resp. between amin and amax).

  ? partitions(4, 2)  \\ parts bounded by 2
  %1 = [Vecsmall([2, 2]), Vecsmall([1, 1, 2]), Vecsmall([1, 1, 1, 1])]
  ? partitions(4,, 2) \\ at most 2 parts
  %2 = [Vecsmall([4]), Vecsmall([1, 3]), Vecsmall([2, 2])]
  ? partitions(4,[0,3], 2) \\ at most 2 parts
  %3 = [Vecsmall([4]), Vecsmall([1, 3]), Vecsmall([2, 2])]

By default, parts are positive and we remove zero entries unless amin <= 0, in which case nmin is ignored and X is of constant length nmax:

  ? partitions(4, [0,3])  \\ parts between 0 and 3
  %1 = [Vecsmall([0, 0, 1, 3]), Vecsmall([0, 0, 2, 2]),\
        Vecsmall([0, 1, 1, 2]), Vecsmall([1, 1, 1, 1])]

The library syntax is GEN partitions(long k, GEN a = NULL, GEN n) = NULL).

polrootsff(x,{p},{a})

Returns the vector of distinct roots of the polynomial x in the field F_q defined by the irreducible polynomial a over F_p. The coefficients of x must be operation-compatible with Z/pZ. Either a or p can omitted (in which case both are ignored) if x has t_FFELT coefficients:

  ? polrootsff(x^2 + 1, 5, y^2+3)  \\ over F_5[y]/(y^2+3) ~ F_25
  %1 = [Mod(Mod(3, 5), Mod(1, 5)*y^2 + Mod(3, 5)),
        Mod(Mod(2, 5), Mod(1, 5)*y^2 + Mod(3, 5))]
  ? t = ffgen(y^2 + Mod(3,5), 't); \\ a generator for F_25 as a t_FFELT
  ? polrootsff(x^2 + 1)   \\ not enough information to determine the base field
   ***   at top-level: polrootsff(x^2+1)
   ***                 ^-----------------
   *** polrootsff: incorrect type in factorff.
  ? polrootsff(x^2 + t^0) \\ make sure one coeff. is a t_FFELT
  %3 = [3, 2]
  ? polrootsff(x^2 + t + 1)
  %4 = [2*t + 1, 3*t + 4]

Notice that the second syntax is easier to use and much more readable.

The library syntax is GEN polrootsff(GEN x, GEN p = NULL, GEN a = NULL).

precprime(x)

Finds the largest pseudoprime (see ispseudoprime) less than or equal to x. x can be of any real type. Returns 0 if x <= 1. Note that if x is a prime, this function returns x and not the largest prime strictly smaller than x. To rigorously prove that the result is prime, use isprime.

The library syntax is GEN precprime(GEN x).

prime(n)

The n-th prime number

  ? prime(10^9)
  %1 = 22801763489

@3Uses checkpointing and a naive O(n) algorithm.

The library syntax is GEN prime(long n).

primepi(x)

The prime counting function. Returns the number of primes p, p <= x.

  ? primepi(10)
  %1 = 4;
  ? primes(5)
  %2 = [2, 3, 5, 7, 11]
  ? primepi(10^11)
  %3 = 4118054813

@3Uses checkpointing and a naive O(x) algorithm.

The library syntax is GEN primepi(GEN x).

primes(n)

Creates a row vector whose components are the first n prime numbers. (Returns the empty vector for n <= 0.) A t_VEC n = [a,b] is also allowed, in which case the primes in [a,b] are returned

  ? primes(10)     \\ the first 10 primes
  %1 = [2, 3, 5, 7, 11, 13, 17, 19, 23, 29]
  ? primes([0,29])  \\ the primes up to 29
  %2 = [2, 3, 5, 7, 11, 13, 17, 19, 23, 29]
  ? primes([15,30])
  %3 = [17, 19, 23, 29]

The library syntax is GEN primes0(GEN n).

qfbclassno(D,{flag = 0})

Ordinary class number of the quadratic order of discriminant D. In the present version 2.7.4, a O(D^{1/2}) algorithm is used for D > 0 (using Euler product and the functional equation) so D should not be too large, say D < 10^8, for the time to be reasonable. On the other hand, for D < 0 one can reasonably compute qfbclassno(D) for |D| < 10^{25}, since the routine uses Shanks's method which is in O(|D|^{1/4}). For larger values of |D|, see quadclassunit.

If flag = 1, compute the class number using Euler products and the functional equation. However, it is in O(|D|^{1/2}).

@3Important warning. For D < 0, this function may give incorrect results when the class group has many cyclic factors, because implementing Shanks's method in full generality slows it down immensely. It is therefore strongly recommended to double-check results using either the version with flag = 1 or the function quadclassunit.

@3Warning. Contrary to what its name implies, this routine does not compute the number of classes of binary primitive forms of discriminant D, which is equal to the narrow class number. The two notions are the same when D < 0 or the fundamental unit varepsilon has negative norm; when D > 0 and Nvarepsilon > 0, the number of classes of forms is twice the ordinary class number. This is a problem which we cannot fix for backward compatibility reasons. Use the following routine if you are only interested in the number of classes of forms:

  QFBclassno(D) =
  qfbclassno(D) * if (D < 0 || norm(quadunit(D)) < 0, 1, 2)

Here are a few examples:

  ? qfbclassno(400000028)
  time = 3,140 ms.
  %1 = 1
  ? quadclassunit(400000028).no
  time = 20 ms. \{ much faster}
  %2 = 1
  ? qfbclassno(-400000028)
  time = 0 ms.
  %3 = 7253 \{ correct, and fast enough}
  ? quadclassunit(-400000028).no
  time = 0 ms.
  %4 = 7253

See also qfbhclassno.

The library syntax is GEN qfbclassno0(GEN D, long flag). The following functions are also available:

GEN classno(GEN D) (flag = 0)

GEN classno2(GEN D) (flag = 1).

@3Finally

GEN hclassno(GEN D) computes the class number of an imaginary quadratic field by counting reduced forms, an O(|D|) algorithm.

qfbcompraw(x,y)

composition of the binary quadratic forms x and y, without reduction of the result. This is useful e.g. to compute a generating element of an ideal. The result is undefined if x and y do not have the same discriminant.

The library syntax is GEN qfbcompraw(GEN x, GEN y).

qfbhclassno(x)

Hurwitz class number of x, where x is non-negative and congruent to 0 or 3 modulo 4. For x > 5. 10^5, we assume the GRH, and use quadclassunit with default parameters.

The library syntax is GEN hclassno(GEN x).

qfbnucomp(x,y,L)

composition of the primitive positive definite binary quadratic forms x and y (type t_QFI) using the NUCOMP and NUDUPL algorithms of Shanks, à la Atkin. L is any positive constant, but for optimal speed, one should take L = |D|^{1/4}, where D is the common discriminant of x and y. When x and y do not have the same discriminant, the result is undefined.

The current implementation is straightforward and in general slower than the generic routine (since the latter takes advantage of asymptotically fast operations and careful optimizations).

The library syntax is GEN nucomp(GEN x, GEN y, GEN L). Also available is GEN nudupl(GEN x, GEN L) when x = y.

qfbnupow(x,n)

n-th power of the primitive positive definite binary quadratic form x using Shanks's NUCOMP and NUDUPL algorithms (see qfbnucomp, in particular the final warning).

The library syntax is GEN nupow(GEN x, GEN n).

qfbpowraw(x,n)

n-th power of the binary quadratic form x, computed without doing any reduction (i.e. using qfbcompraw). Here n must be non-negative and n < 2^{31}.

The library syntax is GEN qfbpowraw(GEN x, long n).

qfbprimeform(x,p)

Prime binary quadratic form of discriminant x whose first coefficient is p, where |p| is a prime number. By abuse of notation, p = +- 1 is also valid and returns the unit form. Returns an error if x is not a quadratic residue mod p, or if x < 0 and p < 0. (Negative definite t_QFI are not implemented.) In the case where x > 0, the ``distance'' component of the form is set equal to zero according to the current precision.

The library syntax is GEN primeform(GEN x, GEN p, long prec).

qfbred(x,{flag = 0},{d},{isd},{sd})

Reduces the binary quadratic form x (updating Shanks's distance function if x is indefinite). The binary digits of flag are toggles meaning

  1: perform a single reduction step

  2: don't update Shanks's distance

The arguments d, isd, sd, if present, supply the values of the discriminant, floor{ sqrt {d}}, and sqrt {d} respectively (no checking is done of these facts). If d < 0 these values are useless, and all references to Shanks's distance are irrelevant.

The library syntax is GEN qfbred0(GEN x, long flag, GEN d = NULL, GEN isd = NULL, GEN sd = NULL). Also available are

GEN redimag(GEN x) (for definite x),

@3and for indefinite forms:

GEN redreal(GEN x)

GEN rhoreal(GEN x) ( = qfbred(x,1)),

GEN redrealnod(GEN x, GEN isd) ( = qfbred(x,2,,isd)),

GEN rhorealnod(GEN x, GEN isd) ( = qfbred(x,3,,isd)).

qfbsolve(Q,p)

Solve the equation Q(x,y) = p over the integers, where Q is a binary quadratic form and p a prime number.

Return [x,y] as a two-components vector, or zero if there is no solution. Note that this function returns only one solution and not all the solutions.

Let D = disc Q. The algorithm used runs in probabilistic polynomial time in p (through the computation of a square root of D modulo p); it is polynomial time in D if Q is imaginary, but exponential time if Q is real (through the computation of a full cycle of reduced forms). In the latter case, note that bnfisprincipal provides a solution in heuristic subexponential time in D assuming the GRH.

The library syntax is GEN qfbsolve(GEN Q, GEN p).

quadclassunit(D,{flag = 0},{tech = []})

Buchmann-McCurley's sub-exponential algorithm for computing the class group of a quadratic order of discriminant D.

This function should be used instead of qfbclassno or quadregula when D < -10^{25}, D > 10^{10}, or when the structure is wanted. It is a special case of bnfinit, which is slower, but more robust.

The result is a vector v whose components should be accessed using member functions:

@3* v.no: the class number

@3* v.cyc: a vector giving the structure of the class group as a product of cyclic groups;

@3* v.gen: a vector giving generators of those cyclic groups (as binary quadratic forms).

@3* v.reg: the regulator, computed to an accuracy which is the maximum of an internal accuracy determined by the program and the current default (note that once the regulator is known to a small accuracy it is trivial to compute it to very high accuracy, see the tutorial).

The flag is obsolete and should be left alone. In older versions, it supposedly computed the narrow class group when D > 0, but this did not work at all; use the general function bnfnarrow.

Optional parameter tech is a row vector of the form [c_1, c_2], where c_1 <= c_2 are non-negative real numbers which control the execution time and the stack size, see se:GRHbnf. The parameter is used as a threshold to balance the relation finding phase against the final linear algebra. Increasing the default c_1 means that relations are easier to find, but more relations are needed and the linear algebra will be harder. The default value for c_1 is 0 and means that it is taken equal to c_2. The parameter c_2 is mostly obsolete and should not be changed, but we still document it for completeness: we compute a tentative class group by generators and relations using a factorbase of prime ideals <= c_1 ( log |D|)^2, then prove that ideals of norm <= c_2 ( log |D|)^2 do not generate a larger group. By default an optimal c_2 is chosen, so that the result is provably correct under the GRH --- a famous result of Bach states that c_2 = 6 is fine, but it is possible to improve on this algorithmically. You may provide a smaller c_2, it will be ignored (we use the provably correct one); you may provide a larger c_2 than the default value, which results in longer computing times for equally correct outputs (under GRH).

The library syntax is GEN quadclassunit0(GEN D, long flag, GEN tech = NULL, long prec). If you really need to experiment with the tech parameter, it is usually more convenient to use GEN Buchquad(GEN D, double c1, double c2, long prec)

quaddisc(x)

Discriminant of the quadratic field Q( sqrt {x}), where x belongs to Q.

The library syntax is GEN quaddisc(GEN x).

quadgen(D)

Creates the quadratic number omega = (a+ sqrt {D})/2 where a = 0 if D = 0 mod 4, a = 1 if D = 1 mod 4, so that (1,omega) is an integral basis for the quadratic order of discriminant D. D must be an integer congruent to 0 or 1 modulo 4, which is not a square.

The library syntax is GEN quadgen(GEN D).

quadhilbert(D)

Relative equation defining the Hilbert class field of the quadratic field of discriminant D.

If D < 0, uses complex multiplication (Schertz's variant).

If D > 0 Stark units are used and (in rare cases) a vector of extensions may be returned whose compositum is the requested class field. See bnrstark for details.

The library syntax is GEN quadhilbert(GEN D, long prec).

quadpoly(D,{v = 'x})

Creates the ``canonical'' quadratic polynomial (in the variable v) corresponding to the discriminant D, i.e. the minimal polynomial of quadgen(D). D must be an integer congruent to 0 or 1 modulo 4, which is not a square.

The library syntax is GEN quadpoly0(GEN D, long v = -1), where v is a variable number.

quadray(D,f)

Relative equation for the ray class field of conductor f for the quadratic field of discriminant D using analytic methods. A bnf for x^2 - D is also accepted in place of D.

For D < 0, uses the sigma function and Schertz's method.

For D > 0, uses Stark's conjecture, and a vector of relative equations may be returned. See bnrstark for more details.

The library syntax is GEN quadray(GEN D, GEN f, long prec).

quadregulator(x)

Regulator of the quadratic field of positive discriminant x. Returns an error if x is not a discriminant (fundamental or not) or if x is a square. See also quadclassunit if x is large.

The library syntax is GEN quadregulator(GEN x, long prec).

quadunit(D)

Fundamental unit of the real quadratic field Q( sqrt D) where D is the positive discriminant of the field. If D is not a fundamental discriminant, this probably gives the fundamental unit of the corresponding order. D must be an integer congruent to 0 or 1 modulo 4, which is not a square; the result is a quadratic number (see Label se:quadgen).

The library syntax is GEN quadunit(GEN D).

\subsec{randomprime({N = 2^{{31}}})} Returns a strong pseudo prime (see ispseudoprime) in [2,N-1]. A t_VEC N = [a,b] is also allowed, with a <= b in which case a pseudo prime a <= p <= b is returned; if no prime exists in the interval, the function will run into an infinite loop. If the upper bound is less than 2^{64} the pseudo prime returned is a proven prime.

The library syntax is GEN randomprime(GEN N = NULL).

removeprimes({x = []})

Removes the primes listed in x from the prime number table. In particular removeprimes(addprimes()) empties the extra prime table. x can also be a single integer. List the current extra primes if x is omitted.

The library syntax is GEN removeprimes(GEN x = NULL).

sigma(x,{k = 1})

Sum of the k-th powers of the positive divisors of |x|. x and k must be of type integer.

The library syntax is GEN sumdivk(GEN x, long k). Also available is GEN sumdiv(GEN n), for k = 1.

sqrtint(x)

Returns the integer square root of x, i.e. the largest integer y such that y^2 <= x, where x a non-negative integer.

  ? N = 120938191237; sqrtint(N)
  %1 = 347761
  ? sqrt(N)
  %2 = 347761.68741970412747602130964414095216

The library syntax is GEN sqrtint(GEN x).

sqrtnint(x,n)

Returns the integer n-th root of x, i.e. the largest integer y such that y^n <= x, where x is a non-negative integer.

  ? N = 120938191237; sqrtnint(N, 5)
  %1 = 164
  ? N^(1/5)
  %2 = 164.63140849829660842958614676939677391

@3The special case n = 2 is sqrtint

The library syntax is GEN sqrtnint(GEN x, long n).

stirling(n,k,{flag = 1})

Stirling number of the first kind s(n,k) (flag = 1, default) or of the second kind S(n,k) (flag = 2), where n, k are non-negative integers. The former is (-1)^{n-k} times the number of permutations of n symbols with exactly k cycles; the latter is the number of ways of partitioning a set of n elements into k non-empty subsets. Note that if all s(n,k) are needed, it is much faster to compute

  sum_k s(n,k) x^k = x(x-1)...(x-n+1).

Similarly, if a large number of S(n,k) are needed for the same k, one should use

  sum_n S(n,k) x^n = (x^k)/((1-x)...(1-kx)).

(Should be implemented using a divide and conquer product.) Here are simple variants for n fixed:

  /* list of s(n,k), k = 1..n */
  vecstirling(n) = Vec( factorback(vector(n-1,i,1-i*'x)) )
  /* list of S(n,k), k = 1..n */
  vecstirling2(n) =
  { my(Q = x^(n-1), t);
    vector(n, i, t = divrem(Q, x-i); Q=t[1]; simplify(t[2]));
  }

The library syntax is GEN stirling(long n, long k, long flag). Also available are GEN stirling1(ulong n, ulong k) (flag = 1) and GEN stirling2(ulong n, ulong k) (flag = 2).

sumdedekind(h,k)

Returns the Dedekind sum associated to the integers h and k, corresponding to a fast implementation of

    s(h,k) = sum(n = 1, k-1, (n/k)*(frac(h*n/k) - 1/2))

The library syntax is GEN sumdedekind(GEN h, GEN k).

sumdigits(n)

Sum of (decimal) digits in the integer n.

  ? sumdigits(123456789)
  %1 = 45

@3Other bases that 10 are not supported. Note that the sum of bits in n is returned by hammingweight.

The library syntax is GEN sumdigits(GEN n).

zncoppersmith(P, N, X, {B = N})

N being an integer and P belongs to Z[X], finds all integers x with |x| <= X such that

  gcd(N, P(x)) >= B,

using Coppersmith's algorithm (a famous application of the LLL algorithm). X must be smaller than exp ( log ^2 B / ( deg (P) log N)): for B = N, this means X < N^{1/ deg (P)}. Some x larger than X may be returned if you are very lucky. The smaller B (or the larger X), the slower the routine will be. The strength of Coppersmith method is the ability to find roots modulo a general composite N: if N is a prime or a prime power, polrootsmod or polrootspadic will be much faster.

We shall now present two simple applications. The first one is finding non-trivial factors of N, given some partial information on the factors; in that case B must obviously be smaller than the largest non-trivial divisor of N.

  setrand(1); \\ to make the example reproducible
  p = nextprime(random(10^30));
  q = nextprime(random(10^30)); N = p*q;
  p0 = p % 10^20; \\ assume we know 1) p > 10^29, 2) the last 19 digits of p
  p1 = zncoppersmith(10^19*x + p0, N, 10^12, 10^29)
  \\ result in 10ms.
  %1 = [35023733690]
  ? gcd(p1[1] * 10^19 + p0, N) == p
  %2 = 1

@3and we recovered p, faster than by trying all possibilities < 10^{12}.

The second application is an attack on RSA with low exponent, when the message x is short and the padding P is known to the attacker. We use the same RSA modulus N as in the first example:

  setrand(1);
  P = random(N);    \\ known padding
  e = 3;            \\ small public encryption exponent
  X = floor(N^0.3); \\ N^(1/e - epsilon)
  x0 = random(X);   \\ unknown short message
  C = lift( (Mod(x0,N) + P)^e ); \\ known ciphertext, with padding P
  zncoppersmith((P + x)^3 - C, N, X)
  \\ result in 244ms.
  %3 = [265174753892462432]
  ? %[1] == x0
  %4 = 1

@3 We guessed an integer of the order of 10^{18}, almost instantly.

The library syntax is GEN zncoppersmith(GEN P, GEN N, GEN X, GEN B = NULL).

znlog(x,g,{o})

Discrete logarithm of x in (Z/NZ)^* in base g. The result is [] when x is not a power of g. If present, o represents the multiplicative order of g, see Label se:DLfun; the preferred format for this parameter is [ord, factor(ord)], where ord is the order of g. This provides a definite speedup when the discrete log problem is simple:

  ? p = nextprime(10^4); g = znprimroot(p); o = [p-1, factor(p-1)];
  ? for(i=1,10^4, znlog(i, g, o))
  time = 205 ms.
  ? for(i=1,10^4, znlog(i, g))
  time = 244 ms. \\ a little slower

The result is undefined if g is not invertible mod N or if the supplied order is incorrect.

This function uses

@3* a combination of generic discrete log algorithms (see below).

@3* in (Z/NZ)^* when N is prime: a linear sieve index calculus method, suitable for N < 10^{50}, say, is used for large prime divisors of the order.

The generic discrete log algorithms are:

@3* Pohlig-Hellman algorithm, to reduce to groups of prime order q, where q | p-1 and p is an odd prime divisor of N,

@3* Shanks baby-step/giant-step (q < 2^{32} is small),

@3* Pollard rho method (q > 2^{32}).

The latter two algorithms require O( sqrt {q}) operations in the group on average, hence will not be able to treat cases where q > 10^{30}, say. In addition, Pollard rho is not able to handle the case where there are no solutions: it will enter an infinite loop.

  ? g = znprimroot(101)
  %1 = Mod(2,101)
  ? znlog(5, g)
  %2 = 24
  ? g^24
  %3 = Mod(5, 101)
  ? G = znprimroot(2 * 101^10)
  %4 = Mod(110462212541120451003, 220924425082240902002)
  ? znlog(5, G)
  %5 = 76210072736547066624
  ? G^% == 5
  %6 = 1
  ? N = 2^4*3^2*5^3*7^4*11; g = Mod(13, N); znlog(g^110, g)
  %7 = 110
  ? znlog(6, Mod(2,3))  \\ no solution
  %8 = []

@3For convenience, g is also allowed to be a p-adic number:

  ? g = 3+O(5^10); znlog(2, g)
  %1 = 1015243
  ? g^%
  %2 = 2 + O(5^10)

The library syntax is GEN znlog(GEN x, GEN g, GEN o = NULL).

znorder(x,{o})

x must be an integer mod n, and the result is the order of x in the multiplicative group (Z/nZ)^*. Returns an error if x is not invertible. The parameter o, if present, represents a non-zero multiple of the order of x, see Label se:DLfun; the preferred format for this parameter is [ord, factor(ord)], where ord = eulerphi(n) is the cardinality of the group.

The library syntax is GEN znorder(GEN x, GEN o = NULL). Also available is GEN order(GEN x).

znprimroot(n)

Returns a primitive root (generator) of (Z/nZ)^*, whenever this latter group is cyclic (n = 4 or n = 2p^k or n = p^k, where p is an odd prime and k >= 0). If the group is not cyclic, the result is undefined. If n is a prime power, then the smallest positive primitive root is returned. This may not be true for n = 2p^k, p odd.

Note that this function requires factoring p-1 for p as above, in order to determine the exact order of elements in (Z/nZ)^*: this is likely to be costly if p is large.

The library syntax is GEN znprimroot(GEN n).

znstar(n)

Gives the structure of the multiplicative group (Z/nZ)^* as a 3-component row vector v, where v[1] = phi(n) is the order of that group, v[2] is a k-component row-vector d of integers d[i] such that d[i] > 1 and d[i] | d[i-1] for i >= 2 and (Z/nZ)^* ~ prod_{i = 1}^k(Z/d[i]Z), and v[3] is a k-component row vector giving generators of the image of the cyclic groups Z/d[i]Z.

  ? G = znstar(40)
  %1 = [16, [4, 2, 2], [Mod(17, 40), Mod(21, 40), Mod(11, 40)]]
  ? G.no   \\ eulerphi(40)
  %2 = 16
  ? G.cyc  \\ cycle structure
  %3 = [4, 2, 2]
  ? G.gen  \\ generators for the cyclic components
  %4 = [Mod(17, 40), Mod(21, 40), Mod(11, 40)]
  ? apply(znorder, G.gen)
  %5 = [4, 2, 2]

@3According to the above definitions, znstar(0) is [2, [2], [-1]], corresponding to Z^*.

The library syntax is GEN znstar(GEN n).


Functions related to elliptic curves

Elliptic curve structures

An elliptic curve is given by a Weierstrass model

   y^2+a_1xy+a_3y = x^3+a_2x^2+a_4x+a_6,

whose discriminant is non-zero. Affine points on E are represented as two-component vectors [x,y]; the point at infinity, i.e. the identity element of the group law, is represented by the one-component vector [0].

Given a vector of coefficients [a_1,a_2,a_3,a_4,a_6], the function ellinit initializes and returns an ell structure. (An additional optional argument allows to specify the base field in case it cannot be inferred from the curve coefficients.) This structure contains data needed by elliptic curve related functions, and is generally passed as a first argument. Expensive data are skipped on initialization: they will be dynamically computed when (and if) needed, and then inserted in the structure. The precise layout of the ell structure is left undefined and should never be used directly. The following member functions are available, depending on the underlying domain.

All domains

@3* a1, a2, a3, a4, a6: coefficients of the elliptic curve.

@3* b2, b4, b6, b8: b-invariants of the curve; in characteristic != 2, for Y = 2y + a_1x+a3, the curve equation becomes

   Y^2 = 4 x^3 + b_2 x^2 + 2b_4 x + b_6 = : g(x).

@3* c4, c6: c-invariants of the curve; in characteristic != 2,3, for X = x + b_2/12 and Y = 2y + a_1x+a3, the curve equation becomes

   Y^2 = 4 X^3 - (c_4/12) X - (c_6/216).

@3* disc: discriminant of the curve. This is only required to be non-zero, not necessarily a unit.

@3* j: j-invariant of the curve.

@3These are used as follows:

  ? E = ellinit([0,0,0, a4,a6]);
  ? E.b4
  %2 = 2*a4
  ? E.disc
  %3 = -64*a4^3 - 432*a6^2

Curves over R

This in particular includes curves defined over Q. All member functions in this section return data, as it is currently stored in the structure, if present; and otherwise compute it to the default accuracy, that was fixed at the time of ellinit (via a t_REAL D domain argument, or realprecision by default). The function ellperiods allows to recompute (and cache) the following data to current realprecision.

@3* area: volume of the complex lattice defining E.

@3* roots is a vector whose three components contain the complex roots of the right hand side g(x) of the associated b-model Y^2 = g(x). If the roots are all real, they are ordered by decreasing value. If only one is real, it is the first component.

@3* omega: [omega_1,omega_2], periods forming a basis of the complex lattice defining E. The first component omega_1 is the (positive) real period, in other words the integral of dx/(2y+a_1x+a_3) over the connected component of the identity component of E(R). The second component omega_2 is a complex period, such that tau = (omega_1)/(omega_2) belongs to Poincaré's half-plane (positive imaginary part); not necessarily to the standard fundamental domain.

@3* eta is a row vector containing the quasi-periods eta_1 and eta_2 such that eta_i = 2zeta(omega_i/2), where zeta is the Weierstrass zeta function associated to the period lattice; see ellzeta. In particular, the Legendre relation holds: eta_2omega_1 - eta_1omega_2 = 2iPi.

@3Warning. As for the orientation of the basis of the period lattice, beware that many sources use the inverse convention where omega_2/omega_1 has positive imaginary part and our omega_2 is the negative of theirs. Our convention tau = omega_1/omega_2 ensures that the action of {PSL}_2 is the natural one:

  [a,b;c,d].tau = (atau+b)/(ctau+d) = (a omega_1 + bomega_2)/(comega_1 + domega_2),

instead of a twisted one. (Our tau is -1/tau in the above inverse convention.)

Curves over Q_p

We advise to input a model defined over Q for such curves. In any case, if you input an approximate model with t_PADIC coefficients, it will be replaced by a lift to Q (an exact model ``close'' to the one that was input) and all quantities will then be computed in terms of this lifted model.

For the time being only curves with multiplicative reduction (split or non-split), i.e. v_p(j) < 0, are supported by non-trivial functions. In this case the curve is analytically isomorphic to \bar{Q}_p^*/q^Z := E_q(\bar{Q}_p), for some p-adic integer q (the Tate period). In particular, we have j(q) = j(E).

@3* p is the residual characteristic

@3* roots is a vector with a single component, equal to the p-adic root e_1 of the right hand side g(x) of the associated b-model Y^2 = g(x). The point (e_1,0) corresponds to -1 belongs to \bar{Q}_p^*/q^Z under the Tate parametrization.

@3* tate returns [u^2,u,q,[a,b]] in the notation of Henniart-Mestre (CRAS t. 308, p. 391--395, 1989): q is as above, u belongs to Q_p( sqrt {-c_6}) is such that phi^* dx/(2y + a_1x+a3) = u dt/t, where phi: E_q\to E is an isomorphism (well defined up to sign) and dt/t is the canonical invariant differential on the Tate curve; u^2 belongs to Q_p does not depend on phi. (Technicality: if u\not belongs to Q_p, it is stored as a quadratic t_POLMOD.) Finally, [a,b] satisfy 4u^2 b.{agm}( sqrt {a/b},1)^2 = 1 as in Theorem 2 (loc. cit.).

Curves over F_q

@3* p is the characteristic of F_q.

@3* no is #E(F_q).

@3* cyc gives the cycle structure of E(F_q).

@3* gen returns the generators of E(F_q).

@3* group returns [no,cyc,gen], i.e. E(F_q) as an abelian group structure.

Curves over Q

All functions should return a correct result, whether the model is minimal or not, but it is a good idea to stick to minimal models whenever gcd(c_4,c_6) is easy to factor (minor speed-up). The construction

    E = ellminimalmodel(E0, &v)

@3replaces the original model E_0 by a minimal model E, and the variable change v allows to go between the two models:

    ellchangepoint(P0, v)
    ellchangepointinv(P, v)

@3respectively map the point P_0 on E_0 to its image on E, and the point P on E to its pre-image on E_0.

A few routines --- namely ellgenerators, ellidentify, ellsearch, forell --- require the optional package elldata (John Cremona's database) to be installed. In that case, the function ellinit will allow alternative inputs, e.g. ellinit("11a1"). Functions using this package need to load chunks of a large database in memory and require at least 2MB stack to avoid stack overflows.

@3* gen returns the generators of E(Q), if known (from John Cremona's database)

ellL1(e, r)

Returns the value at s = 1 of the derivative of order r of the L-function of the elliptic curve e assuming that r is at most the order of vanishing of the L-function at s = 1. (The result is wrong if r is strictly larger than the order of vanishing at 1.)

  ? e = ellinit("11a1"); \\ order of vanishing is 0
  ? ellL1(e, 0)
  %2 = 0.2538418608559106843377589233
  ? e = ellinit("389a1");  \\ order of vanishing is 2
  ? ellL1(e, 0)
  %4 = -5.384067311837218089235032414 E-29
  ? ellL1(e, 1)
  %5 = 0
  ? ellL1(e, 2)
  %6 = 1.518633000576853540460385214

The main use of this function, after computing at low accuracy the order of vanishing using ellanalyticrank, is to compute the leading term at high accuracy to check (or use) the Birch and Swinnerton-Dyer conjecture:

  ? \p18
    realprecision = 18 significant digits
  ? ellanalyticrank(ellinit([0, 0, 1, -7, 6]))
  time = 32 ms.
  %1 = [3, 10.3910994007158041]
  ? \p200
    realprecision = 202 significant digits (200 digits displayed)
  ? ellL1(e, 3)
  time = 23,113 ms.
  %3 = 10.3910994007158041387518505103609170697263563756570092797[...]

The library syntax is GEN ellL1(GEN e, long r, long prec).

elladd(E,z1,z2)

Sum of the points z1 and z2 on the elliptic curve corresponding to E.

The library syntax is GEN elladd(GEN E, GEN z1, GEN z2).

ellak(E,n)

Computes the coefficient a_n of the L-function of the elliptic curve E/Q, i.e. coefficients of a newform of weight 2 by the modularity theorem (Taniyama-Shimura-Weil conjecture). E must be an ell structure over Q as output by ellinit. E must be given by an integral model, not necessarily minimal, although a minimal model will make the function faster.

  ? E = ellinit([0,1]);
  ? ellak(E, 10)
  %2 = 0
  ? e = ellinit([5^4,5^6]); \\ not minimal at 5
  ? ellak(e, 5) \\ wasteful but works
  %3 = -3
  ? E = ellminimalmodel(e); \\ now minimal
  ? ellak(E, 5)
  %5 = -3

@3If the model is not minimal at a number of bad primes, then the function will be slower on those n divisible by the bad primes. The speed should be comparable for other n:

  ? for(i=1,10^6, ellak(E,5))
  time = 820 ms.
  ? for(i=1,10^6, ellak(e,5)) \\ 5 is bad, markedly slower
  time = 1,249 ms.
  ? for(i=1,10^5,ellak(E,5*i))
  time = 977 ms.
  ? for(i=1,10^5,ellak(e,5*i)) \\ still slower but not so much on average
  time = 1,008 ms.

The library syntax is GEN akell(GEN E, GEN n).

ellan(E,n)

Computes the vector of the first n Fourier coefficients a_k corresponding to the elliptic curve E. The curve must be given by an integral model, not necessarily minimal, although a minimal model will make the function faster.

The library syntax is GEN anell(GEN E, long n). Also available is GEN anellsmall(GEN e, long n), which returns a t_VECSMALL instead of a t_VEC, saving on memory.

ellanalyticrank(e, {eps})

Returns the order of vanishing at s = 1 of the L-function of the elliptic curve e and the value of the first non-zero derivative. To determine this order, it is assumed that any value less than eps is zero. If no value of eps is given, a value of half the current precision is used.

  ? e = ellinit("11a1"); \\ rank 0
  ? ellanalyticrank(e)
  %2 = [0, 0.2538418608559106843377589233]
  ? e = ellinit("37a1"); \\ rank 1
  ? ellanalyticrank(e)
  %4 = [1, 0.3059997738340523018204836835]
  ? e = ellinit("389a1"); \\ rank 2
  ? ellanalyticrank(e)
  %6 = [2, 1.518633000576853540460385214]
  ? e = ellinit("5077a1"); \\ rank 3
  ? ellanalyticrank(e)
  %8 = [3, 10.39109940071580413875185035]

The library syntax is GEN ellanalyticrank(GEN e, GEN eps = NULL, long prec).

ellap(E,{p})

Let E be an ell structure as output by ellinit, defined over Q or a finite field F_q. The argument p is best left omitted if the curve is defined over a finite field, and must be a prime number otherwise. This function computes the trace of Frobenius t for the elliptic curve E, defined by the equation #E(F_q) = q+1 - t.

If the curve is defined over Q, p must be explicitly given and the function computes the trace of the reduction over F_p. The trace of Frobenius is also the a_p coefficient in the curve L-series L(E,s) = sum_n a_n n^{-s}, whence the function name. The equation must be integral at p but need not be minimal at p; of course, a minimal model will be more efficient.

  ? E = ellinit([0,1]);  \\ y^2 = x^3 + 0.x + 1, defined over Q
  ? ellap(E, 7) \\ 7 necessary here
  %2 = -4       \\ #E(F_7) = 7+1-(-4) = 12
  ? ellcard(E, 7)
  %3 = 12       \\ OK
  ? E = ellinit([0,1], 11);  \\ defined over F_11
  ? ellap(E)       \\ no need to repeat 11
  %4 = 0
  ? ellap(E, 11)   \\ ... but it also works
  %5 = 0
  ? ellgroup(E, 13) \\ ouch, inconsistent input!
     ***   at top-level: ellap(E,13)
     ***                 ^-----------
     *** ellap: inconsistent moduli in Rg_to_Fp:
       11
       13
  ? Fq = ffgen(ffinit(11,3), 'a); \\ defines F_q := F_{11^3}
  ? E = ellinit([a+1,a], Fq);  \\ y^2 = x^3 + (a+1)x + a, defined over F_q
  ? ellap(E)
  %8 = -3

@3Algorithms used. If E/F_q has CM by a principal imaginary quadratic order we use a fast explicit formula (involving essentially Kronecker symbols and Cornacchia's algorithm), in O( log q)^2. Otherwise, we use Shanks-Mestre's baby-step/giant-step method, which runs in time q(p^{1/4}) using O(q^{1/4}) storage, hence becomes unreasonable when q has about 30 digits. If the seadata package is installed, the SEA algorithm becomes available, heuristically in ~{O}( log q)^4, and primes of the order of 200 digits become feasible. In very small characteristic (2,3,5,7 or 13), we use Harley's algorithm.

The library syntax is GEN ellap(GEN E, GEN p = NULL).

ellbil(E,z1,z2)

If z1 and z2 are points on the elliptic curve E this function computes the value of the canonical bilinear form on z1, z2:

   ( h(E,z1+z2) - h(E,z1) - h(E,z2) ) / 2

where + denotes of course addition on E. In addition, z1 or z2 (but not both) can be vectors or matrices.

The library syntax is GEN bilhell(GEN E, GEN z1, GEN z2, long prec).

ellcard(E,{p})

Let E be an ell structure as output by ellinit, defined over Q or a finite field F_q. The argument p is best left omitted if the curve is defined over a finite field, and must be a prime number otherwise. This function computes the order of the group E(F_q) (as would be computed by ellgroup).

If the curve is defined over Q, p must be explicitly given and the function computes the cardinal of the reduction over F_p; the equation need not be minimal at p, but a minimal model will be more efficient. The reduction is allowed to be singular, and we return the order of the group of non-singular points in this case.

The library syntax is GEN ellcard(GEN E, GEN p = NULL). Also available is GEN ellcard(GEN E, GEN p) where p is not NULL.

ellchangecurve(E,v)

Changes the data for the elliptic curve E by changing the coordinates using the vector v = [u,r,s,t], i.e. if x' and y' are the new coordinates, then x = u^2x'+r, y = u^3y'+su^2x'+t. E must be an ell structure as output by ellinit. The special case v = 1 is also used instead of [1,0,0,0] to denote the trivial coordinate change.

The library syntax is GEN ellchangecurve(GEN E, GEN v).

ellchangepoint(x,v)

Changes the coordinates of the point or vector of points x using the vector v = [u,r,s,t], i.e. if x' and y' are the new coordinates, then x = u^2x'+r, y = u^3y'+su^2x'+t (see also ellchangecurve).

  ? E0 = ellinit([1,1]); P0 = [0,1]; v = [1,2,3,4];
  ? E = ellchangecurve(E0, v);
  ? P = ellchangepoint(P0,v)
  %3 = [-2, 3]
  ? ellisoncurve(E, P)
  %4 = 1
  ? ellchangepointinv(P,v)
  %5 = [0, 1]

The library syntax is GEN ellchangepoint(GEN x, GEN v). The reciprocal function GEN ellchangepointinv(GEN x, GEN ch) inverts the coordinate change.

ellchangepointinv(x,v)

Changes the coordinates of the point or vector of points x using the inverse of the isomorphism associated to v = [u,r,s,t], i.e. if x' and y' are the old coordinates, then x = u^2x'+r, y = u^3y'+su^2x'+t (inverse of ellchangepoint).

  ? E0 = ellinit([1,1]); P0 = [0,1]; v = [1,2,3,4];
  ? E = ellchangecurve(E0, v);
  ? P = ellchangepoint(P0,v)
  %3 = [-2, 3]
  ? ellisoncurve(E, P)
  %4 = 1
  ? ellchangepointinv(P,v)
  %5 = [0, 1]  \\ we get back P0

The library syntax is GEN ellchangepointinv(GEN x, GEN v).

ellconvertname(name)

Converts an elliptic curve name, as found in the elldata database, from a string to a triplet [conductor, isogeny class, index]. It will also convert a triplet back to a curve name. Examples:

  ? ellconvertname("123b1")
  %1 = [123, 1, 1]
  ? ellconvertname(%)
  %2 = "123b1"

The library syntax is GEN ellconvertname(GEN name).

elldivpol(E,n,{v = 'x})

n-division polynomial f_n for the curve E in the variable v. In standard notation, for any affine point P = (X,Y) on the curve, we have

  [n]P = (phi_n(P)psi_n(P) : omega_n(P) : psi_n(P)^3)

for some polynomials phi_n,omega_n,psi_n in Z[a_1,a_2,a_3,a_4,a_6][X,Y]. We have f_n(X) = psi_n(X) for n odd, and f_n(X) = psi_n(X,Y) (2Y + a_1X+a_3) for n even. We have

   f_1 = 1, f_2 = 4X^3 + b_2X^2 + 2b_4 X + b_6, f_3 = 3 X^4 + b_2 X^3 + 3b_4 X^2 + 3 b_6 X + b8,

   f_4 = f_2(2X^6 + b_2 X^5 + 5b_4 X^4 + 10 b_6 X^3 + 10 b_8 X^2 + (b_2b_8-b_4b_6)X + (b_8b_4 - b_6^2)),...

For n >= 2, the roots of f_n are the X-coordinates of points in E[n].

The library syntax is GEN elldivpol(GEN E, long n, long v = -1), where v is a variable number.

elleisnum(w,k,{flag = 0})

k being an even positive integer, computes the numerical value of the Eisenstein series of weight k at the lattice w, as given by ellperiods, namely

   (2i Pi/omega_2)^k (1 + 2/zeta(1-k) sum_{n >= 0} n^{k-1}q^n / (1-q^n)),

where q = exp (2iPi tau) and tau := omega_1/omega_2 belongs to the complex upper half-plane. It is also possible to directly input w = [omega_1,omega_2], or an elliptic curve E as given by ellinit.

  ? w = ellperiods([1,I]);
  ? elleisnum(w, 4)
  %2 = 2268.8726415508062275167367584190557607
  ? elleisnum(w, 6)
  %3 = -3.977978632282564763 E-33
  ? E = ellinit([1, 0]);
  ? elleisnum(E, 4, 1)
  %5 = -47.999999999999999999999999999999999998

When flag is non-zero and k = 4 or 6, returns the elliptic invariants g_2 or g_3, such that

  y^2 = 4x^3 - g_2 x - g_3

is a Weierstrass equation for E.

The library syntax is GEN elleisnum(GEN w, long k, long flag, long prec).

elleta(w)

Returns the quasi-periods [eta_1,eta_2] associated to the lattice basis w = [omega_1, omega_2]. Alternatively, w can be an elliptic curve E as output by ellinit, in which case, the quasi periods associated to the period lattice basis E.omega (namely, E.eta) are returned.

  ? elleta([1, I])
  %1 = [3.141592653589793238462643383, 9.424777960769379715387930149*I]

The library syntax is GEN elleta(GEN w, long prec).

ellfromj(j)

Returns the coefficients [a_1,a_2,a_3,a_4,a_6] of a fixed elliptic curve with j-invariant j.

The library syntax is GEN ellfromj(GEN j).

ellgenerators(E)

If E is an elliptic curve over the rationals, return a Z-basis of the free part of the Mordell-Weil group associated to E. This relies on the elldata database being installed and referencing the curve, and so is only available for curves over Z of small conductors. If E is an elliptic curve over a finite field F_q as output by ellinit, return a minimal set of generators for the group E(F_q).

The library syntax is GEN ellgenerators(GEN E).

ellglobalred(E)

Calculates the arithmetic conductor, the global minimal model of E and the global Tamagawa number c. E must be an ell structure as output by ellinit, defined over Q. The result is a vector [N,v,c,F,L], where

@3* N is the arithmetic conductor of the curve,

@3* v gives the coordinate change for E over Q to the minimal integral model (see ellminimalmodel),

@3* c is the product of the local Tamagawa numbers c_p, a quantity which enters in the Birch and Swinnerton-Dyer conjecture,

@3* F is the factorization of N over Z.

@3* L is a vector, whose i-th entry contains the local data at the i-th prime divisor of N, i.e. L[i] = elllocalred(E,F[i,1]), where the local coordinate change has been deleted, and replaced by a 0.

The library syntax is GEN ellglobalred(GEN E).

ellgroup(E,{p},{flag})

Let E be an ell structure as output by ellinit, defined over Q or a finite field F_q. The argument p is best left omitted if the curve is defined over a finite field, and must be a prime number otherwise. This function computes the structure of the group E(F_q) ~ Z/d_1Z x Z/d_2Z, with d_2 | d_1.

If the curve is defined over Q, p must be explicitly given and the function computes the structure of the reduction over F_p; the equation need not be minimal at p, but a minimal model will be more efficient. The reduction is allowed to be singular, and we return the structure of the (cyclic) group of non-singular points in this case.

If the flag is 0 (default), return [d_1] or [d_1, d_2], if d_2 > 1. If the flag is 1, return a triple [h,cyc,gen], where h is the curve cardinality, cyc gives the group structure as a product of cyclic groups (as per flag = 0). More precisely, if d_2 > 1, the output is [d_1d_2, [d_1,d_2],[P,Q]] where P is of order d_1 and [P,Q] generates the curve. \misctitle{Caution} It is not guaranteed that Q has order d_2, which in the worst case requires an expensive discrete log computation. Only that ellweilpairing(E, P, Q, d1) has order d_2.

  ? E = ellinit([0,1]);  \\ y^2 = x^3 + 0.x + 1, defined over Q
  ? ellgroup(E, 7)
  %2 = [6, 2] \\ Z/6 x Z/2, non-cyclic
  ? E = ellinit([0,1] * Mod(1,11));  \\ defined over F_11
  ? ellgroup(E)   \\ no need to repeat 11
  %4 = [12]
  ? ellgroup(E, 11)   \\ ... but it also works
  %5 = [12]
  ? ellgroup(E, 13) \\ ouch, inconsistent input!
     ***   at top-level: ellgroup(E,13)
     ***                 ^--------------
     *** ellgroup: inconsistent moduli in Rg_to_Fp:
       11
       13
  ? ellgroup(E, 7, 1)
  %6 = [12, [6, 2], [[Mod(2, 7), Mod(4, 7)], [Mod(4, 7), Mod(4, 7)]]]

If E is defined over Q, we allow singular reduction and in this case we return the structure of the group of non-singular points, satisfying #E_{ns}(F_p) = p - a_p.

  ? E = ellinit([0,5]);
  ? ellgroup(E, 5, 1)
  %2 = [5, [5], [[Mod(4, 5), Mod(2, 5)]]]
  ? ellap(E, 5)
  %3 = 0 \\ additive reduction at 5
  ? E = ellinit([0,-1,0,35,0]);
  ? ellgroup(E, 5, 1)
  %5 = [4, [4], [[Mod(2, 5), Mod(2, 5)]]]
  ? ellap(E, 5)
  %6 = 1 \\ split multiplicative reduction at 5
  ? ellgroup(E, 7, 1)
  %7 = [8, [8], [[Mod(3, 7), Mod(5, 7)]]]
  ? ellap(E, 7)
  %8 = -1 \\ non-split multiplicative reduction at 7

The library syntax is GEN ellgroup0(GEN E, GEN p = NULL, long flag). Also available is GEN ellgroup(GEN E, GEN p), corresponding to flag = 0.

ellheegner(E)

Let E be an elliptic curve over the rationals, assumed to be of (analytic) rank 1. This returns a non-torsion rational point on the curve, whose canonical height is equal to the product of the elliptic regulator by the analytic Sha.

This uses the Heegner point method, described in Cohen GTM 239; the complexity is proportional to the product of the square root of the conductor and the height of the point (thus, it is preferable to apply it to strong Weil curves).

  ? E = ellinit([-157^2,0]);
  ? u = ellheegner(E); print(u[1], "\n", u[2])
  69648970982596494254458225/166136231668185267540804
  538962435089604615078004307258785218335/67716816556077455999228495435742408
  ? ellheegner(ellinit([0,1]))         \\ E has rank 0 !
   ***   at top-level: ellheegner(E=ellinit
   ***                 ^--------------------
   *** ellheegner: The curve has even analytic rank.

The library syntax is GEN ellheegner(GEN E).

ellheight(E,x,{flag = 2})

Global Néron-Tate height of the point z on the elliptic curve E (defined over Q), using the normalization in Cremona's Algorithms for modular elliptic curves. E must be an ell as output by ellinit; it needs not be given by a minimal model although the computation will be faster if it is. flag selects the algorithm used to compute the Archimedean local height. If flag = 0, we use sigma and theta-functions and Silverman's trick (Computing heights on elliptic curves, Math. Comp. 51; note that our height is twice Silverman's height). If flag = 1, use Tate's 4^n algorithm. If flag = 2, use Mestre's AGM algorithm. The latter converges quadratically and is much faster than the other two.

The library syntax is GEN ellheight0(GEN E, GEN x, long flag, long prec). Also available is GEN ghell(GEN E, GEN x, long prec) (flag = 2).

ellheightmatrix(E,x)

x being a vector of points, this function outputs the Gram matrix of x with respect to the Néron-Tate height, in other words, the (i,j) component of the matrix is equal to ellbil(E,x[i],x[j]). The rank of this matrix, at least in some approximate sense, gives the rank of the set of points, and if x is a basis of the Mordell-Weil group of E, its determinant is equal to the regulator of E. Note our height normalization follows Cremona's Algorithms for modular elliptic curves: this matrix should be divided by 2 to be in accordance with, e.g., Silverman's normalizations.

The library syntax is GEN mathell(GEN E, GEN x, long prec).

ellidentify(E)

Look up the elliptic curve E, defined by an arbitrary model over Q, in the elldata database. Return [[N, M, G], C] where N is the curve name in Cremona's elliptic curve database, M is the minimal model, G is a Z-basis of the free part of the Mordell-Weil group E(Q) and C is the change of coordinates change, suitable for ellchangecurve.

The library syntax is GEN ellidentify(GEN E).

ellinit(x,{D = 1})

Initialize an ell structure, associated to the elliptic curve E. E is either

@3* a 5-component vector [a_1,a_2,a_3,a_4,a_6] defining the elliptic curve with Weierstrass equation

   Y^2 + a_1 XY + a_3 Y = X^3 + a_2 X^2 + a_4 X + a_6,

@3* a 2-component vector [a_4,a_6] defining the elliptic curve with short Weierstrass equation

   Y^2 = X^3 + a_4 X + a_6,

@3* a character string in Cremona's notation, e.g. "11a1", in which case the curve is retrieved from the elldata database if available.

The optional argument D describes the domain over which the curve is defined:

@3* the t_INT 1 (default): the field of rational numbers Q.

@3* a t_INT p, where p is a prime number: the prime finite field F_p.

@3* an t_INTMOD Mod(a, p), where p is a prime number: the prime finite field F_p.

@3* a t_FFELT, as returned by ffgen: the corresponding finite field F_q.

@3* a t_PADIC, O(p^n): the field Q_p, where p-adic quantities will be computed to a relative accuracy of n digits. We advise to input a model defined over Q for such curves. In any case, if you input an approximate model with t_PADIC coefficients, it will be replaced by a lift to Q (an exact model ``close'' to the one that was input) and all quantities will then be computed in terms of this lifted model, at the given accuracy.

@3* a t_REAL x: the field C of complex numbers, where floating point quantities are by default computed to a relative accuracy of precision(x). If no such argument is given, the value of realprecision at the time ellinit is called will be used.

This argument D is indicative: the curve coefficients are checked for compatibility, possibly changing D; for instance if D = 1 and an t_INTMOD is found. If inconsistencies are detected, an error is raised:

  ? ellinit([1 + O(5), 1], O(7));
   ***   at top-level: ellinit([1+O(5),1],O
   ***                 ^--------------------
   *** ellinit: inconsistent moduli in ellinit: 7 != 5

@3If the curve coefficients are too general to fit any of the above domain categories, only basic operations, such as point addition, will be supported later.

If the curve (seen over the domain D) is singular, fail and return an empty vector [].

  ? E = ellinit([0,0,0,0,1]); \\ y^2 = x^3 + 1, over Q
  ? E = ellinit([0,1]);       \\ the same curve, short form
  ? E = ellinit("36a1");      \\ sill the same curve, Cremona's notations
  ? E = ellinit([0,1], 2)     \\ over F2: singular curve
  %4 = []
  ? E = ellinit(['a4,'a6] * Mod(1,5));  \\ over F_5[a4,a6], basic support !

The result of ellinit is an ell structure. It contains at least the following information in its components:

   a_1,a_2,a_3,a_4,a_6,b_2,b_4,b_6,b_8,c_4,c_6,Delta,j.

All are accessible via member functions. In particular, the discriminant is E.disc, and the j-invariant is E.j.

  ? E = ellinit([a4, a6]);
  ? E.disc
  %2 = -64*a4^3 - 432*a6^2
  ? E.j
  %3 = -6912*a4^3/(-4*a4^3 - 27*a6^2)

Further components contain domain-specific data, which are in general dynamic: only computed when needed, and then cached in the structure.

  ? E = ellinit([2,3], 10^60+7);  \\ E over F_p, p large
  ? ellap(E)
  time = 4,440 ms.
  %2 = -1376268269510579884904540406082
  ? ellcard(E);  \\ now instantaneous !
  time = 0 ms.
  ? ellgenerators(E);
  time = 5,965 ms.
  ? ellgenerators(E); \\ second time instantaneous
  time = 0 ms.

See the description of member functions related to elliptic curves at the beginning of this section.

The library syntax is GEN ellinit(GEN x, GEN D = NULL, long prec).

ellisoncurve(E,z)

Gives 1 (i.e. true) if the point z is on the elliptic curve E, 0 otherwise. If E or z have imprecise coefficients, an attempt is made to take this into account, i.e. an imprecise equality is checked, not a precise one. It is allowed for z to be a vector of points in which case a vector (of the same type) is returned.

The library syntax is GEN ellisoncurve(GEN E, GEN z). Also available is int oncurve(GEN E, GEN z) which does not accept vectors of points.

ellj(x)

Elliptic j-invariant. x must be a complex number with positive imaginary part, or convertible into a power series or a p-adic number with positive valuation.

The library syntax is GEN jell(GEN x, long prec).

elllocalred(E,p)

Calculates the Kodaira type of the local fiber of the elliptic curve E at the prime p. E must be an ell structure as output by ellinit, and is assumed to have all its coefficients a_i in Z. The result is a 4-component vector [f,kod,v,c]. Here f is the exponent of p in the arithmetic conductor of E, and kod is the Kodaira type which is coded as follows:

1 means good reduction (type I_0), 2, 3 and 4 mean types II, III and IV respectively, 4+nu with nu > 0 means type I_nu; finally the opposite values -1, -2, etc. refer to the starred types I_0^*, II^*, etc. The third component v is itself a vector [u,r,s,t] giving the coordinate changes done during the local reduction; u = 1 if and only if the given equation was already minimal at p. Finally, the last component c is the local Tamagawa number c_p.

The library syntax is GEN elllocalred(GEN E, GEN p).

elllog(E,P,G,{o})

Given two points P and G on the elliptic curve E/F_q, returns the discrete logarithm of P in base G, i.e. the smallest non-negative integer n such that P = [n]G. See znlog for the limitations of the underlying discrete log algorithms. If present, o represents the order of G, see Label se:DLfun; the preferred format for this parameter is [N, factor(N)], where N is the order of G.

If no o is given, assume that G generates the curve. The function also assumes that P is a multiple of G.

  ? a = ffgen(ffinit(2,8),'a);
  ? E = ellinit([a,1,0,0,1]);  \\ over F_{2^8}
  ? x = a^3; y = ellordinate(E,x)[1];
  ? P = [x,y]; G = ellmul(E, P, 113);
  ? ord = [242, factor(242)]; \\ P generates a group of order 242. Initialize.
  ? ellorder(E, G, ord)
  %4 = 242
  ? e = elllog(E, P, G, ord)
  %5 = 15
  ? ellmul(E,G,e) == P
  %6 = 1

The library syntax is GEN elllog(GEN E, GEN P, GEN G, GEN o = NULL).

elllseries(E,s,{A = 1})

E being an elliptic curve, given by an arbitrary model over Q as output by ellinit, this function computes the value of the L-series of E at the (complex) point s. This function uses an O(N^{1/2}) algorithm, where N is the conductor.

The optional parameter A fixes a cutoff point for the integral and is best left omitted; the result must be independent of A, up to realprecision, so this allows to check the function's accuracy.

The library syntax is GEN elllseries(GEN E, GEN s, GEN A = NULL, long prec).

ellminimalmodel(E,{&v})

Return the standard minimal integral model of the rational elliptic curve E. If present, sets v to the corresponding change of variables, which is a vector [u,r,s,t] with rational components. The return value is identical to that of ellchangecurve(E, v).

The resulting model has integral coefficients, is everywhere minimal, a_1 is 0 or 1, a_2 is 0, 1 or -1 and a_3 is 0 or 1. Such a model is unique, and the vector v is unique if we specify that u is positive, which we do.

The library syntax is GEN ellminimalmodel(GEN E, GEN *v = NULL).

ellmodulareqn(N,{x},{y})

Return a vector [eqn,t] where eqn is a modular equation of level N, i.e. a bivariate polynomial with integer coefficients; t indicates the type of this equation: either canonical (t = 0) or Atkin (t = 1). This function currently requires the package seadata to be installed and is limited to N < 500, N prime.

Let j be the j-invariant function. The polynomial eqn satisfies the following functional equation, which allows to compute the values of the classical modular polynomial Phi_N of prime level N, such that Phi_N(j(tau), j(Ntau)) = 0, while being much smaller than the latter:

@3* for canonical type: P(f(tau),j(tau)) = P(N^s/f(tau),j(N tau)) = 0, where s = 12/gcd(12,N-1);

@3* for Atkin type: P(f(tau),j(tau)) = P(f(tau),j(N tau)) = 0.

@3In both cases, f is a suitable modular function (see below).

The following GP function returns values of the classical modular polynomial by eliminating f(tau) in the above two equations, for N <= 31 or N belongs to {41,47,59,71}.

  classicaleqn(N, X='X, Y='Y)=
  {
    my(E=ellmodulareqn(N), P=E[1], t=E[2], Q, d);
    if(poldegree(P,'y)>2,error("level unavailable in classicaleqn"));
    if (t == 0,
      my(s = 12/gcd(12,N-1));
      Q = 'x^(N+1) * substvec(P,['x,'y],[N^s/'x,Y]);
      d = N^(s*(2*N+1)) * (-1)^(N+1);
    ,
      Q = subst(P,'y,Y);
      d = (X-Y)^(N+1));
    polresultant(subst(P,'y,X), Q) / d;
  }

More precisely, let W_N(tau) = ({-1})/({N tau}) be the Atkin-Lehner involution; we have j(W_N(tau)) = j(N tau) and the function f also satisfies:

@3* for canonical type: f(W_N(tau)) = N^s/f(tau);

@3* for Atkin type: f(W_N(tau)) = f(tau).

@3Furthermore, for an equation of canonical type, f is the standard eta-quotient

  f(tau) = N^s (eta(N tau) / eta(tau) )^{2 s},

where eta is Dedekind's eta function, which is invariant under Gamma_0(N).

The library syntax is GEN ellmodulareqn(long N, long x = -1, long y = -1), where x, y are variable numbers.

ellmul(E,z,n)

Computes [n]z, where z is a point on the elliptic curve E. The exponent n is in Z, or may be a complex quadratic integer if the curve E has complex multiplication by n (if not, an error message is issued).

  ? Ei = ellinit([1,0]); z = [0,0];
  ? ellmul(Ei, z, 10)
  %2 = [0]     \\ unsurprising: z has order 2
  ? ellmul(Ei, z, I)
  %3 = [0, 0]  \\ Ei has complex multiplication by Z[i]
  ? ellmul(Ei, z, quadgen(-4))
  %4 = [0, 0]  \\ an alternative syntax for the same query
  ? Ej  = ellinit([0,1]); z = [-1,0];
  ? ellmul(Ej, z, I)
    ***   at top-level: ellmul(Ej,z,I)
    ***                 ^--------------
    *** ellmul: not a complex multiplication in ellmul.
  ? ellmul(Ej, z, 1+quadgen(-3))
  %6 = [1 - w, 0]

The simple-minded algorithm for the CM case assumes that we are in characteristic 0, and that the quadratic order to which n belongs has small discriminant.

The library syntax is GEN ellmul(GEN E, GEN z, GEN n).

ellneg(E,z)

Opposite of the point z on elliptic curve E.

The library syntax is GEN ellneg(GEN E, GEN z).

ellorder(E,z,{o})

Gives the order of the point z on the elliptic curve E, defined over Q or a finite field. If the curve is defined over Q, return (the impossible value) zero if the point has infinite order.

  ? E = ellinit([-157^2,0]);  \\ the "157-is-congruent" curve
  ? P = [2,2]; ellorder(E, P)
  %2 = 2
  ? P = ellheegner(E); ellorder(E, P) \\ infinite order
  %3 = 0
  ? E = ellinit(ellfromj(ffgen(5^10)));
  ? ellcard(E)
  %5 = 9762580
  ? P = random(E); ellorder(E, P)
  %6 = 4881290
  ? p = 2^160+7; E = ellinit([1,2], p);
  ? N = ellcard(E)
  %8 = 1461501637330902918203686560289225285992592471152
  ? o = [N, factor(N)];
  ? for(i=1,100, ellorder(E,random(E)))
  time = 260 ms.

The parameter o, is now mostly useless, and kept for backward compatibility. If present, it represents a non-zero multiple of the order of z, see Label se:DLfun; the preferred format for this parameter is [ord, factor(ord)], where ord is the cardinality of the curve. It is no longer needed since PARI is now able to compute it over large finite fields (was restricted to small prime fields at the time this feature was introduced), and caches the result in E so that it is computed and factored only once. Modifying the last example, we see that including this extra parameter provides no improvement:

  ? o = [N, factor(N)];
  ? for(i=1,100, ellorder(E,random(E),o))
  time = 260 ms.

The library syntax is GEN ellorder(GEN E, GEN z, GEN o = NULL). The obsolete form GEN orderell(GEN e, GEN z) should no longer be used.

ellordinate(E,x)

Gives a 0, 1 or 2-component vector containing the y-coordinates of the points of the curve E having x as x-coordinate.

The library syntax is GEN ellordinate(GEN E, GEN x, long prec).

ellperiods(w, {flag = 0})

Let w describe a complex period lattice (w = [w_1,w_2] or an ellinit structure). Returns normalized periods [W_1,W_2] generating the same lattice such that tau := W_1/W_2 has positive imaginary part and lies in the standard fundamental domain for {SL}_2(Z).

If flag = 1, the function returns [[W_1,W_2], [eta_1,eta_2]], where eta_1 and eta_2 are the quasi-periods associated to [W_1,W_2], satisfying eta_1 W_2 - eta_2 W_1 = 2 i Pi.

The output of this function is meant to be used as the first argument given to ellwp, ellzeta, ellsigma or elleisnum. Quasi-periods are needed by ellzeta and ellsigma only.

The library syntax is GEN ellperiods(GEN w, long flag , long prec).

ellpointtoz(E,P)

If E/C ~ C/Lambda is a complex elliptic curve (Lambda = E.omega), computes a complex number z, well-defined modulo the lattice Lambda, corresponding to the point P; i.e. such that P = [ wp _Lambda(z), wp '_Lambda(z)] satisfies the equation

  y^2 = 4x^3 - g_2 x - g_3,

where g_2, g_3 are the elliptic invariants.

If E is defined over R and P belongs to E(R), we have more precisely, 0 <= Re (t) < w1 and 0 <= Im (t) < Im (w2), where (w1,w2) are the real and complex periods of E.

  ? E = ellinit([0,1]); P = [2,3];
  ? z = ellpointtoz(E, P)
  %2 = 3.5054552633136356529375476976257353387
  ? ellwp(E, z)
  %3 = 2.0000000000000000000000000000000000000
  ? ellztopoint(E, z) - P
  %4 = [6.372367644529809109 E-58, 7.646841173435770930 E-57]
  ? ellpointtoz(E, [0]) \\ the point at infinity
  %5 = 0

If E/Q_p has multiplicative reduction, then E/\bar{Q_p} is analytically isomorphic to \bar{Q}_p^*/q^Z (Tate curve) for some p-adic integer q. The behaviour is then as follows:

@3* If the reduction is split (E.tate[2] is a t_PADIC), we have an isomorphism phi: E(Q_p) ~ Q_p^*/q^Z and the function returns phi(P) belongs to Q_p.

@3* If the reduction is not split (E.tate[2] is a t_POLMOD), we only have an isomorphism phi: E(K) ~ K^*/q^Z over the unramified quadratic extension K/Q_p. In this case, the output phi(P) belongs to K is a t_POLMOD.

  ? E = ellinit([0,-1,1,0,0], O(11^5)); P = [0,0];
  ? [u2,u,q] = E.tate; type(u) \\ split multiplicative reduction
  %2 = "t_PADIC"
  ? ellmul(E, P, 5)  \\ P has order 5
  %3 = [0]
  ? z = ellpointtoz(E, [0,0])
  %4 = 3 + 11^2 + 2*11^3 + 3*11^4 + O(11^5)
  ? z^5
  %5 = 1 + O(11^5)
  ? E = ellinit(ellfromj(1/4), O(2^6)); x=1/2; y=ellordinate(E,x)[1];
  ? z = ellpointtoz(E,[x,y]); \\ t_POLMOD of t_POL with t_PADIC coeffs
  ? liftint(z) \\ lift all p-adics
  %8 = Mod(8*u + 7, u^2 + 437)

The library syntax is GEN zell(GEN E, GEN P, long prec).

ellpow(E,z,n)

Deprecated alias for ellmul.

The library syntax is GEN ellmul(GEN E, GEN z, GEN n).

ellrootno(E,{p})

E being an ell structure over Q as output by ellinit, this function computes the local root number of its L-series at the place p (at the infinite place if p = 0). If p is omitted, return the global root number. Note that the global root number is the sign of the functional equation and conjecturally is the parity of the rank of the Mordell-Weil group. The equation for E needs not be minimal at p, but if the model is already minimal the function will run faster.

The library syntax is long ellrootno(GEN E, GEN p = NULL).

ellsearch(N)

This function finds all curves in the elldata database satisfying the constraint defined by the argument N:

@3* if N is a character string, it selects a given curve, e.g. "11a1", or curves in the given isogeny class, e.g. "11a", or curves with given conductor, e.g. "11";

@3* if N is a vector of integers, it encodes the same constraints as the character string above, according to the ellconvertname correspondance, e.g. [11,0,1] for "11a1", [11,0] for "11a" and [11] for "11";

@3* if N is an integer, curves with conductor N are selected.

If N is a full curve name, e.g. "11a1" or [11,0,1], the output format is [N, [a_1,a_2,a_3,a_4,a_6], G] where [a_1,a_2,a_3,a_4,a_6] are the coefficients of the Weierstrass equation of the curve and G is a Z-basis of the free part of the Mordell-Weil group associated to the curve.

  ? ellsearch("11a3")
  %1 = ["11a3", [0, -1, 1, 0, 0], []]
  ? ellsearch([11,0,3])
  %2 = ["11a3", [0, -1, 1, 0, 0], []]

If N is not a full curve name, then the output is a vector of all matching curves in the above format:

  ? ellsearch("11a")
  %1 = [["11a1", [0, -1, 1, -10, -20], []],
        ["11a2", [0, -1, 1, -7820, -263580], []],
        ["11a3", [0, -1, 1, 0, 0], []]]
  ? ellsearch("11b")
  %2 = []

The library syntax is GEN ellsearch(GEN N). Also available is GEN ellsearchcurve(GEN N) that only accepts complete curve names (as t_STR).

ellsigma(L,{z = 'x},{flag = 0})

Computes the value at z of the Weierstrass sigma function attached to the lattice L as given by ellperiods(,1): including quasi-periods is useful, otherwise there are recomputed from scratch for each new z.

   sigma(z, L) = z prod_{omega belongs to L^*} (1 - (z)/(omega))e^{(z)/(omega) + (z^2)/(2omega^2)}.

It is also possible to directly input L = [omega_1,omega_2], or an elliptic curve E as given by ellinit (L = E.omega).

  ? w = ellperiods([1,I], 1);
  ? ellsigma(w, 1/2)
  %2 = 0.47494937998792065033250463632798296855
  ? E = ellinit([1,0]);
  ? ellsigma(E) \\ at 'x, implicitly at default seriesprecision
  %4 = x + 1/60*x^5 - 1/10080*x^9 - 23/259459200*x^13 + O(x^17)

If flag = 1, computes an arbitrary determination of log (sigma(z)).

The library syntax is GEN ellsigma(GEN L, GEN z = NULL, long flag, long prec).

ellsub(E,z1,z2)

Difference of the points z1 and z2 on the elliptic curve corresponding to E.

The library syntax is GEN ellsub(GEN E, GEN z1, GEN z2).

elltaniyama(E, {d = seriesprecision})

Computes the modular parametrization of the elliptic curve E/Q, where E is an ell structure as output by ellinit. This returns a two-component vector [u,v] of power series, given to d significant terms (seriesprecision by default), characterized by the following two properties. First the point (u,v) satisfies the equation of the elliptic curve. Second, let N be the conductor of E and Phi: X_0(N)\to E be a modular parametrization; the pullback by Phi of the Néron differential du/(2v+a_1u+a_3) is equal to 2iPi f(z)dz, a holomorphic differential form. The variable used in the power series for u and v is x, which is implicitly understood to be equal to exp (2iPi z).

The algorithm assumes that E is a strong Weil curve and that the Manin constant is equal to 1: in fact, f(x) = sum_{n > 0} ellan(E, n) x^n.

The library syntax is GEN elltaniyama(GEN E, long precdl).

elltatepairing(E, P, Q, m)

Computes the Tate pairing of the two points P and Q on the elliptic curve E. The point P must be of m-torsion.

The library syntax is GEN elltatepairing(GEN E, GEN P, GEN Q, GEN m).

elltors(E,{flag = 0})

If E is an elliptic curve defined over Q, outputs the torsion subgroup of E as a 3-component vector [t,v1,v2], where t is the order of the torsion group, v1 gives the structure of the torsion group as a product of cyclic groups (sorted by decreasing order), and v2 gives generators for these cyclic groups. E must be an ell structure as output by ellinit, defined over Q.

  ?  E = ellinit([-1,0]);
  ?  elltors(E)
  %1 = [4, [2, 2], [[0, 0], [1, 0]]]

Here, the torsion subgroup is isomorphic to Z/2Z x Z/2Z, with generators [0,0] and [1,0].

If flag = 0, find rational roots of division polynomials.

If flag = 1, use Lutz-Nagell (much slower).

If flag = 2, use Doud's algorithm: bound torsion by computing #E(F_p) for small primes of good reduction, then look for torsion points using Weierstrass wp function (and Mazur's classification). For this variant, E must be an ell.

The library syntax is GEN elltors0(GEN E, long flag). Also available is GEN elltors(GEN E) for elltors(E, 0).

ellweilpairing(E, P, Q, m)

Computes the Weil pairing of the two points of m-torsion P and Q on the elliptic curve E.

The library syntax is GEN ellweilpairing(GEN E, GEN P, GEN Q, GEN m).

ellwp(w,{z = 'x},{flag = 0})

Computes the value at z of the Weierstrass wp function attached to the lattice w as given by ellperiods. It is also possible to directly input w = [omega_1,omega_2], or an elliptic curve E as given by ellinit (w = E.omega).

  ? w = ellperiods([1,I]);
  ? ellwp(w, 1/2)
  %2 = 6.8751858180203728274900957798105571978
  ? E = ellinit([1,1]);
  ? ellwp(E, 1/2)
  %4 = 3.9413112427016474646048282462709151389

@3One can also compute the series expansion around z = 0:

  ? E = ellinit([1,0]);
  ? ellwp(E)              \\ 'x implicitly at default seriesprecision
  %5 = x^-2 - 1/5*x^2 + 1/75*x^6 - 2/4875*x^10 + O(x^14)
  ? ellwp(E, x + O(x^12)) \\ explicit precision
  %6 = x^-2 - 1/5*x^2 + 1/75*x^6 + O(x^9)

Optional flag means 0 (default): compute only wp (z), 1: compute [ wp (z), wp '(z)].

The library syntax is GEN ellwp0(GEN w, GEN z = NULL, long flag, long prec). For flag = 0, we also have GEN ellwp(GEN w, GEN z, long prec), and GEN ellwpseries(GEN E, long v, long precdl) for the power series in variable v.

ellzeta(w,{z = 'x})

Computes the value at z of the Weierstrass zeta function attached to the lattice w as given by ellperiods(,1): including quasi-periods is useful, otherwise there are recomputed from scratch for each new z.

   zeta(z, L) = (1)/(z) + z^2sum_{omega belongs to L^*} (1)/(omega^2(z-omega)).

It is also possible to directly input w = [omega_1,omega_2], or an elliptic curve E as given by ellinit (w = E.omega). The quasi-periods of zeta, such that

  zeta(z + aomega_1 + bomega_2) = zeta(z) + aeta_1 + beta_2

for integers a and b are obtained as eta_i = 2zeta(omega_i/2). Or using directly elleta.

  ? w = ellperiods([1,I],1);
  ? ellzeta(w, 1/2)
  %2 = 1.5707963267948966192313216916397514421
  ? E = ellinit([1,0]);
  ? ellzeta(E, E.omega[1]/2)
  %4 = 0.84721308479397908660649912348219163647

@3One can also compute the series expansion around z = 0 (the quasi-periods are useless in this case):

  ? E = ellinit([0,1]);
  ? ellzeta(E) \\ at 'x, implicitly at default seriesprecision
  %4 = x^-1 + 1/35*x^5 - 1/7007*x^11 + O(x^15)
  ? ellzeta(E, x + O(x^20)) \\ explicit precision
  %5 = x^-1 + 1/35*x^5 - 1/7007*x^11 + 1/1440257*x^17 + O(x^18)

The library syntax is GEN ellzeta(GEN w, GEN z = NULL, long prec).

ellztopoint(E,z)

E being an ell as output by ellinit, computes the coordinates [x,y] on the curve E corresponding to the complex number z. Hence this is the inverse function of ellpointtoz. In other words, if the curve is put in Weierstrass form y^2 = 4x^3 - g_2x - g_3, [x,y] represents the Weierstrass wp -function -function> and its derivative. More precisely, we have

  x = wp (z) - b_2/12, y = wp '(z) - (a_1 x + a_3)/2.

If z is in the lattice defining E over C, the result is the point at infinity [0].

The library syntax is GEN pointell(GEN E, GEN z, long prec).

genus2red(Q,P,{p})

Let Q,P be polynomials with integer coefficients. Determines the reduction at p > 2 of the (proper, smooth) genus 2 curve C/Q, defined by the hyperelliptic equation y^2+Qy = P. (The special fiber X_p of the minimal regular model X of C over Z.) If p is omitted, determines the reduction type for all (odd) prime divisors of the discriminant.

@3This function rewritten from an implementation of Liu's algorithm by Cohen and Liu (1994), genus2reduction-0.3, see http://www.math.u-bordeaux1.fr/~ liu/G2R/.

@3CAVEAT. The function interface may change: for the time being, it returns [N,FaN, T, V] where N is either the local conductor at p or the global conductor, FaN is its factorization, y^2 = T defines a minimal model over Z[1/2] and V describes the reduction type at the various considered p. Unfortunately, the program is not complete for p = 2, and we may return the odd part of the conductor only: this is the case if the factorization includes the (impossible) term 2^{-1}; if the factorization contains another power of 2, then this is the exact local conductor at 2 and N is the global conductor.

  ? default(debuglevel, 1);
  ? genus2red(0,x^6 + 3*x^3 + 63, 3)
  (potential) stable reduction: [1, []]
  reduction at p: [III{9}] page 184, [3, 3], f = 10
  %1 = [59049, Mat([3, 10]), x^6 + 3*x^3 + 63, [3, [1, []],
         ["[III{9}] page 184", [3, 3]]]]
  ? [N, FaN, T, V] = genus2red(x^3-x^2-1, x^2-x);  \\ X_1(13), global reduction
  p = 13
  (potential) stable reduction: [5, [Mod(0, 13), Mod(0, 13)]]
  reduction at p: [I{0}-II-0] page 159, [], f = 2
  ? N
  %3 = 169
  ? FaN
  %4 = Mat([13, 2])   \\ in particular, good reduction at 2 !
  ? T
  %5 = x^6 + 58*x^5 + 1401*x^4 + 18038*x^3 + 130546*x^2 + 503516*x + 808561
  ? V
  %6 = [[13, [5, [Mod(0, 13), Mod(0, 13)]], ["[I{0}-II-0] page 159", []]]]

We now first describe the format of the vector V = V_p in the case where p was specified (local reduction at p): it is a triple [p, stable, red]. The component stable = [type, vecj] contains information about the stable reduction after a field extension; depending on types, the stable reduction is

@3* 1: smooth (i.e. the curve has potentially good reduction). The Jacobian J(C) has potentially good reduction.

@3* 2: an elliptic curve E with an ordinary double point; vecj contains j mod p, the modular invariant of E. The (potential) semi-abelian reduction of J(C) is the extension of an elliptic curve (with modular invariant j mod p) by a torus.

@3* 3: a projective line with two ordinary double points. The Jacobian J(C) has potentially multiplicative reduction.

@3* 4: the union of two projective lines crossing transversally at three points. The Jacobian J(C) has potentially multiplicative reduction.

@3* 5: the union of two elliptic curves E_1 and E_2 intersecting transversally at one point; vecj contains their modular invariants j_1 and j_2, which may live in a quadratic extension of F_p are need not be distinct. The Jacobian J(C) has potentially good reduction, isomorphic to the product of the reductions of E_1 and E_2.

@3* 6: the union of an elliptic curve E and a projective line which has an ordinary double point, and these two components intersect transversally at one point; vecj contains j mod p, the modular invariant of E. The (potential) semi-abelian reduction of J(C) is the extension of an elliptic curve (with modular invariant j mod p) by a torus.

@3* 7: as in type 6, but the two components are both singular. The Jacobian J(C) has potentially multiplicative reduction.

The component red = [NUtype, neron] contains two data concerning the reduction at p without any ramified field extension.

The NUtype is a t_STR describing the reduction at p of C, following Namikawa-Ueno, The complete classification of fibers in pencils of curves of genus two, Manuscripta Math., vol. 9, (1973), pages 143-186. The reduction symbol is followed by the corresponding page number in this article.

The second datum neron is the group of connected components (over an algebraic closure of F_p) of the Néron model of J(C), given as a finite abelian group (vector of elementary divisors).

If p = 2, the red component may be omitted altogether (and replaced by [], in the case where the program could not compute it. When p was not specified, V is the vector of all V_p, for all considered p.

@3Notes about Namikawa-Ueno types.

@3* A lower index is denoted between braces: for instance, [I{ 2}-II-5] means [I_2-II-5].

@3* If K and K' are Kodaira symbols for singular fibers of elliptic curves, [K-K'-m] and [K'-K-m] are the same.

@3* [K-K'--1] is [K'-K-alpha] in the notation of Namikawa-Ueno.

@3* The figure [2I_0-m] in Namikawa-Ueno, page 159, must be denoted by [2I_0-(m+1)].

The library syntax is GEN genus2red(GEN Q, GEN P, GEN p = NULL).


Functions related to general number fields

In this section can be found functions which are used almost exclusively for working in general number fields. Other less specific functions can be found in the next section on polynomials. Functions related to quadratic number fields are found in section Label se:arithmetic (Arithmetic functions).

Number field structures

Let K = Q[X] / (T) a number field, Z_K its ring of integers, T belongs to Z[X] is monic. Three basic number field structures can be associated to K in GP:

@3* nf denotes a number field, i.e. a data structure output by nfinit. This contains the basic arithmetic data associated to the number field: signature, maximal order (given by a basis nf.zk), discriminant, defining polynomial T, etc.

@3* bnf denotes a ``Buchmann's number field'', i.e. a data structure output by bnfinit. This contains nf and the deeper invariants of the field: units U(K), class group Cl (K), as well as technical data required to solve the two associated discrete logarithm problems.

@3* bnr denotes a ``ray number field'', i.e. a data structure output by bnrinit, corresponding to the ray class group structure of the field, for some modulus f. It contains a bnf, the modulus f, the ray class group Cl _f(K) and data associated to the discrete logarithm problem therein.

Algebraic numbers and ideals

@3An algebraic number belonging to K = Q[X]/(T) is given as

@3* a t_INT, t_FRAC or t_POL (implicitly modulo T), or

@3* a t_POLMOD (modulo T), or

@3* a t_COL v of dimension N = [K:Q], representing the element in terms of the computed integral basis, as sum(i = 1, N, v[i] * nf.zk[i]). Note that a t_VEC will not be recognized.

@3An ideal is given in any of the following ways:

@3* an algebraic number in one of the above forms, defining a principal ideal.

@3* a prime ideal, i.e. a 5-component vector in the format output by idealprimedec or idealfactor.

@3* a t_MAT, square and in Hermite Normal Form (or at least upper triangular with non-negative coefficients), whose columns represent a Z-basis of the ideal.

One may use idealhnf to convert any ideal to the last (preferred) format.

@3* an extended ideal is a 2-component vector [I, t], where I is an ideal as above and t is an algebraic number, representing the ideal (t)I. This is useful whenever idealred is involved, implicitly working in the ideal class group, while keeping track of principal ideals. Ideal operations suitably update the principal part when it makes sense (in a multiplicative context), e.g. using idealmul on [I,t], [J,u], we obtain [IJ, tu]. When it does not make sense, the extended part is silently discarded, e.g. using idealadd with the above input produces I+J.

The ``principal part'' t in an extended ideal may be represented in any of the above forms, and also as a factorization matrix (in terms of number field elements, not ideals!), possibly the empty matrix [;] representing 1. In the latter case, elements stay in factored form, or famat for factorization matrix, which is a convenient way to avoid coefficient explosion. To recover the conventional expanded form, try nffactorback; but many functions already accept famats as input, for instance ideallog, so expanding huge elements should never be necessary.

Finite abelian groups

A finite abelian group G in user-readable format is given by its Smith Normal Form as a pair [h,d] or triple [h,d,g]. Here h is the cardinality of G, (d_i) is the vector of elementary divisors, and (g_i) is a vector of generators. In short, G = oplus _{i <= n} (Z/d_iZ) g_i, with d_n | ... | d_2 | d_1 and prod d_i = h. This information can also be retrieved as G.no, G.cyc and G.gen.

@3* a character on the abelian group oplus (Z/d_iZ) g_i is given by a row vector chi = [a_1,...,a_n] such that chi(prod g_i^{n_i}) = exp (2iPisum a_i n_i / d_i).

@3* given such a structure, a subgroup H is input as a square matrix in HNF, whose columns express generators of H on the given generators g_i. Note that the determinant of that matrix is equal to the index (G:H).

Relative extensions

We now have a look at data structures associated to relative extensions of number fields L/K, and to projective Z_K-modules. When defining a relative extension L/K, the nf associated to the base field K must be defined by a variable having a lower priority (see Label se:priority) than the variable defining the extension. For example, you may use the variable name y to define the base field K, and x to define the relative extension L/K.

Basic definitions\label{se

ZKmodules}

@3* rnf denotes a relative number field, i.e. a data structure output by rnfinit, associated to the extension L/K. The nf associated to be base field K is rnf.nf.

@3* A relative matrix is an m x n matrix whose entries are elements of K, in any form. Its m columns A_j represent elements in K^n.

@3* An ideal list is a row vector of fractional ideals of the number field nf.

@3* A pseudo-matrix is a 2-component row vector (A,I) where A is a relative m x n matrix and I an ideal list of length n. If I = {{a}_1,..., {a}_n} and the columns of A are (A_1,..., A_n), this data defines the torsion-free (projective) Z_K-module {a}_1 A_1 oplus {a}_n A_n.

@3* An integral pseudo-matrix is a 3-component row vector w(A,I,J) where A = (a_{i,j}) is an m x n relative matrix and I = ({b}_1,..., {b}_m), J = ({a}_1,..., {a}_n) are ideal lists, such that a_{i,j} belongs to {b}_i {a}_j^{-1} for all i,j. This data defines two abstract projective Z_K-modules N = {a_1}omega_1 oplus ... oplus {a_n}omega_n in K^n, P = {b_1}eta_1 oplus ... oplus {b_m}eta_m in K^m, and a Z_K-linear map f:N\to P given by

   f(sum alpha_jomega_j) = sum_i (a_{i,j}alpha_j) eta_i.

This data defines the Z_K-module M = P/f(N).

@3* Any projective Z_K-module\varsidx{projective module} M of finite type in K^m can be given by a pseudo matrix (A,I).

@3* An arbitrary Z_K modules of finite type in K^m, with non-trivial torsion, is given by an integral pseudo-matrix (A,I,J)

Pseudo-bases, determinant

@3* The pair (A,I) is a pseudo-basis of the module it generates if the {a_j} are non-zero, and the A_j are K-linearly independent. We call n the size of the pseudo-basis. If A is a relative matrix, the latter condition means it is square with non-zero determinant; we say that it is in Hermite Normal Form (HNF) if it is upper triangular and all the elements of the diagonal are equal to 1.

@3* For instance, the relative integer basis rnf.zk is a pseudo-basis (A,I) of Z_L, where A = rnf.zk[1] is a vector of elements of L, which are K-linearly independent. Most rnf routines return and handle Z_K-modules contained in L (e.g. Z_L-ideals) via a pseudo-basis (A',I'), where A' is a relative matrix representing a vector of elements of L in terms of the fixed basis rnf.zk[1]

@3* The determinant of a pseudo-basis (A,I) is the ideal equal to the product of the determinant of A by all the ideals of I. The determinant of a pseudo-matrix is the determinant of any pseudo-basis of the module it generates.

Class field theory

A modulus, in the sense of class field theory, is a divisor supported on the non-complex places of K. In PARI terms, this means either an ordinary ideal I as above (no Archimedean component), or a pair [I,a], where a is a vector with r_1 {0,1}-components, corresponding to the infinite part of the divisor. More precisely, the i-th component of a corresponds to the real embedding associated to the i-th real root of K.roots. (That ordering is not canonical, but well defined once a defining polynomial for K is chosen.) For instance, [1, [1,1]] is a modulus for a real quadratic field, allowing ramification at any of the two places at infinity, and nowhere else.

A bid or ``big ideal'' is a structure output by idealstar needed to compute in (Z_K/I)^*, where I is a modulus in the above sense. It is a finite abelian group as described above, supplemented by technical data needed to solve discrete log problems.

Finally we explain how to input ray number fields (or bnr), using class field theory. These are defined by a triple A, B, C, where the defining set [A,B,C] can have any of the following forms: [bnr], [bnr,subgroup], [bnf,mod], [bnf,mod,subgroup]. The last two forms are kept for backward compatibility, but no longer serve any real purpose (see example below); no newly written function will accept them.

@3* bnf is as output by bnfinit, where units are mandatory unless the modulus is trivial; bnr is as output by bnrinit. This is the ground field K.

@3* mod is a modulus f, as described above.

@3* subgroup a subgroup of the ray class group modulo f of K. As described above, this is input as a square matrix expressing generators of a subgroup of the ray class group bnr.clgp on the given generators.

The corresponding bnr is the subfield of the ray class field of K modulo f, fixed by the given subgroup.

    ? K = bnfinit(y^2+1);
    ? bnr = bnrinit(K, 13)
    ? %.clgp
    %3 = [36, [12, 3]]
    ? bnrdisc(bnr); \\ discriminant of the full ray class field
    ? bnrdisc(bnr, [3,1;0,1]); \\ discriminant of cyclic cubic extension of K

We could have written directly

    ? bnrdisc(K, 13);
    ? bnrdisc(K, 13, [3,1;0,1]);

avoiding one bnrinit, but this would actually be slower since the bnrinit is called internally anyway. And now twice!

General use

All the functions which are specific to relative extensions, number fields, Buchmann's number fields, Buchmann's number rays, share the prefix rnf, nf, bnf, bnr respectively. They take as first argument a number field of that precise type, respectively output by rnfinit, nfinit, bnfinit, and bnrinit.

However, and even though it may not be specified in the descriptions of the functions below, it is permissible, if the function expects a nf, to use a bnf instead, which contains much more information. On the other hand, if the function requires a bnf, it will not launch bnfinit for you, which is a costly operation. Instead, it will give you a specific error message. In short, the types

   nf <= bnf <= bnr

are ordered, each function requires a minimal type to work properly, but you may always substitute a larger type.

The data types corresponding to the structures described above are rather complicated. Thus, as we already have seen it with elliptic curves, GP provides ``member functions'' to retrieve data from these structures (once they have been initialized of course). The relevant types of number fields are indicated between parentheses:

 bid (bnr ) : bid ideal structure.

 bnf (bnr, bnf ) : Buchmann's number field.

 clgp (bnr, bnf ) : classgroup. This one admits the following three subclasses:

  cyc : cyclic decomposition (SNF).

  gen : generators.

  no : number of elements.

 diff (bnr, bnf, nf ) : the different ideal.

 codiff (bnr, bnf, nf ) : the codifferent (inverse of the different in the ideal group).

 disc (bnr, bnf, nf ) : discriminant.

 fu (bnr, bnf ) : fundamental units.

 index (bnr, bnf, nf ) : index of the power order in the ring of integers.

 mod (bnr ) : modulus.

 nf (bnr, bnf, nf ) : number field.

 pol (bnr, bnf, nf ) : defining polynomial.

 r1 (bnr, bnf, nf ) : the number of real embeddings.

 r2 (bnr, bnf, nf ) : the number of pairs of complex embeddings.

 reg (bnr, bnf ) : regulator.

 roots (bnr, bnf, nf ) : roots of the polynomial generating the field.

 sign (bnr, bnf, nf ) : signature [r1,r2].

 t2 (bnr, bnf, nf ) : the T_2 matrix (see nfinit).

 tu (bnr, bnf ) : a generator for the torsion units.

 zk (bnr, bnf, nf ) : integral basis, i.e. a Z-basis of the maximal order.

 zkst (bnr ) : structure of (Z_K/m)^*.

@3Deprecated. The following member functions are still available, but deprecated and should not be used in new scripts :

 futu (bnr, bnf, ) : [u_1,...,u_r,w], (u_i) is a vector of fundamental units,

  w generates the torsion units.

 tufu (bnr, bnf, ) : [w,u_1,...,u_r], (u_i) is a vector of fundamental units,

  w generates the torsion units.

For instance, assume that bnf = bnfinit(pol), for some polynomial. Then bnf.clgp retrieves the class group, and bnf.clgp.no the class number. If we had set bnf = nfinit(pol), both would have output an error message. All these functions are completely recursive, thus for instance bnr.bnf.nf.zk will yield the maximal order of bnr, which you could get directly with a simple bnr.zk.

Class group, units, and the GRH

Some of the functions starting with bnf are implementations of the sub-exponential algorithms for finding class and unit groups under GRH, due to Hafner-McCurley, Buchmann and Cohen-Diaz-Olivier. The general call to the functions concerning class groups of general number fields (i.e. excluding quadclassunit) involves a polynomial P and a technical vector

  tech = [c_1, c_2, nrpid ],

where the parameters are to be understood as follows:

P is the defining polynomial for the number field, which must be in Z[X], irreducible and monic. In fact, if you supply a non-monic polynomial at this point, gp issues a warning, then transforms your polynomial so that it becomes monic. The nfinit routine will return a different result in this case: instead of res, you get a vector [res,Mod(a,Q)], where Mod(a,Q) = Mod(X,P) gives the change of variables. In all other routines, the variable change is simply lost.

The tech interface is obsolete and you should not tamper with these parameters. Indeed, from version 2.4.0 on,

@3* the results are always rigorous under GRH (before that version, they relied on a heuristic strengthening, hence the need for overrides).

@3* the influence of these parameters on execution time and stack size is marginal. They can be useful to fine-tune and experiment with the bnfinit code, but you will be better off modifying all tuning parameters in the C code (there are many more than just those three). We nevertheless describe it for completeness.

The numbers c_1 <= c_2 are non-negative real numbers. By default they are chosen so that the result is correct under GRH. For i = 1,2, let B_i = c_i( log |d_K|)^2, and denote by S(B) the set of maximal ideals of K whose norm is less than B. We want S(B_1) to generate Cl (K) and hope that S(B_2) can be proven to generate Cl (K).

More precisely, S(B_1) is a factorbase used to compute a tentative Cl (K) by generators and relations. We then check explicitly, using essentially bnfisprincipal, that the elements of S(B_2) belong to the span of S(B_1). Under the assumption that S(B_2) generates Cl (K), we are done. User-supplied c_i are only used to compute initial guesses for the bounds B_i, and the algorithm increases them until one can prove under GRH that S(B_2) generates Cl (K). A uniform result of Bach says that c_2 = 12 is always suitable, but this bound is very pessimistic and a direct algorithm due to Belabas-Diaz-Friedman is used to check the condition, assuming GRH. The default values are c_1 = c_2 = 0. When c_1 is equal to 0 the algorithm takes it equal to c_2.

nrpid is the maximal number of small norm relations associated to each ideal in the factor base. Set it to 0 to disable the search for small norm relations. Otherwise, reasonable values are between 4 and 20. The default is 4.

@3Warning. Make sure you understand the above! By default, most of the bnf routines depend on the correctness of the GRH. In particular, any of the class number, class group structure, class group generators, regulator and fundamental units may be wrong, independently of each other. Any result computed from such a bnf may be wrong. The only guarantee is that the units given generate a subgroup of finite index in the full unit group. You must use bnfcertify to certify the computations unconditionally.

@3Remarks.

You do not need to supply the technical parameters (under the library you still need to send at least an empty vector, coded as NULL). However, should you choose to set some of them, they must be given in the requested order. For example, if you want to specify a given value of nrpid, you must give some values as well for c_1 and c_2, and provide a vector [c_1,c_2,nrpid].

Note also that you can use an nf instead of P, which avoids recomputing the integral basis and analogous quantities.

bnfcertify(bnf,{flag = 0})

bnf being as output by bnfinit, checks whether the result is correct, i.e. whether it is possible to remove the assumption of the Generalized Riemann Hypothesis. It is correct if and only if the answer is 1. If it is incorrect, the program may output some error message, or loop indefinitely. You can check its progress by increasing the debug level. The bnf structure must contain the fundamental units:

  ? K = bnfinit(x^3+2^2^3+1); bnfcertify(K)
    ***   at top-level: K=bnfinit(x^3+2^2^3+1);bnfcertify(K)
    ***                                        ^-------------
    *** bnfcertify: missing units in bnf.
  ? K = bnfinit(x^3+2^2^3+1, 1); \\ include units
  ? bnfcertify(K)
  %3 = 1

If flag is present, only certify that the class group is a quotient of the one computed in bnf (much simpler in general); likewise, the computed units may form a subgroup of the full unit group. In this variant, the units are no longer needed:

  ? K = bnfinit(x^3+2^2^3+1); bnfcertify(K, 1)
  %4 = 1

The library syntax is long bnfcertify0(GEN bnf, long flag ). Also available is GEN bnfcertify(GEN bnf) (flag = 0).

bnfcompress(bnf)

Computes a compressed version of bnf (from bnfinit), a ``small Buchmann's number field'' (or sbnf for short) which contains enough information to recover a full bnf vector very rapidly, but which is much smaller and hence easy to store and print. Calling bnfinit on the result recovers a true bnf, in general different from the original. Note that an snbf is useless for almost all purposes besides storage, and must be converted back to bnf form before use; for instance, no nf*, bnf* or member function accepts them.

An sbnf is a 12 component vector v, as follows. Let bnf be the result of a full bnfinit, complete with units. Then v[1] is bnf.pol, v[2] is the number of real embeddings bnf.sign[1], v[3] is bnf.disc, v[4] is bnf.zk, v[5] is the list of roots bnf.roots, v[7] is the matrix W = bnf[1], v[8] is the matrix matalpha = bnf[2], v[9] is the prime ideal factor base bnf[5] coded in a compact way, and ordered according to the permutation bnf[6], v[10] is the 2-component vector giving the number of roots of unity and a generator, expressed on the integral basis, v[11] is the list of fundamental units, expressed on the integral basis, v[12] is a vector containing the algebraic numbers alpha corresponding to the columns of the matrix matalpha, expressed on the integral basis.

All the components are exact (integral or rational), except for the roots in v[5].

The library syntax is GEN bnfcompress(GEN bnf).

bnfdecodemodule(nf,m)

If m is a module as output in the first component of an extension given by bnrdisclist, outputs the true module.

  ? K = bnfinit(x^2+23); L = bnrdisclist(K, 10); s = L[1][2]
  %1 = [[Mat([8, 1]), [[0, 0, 0]]], [Mat([9, 1]), [[0, 0, 0]]]]
  ? bnfdecodemodule(K, s[1][1])
  %2 =
  [2 0]
  [0 1]

The library syntax is GEN decodemodule(GEN nf, GEN m).

bnfinit(P,{flag = 0},{tech = []})

Initializes a bnf structure. Used in programs such as bnfisprincipal, bnfisunit or bnfnarrow. By default, the results are conditional on the GRH, see se:GRHbnf. The result is a 10-component vector bnf.

This implements Buchmann's sub-exponential algorithm for computing the class group, the regulator and a system of fundamental units of the general algebraic number field K defined by the irreducible polynomial P with integer coefficients.

If the precision becomes insufficient, gp does not strive to compute the units by default (flag = 0).

When flag = 1, we insist on finding the fundamental units exactly. Be warned that this can take a very long time when the coefficients of the fundamental units on the integral basis are very large. If the fundamental units are simply too large to be represented in this form, an error message is issued. They could be obtained using the so-called compact representation of algebraic numbers as a formal product of algebraic integers. The latter is implemented internally but not publicly accessible yet.

tech is a technical vector (empty by default, see se:GRHbnf). Careful use of this parameter may speed up your computations, but it is mostly obsolete and you should leave it alone.

The components of a bnf or sbnf are technical and never used by the casual user. In fact: never access a component directly, always use a proper member function. However, for the sake of completeness and internal documentation, their description is as follows. We use the notations explained in the book by H. Cohen, A Course in Computational Algebraic Number Theory, Graduate Texts in Maths 138, Springer-Verlag, 1993, Section 6.5, and subsection 6.5.5 in particular.

bnf[1] contains the matrix W, i.e. the matrix in Hermite normal form giving relations for the class group on prime ideal generators (p_i)_{1 <= i <= r}.

bnf[2] contains the matrix B, i.e. the matrix containing the expressions of the prime ideal factorbase in terms of the p_i. It is an r x c matrix.

bnf[3] contains the complex logarithmic embeddings of the system of fundamental units which has been found. It is an (r_1+r_2) x (r_1+r_2-1) matrix.

bnf[4] contains the matrix M''_C of Archimedean components of the relations of the matrix (W|B).

bnf[5] contains the prime factor base, i.e. the list of prime ideals used in finding the relations.

bnf[6] used to contain a permutation of the prime factor base, but has been obsoleted. It contains a dummy 0.

bnf[7] or bnf.nf is equal to the number field data nf as would be given by nfinit.

bnf[8] is a vector containing the classgroup bnf.clgp as a finite abelian group, the regulator bnf.reg, a 1 (used to contain an obsolete ``check number''), the number of roots of unity and a generator bnf.tu, the fundamental units bnf.fu.

bnf[9] is a 3-element row vector used in bnfisprincipal only and obtained as follows. Let D = U W V obtained by applying the Smith normal form algorithm to the matrix W ( = bnf[1]) and let U_r be the reduction of U modulo D. The first elements of the factorbase are given (in terms of bnf.gen) by the columns of U_r, with Archimedean component g_a; let also GD_a be the Archimedean components of the generators of the (principal) ideals defined by the bnf.gen[i]^bnf.cyc[i]. Then bnf[9] = [U_r, g_a, GD_a].

bnf[10] is by default unused and set equal to 0. This field is used to store further information about the field as it becomes available, which is rarely needed, hence would be too expensive to compute during the initial bnfinit call. For instance, the generators of the principal ideals bnf.gen[i]^bnf.cyc[i] (during a call to bnrisprincipal), or those corresponding to the relations in W and B (when the bnf internal precision needs to be increased).

The library syntax is GEN bnfinit0(GEN P, long flag, GEN tech = NULL, long prec).

Also available is GEN Buchall(GEN P, long flag, long prec), corresponding to tech = NULL, where flag is either 0 (default) or nf_FORCE (insist on finding fundamental units). The function GEN Buchall_param(GEN P, double c1, double c2, long nrpid, long flag, long prec) gives direct access to the technical parameters.

bnfisintnorm(bnf,x)

Computes a complete system of solutions (modulo units of positive norm) of the absolute norm equation Norm (a) = x, where a is an integer in bnf. If bnf has not been certified, the correctness of the result depends on the validity of GRH.

See also bnfisnorm.

The library syntax is GEN bnfisintnorm(GEN bnf, GEN x). The function GEN bnfisintnormabs(GEN bnf, GEN a) returns a complete system of solutions modulo units of the absolute norm equation | Norm (x) |= |a|. As fast as bnfisintnorm, but solves the two equations Norm (x) = +- a simultaneously.

bnfisnorm(bnf,x,{flag = 1})

Tries to tell whether the rational number x is the norm of some element y in bnf. Returns a vector [a,b] where x = Norm(a)*b. Looks for a solution which is an S-unit, with S a certain set of prime ideals containing (among others) all primes dividing x. If bnf is known to be Galois, set flag = 0 (in this case, x is a norm iff b = 1). If flag is non zero the program adds to S the following prime ideals, depending on the sign of flag. If flag > 0, the ideals of norm less than flag. And if flag < 0 the ideals dividing flag.

Assuming GRH, the answer is guaranteed (i.e. x is a norm iff b = 1), if S contains all primes less than 12 log ( disc (Bnf))^2, where Bnf is the Galois closure of bnf.

See also bnfisintnorm.

The library syntax is GEN bnfisnorm(GEN bnf, GEN x, long flag).

bnfisprincipal(bnf,x,{flag = 1})

bnf being the number field data output by bnfinit, and x being an ideal, this function tests whether the ideal is principal or not. The result is more complete than a simple true/false answer and solves general discrete logarithm problem. Assume the class group is oplus (Z/d_iZ)g_i (where the generators g_i and their orders d_i are respectively given by bnf.gen and bnf.cyc). The routine returns a row vector [e,t], where e is a vector of exponents 0 <= e_i < d_i, and t is a number field element such that

   x = (t) prod_i g_i^{e_i}.

For given g_i (i.e. for a given bnf), the e_i are unique, and t is unique modulo units.

In particular, x is principal if and only if e is the zero vector. Note that the empty vector, which is returned when the class number is 1, is considered to be a zero vector (of dimension 0).

  ? K = bnfinit(y^2+23);
  ? K.cyc
  %2 = [3]
  ? K.gen
  %3 = [[2, 0; 0, 1]]          \\ a prime ideal above 2
  ? P = idealprimedec(K,3)[1]; \\ a prime ideal above 3
  ? v = bnfisprincipal(K, P)
  %5 = [[2]~, [3/4, 1/4]~]
  ? idealmul(K, v[2], idealfactorback(K, K.gen, v[1]))
  %6 =
  [3 0]
  [0 1]
  ? % == idealhnf(K, P)
  %7 = 1

@3The binary digits of flag mean:

@3* 1: If set, outputs [e,t] as explained above, otherwise returns only e, which is much easier to compute. The following idiom only tests whether an ideal is principal:

    is_principal(bnf, x) = !bnfisprincipal(bnf,x,0);

@3* 2: It may not be possible to recover t, given the initial accuracy to which bnf was computed. In that case, a warning is printed and t is set equal to the empty vector []~. If this bit is set, increase the precision and recompute needed quantities until t can be computed. Warning: setting this may induce very lengthy computations.

The library syntax is GEN bnfisprincipal0(GEN bnf, GEN x, long flag). Instead of the above hardcoded numerical flags, one should rather use an or-ed combination of the symbolic flags nf_GEN (include generators, possibly a place holder if too difficult) and nf_FORCE (insist on finding the generators).

bnfissunit(bnf,sfu,x)

bnf being output by bnfinit, sfu by bnfsunit, gives the column vector of exponents of x on the fundamental S-units and the roots of unity. If x is not a unit, outputs an empty vector.

The library syntax is GEN bnfissunit(GEN bnf, GEN sfu, GEN x).

bnfisunit(bnf,x)

bnf being the number field data output by bnfinit and x being an algebraic number (type integer, rational or polmod), this outputs the decomposition of x on the fundamental units and the roots of unity if x is a unit, the empty vector otherwise. More precisely, if u_1,...,u_r are the fundamental units, and zeta is the generator of the group of roots of unity (bnf.tu), the output is a vector [x_1,...,x_r,x_{r+1}] such that x = u_1^{x_1}... u_r^{x_r}.zeta^{x_{r+1}}. The x_i are integers for i <= r and is an integer modulo the order of zeta for i = r+1.

Note that bnf need not contain the fundamental unit explicitly:

  ? setrand(1); bnf = bnfinit(x^2-x-100000);
  ? bnf.fu
    ***   at top-level: bnf.fu
    ***                     ^--
    *** _.fu: missing units in .fu.
  ? u = [119836165644250789990462835950022871665178127611316131167, \
         379554884019013781006303254896369154068336082609238336]~;
  ? bnfisunit(bnf, u)
  %3 = [-1, Mod(0, 2)]~

@3The given u is the inverse of the fundamental unit implicitly stored in bnf. In this case, the fundamental unit was not computed and stored in algebraic form since the default accuracy was too low. (Re-run the command at \g1 or higher to see such diagnostics.)

The library syntax is GEN bnfisunit(GEN bnf, GEN x).

bnfnarrow(bnf)

bnf being as output by bnfinit, computes the narrow class group of bnf. The output is a 3-component row vector v analogous to the corresponding class group component bnf.clgp (bnf[8][1]): the first component is the narrow class number v.no, the second component is a vector containing the SNF cyclic components v.cyc of the narrow class group, and the third is a vector giving the generators of the corresponding v.gen cyclic groups. Note that this function is a special case of bnrinit.

The library syntax is GEN buchnarrow(GEN bnf).

bnfsignunit(bnf)

bnf being as output by bnfinit, this computes an r_1 x (r_1+r_2-1) matrix having +-1 components, giving the signs of the real embeddings of the fundamental units. The following functions compute generators for the totally positive units:

  /* exponents of totally positive units generators on bnf.tufu */
  tpuexpo(bnf)=
  { my(S,d,K);
    S = bnfsignunit(bnf); d = matsize(S);
    S = matrix(d[1],d[2], i,j, if (S[i,j] < 0, 1,0));
    S = concat(vectorv(d[1],i,1), S);   \\ add sign(-1)
    K = lift(matker(S * Mod(1,2)));
    if (K, mathnfmodid(K, 2), 2*matid(d[1]))
  }
  /* totally positive units */
  tpu(bnf)=
  { my(vu = bnf.tufu, ex = tpuexpo(bnf));
    vector(#ex-1, i, factorback(vu, ex[,i+1]))  \\ ex[,1] is 1
  }

The library syntax is GEN signunits(GEN bnf).

bnfsunit(bnf,S)

Computes the fundamental S-units of the number field bnf (output by bnfinit), where S is a list of prime ideals (output by idealprimedec). The output is a vector v with 6 components.

v[1] gives a minimal system of (integral) generators of the S-unit group modulo the unit group.

v[2] contains technical data needed by bnfissunit.

v[3] is an empty vector (used to give the logarithmic embeddings of the generators in v[1] in version 2.0.16).

v[4] is the S-regulator (this is the product of the regulator, the determinant of v[2] and the natural logarithms of the norms of the ideals in S).

v[5] gives the S-class group structure, in the usual format (a row vector whose three components give in order the S-class number, the cyclic components and the generators).

v[6] is a copy of S.

The library syntax is GEN bnfsunit(GEN bnf, GEN S, long prec).

bnrL1(bnr, {H}, {flag = 0})

Let bnr be the number field data output by bnrinit(,,1) and H be a square matrix defining a congruence subgroup of the ray class group corresponding to bnr (the trivial congruence subgroup if omitted). This function returns, for each character chi of the ray class group which is trivial on H, the value at s = 1 (or s = 0) of the abelian L-function associated to chi. For the value at s = 0, the function returns in fact for each chi a vector [r_chi, c_chi] where

  L(s, chi) = c.s^r + O(s^{r + 1})

@3near 0.

The argument flag is optional, its binary digits mean 1: compute at s = 0 if unset or s = 1 if set, 2: compute the primitive L-function associated to chi if unset or the L-function with Euler factors at prime ideals dividing the modulus of bnr removed if set (that is L_S(s, chi), where S is the set of infinite places of the number field together with the finite prime ideals dividing the modulus of bnr), 3: return also the character if set.

  K = bnfinit(x^2-229);
  bnr = bnrinit(K,1,1);
  bnrL1(bnr)

returns the order and the first non-zero term of L(s, chi) at s = 0 where chi runs through the characters of the class group of K = Q( sqrt {229}). Then

  bnr2 = bnrinit(K,2,1);
  bnrL1(bnr2,,2)

returns the order and the first non-zero terms of L_S(s, chi) at s = 0 where chi runs through the characters of the class group of K and S is the set of infinite places of K together with the finite prime 2. Note that the ray class group modulo 2 is in fact the class group, so bnrL1(bnr2,0) returns the same answer as bnrL1(bnr,0).

This function will fail with the message

   *** bnrL1: overflow in zeta_get_N0 [need too many primes].

@3if the approximate functional equation requires us to sum too many terms (if the discriminant of K is too large).

The library syntax is GEN bnrL1(GEN bnr, GEN H = NULL, long flag, long prec).

bnrclassno(A,{B},{C})

Let A, B, C define a class field L over a ground field K (of type [bnr], [bnr, subgroup], or [bnf, modulus], or [bnf, modulus,subgroup], Label se:CFT); this function returns the relative degree [L:K].

In particular if A is a bnf (with units), and B a modulus, this function returns the corresponding ray class number modulo B. One can input the associated bid (with generators if the subgroup C is non trivial) for B instead of the module itself, saving some time.

This function is faster than bnrinit and should be used if only the ray class number is desired. See bnrclassnolist if you need ray class numbers for all moduli less than some bound.

The library syntax is GEN bnrclassno0(GEN A, GEN B = NULL, GEN C = NULL). Also available is GEN bnrclassno(GEN bnf,GEN f) to compute the ray class number modulo f.

bnrclassnolist(bnf,list)

bnf being as output by bnfinit, and list being a list of moduli (with units) as output by ideallist or ideallistarch, outputs the list of the class numbers of the corresponding ray class groups. To compute a single class number, bnrclassno is more efficient.

  ? bnf = bnfinit(x^2 - 2);
  ? L = ideallist(bnf, 100, 2);
  ? H = bnrclassnolist(bnf, L);
  ? H[98]
  %4 = [1, 3, 1]
  ? l = L[1][98]; ids = vector(#l, i, l[i].mod[1])
  %5 = [[98, 88; 0, 1], [14, 0; 0, 7], [98, 10; 0, 1]]

The weird l[i].mod[1], is the first component of l[i].mod, i.e. the finite part of the conductor. (This is cosmetic: since by construction the Archimedean part is trivial, I do not want to see it). This tells us that the ray class groups modulo the ideals of norm 98 (printed as %5) have respectively order 1, 3 and 1. Indeed, we may check directly:

  ? bnrclassno(bnf, ids[2])
  %6 = 3

The library syntax is GEN bnrclassnolist(GEN bnf, GEN list).

bnrconductor(A,{B},{C},{flag = 0})

Conductor f of the subfield of a ray class field as defined by [A,B,C] (of type [bnr], [bnr, subgroup], [bnf, modulus] or [bnf, modulus, subgroup], Label se:CFT)

If flag = 0, returns f.

If flag = 1, returns [f, Cl_f, H], where Cl_f is the ray class group modulo f, as a finite abelian group; finally H is the subgroup of Cl_f defining the extension.

If flag = 2, returns [f, bnr(f), H], as above except Cl_f is replaced by a bnr structure, as output by bnrinit(,f,1).

The library syntax is GEN bnrconductor0(GEN A, GEN B = NULL, GEN C = NULL, long flag).

Also available is GEN bnrconductor(GEN bnr, GEN H, long flag)

bnrconductorofchar(bnr,chi)

bnr being a big ray number field as output by bnrinit, and chi being a row vector representing a character as expressed on the generators of the ray class group, gives the conductor of this character as a modulus.

The library syntax is GEN bnrconductorofchar(GEN bnr, GEN chi).

bnrdisc(A,{B},{C},{flag = 0})

A, B, C defining a class field L over a ground field K (of type [bnr], [bnr, subgroup], [bnf, modulus] or [bnf, modulus, subgroup], Label se:CFT), outputs data [N,r_1,D] giving the discriminant and signature of L, depending on the binary digits of flag:

@3* 1: if this bit is unset, output absolute data related to L/Q: N is the absolute degree [L:Q], r_1 the number of real places of L, and D the discriminant of L/Q. Otherwise, output relative data for L/K: N is the relative degree [L:K], r_1 is the number of real places of K unramified in L (so that the number of real places of L is equal to r_1 times N), and D is the relative discriminant ideal of L/K.

@3* 2: if this bit is set and if the modulus is not the conductor of L, only return 0.

The library syntax is GEN bnrdisc0(GEN A, GEN B = NULL, GEN C = NULL, long flag).

bnrdisclist(bnf,bound,{arch})

bnf being as output by bnfinit (with units), computes a list of discriminants of Abelian extensions of the number field by increasing modulus norm up to bound bound. The ramified Archimedean places are given by arch; all possible values are taken if arch is omitted.

The alternative syntax bnrdisclist(bnf,list) is supported, where list is as output by ideallist or ideallistarch (with units), in which case arch is disregarded.

The output v is a vector of vectors, where v[i][j] is understood to be in fact V[2^{15}(i-1)+j] of a unique big vector V. (This awkward scheme allows for larger vectors than could be otherwise represented.)

V[k] is itself a vector W, whose length is the number of ideals of norm k. We consider first the case where arch was specified. Each component of W corresponds to an ideal m of norm k, and gives invariants associated to the ray class field L of bnf of conductor [m, arch]. Namely, each contains a vector [m,d,r,D] with the following meaning: m is the prime ideal factorization of the modulus, d = [L:Q] is the absolute degree of L, r is the number of real places of L, and D is the factorization of its absolute discriminant. We set d = r = D = 0 if m is not the finite part of a conductor.

If arch was omitted, all t = 2^{r_1} possible values are taken and a component of W has the form [m, [[d_1,r_1,D_1],..., [d_t,r_t,D_t]]], where m is the finite part of the conductor as above, and [d_i,r_i,D_i] are the invariants of the ray class field of conductor [m,v_i], where v_i is the i-th Archimedean component, ordered by inverse lexicographic order; so v_1 = [0,...,0], v_2 = [1,0...,0], etc. Again, we set d_i = r_i = D_i = 0 if [m,v_i] is not a conductor.

Finally, each prime ideal pr = [p,alpha,e,f,beta] in the prime factorization m is coded as the integer p.n^2+(f-1).n+(j-1), where n is the degree of the base field and j is such that

pr = idealprimedec(nf,p)[j].

@3m can be decoded using bnfdecodemodule.

Note that to compute such data for a single field, either bnrclassno or bnrdisc is more efficient.

The library syntax is GEN bnrdisclist0(GEN bnf, GEN bound, GEN arch = NULL).

bnrinit(bnf,f,{flag = 0})

bnf is as output by bnfinit, f is a modulus, initializes data linked to the ray class group structure corresponding to this module, a so-called bnr structure. One can input the associated bid with generators for f instead of the module itself, saving some time. (As in idealstar, the finite part of the conductor may be given by a factorization into prime ideals, as produced by idealfactor.)

The following member functions are available on the result: .bnf is the underlying bnf, .mod the modulus, .bid the bid structure associated to the modulus; finally, .clgp, .no, .cyc, .gen refer to the ray class group (as a finite abelian group), its cardinality, its elementary divisors, its generators (only computed if flag = 1).

The last group of functions are different from the members of the underlying bnf, which refer to the class group; use bnr.bnf.xxx to access these, e.g. bnr.bnf.cyc to get the cyclic decomposition of the class group.

They are also different from the members of the underlying bid, which refer to (Z_K/f)^*; use bnr.bid.xxx to access these, e.g. bnr.bid.no to get phi(f).

If flag = 0 (default), the generators of the ray class group are not computed, which saves time. Hence bnr.gen would produce an error.

If flag = 1, as the default, except that generators are computed.

The library syntax is GEN bnrinit0(GEN bnf, GEN f, long flag). Instead the above hardcoded numerical flags, one should rather use GEN Buchray(GEN bnf, GEN module, long flag) where flag is an or-ed combination of nf_GEN (include generators) and nf_INIT (if omitted, return just the cardinal of the ray class group and its structure), possibly 0.

bnrisconductor(A,{B},{C})

A, B, C represent an extension of the base field, given by class field theory (see Label se:CFT). Outputs 1 if this modulus is the conductor, and 0 otherwise. This is slightly faster than bnrconductor.

The library syntax is long bnrisconductor0(GEN A, GEN B = NULL, GEN C = NULL).

bnrisprincipal(bnr,x,{flag = 1})

bnr being the number field data which is output by bnrinit(,,1) and x being an ideal in any form, outputs the components of x on the ray class group generators in a way similar to bnfisprincipal. That is a 2-component vector v where v[1] is the vector of components of x on the ray class group generators, v[2] gives on the integral basis an element alpha such that x = alphaprod_ig_i^{x_i}.

If flag = 0, outputs only v_1. In that case, bnr need not contain the ray class group generators, i.e. it may be created with bnrinit(,,0) If x is not coprime to the modulus of bnr the result is undefined.

The library syntax is GEN bnrisprincipal(GEN bnr, GEN x, long flag). Instead of hardcoded numerical flags, one should rather use GEN isprincipalray(GEN bnr, GEN x) for flag = 0, and if you want generators:

    bnrisprincipal(bnr, x, nf_GEN)

bnrrootnumber(bnr,chi,{flag = 0})

If chi = chi is a character over bnr, not necessarily primitive, let L(s,chi) = sum_{id} chi(id) N(id)^{-s} be the associated Artin L-function. Returns the so-called Artin root number, i.e. the complex number W(chi) of modulus 1 such that

  Lambda(1-s,chi) = W(chi) Lambda(s,\overline{chi})

@3where Lambda(s,chi) = A(chi)^{s/2}gamma_chi(s) L(s,chi) is the enlarged L-function associated to L.

The generators of the ray class group are needed, and you can set flag = 1 if the character is known to be primitive. Example:

  bnf = bnfinit(x^2 - x - 57);
  bnr = bnrinit(bnf, [7,[1,1]], 1);
  bnrrootnumber(bnr, [2,1])

returns the root number of the character chi of Cl _{7 oo _1 oo _2}(Q( sqrt {229})) defined by chi(g_1^ag_2^b) = zeta_1^{2a}zeta_2^b. Here g_1, g_2 are the generators of the ray-class group given by bnr.gen and zeta_1 = e^{2iPi/N_1}, zeta_2 = e^{2iPi/N_2} where N_1, N_2 are the orders of g_1 and g_2 respectively (N_1 = 6 and N_2 = 3 as bnr.cyc readily tells us).

The library syntax is GEN bnrrootnumber(GEN bnr, GEN chi, long flag, long prec).

bnrstark(bnr,{subgroup})

bnr being as output by bnrinit(,,1), finds a relative equation for the class field corresponding to the modulus in bnr and the given congruence subgroup (as usual, omit subgroup if you want the whole ray class group).

The main variable of bnr must not be x, and the ground field and the class field must be totally real. When the base field is Q, the vastly simpler galoissubcyclo is used instead. Here is an example:

  bnf = bnfinit(y^2 - 3);
  bnr = bnrinit(bnf, 5, 1);
  bnrstark(bnr)

returns the ray class field of Q( sqrt {3}) modulo 5. Usually, one wants to apply to the result one of

  rnfpolredabs(bnf, pol, 16)     \\ compute a reduced relative polynomial
  rnfpolredabs(bnf, pol, 16 + 2) \\ compute a reduced absolute polynomial

The routine uses Stark units and needs to find a suitable auxiliary conductor, which may not exist when the class field is not cyclic over the base. In this case bnrstark is allowed to return a vector of polynomials defining independent relative extensions, whose compositum is the requested class field. It was decided that it was more useful to keep the extra information thus made available, hence the user has to take the compositum herself.

Even if it exists, the auxiliary conductor may be so large that later computations become unfeasible. (And of course, Stark's conjecture may simply be wrong.) In case of difficulties, try rnfkummer:

  ? bnr = bnrinit(bnfinit(y^8-12*y^6+36*y^4-36*y^2+9,1), 2, 1);
  ? bnrstark(bnr)
    ***   at top-level: bnrstark(bnr)
    ***                 ^-------------
    *** bnrstark: need 3919350809720744 coefficients in initzeta.
    *** Computation impossible.
  ? lift( rnfkummer(bnr) )
  time = 24 ms.
  %2 = x^2 + (1/3*y^6 - 11/3*y^4 + 8*y^2 - 5)

The library syntax is GEN bnrstark(GEN bnr, GEN subgroup = NULL, long prec).

dirzetak(nf,b)

Gives as a vector the first b coefficients of the Dedekind zeta function of the number field nf considered as a Dirichlet series.

The library syntax is GEN dirzetak(GEN nf, GEN b).

factornf(x,t)

Factorization of the univariate polynomial x over the number field defined by the (univariate) polynomial t. x may have coefficients in Q or in the number field. The algorithm reduces to factorization over Q (Trager's trick). The direct approach of nffactor, which uses van Hoeij's method in a relative setting, is in general faster.

The main variable of t must be of lower priority than that of x (see Label se:priority). However if non-rational number field elements occur (as polmods or polynomials) as coefficients of x, the variable of these polmods must be the same as the main variable of t. For example

  ? factornf(x^2 + Mod(y, y^2+1), y^2+1);
  ? factornf(x^2 + y, y^2+1); \\ these two are OK
  ? factornf(x^2 + Mod(z,z^2+1), y^2+1)
    ***   at top-level: factornf(x^2+Mod(z,z
    ***                 ^--------------------
    *** factornf: inconsistent data in rnf function.
  ? factornf(x^2 + z, y^2+1)
    ***   at top-level: factornf(x^2+z,y^2+1
    ***                 ^--------------------
    *** factornf: incorrect variable in rnf function.

The library syntax is GEN polfnf(GEN x, GEN t).

galoisexport(gal,{flag})

gal being be a Galois group as output by galoisinit, export the underlying permutation group as a string suitable for (no flags or flag = 0) GAP or (flag = 1) Magma. The following example compute the index of the underlying abstract group in the GAP library:

  ? G = galoisinit(x^6+108);
  ? s = galoisexport(G)
  %2 = "Group((1, 2, 3)(4, 5, 6), (1, 4)(2, 6)(3, 5))"
  ? extern("echo \"IdGroup("s");\" | gap -q")
  %3 = [6, 1]
  ? galoisidentify(G)
  %4 = [6, 1]

This command also accepts subgroups returned by galoissubgroups.

To import a GAP permutation into gp (for galoissubfields for instance), the following GAP function may be useful:

  PermToGP := function(p, n)
    return Permuted([1..n],p);
  end;
  gap> p:= (1,26)(2,5)(3,17)(4,32)(6,9)(7,11)(8,24)(10,13)(12,15)(14,27)
    (16,22)(18,28)(19,20)(21,29)(23,31)(25,30)
  gap> PermToGP(p,32);
  [ 26, 5, 17, 32, 2, 9, 11, 24, 6, 13, 7, 15, 10, 27, 12, 22, 3, 28, 20, 19,
    29, 16, 31, 8, 30, 1, 14, 18, 21, 25, 23, 4 ]

The library syntax is GEN galoisexport(GEN gal, long flag).

galoisfixedfield(gal,perm,{flag},{v = y})

gal being be a Galois group as output by galoisinit and perm an element of gal.group, a vector of such elements or a subgroup of gal as returned by galoissubgroups, computes the fixed field of gal by the automorphism defined by the permutations perm of the roots gal.roots. P is guaranteed to be squarefree modulo gal.p.

If no flags or flag = 0, output format is the same as for nfsubfield, returning [P,x] such that P is a polynomial defining the fixed field, and x is a root of P expressed as a polmod in gal.pol.

If flag = 1 return only the polynomial P.

If flag = 2 return [P,x,F] where P and x are as above and F is the factorization of gal.pol over the field defined by P, where variable v (y by default) stands for a root of P. The priority of v must be less than the priority of the variable of gal.pol (see Label se:priority). Example:

  ? G = galoisinit(x^4+1);
  ? galoisfixedfield(G,G.group[2],2)
  %2 = [x^2 + 2, Mod(x^3 + x, x^4 + 1), [x^2 - y*x - 1, x^2 + y*x - 1]]

computes the factorization x^4+1 = (x^2- sqrt {-2}x-1)(x^2+ sqrt {-2}x-1)

The library syntax is GEN galoisfixedfield(GEN gal, GEN perm, long flag, long v = -1), where v is a variable number.

galoisgetpol(a,{b},{s})

Query the galpol package for a polynomial with Galois group isomorphic to GAP4(a,b), totally real if s = 1 (default) and totally complex if s = 2. The output is a vector [pol, den] where

@3* pol is the polynomial of degree a

@3* den is the denominator of nfgaloisconj(pol). Pass it as an optional argument to galoisinit or nfgaloisconj to speed them up:

  ? [pol,den] = galoisgetpol(64,4,1);
  ? G = galoisinit(pol);
  time = 352ms
  ? galoisinit(pol, den);  \\ passing 'den' speeds up the computation
  time = 264ms
  ? % == %`
  %4 = 1  \\ same answer

If b and s are omitted, return the number of isomorphism classes of groups of order a.

The library syntax is GEN galoisgetpol(long a, long b, long s). Also available is GEN galoisnbpol(long a) when b and s are omitted.

galoisidentify(gal)

gal being be a Galois group as output by galoisinit, output the isomorphism class of the underlying abstract group as a two-components vector [o,i], where o is the group order, and i is the group index in the GAP4 Small Group library, by Hans Ulrich Besche, Bettina Eick and Eamonn O'Brien.

This command also accepts subgroups returned by galoissubgroups.

The current implementation is limited to degree less or equal to 127. Some larger ``easy'' orders are also supported.

The output is similar to the output of the function IdGroup in GAP4. Note that GAP4 IdGroup handles all groups of order less than 2000 except 1024, so you can use galoisexport and GAP4 to identify large Galois groups.

The library syntax is GEN galoisidentify(GEN gal).

galoisinit(pol,{den})

Computes the Galois group and all necessary information for computing the fixed fields of the Galois extension K/Q where K is the number field defined by pol (monic irreducible polynomial in Z[X] or a number field as output by nfinit). The extension K/Q must be Galois with Galois group ``weakly'' super-solvable, see below; returns 0 otherwise. Hence this permits to quickly check whether a polynomial of order strictly less than 36 is Galois or not.

The algorithm used is an improved version of the paper ``An efficient algorithm for the computation of Galois automorphisms'', Bill Allombert, Math. Comp, vol. 73, 245, 2001, pp. 359--375.

A group G is said to be ``weakly'' super-solvable if there exists a normal series

{1} = H_0 \triangleleft H_1 \triangleleft...\triangleleft H_{n-1} \triangleleft H_n

such that each H_i is normal in G and for i < n, each quotient group H_{i+1}/H_i is cyclic, and either H_n = G (then G is super-solvable) or G/H_n is isomorphic to either A_4 or S_4.

In practice, almost all small groups are WKSS, the exceptions having order 36(1 exception), 48(2), 56(1), 60(1), 72(5), 75(1), 80(1), 96(10) and >= 108.

This function is a prerequisite for most of the galoisxxx routines. For instance:

  P = x^6 + 108;
  G = galoisinit(P);
  L = galoissubgroups(G);
  vector(#L, i, galoisisabelian(L[i],1))
  vector(#L, i, galoisidentify(L[i]))

The output is an 8-component vector gal.

gal[1] contains the polynomial pol (gal.pol).

gal[2] is a three-components vector [p,e,q] where p is a prime number (gal.p) such that pol totally split modulo p , e is an integer and q = p^e (gal.mod) is the modulus of the roots in gal.roots.

gal[3] is a vector L containing the p-adic roots of pol as integers implicitly modulo gal.mod. (gal.roots).

gal[4] is the inverse of the Vandermonde matrix of the p-adic roots of pol, multiplied by gal[5].

gal[5] is a multiple of the least common denominator of the automorphisms expressed as polynomial in a root of pol.

gal[6] is the Galois group G expressed as a vector of permutations of L (gal.group).

gal[7] is a generating subset S = [s_1,...,s_g] of G expressed as a vector of permutations of L (gal.gen).

gal[8] contains the relative orders [o_1,...,o_g] of the generators of S (gal.orders).

Let H_n be as above, we have the following properties:

  * if G/H_n ~ A_4 then [o_1,...,o_g] ends by [2,2,3].

  * if G/H_n ~ S_4 then [o_1,...,o_g] ends by [2,2,3,2].

  * for 1 <= i <= g the subgroup of G generated by [s_1,...,s_g] is normal, with the exception of i = g-2 in the A_4 case and of i = g-3 in the S_A case.

  * the relative order o_i of s_i is its order in the quotient group G/<s_1,...,s_{i-1}>, with the same exceptions.

  * for any x belongs to G there exists a unique family [e_1,...,e_g] such that (no exceptions):

-- for 1 <= i <= g we have 0 <= e_i < o_i

-- x = g_1^{e_1}g_2^{e_2}...g_n^{e_n}

If present den must be a suitable value for gal[5].

The library syntax is GEN galoisinit(GEN pol, GEN den = NULL).

galoisisabelian(gal,{flag = 0})

gal being as output by galoisinit, return 0 if gal is not an abelian group, and the HNF matrix of gal over gal.gen if fl = 0, 1 if fl = 1.

This command also accepts subgroups returned by galoissubgroups.

The library syntax is GEN galoisisabelian(GEN gal, long flag).

galoisisnormal(gal,subgrp)

gal being as output by galoisinit, and subgrp a subgroup of gal as output by galoissubgroups,return 1 if subgrp is a normal subgroup of gal, else return 0.

This command also accepts subgroups returned by galoissubgroups.

The library syntax is long galoisisnormal(GEN gal, GEN subgrp).

galoispermtopol(gal,perm)

gal being a Galois group as output by galoisinit and perm a element of gal.group, return the polynomial defining the Galois automorphism, as output by nfgaloisconj, associated with the permutation perm of the roots gal.roots. perm can also be a vector or matrix, in this case, galoispermtopol is applied to all components recursively.

@3Note that

  G = galoisinit(pol);
  galoispermtopol(G, G[6])~

is equivalent to nfgaloisconj(pol), if degree of pol is greater or equal to 2.

The library syntax is GEN galoispermtopol(GEN gal, GEN perm).

galoissubcyclo(N,H,{fl = 0},{v})

Computes the subextension of Q(zeta_n) fixed by the subgroup H \subset (Z/nZ)^*. By the Kronecker-Weber theorem, all abelian number fields can be generated in this way (uniquely if n is taken to be minimal).

@3The pair (n, H) is deduced from the parameters (N, H) as follows

@3* N an integer: then n = N; H is a generator, i.e. an integer or an integer modulo n; or a vector of generators.

@3* N the output of znstar(n). H as in the first case above, or a matrix, taken to be a HNF left divisor of the SNF for (Z/nZ)^* (of type N.cyc), giving the generators of H in terms of N.gen.

@3* N the output of bnrinit(bnfinit(y), m, 1) where m is a module. H as in the first case, or a matrix taken to be a HNF left divisor of the SNF for the ray class group modulo m (of type N.cyc), giving the generators of H in terms of N.gen.

In this last case, beware that H is understood relatively to N; in particular, if the infinite place does not divide the module, e.g if m is an integer, then it is not a subgroup of (Z/nZ)^*, but of its quotient by {+- 1}.

If fl = 0, compute a polynomial (in the variable v) defining the the subfield of Q(zeta_n) fixed by the subgroup H of (Z/nZ)^*.

If fl = 1, compute only the conductor of the abelian extension, as a module.

If fl = 2, output [pol, N], where pol is the polynomial as output when fl = 0 and N the conductor as output when fl = 1.

The following function can be used to compute all subfields of Q(zeta_n) (of exact degree d, if d is set):

  polsubcyclo(n, d = -1)=
  { my(bnr,L,IndexBound);
    IndexBound = if (d < 0, n, [d]);
    bnr = bnrinit(bnfinit(y), [n,[1]], 1);
    L = subgrouplist(bnr, IndexBound, 1);
    vector(#L,i, galoissubcyclo(bnr,L[i]));
  }

Setting L = subgrouplist(bnr, IndexBound) would produce subfields of exact conductor n oo .

The library syntax is GEN galoissubcyclo(GEN N, GEN H = NULL, long fl, long v = -1), where v is a variable number.

galoissubfields(G,{flags = 0},{v})

Outputs all the subfields of the Galois group G, as a vector. This works by applying galoisfixedfield to all subgroups. The meaning of the flag fl is the same as for galoisfixedfield.

The library syntax is GEN galoissubfields(GEN G, long flags, long v = -1), where v is a variable number.

galoissubgroups(G)

Outputs all the subgroups of the Galois group gal. A subgroup is a vector [gen, orders], with the same meaning as for gal.gen and gal.orders. Hence gen is a vector of permutations generating the subgroup, and orders is the relatives orders of the generators. The cardinal of a subgroup is the product of the relative orders. Such subgroup can be used instead of a Galois group in the following command: galoisisabelian, galoissubgroups, galoisexport and galoisidentify.

To get the subfield fixed by a subgroup sub of gal, use

  galoisfixedfield(gal,sub[1])

The library syntax is GEN galoissubgroups(GEN G).

idealadd(nf,x,y)

Sum of the two ideals x and y in the number field nf. The result is given in HNF.

   ? K = nfinit(x^2 + 1);
   ? a = idealadd(K, 2, x + 1)  \\ ideal generated by 2 and 1+I
   %2 =
   [2 1]
   [0 1]
   ? pr = idealprimedec(K, 5)[1];  \\ a prime ideal above 5
   ? idealadd(K, a, pr)     \\ coprime, as expected
   %4 =
   [1 0]
   [0 1]

@3 This function cannot be used to add arbitrary Z-modules, since it assumes that its arguments are ideals:

    ? b = Mat([1,0]~);
    ? idealadd(K, b, b)     \\ only square t_MATs represent ideals
    *** idealadd: non-square t_MAT in idealtyp.
    ? c = [2, 0; 2, 0]; idealadd(K, c, c)   \\ non-sense
    %6 =
    [2 0]
    [0 2]
    ? d = [1, 0; 0, 2]; idealadd(K, d, d)   \\ non-sense
    %7 =
    [1 0]
    [0 1]

@3In the last two examples, we get wrong results since the matrices c and d do not correspond to an ideal: the Z-span of their columns (as usual interpreted as coordinates with respect to the integer basis K.zk) is not an O_K-module. To add arbitrary Z-modules generated by the columns of matrices A and B, use mathnf(concat(A,B)).

The library syntax is GEN idealadd(GEN nf, GEN x, GEN y).

idealaddtoone(nf,x,{y})

x and y being two co-prime integral ideals (given in any form), this gives a two-component row vector [a,b] such that a belongs to x, b belongs to y and a+b = 1.

The alternative syntax idealaddtoone(nf,v), is supported, where v is a k-component vector of ideals (given in any form) which sum to Z_K. This outputs a k-component vector e such that e[i] belongs to x[i] for 1 <= i <= k and sum_{1 <= i <= k}e[i] = 1.

The library syntax is GEN idealaddtoone0(GEN nf, GEN x, GEN y = NULL).

idealappr(nf,x,{flag = 0})

If x is a fractional ideal (given in any form), gives an element alpha in nf such that for all prime ideals p such that the valuation of x at p is non-zero, we have v_{p}(alpha) = v_{p}(x), and v_{p}(alpha) >= 0 for all other p.

If flag is non-zero, x must be given as a prime ideal factorization, as output by idealfactor, but possibly with zero or negative exponents. This yields an element alpha such that for all prime ideals p occurring in x, v_{p}(alpha) is equal to the exponent of p in x, and for all other prime ideals, v_{p}(alpha) >= 0. This generalizes idealappr(nf,x,0) since zero exponents are allowed. Note that the algorithm used is slightly different, so that

    idealappr(nf, idealfactor(nf,x))

may not be the same as idealappr(nf,x,1).

The library syntax is GEN idealappr0(GEN nf, GEN x, long flag).

idealchinese(nf,x,y)

x being a prime ideal factorization (i.e. a 2 by 2 matrix whose first column contains prime ideals, and the second column integral exponents), y a vector of elements in nf indexed by the ideals in x, computes an element b such that

v_{p}(b - y_{p}) >= v_{p}(x) for all prime ideals in x and v_{p}(b) >= 0 for all other p.

The library syntax is GEN idealchinese(GEN nf, GEN x, GEN y).

idealcoprime(nf,x,y)

Given two integral ideals x and y in the number field nf, returns a beta in the field, such that beta.x is an integral ideal coprime to y.

The library syntax is GEN idealcoprime(GEN nf, GEN x, GEN y).

idealdiv(nf,x,y,{flag = 0})

Quotient x.y^{-1} of the two ideals x and y in the number field nf. The result is given in HNF.

If flag is non-zero, the quotient x.y^{-1} is assumed to be an integral ideal. This can be much faster when the norm of the quotient is small even though the norms of x and y are large.

The library syntax is GEN idealdiv0(GEN nf, GEN x, GEN y, long flag). Also available are GEN idealdiv(GEN nf, GEN x, GEN y) (flag = 0) and GEN idealdivexact(GEN nf, GEN x, GEN y) (flag = 1).

idealfactor(nf,x)

Factors into prime ideal powers the ideal x in the number field nf. The output format is similar to the factor function, and the prime ideals are represented in the form output by the idealprimedec function, i.e. as 5-element vectors.

The library syntax is GEN idealfactor(GEN nf, GEN x).

idealfactorback(nf,f,{e},{flag = 0})

Gives back the ideal corresponding to a factorization. The integer 1 corresponds to the empty factorization. If e is present, e and f must be vectors of the same length (e being integral), and the corresponding factorization is the product of the f[i]^{e[i]}.

If not, and f is vector, it is understood as in the preceding case with e a vector of 1s: we return the product of the f[i]. Finally, f can be a regular factorization, as produced by idealfactor.

  ? nf = nfinit(y^2+1); idealfactor(nf, 4 + 2*y)
  %1 =
  [[2, [1, 1]~, 2, 1, [1, 1]~] 2]
  [[5, [2, 1]~, 1, 1, [-2, 1]~] 1]
  ? idealfactorback(nf, %)
  %2 =
  [10 4]
  [0  2]
  ? f = %1[,1]; e = %1[,2]; idealfactorback(nf, f, e)
  %3 =
  [10 4]
  [0  2]
  ? % == idealhnf(nf, 4 + 2*y)
  %4 = 1

If flag is non-zero, perform ideal reductions (idealred) along the way. This is most useful if the ideals involved are all extended ideals (for instance with trivial principal part), so that the principal parts extracted by idealred are not lost. Here is an example:

  ? f = vector(#f, i, [f[i], [;]]);  \\ transform to extended ideals
  ? idealfactorback(nf, f, e, 1)
  %6 = [[1, 0; 0, 1], [2, 1; [2, 1]~, 1]]
  ? nffactorback(nf, %[2])
  %7 = [4, 2]~

The extended ideal returned in %6 is the trivial ideal 1, extended with a principal generator given in factored form. We use nffactorback to recover it in standard form.

The library syntax is GEN idealfactorback(GEN nf, GEN f, GEN e = NULL, long flag ).

idealfrobenius(nf,gal,pr)

Let K be the number field defined by nf and assume K/Q be a Galois extension with Galois group given gal = galoisinit(nf), and that pr is the prime ideal P in prid format, and that P is unramified. This function returns a permutation of gal.group which defines the automorphism sigma = (P\over K/Q ), i.e the Frobenius element associated to P. If p is the unique prime number in P, then sigma(x) = x^p mod \P for all x belongs to Z_K.

  ? nf = nfinit(polcyclo(31));
  ? gal = galoisinit(nf);
  ? pr = idealprimedec(nf,101)[1];
  ? g = idealfrobenius(nf,gal,pr);
  ? galoispermtopol(gal,g)
  %5 = x^8

@3This is correct since 101 = 8 mod 31.

The library syntax is GEN idealfrobenius(GEN nf, GEN gal, GEN pr).

idealhnf(nf,u,{v})

Gives the Hermite normal form of the ideal uZ_K+vZ_K, where u and v are elements of the number field K defined by nf.

  ? nf = nfinit(y^3 - 2);
  ? idealhnf(nf, 2, y+1)
  %2 =
  [1 0 0]
  [0 1 0]
  [0 0 1]
  ? idealhnf(nf, y/2, [0,0,1/3]~)
  %3 =
  [1/3 0 0]
  [0 1/6 0]
  [0 0 1/6]

If b is omitted, returns the HNF of the ideal defined by u: u may be an algebraic number (defining a principal ideal), a maximal ideal (as given by idealprimedec or idealfactor), or a matrix whose columns give generators for the ideal. This last format is a little complicated, but useful to reduce general modules to the canonical form once in a while:

@3* if strictly less than N = [K:Q] generators are given, u is the Z_K-module they generate,

@3* if N or more are given, it is assumed that they form a Z-basis of the ideal, in particular that the matrix has maximal rank N. This acts as mathnf since the Z_K-module structure is (taken for granted hence) not taken into account in this case.

  ? idealhnf(nf, idealprimedec(nf,2)[1])
  %4 =
  [2 0 0]
  [0 1 0]
  [0 0 1]
  ? idealhnf(nf, [1,2;2,3;3,4])
  %5 =
  [1 0 0]
  [0 1 0]
  [0 0 1]

@3Finally, when K is quadratic with discriminant D_K, we allow u = Qfb(a,b,c), provided b^2 - 4ac = D_K. As usual, this represents the ideal a Z + (1/2)(-b + sqrt {D_K}) Z.

  ? K = nfinit(x^2 - 60); K.disc
  %1 = 60
  ? idealhnf(K, qfbprimeform(60,2))
  %2 =
  [2 1]
  [0 1]
  ? idealhnf(K, Qfb(1,2,3))
    ***   at top-level: idealhnf(K,Qfb(1,2,3
    ***                 ^--------------------
    *** idealhnf: Qfb(1, 2, 3) has discriminant != 60 in idealhnf.

The library syntax is GEN idealhnf0(GEN nf, GEN u, GEN v = NULL). Also available is GEN idealhnf(GEN nf, GEN a).

idealintersect(nf,A,B)

Intersection of the two ideals A and B in the number field nf. The result is given in HNF.

  ? nf = nfinit(x^2+1);
  ? idealintersect(nf, 2, x+1)
  %2 =
  [2 0]
  [0 2]

This function does not apply to general Z-modules, e.g. orders, since its arguments are replaced by the ideals they generate. The following script intersects Z-modules A and B given by matrices of compatible dimensions with integer coefficients:

  ZM_intersect(A,B) =
  { my(Ker = matkerint(concat(A,B)));
    mathnf( A * Ker[1..#A,] )
  }

The library syntax is GEN idealintersect(GEN nf, GEN A, GEN B).

idealinv(nf,x)

Inverse of the ideal x in the number field nf, given in HNF. If x is an extended ideal, its principal part is suitably updated: i.e. inverting [I,t], yields [I^{-1}, 1/t].

The library syntax is GEN idealinv(GEN nf, GEN x).

ideallist(nf,bound,{flag = 4})

Computes the list of all ideals of norm less or equal to bound in the number field nf. The result is a row vector with exactly bound components. Each component is itself a row vector containing the information about ideals of a given norm, in no specific order, depending on the value of flag:

The possible values of flag are:

  0: give the bid associated to the ideals, without generators.

  1: as 0, but include the generators in the bid.

  2: in this case, nf must be a bnf with units. Each component is of the form [bid,U], where bid is as case 0 and U is a vector of discrete logarithms of the units. More precisely, it gives the ideallogs with respect to bid of bnf.tufu. This structure is technical, and only meant to be used in conjunction with bnrclassnolist or bnrdisclist.

  3: as 2, but include the generators in the bid.

  4: give only the HNF of the ideal.

  ? nf = nfinit(x^2+1);
  ? L = ideallist(nf, 100);
  ? L[1]
  %3 = [[1, 0; 0, 1]]  \\ A single ideal of norm 1
  ? #L[65]
  %4 = 4               \\ There are 4 ideals of norm 4 in B<Z>[i]

If one wants more information, one could do instead:

  ? nf = nfinit(x^2+1);
  ? L = ideallist(nf, 100, 0);
  ? l = L[25]; vector(#l, i, l[i].clgp)
  %3 = [[20, [20]], [16, [4, 4]], [20, [20]]]
  ? l[1].mod
  %4 = [[25, 18; 0, 1], []]
  ? l[2].mod
  %5 = [[5, 0; 0, 5], []]
  ? l[3].mod
  %6 = [[25, 7; 0, 1], []]

@3where we ask for the structures of the (Z[i]/I)^* for all three ideals of norm 25. In fact, for all moduli with finite part of norm 25 and trivial Archimedean part, as the last 3 commands show. See ideallistarch to treat general moduli.

The library syntax is GEN ideallist0(GEN nf, long bound, long flag).

ideallistarch(nf,list,arch)

list is a vector of vectors of bid's, as output by ideallist with flag 0 to 3. Return a vector of vectors with the same number of components as the original list. The leaves give information about moduli whose finite part is as in original list, in the same order, and Archimedean part is now arch (it was originally trivial). The information contained is of the same kind as was present in the input; see ideallist, in particular the meaning of flag.

  ? bnf = bnfinit(x^2-2);
  ? bnf.sign
  %2 = [2, 0]                         \\ two places at infinity
  ? L = ideallist(bnf, 100, 0);
  ? l = L[98]; vector(#l, i, l[i].clgp)
  %4 = [[42, [42]], [36, [6, 6]], [42, [42]]]
  ? La = ideallistarch(bnf, L, [1,1]); \\ add them to the modulus
  ? l = La[98]; vector(#l, i, l[i].clgp)
  %6 = [[168, [42, 2, 2]], [144, [6, 6, 2, 2]], [168, [42, 2, 2]]]

Of course, the results above are obvious: adding t places at infinity will add t copies of Z/2Z to the ray class group. The following application is more typical:

  ? L = ideallist(bnf, 100, 2);        \\ units are required now
  ? La = ideallistarch(bnf, L, [1,1]);
  ? H = bnrclassnolist(bnf, La);
  ? H[98];
  %6 = [2, 12, 2]

The library syntax is GEN ideallistarch(GEN nf, GEN list, GEN arch).

ideallog(nf,x,bid)

nf is a number field, bid is as output by idealstar(nf, D,...) and x a non-necessarily integral element of nf which must have valuation equal to 0 at all prime ideals in the support of D. This function computes the discrete logarithm of x on the generators given in bid.gen. In other words, if g_i are these generators, of orders d_i respectively, the result is a column vector of integers (x_i) such that 0 <= x_i < d_i and

  x = prod_i g_i^{x_i} (mod ^*D) .

Note that when the support of D contains places at infinity, this congruence implies also sign conditions on the associated real embeddings. See znlog for the limitations of the underlying discrete log algorithms.

The library syntax is GEN ideallog(GEN nf, GEN x, GEN bid).

idealmin(nf,ix,{vdir})

This function is useless and kept for backward compatibility only, use idealred. Computes a pseudo-minimum of the ideal x in the direction vdir in the number field nf.

The library syntax is GEN idealmin(GEN nf, GEN ix, GEN vdir = NULL).

idealmul(nf,x,y,{flag = 0})

Ideal multiplication of the ideals x and y in the number field nf; the result is the ideal product in HNF. If either x or y are extended ideals, their principal part is suitably updated: i.e. multiplying [I,t], [J,u] yields [IJ, tu]; multiplying I and [J, u] yields [IJ, u].

  ? nf = nfinit(x^2 + 1);
  ? idealmul(nf, 2, x+1)
  %2 =
  [4 2]
  [0 2]
  ? idealmul(nf, [2, x], x+1)        \\ extended ideal * ideal
  %4 = [[4, 2; 0, 2], x]
  ? idealmul(nf, [2, x], [x+1, x])   \\ two extended ideals
  %5 = [[4, 2; 0, 2], [-1, 0]~]

@3 If flag is non-zero, reduce the result using idealred.

The library syntax is GEN idealmul0(GEN nf, GEN x, GEN y, long flag).

@3See also GEN idealmul(GEN nf, GEN x, GEN y) (flag = 0) and GEN idealmulred(GEN nf, GEN x, GEN y) (flag != 0).

idealnorm(nf,x)

Computes the norm of the ideal x in the number field nf.

The library syntax is GEN idealnorm(GEN nf, GEN x).

idealnumden(nf,x)

Returns [A,B], where A,B are coprime integer ideals such that x = A/B, in the number field nf.

  ? nf = nfinit(x^2+1);
  ? idealnumden(nf, (x+1)/2)
  %2 = [[1, 0; 0, 1], [2, 1; 0, 1]]

The library syntax is GEN idealnumden(GEN nf, GEN x).

idealpow(nf,x,k,{flag = 0})

Computes the k-th power of the ideal x in the number field nf; k belongs to Z. If x is an extended ideal, its principal part is suitably updated: i.e. raising [I,t] to the k-th power, yields [I^k, t^k].

If flag is non-zero, reduce the result using idealred, throughout the (binary) powering process; in particular, this is not the same as as idealpow(nf,x,k) followed by reduction.

The library syntax is GEN idealpow0(GEN nf, GEN x, GEN k, long flag).

@3See also GEN idealpow(GEN nf, GEN x, GEN k) and GEN idealpows(GEN nf, GEN x, long k) (flag = 0). Corresponding to flag = 1 is GEN idealpowred(GEN nf, GEN vp, GEN k).

idealprimedec(nf,p)

Computes the prime ideal decomposition of the (positive) prime number p in the number field K represented by nf. If a non-prime p is given the result is undefined.

The result is a vector of prid structures, each representing one of the prime ideals above p in the number field nf. The representation pr = [p,a,e,f,mb] of a prime ideal means the following: a and is an algebraic integer in the maximal order Z_K and the prime ideal is equal to p = pZ_K + aZ_K; e is the ramification index; f is the residual index; finally, mb is the multiplication table associated to the algebraic integer b is such that p^{-1} = Z_K+ b/ pZ_K, which is used internally to compute valuations. In other words if p is inert, then mb is the integer 1, and otherwise it's a square t_MAT whose j-th column is b.nf.zk[j].

The algebraic number a is guaranteed to have a valuation equal to 1 at the prime ideal (this is automatic if e > 1).

The components of pr should be accessed by member functions: pr.p, pr.e, pr.f, and pr.gen (returns the vector [p,a]):

  ? K = nfinit(x^3-2);
  ? L = idealprimedec(K, 5);
  ? #L       \\ 2 primes above 5 in Q(2^(1/3))
  %3 = 2
  ? p1 = L[1]; p2 = L[2];
  ? [p1.e, p1.f]    \\ the first is unramified of degree 1
  %4 = [1, 1]
  ? [p2.e, p2.f]    \\ the second is unramified of degree 2
  %5 = [1, 2]
  ? p1.gen
  %6 = [5, [2, 1, 0]~]
  ? nfbasistoalg(K, %[2])  \\ a uniformizer for p1
  %7 = Mod(x + 2, x^3 - 2)

The library syntax is GEN idealprimedec(GEN nf, GEN p).

idealprincipalunits(nf,pr,k)

Given a prime ideal in idealprimedec format, returns the multiplicative group (1 + pr) / (1 + pr^k) as an abelian group. This function is much faster than idealstar when the norm of pr is large, since it avoids (useless) work in the multiplicative group of the residue field.

  ? K = nfinit(y^2+1);
  ? P = idealprimedec(K,2)[1];
  ? G = idealprincipalunits(K, P, 20);
  ? G.cyc
  [512, 256, 4]   \\ Z/512 x Z/256 x Z/4
  ? G.gen
  %5 = [[-1, -2]~, 1021, [0, -1]~] \\ minimal generators of given order

The library syntax is GEN idealprincipalunits(GEN nf, GEN pr, long k).

idealramgroups(nf,gal,pr)

Let K be the number field defined by nf and assume that K/Q is Galois with Galois group G given by gal = galoisinit(nf). Let pr be the prime ideal P in prid format. This function returns a vector g of subgroups of gal as follow:

@3* g[1] is the decomposition group of P,

@3* g[2] is G_0(P), the inertia group of P,

and for i >= 2,

@3* g[i] is G_{i-2}(P), the i-2-th ramification group of P.

@3The length of g is the number of non-trivial groups in the sequence, thus is 0 if e = 1 and f = 1, and 1 if f > 1 and e = 1. The following function computes the cardinality of a subgroup of G, as given by the components of g:

  card(H) =my(o=H[2]); prod(i=1,#o,o[i]);
  ? nf=nfinit(x^6+3); gal=galoisinit(nf); pr=idealprimedec(nf,3)[1];
  ? g = idealramgroups(nf, gal, pr);
  ? apply(card,g)
  %4 = [6, 6, 3, 3, 3] \\ cardinalities of the G_i
  ? nf=nfinit(x^6+108); gal=galoisinit(nf); pr=idealprimedec(nf,2)[1];
  ? iso=idealramgroups(nf,gal,pr)[2]
  %4 = [[Vecsmall([2, 3, 1, 5, 6, 4])], Vecsmall([3])]
  ? nfdisc(galoisfixedfield(gal,iso,1))
  %5 = -3

@3The field fixed by the inertia group of 2 is not ramified at 2.

The library syntax is GEN idealramgroups(GEN nf, GEN gal, GEN pr).

idealred(nf,I,{v = 0})

LLL reduction of the ideal I in the number field nf, along the direction v. The v parameter is best left omitted, but if it is present, it must be an nf.r1 + nf.r2-component vector of non-negative integers. (What counts is the relative magnitude of the entries: if all entries are equal, the effect is the same as if the vector had been omitted.)

This function finds a ``small'' a in I (see the end for technical details). The result is the Hermite normal form of the ``reduced'' ideal J = r I/a, where r is the unique rational number such that J is integral and primitive. (This is usually not a reduced ideal in the sense of Buchmann.)

  ? K = nfinit(y^2+1);
  ? P = idealprimedec(K,5)[1];
  ? idealred(K, P)
  %3 =
  [1 0]
  [0 1]

@3More often than not, a principal ideal yields the unit ideal as above. This is a quick and dirty way to check if ideals are principal, but it is not a necessary condition: a non-trivial result does not prove that the ideal is non-principal. For guaranteed results, see bnfisprincipal, which requires the computation of a full bnf structure.

If the input is an extended ideal [I,s], the output is [J,sa/r]; this way, one can keep track of the principal ideal part:

  ? idealred(K, [P, 1])
  %5 = [[1, 0; 0, 1], [-2, 1]~]

meaning that P is generated by [-2, 1] . The number field element in the extended part is an algebraic number in any form or a factorization matrix (in terms of number field elements, not ideals!). In the latter case, elements stay in factored form, which is a convenient way to avoid coefficient explosion; see also idealpow.

@3Technical note. The routine computes an LLL-reduced basis for the lattice I equipped with the quadratic form

  || x ||_v^2 = sum_{i = 1}^{r_1+r_2} 2^{v_i}varepsilon_i|sigma_i(x)|^2,

where as usual the sigma_i are the (real and) complex embeddings and varepsilon_i = 1, resp. 2, for a real, resp. complex place. The element a is simply the first vector in the LLL basis. The only reason you may want to try to change some directions and set some v_i != 0 is to randomize the elements found for a fixed ideal, which is heuristically useful in index calculus algorithms like bnfinit and bnfisprincipal.

@3Even more technical note. In fact, the above is a white lie. We do not use ||.||_v exactly but a rescaled rounded variant which gets us faster and simpler LLLs. There's no harm since we are not using any theoretical property of a after all, except that it belongs to I and is ``expected to be small''.

The library syntax is GEN idealred0(GEN nf, GEN I, GEN v = NULL).

idealstar(nf,I,{flag = 1})

Outputs a bid structure, necessary for computing in the finite abelian group G = (Z_K/I)^*. Here, nf is a number field and I is a modulus: either an ideal in any form, or a row vector whose first component is an ideal and whose second component is a row vector of r_1 0 or 1. Ideals can also be given by a factorization into prime ideals, as produced by idealfactor.

This bid is used in ideallog to compute discrete logarithms. It also contains useful information which can be conveniently retrieved as bid.mod (the modulus), bid.clgp (G as a finite abelian group), bid.no (the cardinality of G), bid.cyc (elementary divisors) and bid.gen (generators).

If flag = 1 (default), the result is a bid structure without generators.

If flag = 2, as flag = 1, but including generators, which wastes some time.

If flag = 0, only outputs (Z_K/I)^* as an abelian group, i.e as a 3-component vector [h,d,g]: h is the order, d is the vector of SNF cyclic components and g the corresponding generators.

The library syntax is GEN idealstar0(GEN nf, GEN I, long flag). Instead the above hardcoded numerical flags, one should rather use GEN Idealstar(GEN nf, GEN ideal, long flag), where flag is an or-ed combination of nf_GEN (include generators) and nf_INIT (return a full bid, not a group), possibly 0. This offers one more combination: gen, but no init.

idealtwoelt(nf,x,{a})

Computes a two-element representation of the ideal x in the number field nf, combining a random search and an approximation theorem; x is an ideal in any form (possibly an extended ideal, whose principal part is ignored)

@3* When called as idealtwoelt(nf,x), the result is a row vector [a,alpha] with two components such that x = aZ_K+alphaZ_K and a is chosen to be the positive generator of x cap Z, unless x was given as a principal ideal (in which case we may choose a = 0). The algorithm uses a fast lazy factorization of x cap Z and runs in randomized polynomial time.

@3* When called as idealtwoelt(nf,x,a) with an explicit non-zero a supplied as third argument, the function assumes that a belongs to x and returns alpha belongs to x such that x = aZ_K + alphaZ_K. Note that we must factor a in this case, and the algorithm is generally much slower than the default variant.

The library syntax is GEN idealtwoelt0(GEN nf, GEN x, GEN a = NULL). Also available are GEN idealtwoelt(GEN nf, GEN x) and GEN idealtwoelt2(GEN nf, GEN x, GEN a).

idealval(nf,x,pr)

Gives the valuation of the ideal x at the prime ideal pr in the number field nf, where pr is in idealprimedec format.

The library syntax is long idealval(GEN nf, GEN x, GEN pr).

matalgtobasis(nf,x)

nf being a number field in nfinit format, and x a (row or column) vector or matrix, apply nfalgtobasis to each entry of x.

The library syntax is GEN matalgtobasis(GEN nf, GEN x).

matbasistoalg(nf,x)

nf being a number field in nfinit format, and x a (row or column) vector or matrix, apply nfbasistoalg to each entry of x.

The library syntax is GEN matbasistoalg(GEN nf, GEN x).

modreverse(z)

Let z = Mod(A, T) be a polmod, and Q be its minimal polynomial, which must satisfy {deg}(Q) = {deg}(T). Returns a ``reverse polmod'' Mod(B, Q), which is a root of T.

This is quite useful when one changes the generating element in algebraic extensions:

  ? u = Mod(x, x^3 - x -1); v = u^5;
  ? w = modreverse(v)
  %2 = Mod(x^2 - 4*x + 1, x^3 - 5*x^2 + 4*x - 1)

which means that x^3 - 5x^2 + 4x -1 is another defining polynomial for the cubic field

  Q(u) = Q[x]/(x^3 - x - 1) = Q[x]/(x^3 - 5x^2 + 4x - 1) = Q(v),

and that u \to v^2 - 4v + 1 gives an explicit isomorphism. From this, it is easy to convert elements between the A(u) belongs to Q(u) and B(v) belongs to Q(v) representations:

  ? A = u^2 + 2*u + 3; subst(lift(A), 'x, w)
  %3 = Mod(x^2 - 3*x + 3, x^3 - 5*x^2 + 4*x - 1)
  ? B = v^2 + v + 1;   subst(lift(B), 'x, v)
  %4 = Mod(26*x^2 + 31*x + 26, x^3 - x - 1)

If the minimal polynomial of z has lower degree than expected, the routine fails

  ? u = Mod(-x^3 + 9*x, x^4 - 10*x^2 + 1)
  ? modreverse(u)
   *** modreverse: domain error in modreverse: deg(minpoly(z)) < 4
   ***   Break loop: type 'break' to go back to GP prompt
  break> Vec( dbg_err() ) \\ ask for more info
  ["e_DOMAIN", "modreverse", "deg(minpoly(z))", "<", 4,
    Mod(-x^3 + 9*x, x^4 - 10*x^2 + 1)]
  break> minpoly(u)
  x^2 - 8

The library syntax is GEN modreverse(GEN z).

newtonpoly(x,p)

Gives the vector of the slopes of the Newton polygon of the polynomial x with respect to the prime number p. The n components of the vector are in decreasing order, where n is equal to the degree of x. Vertical slopes occur iff the constant coefficient of x is zero and are denoted by LONG_MAX, the biggest single precision integer representable on the machine (2^{31}-1 (resp. 2^{63}-1) on 32-bit (resp. 64-bit) machines), see Label se:valuation.

The library syntax is GEN newtonpoly(GEN x, GEN p).

nfalgtobasis(nf,x)

Given an algebraic number x in the number field nf, transforms it to a column vector on the integral basis nf.zk.

  ? nf = nfinit(y^2 + 4);
  ? nf.zk
  %2 = [1, 1/2*y]
  ? nfalgtobasis(nf, [1,1]~)
  %3 = [1, 1]~
  ? nfalgtobasis(nf, y)
  %4 = [0, 2]~
  ? nfalgtobasis(nf, Mod(y, y^2+4))
  %4 = [0, 2]~

This is the inverse function of nfbasistoalg.

The library syntax is GEN algtobasis(GEN nf, GEN x).

nfbasis(T)

Let T(X) be an irreducible polynomial with integral coefficients. This function returns an integral basis of the number field defined by T, that is a Z-basis of its maximal order. The basis elements are given as elements in Q[X]/(T):

  ? nfbasis(x^2 + 1)
  %1 = [1, x]

This function uses a modified version of the round 4 algorithm, due to David Ford, Sebastian Pauli and Xavier Roblot.

@3Local basis, orders maximal at certain primes.

Obtaining the maximal order is hard: it requires factoring the discriminant D of T. Obtaining an order which is maximal at a finite explicit set of primes is easy, but if may then be a strict suborder of the maximal order. To specify that we are interested in a given set of places only, we can replace the argument T by an argument [T,listP], where listP encodes the primes we are interested in: it must be a factorization matrix, a vector of integers or a single integer.

@3* Vector: we assume that it contains distinct prime numbers.

@3* Matrix: we assume that it is a two-column matrix of a (partial) factorization of D; namely the first column contains primes and the second one the valuation of D at each of these primes.

@3* Integer B: this is replaced by the vector of primes up to B. Note that the function will use at least O(B) time: a small value, about 10^5, should be enough for most applications. Values larger than 2^{32} are not supported.

In all these cases, the primes may or may not divide the discriminant D of T. The function then returns a Z-basis of an order whose index is not divisible by any of these prime numbers. The result is actually a global integral basis if all prime divisors of the field discriminant are included! Note that nfinit has built-in support for such a check:

  ? K = nfinit([T, listP]);
  ? nfcertify(K)   \\ we computed an actual maximal order
  %2 = [];

@3The first line initializes a number field structure incorporating nfbasis([T, listP] in place of a proven integral basis. The second line certifies that the resulting structure is correct. This allows to create an nf structure associated to the number field K = Q[X]/(T), when the discriminant of T cannot be factored completely, whereas the prime divisors of disc K are known.

Of course, if listP contains a single prime number p, the function returns a local integral basis for Z_p[X]/(T):

  ? nfbasis(x^2+x-1001)
  %1 = [1, 1/3*x - 1/3]
  ? nfbasis( [x^2+x-1001, [2]] )
  %2 = [1, x]

@3The Buchmann-Lenstra algorithm.

We now complicate the picture: it is in fact allowed to include composite numbers instead of primes in listP (Vector or Matrix case), provided they are pairwise coprime. The result will still be a correct integral basis if the field discriminant factors completely over the actual primes in the list. Adding a composite C such that C^2 divides D may help because when we consider C as a prime and run the algorithm, two good things can happen: either we succeed in proving that no prime dividing C can divide the index (without actually needing to find those primes), or the computation exhibits a non-trivial zero divisor, thereby factoring C and we go on with the refined factorization. (Note that including a C such that C^2 does not divide D is useless.) If neither happen, then the computed basis need not generate the maximal order. Here is an example:

  ? B = 10^5;
  ? P = factor(poldisc(T), B)[,1]; \\ primes <= B dividing D + cofactor
  ? basis = nfbasis([T, listP])
  ? disc = nfdisc([T, listP])

@3We obtain the maximal order and its discriminant if the field discriminant factors completely over the primes less than B (together with the primes contained in the addprimes table). This can be tested as follows:

    check = factor(disc, B);
    lastp = check[-1..-1,1];
    if (lastp > B && !setsearch(addprimes(), lastp),
      warning("nf may be incorrect!"))

This is a sufficient but not a necessary condition, hence the warning, instead of an error. N.B. lastp is the last entry in the first column of the check matrix, i.e. the largest prime dividing nf.disc if <= B or if it belongs to the prime table.

The function nfcertify speeds up and automates the above process:

  ? B = 10^5;
  ? nf = nfinit([T, B]);
  ? nfcertify(nf)
  %3 = []      \\ nf is unconditionally correct
  ? basis = nf.zk;
  ? disc = nf.disc;

The library syntax is nfbasis(GEN T, GEN *d, GEN listP = NULL), which returns the order basis, and where *d receives the order discriminant.

nfbasistoalg(nf,x)

Given an algebraic number x in the number field nf, transforms it into t_POLMOD form.

  ? nf = nfinit(y^2 + 4);
  ? nf.zk
  %2 = [1, 1/2*y]
  ? nfbasistoalg(nf, [1,1]~)
  %3 = Mod(1/2*y + 1, y^2 + 4)
  ? nfbasistoalg(nf, y)
  %4 = Mod(y, y^2 + 4)
  ? nfbasistoalg(nf, Mod(y, y^2+4))
  %4 = Mod(y, y^2 + 4)

This is the inverse function of nfalgtobasis.

The library syntax is GEN basistoalg(GEN nf, GEN x).

nfcertify(nf)

nf being as output by nfinit, checks whether the integer basis is known unconditionally. This is in particular useful when the argument to nfinit was of the form [T, listP], specifying a finite list of primes when p-maximality had to be proven.

The function returns a vector of composite integers. If this vector is empty, then nf.zk and nf.disc are correct. Otherwise, the result is dubious. In order to obtain a certified result, one must completely factor each of the given integers, then addprime each of them, then check whether nfdisc(nf.pol) is equal to nf.disc.

The library syntax is GEN nfcertify(GEN nf).

nfdetint(nf,x)

Given a pseudo-matrix x, computes a non-zero ideal contained in (i.e. multiple of) the determinant of x. This is particularly useful in conjunction with nfhnfmod.

The library syntax is GEN nfdetint(GEN nf, GEN x).

nfdisc(T)

field discriminant of the number field defined by the integral, preferably monic, irreducible polynomial T(X). Returns the discriminant of the number field Q[X]/(T), using the Round 4 algorithm.

@3Local discriminants, valuations at certain primes.

As in nfbasis, the argument T can be replaced by [T,listP], where listP is as in nfbasis: a vector of pairwise coprime integers (usually distinct primes), a factorization matrix, or a single integer. In that case, the function returns the discriminant of an order whose basis is given by nfbasis(T,listP), which need not be the maximal order, and whose valuation at a prime entry in listP is the same as the valuation of the field discriminant.

In particular, if listP is [p] for a prime p, we can return the p-adic discriminant of the maximal order of Z_p[X]/(T), as a power of p, as follows:

  ? padicdisc(T,p) = p^valuation(nfdisc(T,[p]), p);
  ? nfdisc(x^2 + 6)
  %1 = -24
  ? padicdisc(x^2 + 6, 2)
  %2 = 8
  ? padicdisc(x^2 + 6, 3)
  %3 = 3

The library syntax is nfdisc(GEN T) (listP = NULL). Also available is GEN nfbasis(GEN T, GEN *d, GEN listP = NULL), which returns the order basis, and where *d receives the order discriminant.

nfeltadd(nf,x,y)

Given two elements x and y in nf, computes their sum x+y in the number field nf.

The library syntax is GEN nfadd(GEN nf, GEN x, GEN y).

nfeltdiv(nf,x,y)

Given two elements x and y in nf, computes their quotient x/y in the number field nf.

The library syntax is GEN nfdiv(GEN nf, GEN x, GEN y).

nfeltdiveuc(nf,x,y)

Given two elements x and y in nf, computes an algebraic integer q in the number field nf such that the components of x-qy are reasonably small. In fact, this is functionally identical to round(nfdiv(nf,x,y)).

The library syntax is GEN nfdiveuc(GEN nf, GEN x, GEN y).

nfeltdivmodpr(nf,x,y,pr)

Given two elements x and y in nf and pr a prime ideal in modpr format (see nfmodprinit), computes their quotient x / y modulo the prime ideal pr.

The library syntax is GEN nfdivmodpr(GEN nf, GEN x, GEN y, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nf_to_Fq, then work there.

nfeltdivrem(nf,x,y)

Given two elements x and y in nf, gives a two-element row vector [q,r] such that x = qy+r, q is an algebraic integer in nf, and the components of r are reasonably small.

The library syntax is GEN nfdivrem(GEN nf, GEN x, GEN y).

nfeltmod(nf,x,y)

Given two elements x and y in nf, computes an element r of nf of the form r = x-qy with q and algebraic integer, and such that r is small. This is functionally identical to

  x - nfmul(nf,round(nfdiv(nf,x,y)),y).

The library syntax is GEN nfmod(GEN nf, GEN x, GEN y).

nfeltmul(nf,x,y)

Given two elements x and y in nf, computes their product x*y in the number field nf.

The library syntax is GEN nfmul(GEN nf, GEN x, GEN y).

nfeltmulmodpr(nf,x,y,pr)

Given two elements x and y in nf and pr a prime ideal in modpr format (see nfmodprinit), computes their product x*y modulo the prime ideal pr.

The library syntax is GEN nfmulmodpr(GEN nf, GEN x, GEN y, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nf_to_Fq, then work there.

nfeltnorm(nf,x)

Returns the absolute norm of x.

The library syntax is GEN nfnorm(GEN nf, GEN x).

nfeltpow(nf,x,k)

Given an element x in nf, and a positive or negative integer k, computes x^k in the number field nf.

The library syntax is GEN nfpow(GEN nf, GEN x, GEN k). GEN nfinv(GEN nf, GEN x) correspond to k = -1, and GEN nfsqr(GEN nf,GEN x) to k = 2.

nfeltpowmodpr(nf,x,k,pr)

Given an element x in nf, an integer k and a prime ideal pr in modpr format (see nfmodprinit), computes x^k modulo the prime ideal pr.

The library syntax is GEN nfpowmodpr(GEN nf, GEN x, GEN k, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nf_to_Fq, then work there.

nfeltreduce(nf,a,id)

Given an ideal id in Hermite normal form and an element a of the number field nf, finds an element r in nf such that a-r belongs to the ideal and r is small.

The library syntax is GEN nfreduce(GEN nf, GEN a, GEN id).

nfeltreducemodpr(nf,x,pr)

Given an element x of the number field nf and a prime ideal pr in modpr format compute a canonical representative for the class of x modulo pr.

The library syntax is GEN nfreducemodpr(GEN nf, GEN x, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nf_to_Fq, then work there.

nfelttrace(nf,x)

Returns the absolute trace of x.

The library syntax is GEN nftrace(GEN nf, GEN x).

nfeltval(nf,x,pr)

Given an element x in nf and a prime ideal pr in the format output by idealprimedec, computes the valuation at pr of the element x. The same result can be obtained using idealval(nf,x,pr), since x is then converted to a principal ideal.

The library syntax is long nfval(GEN nf, GEN x, GEN pr).

nffactor(nf,T)

Factorization of the univariate polynomial T over the number field nf given by nfinit; T has coefficients in nf (i.e. either scalar, polmod, polynomial or column vector). The factors are sorted by increasing degree.

The main variable of nf must be of lower priority than that of T, see Label se:priority. However if the polynomial defining the number field occurs explicitly in the coefficients of T as modulus of a t_POLMOD or as a t_POL coefficient, its main variable must be the same as the main variable of T. For example,

  ? nf = nfinit(y^2 + 1);
  ? nffactor(nf, x^2 + y); \\ OK
  ? nffactor(nf, x^2 + Mod(y, y^2+1)); \\  OK
  ? nffactor(nf, x^2 + Mod(z, z^2+1)); \\  WRONG

It is possible to input a defining polynomial for nf instead, but this is in general less efficient since parts of an nf structure will then be computed internally. This is useful in two situations: when you do not need the nf elsewhere, or when you cannot compute the field discriminant due to integer factorization difficulties. In the latter case, if you must use a partial discriminant factorization (as allowed by both nfdisc or nfbasis) to build a partially correct nf structure, always input nf.pol to nffactor, and not your makeshift nf: otherwise factors could be missed.

The library syntax is GEN nffactor(GEN nf, GEN T).

nffactorback(nf,f,{e})

Gives back the nf element corresponding to a factorization. The integer 1 corresponds to the empty factorization.

If e is present, e and f must be vectors of the same length (e being integral), and the corresponding factorization is the product of the f[i]^{e[i]}.

If not, and f is vector, it is understood as in the preceding case with e a vector of 1s: we return the product of the f[i]. Finally, f can be a regular factorization matrix.

  ? nf = nfinit(y^2+1);
  ? nffactorback(nf, [3, y+1, [1,2]~], [1, 2, 3])
  %2 = [12, -66]~
  ? 3 * (I+1)^2 * (1+2*I)^3
  %3 = 12 - 66*I

The library syntax is GEN nffactorback(GEN nf, GEN f, GEN e = NULL).

nffactormod(nf,Q,pr)

Factors the univariate polynomial Q modulo the prime ideal pr in the number field nf. The coefficients of Q belong to the number field (scalar, polmod, polynomial, even column vector) and the main variable of nf must be of lower priority than that of Q (see Label se:priority). The prime ideal pr is either in idealprimedec or (preferred) modprinit format. The coefficients of the polynomial factors are lifted to elements of nf:

  ? K = nfinit(y^2+1);
  ? P = idealprimedec(K, 3)[1];
  ? nffactormod(K, x^2 + y*x + 18*y+1, P)
  %3 =
  [x + (2*y + 1) 1]
  [x + (2*y + 2) 1]
  ? P = nfmodprinit(K, P);  \\ convert to nfmodprinit format
  ? nffactormod(K, x^2 + y*x + 18*y+1)
  [x + (2*y + 1) 1]
  [x + (2*y + 2) 1]

@3Same result, of course, here about 10% faster due to the precomputation.

The library syntax is GEN nffactormod(GEN nf, GEN Q, GEN pr).

nfgaloisapply(nf,aut,x)

Let nf be a number field as output by nfinit, and let aut be a Galois automorphism of nf expressed by its image on the field generator (such automorphisms can be found using nfgaloisconj). The function computes the action of the automorphism aut on the object x in the number field; x can be a number field element, or an ideal (possibly extended). Because of possible confusion with elements and ideals, other vector or matrix arguments are forbidden.

   ? nf = nfinit(x^2+1);
   ? L = nfgaloisconj(nf)
   %2 = [-x, x]~
   ? aut = L[1]; /* the non-trivial automorphism */
   ? nfgaloisapply(nf, aut, x)
   %4 = Mod(-x, x^2 + 1)
   ? P = idealprimedec(nf,5); /* prime ideals above 5 */
   ? nfgaloisapply(nf, aut, P[2]) == P[1]
   %7 = 0 \\ !!!!
   ? idealval(nf, nfgaloisapply(nf, aut, P[2]), P[1])
   %8 = 1

@3The surprising failure of the equality test (%7) is due to the fact that although the corresponding prime ideals are equal, their representations are not. (A prime ideal is specified by a uniformizer, and there is no guarantee that applying automorphisms yields the same elements as a direct idealprimedec call.)

The automorphism can also be given as a column vector, representing the image of Mod(x, nf.pol) as an algebraic number. This last representation is more efficient and should be preferred if a given automorphism must be used in many such calls.

   ? nf = nfinit(x^3 - 37*x^2 + 74*x - 37);
   ? l = nfgaloisconj(nf); aut = l[2] \\  automorphisms in basistoalg form
   %2 = -31/11*x^2 + 1109/11*x - 925/11
   ? L = matalgtobasis(nf, l); AUT = L[2] \\  same in algtobasis form
   %3 = [16, -6, 5]~
   ? v = [1, 2, 3]~; nfgaloisapply(nf, aut, v) == nfgaloisapply(nf, AUT, v)
   %4 = 1 \\  same result...
   ? for (i=1,10^5, nfgaloisapply(nf, aut, v))
   time = 1,451 ms.
   ? for (i=1,10^5, nfgaloisapply(nf, AUT, v))
   time = 1,045 ms.  \\  but the latter is faster

The library syntax is GEN galoisapply(GEN nf, GEN aut, GEN x).

nfgaloisconj(nf,{flag = 0},{d})

nf being a number field as output by nfinit, computes the conjugates of a root r of the non-constant polynomial x = nf[1] expressed as polynomials in r. This also makes sense when the number field is not Galois since some conjugates may lie in the field. nf can simply be a polynomial.

If no flags or flag = 0, use a combination of flag 4 and 1 and the result is always complete. There is no point whatsoever in using the other flags.

If flag = 1, use nfroots: a little slow, but guaranteed to work in polynomial time.

If flag = 2 (OBSOLETE), use complex approximations to the roots and an integral LLL. The result is not guaranteed to be complete: some conjugates may be missing (a warning is issued if the result is not proved complete), especially so if the corresponding polynomial has a huge index, and increasing the default precision may help. This variant is slow and unreliable: don't use it.

If flag = 4, use galoisinit: very fast, but only applies to (most) Galois fields. If the field is Galois with weakly super-solvable Galois group (see galoisinit), return the complete list of automorphisms, else only the identity element. If present, d is assumed to be a multiple of the least common denominator of the conjugates expressed as polynomial in a root of pol.

This routine can only compute Q-automorphisms, but it may be used to get K-automorphism for any base field K as follows:

  rnfgaloisconj(nfK, R) = \\ K-automorphisms of L = K[X] / (R)
  { my(polabs, N);
    R *= Mod(1, nfK.pol);             \\ convert coeffs to polmod elts of K
    polabs = rnfequation(nfK, R);
    N = nfgaloisconj(polabs) % R;     \\ Q-automorphisms of L
    \\ select the ones that fix K
    select(s->subst(R, variable(R), Mod(s,R)) == 0, N);
  }
  K  = nfinit(y^2 + 7);
  rnfgaloisconj(K, x^4 - y*x^3 - 3*x^2 + y*x + 1)  \\ K-automorphisms of L

The library syntax is GEN galoisconj0(GEN nf, long flag, GEN d = NULL, long prec). Use directly GEN galoisconj(GEN nf, GEN d), corresponding to flag = 0, the others only have historical interest.

nfhilbert(nf,a,b,{pr})

If pr is omitted, compute the global quadratic Hilbert symbol (a,b) in nf, that is 1 if x^2 - a y^2 - b z^2 has a non trivial solution (x,y,z) in nf, and -1 otherwise. Otherwise compute the local symbol modulo the prime ideal pr, as output by idealprimedec.

The library syntax is long nfhilbert0(GEN nf, GEN a, GEN b, GEN pr = NULL).

Also available is long nfhilbert(GEN bnf,GEN a,GEN b) (global quadratic Hilbert symbol).

nfhnf(nf,x)

Given a pseudo-matrix (A,I), finds a pseudo-basis in Hermite normal form of the module it generates.

The library syntax is GEN nfhnf(GEN nf, GEN x). Also available:

GEN rnfsimplifybasis(GEN bnf, GEN x) simplifies the pseudo-basis given by x = (A,I). The ideals in the list I are integral, primitive and either trivial (equal to the full ring of integer) or non-principal.

nfhnfmod(nf,x,detx)

Given a pseudo-matrix (A,I) and an ideal detx which is contained in (read integral multiple of) the determinant of (A,I), finds a pseudo-basis in Hermite normal form of the module generated by (A,I). This avoids coefficient explosion. detx can be computed using the function nfdetint.

The library syntax is GEN nfhnfmod(GEN nf, GEN x, GEN detx).

nfinit(pol,{flag = 0})

pol being a non-constant, preferably monic, irreducible polynomial in Z[X], initializes a number field structure (nf) associated to the field K defined by pol. As such, it's a technical object passed as the first argument to most nfxxx functions, but it contains some information which may be directly useful. Access to this information via member functions is preferred since the specific data organization specified below may change in the future. Currently, nf is a row vector with 9 components:

nf[1] contains the polynomial pol (nf.pol).

nf[2] contains [r1,r2] (nf.sign, nf.r1, nf.r2), the number of real and complex places of K.

nf[3] contains the discriminant d(K) (nf.disc) of K.

nf[4] contains the index of nf[1] (nf.index), i.e. [Z_K : Z[theta]], where theta is any root of nf[1].

nf[5] is a vector containing 7 matrices M, G, roundG, T, MD, TI, MDI useful for certain computations in the number field K.

  * M is the (r1+r2) x n matrix whose columns represent the numerical values of the conjugates of the elements of the integral basis.

  * G is an n x n matrix such that T2 = ^t G G, where T2 is the quadratic form T_2(x) = sum |sigma(x)|^2, sigma running over the embeddings of K into C.

  * roundG is a rescaled copy of G, rounded to nearest integers.

  * T is the n x n matrix whose coefficients are {Tr}(omega_iomega_j) where the omega_i are the elements of the integral basis. Note also that det (T) is equal to the discriminant of the field K. Also, when understood as an ideal, the matrix T^{-1} generates the codifferent ideal.

  * The columns of MD (nf.diff) express a Z-basis of the different of K on the integral basis.

  * TI is equal to the primitive part of T^{-1}, which has integral coefficients.

  * Finally, MDI is a two-element representation (for faster ideal product) of d(K) times the codifferent ideal (nf.disc*nf.codiff, which is an integral ideal). MDI is only used in idealinv.

nf[6] is the vector containing the r1+r2 roots (nf.roots) of nf[1] corresponding to the r1+r2 embeddings of the number field into C (the first r1 components are real, the next r2 have positive imaginary part).

nf[7] is an integral basis for Z_K (nf.zk) expressed on the powers of theta. Its first element is guaranteed to be 1. This basis is LLL-reduced with respect to T_2 (strictly speaking, it is a permutation of such a basis, due to the condition that the first element be 1).

nf[8] is the n x n integral matrix expressing the power basis in terms of the integral basis, and finally

nf[9] is the n x n^2 matrix giving the multiplication table of the integral basis.

If a non monic polynomial is input, nfinit will transform it into a monic one, then reduce it (see flag = 3). It is allowed, though not very useful given the existence of nfnewprec, to input a nf or a bnf instead of a polynomial.

  ? nf = nfinit(x^3 - 12); \\ initialize number field Q[X] / (X^3 - 12)
  ? nf.pol   \\ defining polynomial
  %2 = x^3 - 12
  ? nf.disc  \\ field discriminant
  %3 = -972
  ? nf.index \\ index of power basis order in maximal order
  %4 = 2
  ? nf.zk    \\ integer basis, lifted to Q[X]
  %5 = [1, x, 1/2*x^2]
  ? nf.sign  \\ signature
  %6 = [1, 1]
  ? factor(abs(nf.disc ))  \\ determines ramified primes
  %7 =
  [2 2]
  [3 5]
  ? idealfactor(nf, 2)
  %8 =
  [[2, [0, 0, -1]~, 3, 1, [0, 1, 0]~] 3]  \\  B<p>_2^3

@3Huge discriminants, helping nfdisc.

In case pol has a huge discriminant which is difficult to factor, it is hard to compute from scratch the maximal order. The special input format [pol, B] is also accepted where pol is a polynomial as above and B has one of the following forms

@3* an integer basis, as would be computed by nfbasis: a vector of polynomials with first element 1. This is useful if the maximal order is known in advance.

@3* an argument listP which specifies a list of primes (see nfbasis). Instead of the maximal order, nfinit then computes an order which is maximal at these particular primes as well as the primes contained in the private prime table (see addprimes). The result is unconditionaly correct when the discriminant nf.disc factors completely over this set of primes. The function nfcertify automates this:

  ? pol = polcompositum(x^5 - 101, polcyclo(7))[1];
  ? nf = nfinit( [pol, 10^3] );
  ? nfcertify(nf)
  %3 = []

@3A priori, nf.zk defines an order which is only known to be maximal at all primes <= 10^3 (no prime <= 10^3 divides nf.index). The certification step proves the correctness of the computation.

If flag = 2: pol is changed into another polynomial P defining the same number field, which is as simple as can easily be found using the polredbest algorithm, and all the subsequent computations are done using this new polynomial. In particular, the first component of the result is the modified polynomial.

If flag = 3, apply polredbest as in case 2, but outputs [nf,Mod(a,P)], where nf is as before and Mod(a,P) = Mod(x,pol) gives the change of variables. This is implicit when pol is not monic: first a linear change of variables is performed, to get a monic polynomial, then polredbest.

The library syntax is GEN nfinit0(GEN pol, long flag, long prec). Also available are GEN nfinit(GEN x, long prec) (flag = 0), GEN nfinitred(GEN x, long prec) (flag = 2), GEN nfinitred2(GEN x, long prec) (flag = 3). Instead of the above hardcoded numerical flags in nfinit0, one should rather use

GEN nfinitall(GEN x, long flag, long prec), where flag is an or-ed combination of

@3* nf_RED: find a simpler defining polynomial,

@3* nf_ORIG: if nf_RED set, also return the change of variable,

@3* nf_ROUND2: Deprecated. Slow down the routine by using an obsolete normalization algorithm (do not use this one!),

@3* nf_PARTIALFACT: Deprecated. Lazy factorization of the polynomial discriminant. Result is conditional unless nfcertify can certify it.

nfisideal(nf,x)

Returns 1 if x is an ideal in the number field nf, 0 otherwise.

The library syntax is long isideal(GEN nf, GEN x).

nfisincl(x,y)

Tests whether the number field K defined by the polynomial x is conjugate to a subfield of the field L defined by y (where x and y must be in Q[X]). If they are not, the output is the number 0. If they are, the output is a vector of polynomials, each polynomial a representing an embedding of K into L, i.e. being such that y | x o a.

If y is a number field (nf), a much faster algorithm is used (factoring x over y using nffactor). Before version 2.0.14, this wasn't guaranteed to return all the embeddings, hence was triggered by a special flag. This is no more the case.

The library syntax is GEN nfisincl(GEN x, GEN y).

nfisisom(x,y)

As nfisincl, but tests for isomorphism. If either x or y is a number field, a much faster algorithm will be used.

The library syntax is GEN nfisisom(GEN x, GEN y).

nfkermodpr(nf,x,pr)

Kernel of the matrix a in Z_K/pr, where pr is in modpr format (see nfmodprinit).

The library syntax is GEN nfkermodpr(GEN nf, GEN x, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nfM_to_FqM, then work there.

nfmodprinit(nf,pr)

Transforms the prime ideal pr into modpr format necessary for all operations modulo pr in the number field nf.

The library syntax is GEN nfmodprinit(GEN nf, GEN pr).

nfnewprec(nf)

Transforms the number field nf into the corresponding data using current (usually larger) precision. This function works as expected if nf is in fact a bnf (update bnf to current precision) but may be quite slow (many generators of principal ideals have to be computed).

The library syntax is GEN nfnewprec(GEN nf, long prec). See also GEN bnfnewprec(GEN bnf, long prec) and GEN bnrnewprec(GEN bnr, long prec).

nfroots({nf},x)

Roots of the polynomial x in the number field nf given by nfinit without multiplicity (in Q if nf is omitted). x has coefficients in the number field (scalar, polmod, polynomial, column vector). The main variable of nf must be of lower priority than that of x (see Label se:priority). However if the coefficients of the number field occur explicitly (as polmods) as coefficients of x, the variable of these polmods must be the same as the main variable of t (see nffactor).

It is possible to input a defining polynomial for nf instead, but this is in general less efficient since parts of an nf structure will be computed internally. This is useful in two situations: when you don't need the nf, or when you can't compute its discriminant due to integer factorization difficulties. In the latter case, addprimes is a possibility but a dangerous one: roots will probably be missed if the (true) field discriminant and an addprimes entry are strictly divisible by some prime. If you have such an unsafe nf, it is safer to input nf.pol.

The library syntax is GEN nfroots(GEN nf = NULL, GEN x). See also GEN nfrootsQ(GEN x), corresponding to nf = NULL.

nfrootsof1(nf)

Returns a two-component vector [w,z] where w is the number of roots of unity in the number field nf, and z is a primitive w-th root of unity.

  ? K = nfinit(polcyclo(11));
  ? nfrootsof1(K)
  %2 = [22, [0, 0, 0, 0, 0, -1, 0, 0, 0, 0]~]
  ? z = nfbasistoalg(K, %[2])   \\ in algebraic form
  %3 = Mod(-x^5, x^10 + x^9 + x^8 + x^7 + x^6 + x^5 + x^4 + x^3 + x^2 + x + 1)
  ? [lift(z^11), lift(z^2)]     \\ proves that the order of z is 22
  %4 = [-1, -x^9 - x^8 - x^7 - x^6 - x^5 - x^4 - x^3 - x^2 - x - 1]

This function guesses the number w as the gcd of the #k(v)^* for unramified v above odd primes, then computes the roots in nf of the w-th cyclotomic polynomial: the algorithm is polynomial time with respect to the field degree and the bitsize of the multiplication table in nf (both of them polynomially bounded in terms of the size of the discriminant). Fields of degree up to 100 or so should require less than one minute.

The library syntax is GEN rootsof1(GEN nf). Also available is GEN rootsof1_kannan(GEN nf), that computes all algebraic integers of T_2 norm equal to the field degree (all roots of 1, by Kronecker's theorem). This is in general a little faster than the default when there are roots of 1 in the field (say twice faster), but can be much slower (say, days slower), since the algorithm is a priori exponential in the field degree.

nfsnf(nf,x)

Given a Z_K-module x associated to the integral pseudo-matrix (A,I,J), returns an ideal list d_1,...,d_n which is the Smith normal form of x. In other words, x is isomorphic to Z_K/d_1 oplus ... oplus Z_K/d_n and d_i divides d_{i-1} for i >= 2.

See Label se:ZKmodules for the definition of integral pseudo-matrix; briefly, it is input as a 3-component row vector [A,I,J] where I = [b_1,...,b_n] and J = [a_1,...,a_n] are two ideal lists, and A is a square n x n matrix with columns (A_1,...,A_n), seen as elements in K^n (with canonical basis (e_1,...,e_n)). This data defines the Z_K module x given by

   (b_1e_1 oplus ... oplus b_ne_n) / (a_1A_1 oplus ... oplus a_nA_n) ,

The integrality condition is a_{i,j} belongs to b_i a_j^{-1} for all i,j. If it is not satisfied, then the d_i will not be integral. Note that every finitely generated torsion module is isomorphic to a module of this form and even with b_i = Z_K for all i.

The library syntax is GEN nfsnf(GEN nf, GEN x).

nfsolvemodpr(nf,a,b,P)

Let P be a prime ideal in modpr format (see nfmodprinit), let a be a matrix, invertible over the residue field, and let b be a column vector or matrix. This function returns a solution of a.x = b; the coefficients of x are lifted to nf elements.

  ? K = nfinit(y^2+1);
  ? P = idealprimedec(K, 3)[1];
  ? P = nfmodprinit(K, P);
  ? a = [y+1, y; y, 0]; b = [1, y]~
  ? nfsolvemodpr(K, a,b, P)
  %5 = [1, 2]~

The library syntax is GEN nfsolvemodpr(GEN nf, GEN a, GEN b, GEN P). This function is normally useless in library mode. Project your inputs to the residue field using nfM_to_FqM, then work there.

nfsubfields(pol,{d = 0})

Finds all subfields of degree d of the number field defined by the (monic, integral) polynomial pol (all subfields if d is null or omitted). The result is a vector of subfields, each being given by [g,h], where g is an absolute equation and h expresses one of the roots of g in terms of the root x of the polynomial defining nf. This routine uses J. Klüners's algorithm in the general case, and B. Allombert's galoissubfields when nf is Galois (with weakly supersolvable Galois group).

The library syntax is GEN nfsubfields(GEN pol, long d).

polcompositum(P,Q,{flag = 0})

P and Q being squarefree polynomials in Z[X] in the same variable, outputs the simple factors of the étale Q-algebra A = Q(X, Y) / (P(X), Q(Y)). The factors are given by a list of polynomials R in Z[X], associated to the number field Q(X)/ (R), and sorted by increasing degree (with respect to lexicographic ordering for factors of equal degrees). Returns an error if one of the polynomials is not squarefree.

Note that it is more efficient to reduce to the case where P and Q are irreducible first. The routine will not perform this for you, since it may be expensive, and the inputs are irreducible in most applications anyway. In this case, there will be a single factor R if and only if the number fields defined by P and Q are disjoint.

Assuming P is irreducible (of smaller degree than Q for efficiency), it is in general much faster to proceed as follows

  nf = nfinit(P); L = nffactor(nf, Q)[,1];
  vector(#L, i, rnfequation(nf, L[i]))

to obtain the same result. If you are only interested in the degrees of the simple factors, the rnfequation instruction can be replaced by a trivial poldegree(P) * poldegree(L[i]).

If flag = 1, outputs a vector of 4-component vectors [R,a,b,k], where R ranges through the list of all possible compositums as above, and a (resp. b) expresses the root of P (resp. Q) as an element of Q(X)/(R). Finally, k is a small integer such that b + ka = X modulo R.

A compositum is often defined by a complicated polynomial, which it is advisable to reduce before further work. Here is an example involving the field Q(zeta_5, 5^{1/5}):

  ? L = polcompositum(x^5 - 5, polcyclo(5), 1); \\ list of [R,a,b,k]
  ? [R, a] = L[1];  \\ pick the single factor, extract R,a (ignore b,k)
  ? R               \\ defines the compositum
  %3 = x^20 + 5*x^19 + 15*x^18 + 35*x^17 + 70*x^16 + 141*x^15 + 260*x^14\
  + 355*x^13 + 95*x^12 - 1460*x^11 - 3279*x^10 - 3660*x^9 - 2005*x^8    \
  + 705*x^7 + 9210*x^6 + 13506*x^5 + 7145*x^4 - 2740*x^3 + 1040*x^2     \
  - 320*x + 256
  ? a^5 - 5         \\ a fifth root of 5
  %4 = 0
  ? [T, X] = polredbest(R, 1);
  ? T     \\ simpler defining polynomial for B<Q>[x]/(R)
  %6 = x^20 + 25*x^10 + 5
  ? X     \\  root of R in B<Q>[y]/(T(y))
  %7 = Mod(-1/11*x^15 - 1/11*x^14 + 1/22*x^10 - 47/22*x^5 - 29/11*x^4 + 7/22,\
  x^20 + 25*x^10 + 5)
  ? a = subst(a.pol, 'x, X)  \\ C<a> in the new coordinates
  %8 = Mod(1/11*x^14 + 29/11*x^4, x^20 + 25*x^10 + 5)
  ? a^5 - 5
  %9 = 0

The library syntax is GEN polcompositum0(GEN P, GEN Q, long flag). Also available are GEN compositum(GEN P, GEN Q) (flag = 0) and GEN compositum2(GEN P, GEN Q) (flag = 1).

polgalois(T)

Galois group of the non-constant polynomial T belongs to Q[X]. In the present version 2.7.4, T must be irreducible and the degree d of T must be less than or equal to 7. If the galdata package has been installed, degrees 8, 9, 10 and 11 are also implemented. By definition, if K = Q[x]/(T), this computes the action of the Galois group of the Galois closure of K on the d distinct roots of T, up to conjugacy (corresponding to different root orderings).

The output is a 4-component vector [n,s,k,name] with the following meaning: n is the cardinality of the group, s is its signature (s = 1 if the group is a subgroup of the alternating group A_d, s = -1 otherwise) and name is a character string containing name of the transitive group according to the GAP 4 transitive groups library by Alexander Hulpke.

k is more arbitrary and the choice made up to version 2.2.3 of PARI is rather unfortunate: for d > 7, k is the numbering of the group among all transitive subgroups of S_d, as given in ``The transitive groups of degree up to eleven'', G. Butler and J. McKay, Communications in Algebra, vol. 11, 1983, pp. 863--911 (group k is denoted T_k there). And for d <= 7, it was ad hoc, so as to ensure that a given triple would denote a unique group. Specifically, for polynomials of degree d <= 7, the groups are coded as follows, using standard notations

In degree 1: S_1 = [1,1,1].

In degree 2: S_2 = [2,-1,1].

In degree 3: A_3 = C_3 = [3,1,1], S_3 = [6,-1,1].

In degree 4: C_4 = [4,-1,1], V_4 = [4,1,1], D_4 = [8,-1,1], A_4 = [12,1,1], S_4 = [24,-1,1].

In degree 5: C_5 = [5,1,1], D_5 = [10,1,1], M_{20} = [20,-1,1], A_5 = [60,1,1], S_5 = [120,-1,1].

In degree 6: C_6 = [6,-1,1], S_3 = [6,-1,2], D_6 = [12,-1,1], A_4 = [12,1,1], G_{18} = [18,-1,1], S_4^ -= [24,-1,1], A_4 x C_2 = [24,-1,2], S_4^ += [24,1,1], G_{36}^ -= [36,-1,1], G_{36}^ += [36,1,1], S_4 x C_2 = [48,-1,1], A_5 = PSL_2(5) = [60,1,1], G_{72} = [72,-1,1], S_5 = PGL_2(5) = [120,-1,1], A_6 = [360,1,1], S_6 = [720,-1,1].

In degree 7: C_7 = [7,1,1], D_7 = [14,-1,1], M_{21} = [21,1,1], M_{42} = [42,-1,1], PSL_2(7) = PSL_3(2) = [168,1,1], A_7 = [2520,1,1], S_7 = [5040,-1,1].

This is deprecated and obsolete, but for reasons of backward compatibility, we cannot change this behavior yet. So you can use the default new_galois_format to switch to a consistent naming scheme, namely k is always the standard numbering of the group among all transitive subgroups of S_n. If this default is in effect, the above groups will be coded as:

In degree 1: S_1 = [1,1,1].

In degree 2: S_2 = [2,-1,1].

In degree 3: A_3 = C_3 = [3,1,1], S_3 = [6,-1,2].

In degree 4: C_4 = [4,-1,1], V_4 = [4,1,2], D_4 = [8,-1,3], A_4 = [12,1,4], S_4 = [24,-1,5].

In degree 5: C_5 = [5,1,1], D_5 = [10,1,2], M_{20} = [20,-1,3], A_5 = [60,1,4], S_5 = [120,-1,5].

In degree 6: C_6 = [6,-1,1], S_3 = [6,-1,2], D_6 = [12,-1,3], A_4 = [12,1,4], G_{18} = [18,-1,5], A_4 x C_2 = [24,-1,6], S_4^ += [24,1,7], S_4^ -= [24,-1,8], G_{36}^ -= [36,-1,9], G_{36}^ += [36,1,10], S_4 x C_2 = [48,-1,11], A_5 = PSL_2(5) = [60,1,12], G_{72} = [72,-1,13], S_5 = PGL_2(5) = [120,-1,14], A_6 = [360,1,15], S_6 = [720,-1,16].

In degree 7: C_7 = [7,1,1], D_7 = [14,-1,2], M_{21} = [21,1,3], M_{42} = [42,-1,4], PSL_2(7) = PSL_3(2) = [168,1,5], A_7 = [2520,1,6], S_7 = [5040,-1,7].

@3Warning. The method used is that of resolvent polynomials and is sensitive to the current precision. The precision is updated internally but, in very rare cases, a wrong result may be returned if the initial precision was not sufficient.

The library syntax is GEN polgalois(GEN T, long prec). To enable the new format in library mode, set the global variable new_galois_format to 1.

polred(T,{flag = 0})

This function is deprecated, use polredbest instead. Finds polynomials with reasonably small coefficients defining subfields of the number field defined by T. One of the polynomials always defines Q (hence is equal to x-1), and another always defines the same number field as T if T is irreducible.

All T accepted by nfinit are also allowed here; in particular, the format [T, listP] is recommended, e.g. with listP = 10^5 or a vector containing all ramified primes. Otherwise, the maximal order of Q[x]/(T) must be computed.

The following binary digits of flag are significant:

1: Possibly use a suborder of the maximal order. The primes dividing the index of the order chosen are larger than primelimit or divide integers stored in the addprimes table. This flag is deprecated, the [T, listP] format is more flexible.

2: gives also elements. The result is a two-column matrix, the first column giving primitive elements defining these subfields, the second giving the corresponding minimal polynomials.

  ? M = polred(x^4 + 8, 2)
  %1 =
  [1 x - 1]
  [1/2*x^2 x^2 + 2]
  [1/4*x^3 x^4 + 2]
  [x x^4 + 8]
  ? minpoly(Mod(M[2,1], x^4+8))
  %2 = x^2 + 2

The library syntax is polred(GEN T) (flag = 0). Also available is GEN polred2(GEN T) (flag = 2). The function polred0 is deprecated, provided for backward compatibility.

polredabs(T,{flag = 0})

Returns a canonical defining polynomial P for the number field Q[X]/(T) defined by T, such that the sum of the squares of the modulus of the roots (i.e. the T_2-norm) is minimal. Different T defining isomorphic number fields will yield the same P. All T accepted by nfinit are also allowed here, e.g. non-monic polynomials, or pairs [T, listP] specifying that a non-maximal order may be used.

@3Warning 1. Using a t_POL T requires fully factoring the discriminant of T, which may be very hard. The format [T, listP] computes only a suborder of the maximal order and replaces this part of the algorithm by a polynomial time computation. In that case the polynomial P is a priori no longer canonical, and it may happen that it does not have minimal T_2 norm. The routine attempts to certify the result independently of this order computation (as per nfcertify: we try to prove that the order is maximal); if it fails, the routine returns 0 instead of P. In order to force an output in that case as well, you may either use polredbest, or polredabs(,16), or

    polredabs([T, nfbasis([T, listP])])

@3(In all three cases, the result is no longer canonical.)

@3Warning 2. Apart from the factorization of the discriminant of T, this routine runs in polynomial time for a fixed degree. But the complexity is exponential in the degree: this routine may be exceedingly slow when the number field has many subfields, hence a lot of elements of small T_2-norm. If you do not need a canonical polynomial, the function polredbest is in general much faster (it runs in polynomial time), and tends to return polynomials with smaller discriminants.

The binary digits of flag mean

1: outputs a two-component row vector [P,a], where P is the default output and Mod(a, P) is a root of the original T.

4: gives all polynomials of minimal T_2 norm; of the two polynomials P(x) and +- P(-x), only one is given.

16: Possibly use a suborder of the maximal order, without attempting to certify the result as in Warning 1: we always return a polynomial and never 0. The result is a priori not canonical.

  ? T = x^16 - 136*x^14 + 6476*x^12 - 141912*x^10 + 1513334*x^8 \
        - 7453176*x^6 + 13950764*x^4 - 5596840*x^2 + 46225
  ? T1 = polredabs(T); T2 = polredbest(T);
  ? [ norml2(polroots(T1)), norml2(polroots(T2)) ]
  %3 = [88.0000000, 120.000000]
  ? [ sizedigit(poldisc(T1)), sizedigit(poldisc(T2)) ]
  %4 = [75, 67]

The library syntax is GEN polredabs0(GEN T, long flag). Instead of the above hardcoded numerical flags, one should use an or-ed combination of

@3* nf_PARTIALFACT: possibly use a suborder of the maximal order, without attempting to certify the result.

@3* nf_ORIG: return [P, a], where Mod(a, P) is a root of T.

@3* nf_RAW: return [P, b], where Mod(b, T) is a root of P. The algebraic integer b is the raw result produced by the small vectors enumeration in the maximal order; P was computed as the characteristic polynomial of Mod(b, T). Mod(a, P) as in nf_ORIG is obtained with modreverse.

@3* nf_ADDZK: if r is the result produced with some of the above flags (of the form P or [P,c]), return [r,zk], where zk is a Z-basis for the maximal order of Q[X]/(P).

@3* nf_ALL: return a vector of results of the above form, for all polynomials of minimal T_2-norm.

polredbest(T,{flag = 0})

Finds a polynomial with reasonably small coefficients defining the same number field as T. All T accepted by nfinit are also allowed here (e.g. non-monic polynomials, nf, bnf, [T,Z_K_basis]). Contrary to polredabs, this routine runs in polynomial time, but it offers no guarantee as to the minimality of its result.

This routine computes an LLL-reduced basis for the ring of integers of Q[X]/(T), then examines small linear combinations of the basis vectors, computing their characteristic polynomials. It returns the separable P polynomial of smallest discriminant (the one with lexicographically smallest abs(Vec(P)) in case of ties). This is a good candidate for subsequent number field computations, since it guarantees that the denominators of algebraic integers, when expressed in the power basis, are reasonably small. With no claim of minimality, though.

It can happen that iterating this functions yields better and better polynomials, until it stabilizes:

  ? \p5
  ? P = X^12+8*X^8-50*X^6+16*X^4-3069*X^2+625;
  ? poldisc(P)*1.
  %2 = 1.2622 E55
  ? P = polredbest(P);
  ? poldisc(P)*1.
  %4 = 2.9012 E51
  ? P = polredbest(P);
  ? poldisc(P)*1.
  %6 = 8.8704 E44

@3In this example, the initial polynomial P is the one returned by polredabs, and the last one is stable.

If flag = 1: outputs a two-component row vector [P,a], where P is the default output and Mod(a, P) is a root of the original T.

  ? [P,a] = polredbest(x^4 + 8, 1)
  %1 = [x^4 + 2, Mod(x^3, x^4 + 2)]
  ? charpoly(a)
  %2 = x^4 + 8

@3In particular, the map Q[x]/(T) \to Q[x]/(P), x|--->Mod(a,P) defines an isomorphism of number fields, which can be computed as

    subst(lift(Q), 'x, a)

@3if Q is a t_POLMOD modulo T; b = modreverse(a) returns a t_POLMOD giving the inverse of the above map (which should be useless since Q[x]/(P) is a priori a better representation for the number field and its elements).

The library syntax is GEN polredbest(GEN T, long flag).

polredord(x)

Finds polynomials with reasonably small coefficients and of the same degree as that of x defining suborders of the order defined by x. One of the polynomials always defines Q (hence is equal to (x-1)^n, where n is the degree), and another always defines the same order as x if x is irreducible. Useless function: try polredbest.

The library syntax is GEN polredord(GEN x).

poltschirnhaus(x)

Applies a random Tschirnhausen transformation to the polynomial x, which is assumed to be non-constant and separable, so as to obtain a new equation for the étale algebra defined by x. This is for instance useful when computing resolvents, hence is used by the polgalois function.

The library syntax is GEN tschirnhaus(GEN x).

rnfalgtobasis(rnf,x)

Expresses x on the relative integral basis. Here, rnf is a relative number field extension L/K as output by rnfinit, and x an element of L in absolute form, i.e. expressed as a polynomial or polmod with polmod coefficients, not on the relative integral basis.

The library syntax is GEN rnfalgtobasis(GEN rnf, GEN x).

rnfbasis(bnf,M)

Let K the field represented by bnf, as output by bnfinit. M is a projective Z_K-module of rank n (M\otimes K is an n-dimensional K-vector space), given by a pseudo-basis of size n. The routine returns either a true Z_K-basis of M (of size n) if it exists, or an n+1-element generating set of M if not.

It is allowed to use an irreducible polynomial P in K[X] instead of M, in which case, M is defined as the ring of integers of K[X]/(P), viewed as a Z_K-module.

The library syntax is GEN rnfbasis(GEN bnf, GEN M).

rnfbasistoalg(rnf,x)

Computes the representation of x as a polmod with polmods coefficients. Here, rnf is a relative number field extension L/K as output by rnfinit, and x an element of L expressed on the relative integral basis.

The library syntax is GEN rnfbasistoalg(GEN rnf, GEN x).

rnfcharpoly(nf,T,a,{var = 'x})

Characteristic polynomial of a over nf, where a belongs to the algebra defined by T over nf, i.e. nf[X]/(T). Returns a polynomial in variable v (x by default).

  ? nf = nfinit(y^2+1);
  ? rnfcharpoly(nf, x^2+y*x+1, x+y)
  %2 = x^2 + Mod(-y, y^2 + 1)*x + 1

The library syntax is GEN rnfcharpoly(GEN nf, GEN T, GEN a, long var = -1), where var is a variable number.

rnfconductor(bnf,pol)

Given bnf as output by bnfinit, and pol a relative polynomial defining an Abelian extension, computes the class field theory conductor of this Abelian extension. The result is a 3-component vector [conductor,rayclgp,subgroup], where conductor is the conductor of the extension given as a 2-component row vector [f_0,f_ oo ], rayclgp is the full ray class group corresponding to the conductor given as a 3-component vector [h,cyc,gen] as usual for a group, and subgroup is a matrix in HNF defining the subgroup of the ray class group on the given generators gen.

The library syntax is GEN rnfconductor(GEN bnf, GEN pol).

rnfdedekind(nf,pol,{pr},{flag = 0})

Given a number field K coded by nf and a monic polynomial P belongs to Z_K[X], irreducible over K and thus defining a relative extension L of K, applies Dedekind's criterion to the order Z_K[X]/(P), at the prime ideal pr. It is possible to set pr to a vector of prime ideals (test maximality at all primes in the vector), or to omit altogether, in which case maximality at all primes is tested; in this situation flag is automatically set to 1.

The default historic behavior (flag is 0 or omitted and pr is a single prime ideal) is not so useful since rnfpseudobasis gives more information and is generally not that much slower. It returns a 3-component vector [max, basis, v]:

@3* basis is a pseudo-basis of an enlarged order O produced by Dedekind's criterion, containing the original order Z_K[X]/(P) with index a power of pr. Possibly equal to the original order.

@3* max is a flag equal to 1 if the enlarged order O could be proven to be pr-maximal and to 0 otherwise; it may still be maximal in the latter case if pr is ramified in L,

@3* v is the valuation at pr of the order discriminant.

If flag is non-zero, on the other hand, we just return 1 if the order Z_K[X]/(P) is pr-maximal (resp. maximal at all relevant primes, as described above), and 0 if not. This is much faster than the default, since the enlarged order is not computed.

  ? nf = nfinit(y^2-3); P = x^3 - 2*y;
  ? pr3 = idealprimedec(nf,3)[1];
  ? rnfdedekind(nf, P, pr3)
  %2 = [1, [[1, 0, 0; 0, 1, 0; 0, 0, 1], [1, 1, 1]], 8]
  ? rnfdedekind(nf, P, pr3, 1)
  %3 = 1

@3In this example, pr3 is the ramified ideal above 3, and the order generated by the cube roots of y is already pr3-maximal. The order-discriminant has valuation 8. On the other hand, the order is not maximal at the prime above 2:

  ? pr2 = idealprimedec(nf,2)[1];
  ? rnfdedekind(nf, P, pr2, 1)
  %5 = 0
  ? rnfdedekind(nf, P, pr2)
  %6 = [0, [[2, 0, 0; 0, 1, 0; 0, 0, 1], [[1, 0; 0, 1], [1, 0; 0, 1],
       [1, 1/2; 0, 1/2]]], 2]

The enlarged order is not proven to be pr2-maximal yet. In fact, it is; it is in fact the maximal order:

  ? B = rnfpseudobasis(nf, P)
  %7 = [[1, 0, 0; 0, 1, 0; 0, 0, 1], [1, 1, [1, 1/2; 0, 1/2]],
       [162, 0; 0, 162], -1]
  ? idealval(nf,B[3], pr2)
  %4 = 2

It is possible to use this routine with non-monic P = sum_{i <= n} a_i X^i belongs to Z_K[X] if flag = 1; in this case, we test maximality of Dedekind's order generated by

  1, a_n alpha, a_nalpha^2 + a_{n-1}alpha,..., a_nalpha^{n-1} + a_{n-1}alpha^{n-2} +...+ a_1alpha.

The routine will fail if P is 0 on the projective line over the residue field Z_K/pr (FIXME).

The library syntax is GEN rnfdedekind(GEN nf, GEN pol, GEN pr = NULL, long flag).

rnfdet(nf,M)

Given a pseudo-matrix M over the maximal order of nf, computes its determinant.

The library syntax is GEN rnfdet(GEN nf, GEN M).

rnfdisc(nf,pol)

Given a number field nf as output by nfinit and a polynomial pol with coefficients in nf defining a relative extension L of nf, computes the relative discriminant of L. This is a two-element row vector [D,d], where D is the relative ideal discriminant and d is the relative discriminant considered as an element of nf^*/{nf^*}^2. The main variable of nf must be of lower priority than that of pol, see Label se:priority.

The library syntax is GEN rnfdiscf(GEN nf, GEN pol).

rnfeltabstorel(rnf,x)

rnf being a relative number field extension L/K as output by rnfinit and x being an element of L expressed as a polynomial modulo the absolute equation rnf.pol, computes x as an element of the relative extension L/K as a polmod with polmod coefficients.

  ? K = nfinit(y^2+1); L = rnfinit(K, x^2-y);
  ? L.pol
  %2 = x^4 + 1
  ? rnfeltabstorel(L, Mod(x, L.pol))
  %3 = Mod(x, x^2 + Mod(-y, y^2 + 1))
  ? rnfeltabstorel(L, Mod(2, L.pol))
  %4 = 2
  ? rnfeltabstorel(L, Mod(x, x^2-y))
   ***   at top-level: rnfeltabstorel(L,Mod
   ***                 ^--------------------
   *** rnfeltabstorel: inconsistent moduli in rnfeltabstorel: x^2-y != x^4+1

The library syntax is GEN rnfeltabstorel(GEN rnf, GEN x).

rnfeltdown(rnf,x)

rnf being a relative number field extension L/K as output by rnfinit and x being an element of L expressed as a polynomial or polmod with polmod coefficients, computes x as an element of K as a polmod, assuming x is in K (otherwise a domain error occurs).

  ? K = nfinit(y^2+1); L = rnfinit(K, x^2-y);
  ? L.pol
  %2 = x^4 + 1
  ? rnfeltdown(L, Mod(x^2, L.pol))
  %3 = Mod(y, y^2 + 1)
  ? rnfeltdown(L, Mod(y, x^2-y))
  %4 = Mod(y, y^2 + 1)
  ? rnfeltdown(L, Mod(y,K.pol))
  %5 = Mod(y, y^2 + 1)
  ? rnfeltdown(L, Mod(x, L.pol))
   ***   at top-level: rnfeltdown(L,Mod(x,x
   ***                 ^--------------------
   *** rnfeltdown: domain error in rnfeltdown: element not in the base field

The library syntax is GEN rnfeltdown(GEN rnf, GEN x).

rnfeltnorm(rnf,x)

rnf being a relative number field extension L/K as output by rnfinit and x being an element of L, returns the relative norm N_{L/K}(x) as an element of K.

  ? K = nfinit(y^2+1); L = rnfinit(K, x^2-y);
  ? rnfeltnorm(L, Mod(x, L.pol))
  %2 = Mod(x, x^2 + Mod(-y, y^2 + 1))
  ? rnfeltnorm(L, 2)
  %3 = 4
  ? rnfeltnorm(L, Mod(x, x^2-y))

The library syntax is GEN rnfeltnorm(GEN rnf, GEN x).

rnfeltreltoabs(rnf,x)

rnf being a relative number field extension L/K as output by rnfinit and x being an element of L expressed as a polynomial or polmod with polmod coefficients, computes x as an element of the absolute extension L/Q as a polynomial modulo the absolute equation rnf.pol.

  ? K = nfinit(y^2+1); L = rnfinit(K, x^2-y);
  ? L.pol
  %2 = x^4 + 1
  ? rnfeltreltoabs(L, Mod(x, L.pol))
  %3 = Mod(x, x^4 + 1)
  ? rnfeltreltoabs(L, Mod(y, x^2-y))
  %4 = Mod(x^2, x^4 + 1)
  ? rnfeltreltoabs(L, Mod(y,K.pol))
  %5 = Mod(x^2, x^4 + 1)

The library syntax is GEN rnfeltreltoabs(GEN rnf, GEN x).

rnfelttrace(rnf,x)

rnf being a relative number field extension L/K as output by rnfinit and x being an element of L, returns the relative trace N_{L/K}(x) as an element of K.

  ? K = nfinit(y^2+1); L = rnfinit(K, x^2-y);
  ? rnfelttrace(L, Mod(x, L.pol))
  %2 = 0
  ? rnfelttrace(L, 2)
  %3 = 4
  ? rnfelttrace(L, Mod(x, x^2-y))

The library syntax is GEN rnfelttrace(GEN rnf, GEN x).

rnfeltup(rnf,x)

rnf being a relative number field extension L/K as output by rnfinit and x being an element of K, computes x as an element of the absolute extension L/Q as a polynomial modulo the absolute equation rnf.pol.

  ? K = nfinit(y^2+1); L = rnfinit(K, x^2-y);
  ? L.pol
  %2 = x^4 + 1
  ? rnfeltup(L, Mod(y, K.pol))
  %4 = Mod(x^2, x^4 + 1)
  ? rnfeltup(L, y)
  %5 = Mod(x^2, x^4 + 1)
  ? rnfeltup(L, [1,2]~) \\ in terms of K.zk
  %6 = Mod(2*x^2 + 1, x^4 + 1)

The library syntax is GEN rnfeltup(GEN rnf, GEN x).

rnfequation(nf,pol,{flag = 0})

Given a number field nf as output by nfinit (or simply a polynomial) and a polynomial pol with coefficients in nf defining a relative extension L of nf, computes an absolute equation of L over Q.

The main variable of nf must be of lower priority than that of pol (see Label se:priority). Note that for efficiency, this does not check whether the relative equation is irreducible over nf, but only if it is squarefree. If it is reducible but squarefree, the result will be the absolute equation of the étale algebra defined by pol. If pol is not squarefree, raise an e_DOMAIN exception.

  ? rnfequation(y^2+1, x^2 - y)
  %1 = x^4 + 1
  ? T = y^3-2; rnfequation(nfinit(T), (x^3-2)/(x-Mod(y,T)))
  %2 = x^6 + 108  \\ Galois closure of Q(2^(1/3))

If flag is non-zero, outputs a 3-component row vector [z,a,k], where

@3* z is the absolute equation of L over Q, as in the default behavior,

@3* a expresses as a t_POLMOD modulo z a root alpha of the polynomial defining the base field nf,

@3* k is a small integer such that theta = beta+kalpha is a root of z, where beta is a root of pol.

  ? T = y^3-2; pol = x^2 +x*y + y^2;
  ? [z,a,k] = rnfequation(T, pol, 1);
  ? z
  %4 = x^6 + 108
  ? subst(T, y, a)
  %5 = 0
  ? alpha= Mod(y, T);
  ? beta = Mod(x*Mod(1,T), pol);
  ? subst(z, x, beta + k*alpha)
  %8 = 0

The library syntax is GEN rnfequation0(GEN nf, GEN pol, long flag). Also available are GEN rnfequation(GEN nf, GEN pol) (flag = 0) and GEN rnfequation2(GEN nf, GEN pol) (flag = 1).

rnfhnfbasis(bnf,x)

Given bnf as output by bnfinit, and either a polynomial x with coefficients in bnf defining a relative extension L of bnf, or a pseudo-basis x of such an extension, gives either a true bnf-basis of L in upper triangular Hermite normal form, if it exists, and returns 0 otherwise.

The library syntax is GEN rnfhnfbasis(GEN bnf, GEN x).

rnfidealabstorel(rnf,x)

Let rnf be a relative number field extension L/K as output by rnfinit and x be an ideal of the absolute extension L/Q given by a Z-basis of elements of L. Returns the relative pseudo-matrix in HNF giving the ideal x considered as an ideal of the relative extension L/K, i.e. as a Z_K-module.

The reason why the input does not use the customary HNF in terms of a fixed Z-basis for Z_L is precisely that no such basis has been explicitly specified. On the other hand, if you already computed an (absolute) nf structure Labs associated to L, and m is in HNF, defining an (absolute) ideal with respect to the Z-basis Labs.zk, then Labs.zk * m is a suitable Z-basis for the ideal, and

    rnfidealabstorel(rnf, Labs.zk * m)

@3converts m to a relative ideal.

  ? K = nfinit(y^2+1); L = rnfinit(K, x^2-y); Labs = nfinit(L.pol);
  ? m = idealhnf(Labs, 17, x^3+2);
  ? B = rnfidealabstorel(L, Labs.zk * m)
  %3 = [[1, 8; 0, 1], [[17, 4; 0, 1], 1]]  \\ pseudo-basis for m as Z_K-module
  ? A = rnfidealreltoabs(L, B)
  %4 = [17, x^2 + 4, x + 8, x^3 + 8*x^2]   \\ Z-basis for m in Q[x]/(L.pol)
  ? mathnf(matalgtobasis(Labs, A))
  %5 =
  [17 8 4 2]
  [ 0 1 0 0]
  [ 0 0 1 0]
  [ 0 0 0 1]
  ? % == m
  %6 = 1

The library syntax is GEN rnfidealabstorel(GEN rnf, GEN x).

rnfidealdown(rnf,x)

Let rnf be a relative number field extension L/K as output by rnfinit, and x an ideal of L, given either in relative form or by a Z-basis of elements of L (see Label se:rnfidealabstorel). This function returns the ideal of K below x, i.e. the intersection of x with K.

The library syntax is GEN rnfidealdown(GEN rnf, GEN x).

rnfidealhnf(rnf,x)

rnf being a relative number field extension L/K as output by rnfinit and x being a relative ideal (which can be, as in the absolute case, of many different types, including of course elements), computes the HNF pseudo-matrix associated to x, viewed as a Z_K-module.

The library syntax is GEN rnfidealhnf(GEN rnf, GEN x).

rnfidealmul(rnf,x,y)

rnf being a relative number field extension L/K as output by rnfinit and x and y being ideals of the relative extension L/K given by pseudo-matrices, outputs the ideal product, again as a relative ideal.

The library syntax is GEN rnfidealmul(GEN rnf, GEN x, GEN y).

rnfidealnormabs(rnf,x)

Let rnf be a relative number field extension L/K as output by rnfinit and let x be a relative ideal (which can be, as in the absolute case, of many different types, including of course elements). This function computes the norm of the x considered as an ideal of the absolute extension L/Q. This is identical to

     idealnorm(rnf, rnfidealnormrel(rnf,x))

@3but faster.

The library syntax is GEN rnfidealnormabs(GEN rnf, GEN x).

rnfidealnormrel(rnf,x)

Let rnf be a relative number field extension L/K as output by rnfinit and let x be a relative ideal (which can be, as in the absolute case, of many different types, including of course elements). This function computes the relative norm of x as an ideal of K in HNF.

The library syntax is GEN rnfidealnormrel(GEN rnf, GEN x).

rnfidealreltoabs(rnf,x)

Let rnf be a relative number field extension L/K as output by rnfinit and let x be a relative ideal, given as a Z_K-module by a pseudo matrix [A,I]. This function returns the ideal x as an absolute ideal of L/Q in the form of a Z-basis, given by a vector of polynomials (modulo rnf.pol).

The reason why we do not return the customary HNF in terms of a fixed Z-basis for Z_L is precisely that no such basis has been explicitly specified. On the other hand, if you already computed an (absolute) nf structure Labs associated to L, then

    xabs = rnfidealreltoabs(L, x);
    xLabs = mathnf(matalgtobasis(Labs, xabs));

@3computes a traditional HNF xLabs for x in terms of the fixed Z-basis Labs.zk.

The library syntax is GEN rnfidealreltoabs(GEN rnf, GEN x).

rnfidealtwoelt(rnf,x)

rnf being a relative number field extension L/K as output by rnfinit and x being an ideal of the relative extension L/K given by a pseudo-matrix, gives a vector of two generators of x over Z_L expressed as polmods with polmod coefficients.

The library syntax is GEN rnfidealtwoelement(GEN rnf, GEN x).

rnfidealup(rnf,x)

Let rnf be a relative number field extension L/K as output by rnfinit and let x be an ideal of K. This function returns the ideal xZ_L as an absolute ideal of L/Q, in the form of a Z-basis, given by a vector of polynomials (modulo rnf.pol).

The reason why we do not return the customary HNF in terms of a fixed Z-basis for Z_L is precisely that no such basis has been explicitly specified. On the other hand, if you already computed an (absolute) nf structure Labs associated to L, then

    xabs = rnfidealup(L, x);
    xLabs = mathnf(matalgtobasis(Labs, xabs));

@3computes a traditional HNF xLabs for x in terms of the fixed Z-basis Labs.zk.

The library syntax is GEN rnfidealup(GEN rnf, GEN x).

rnfinit(nf,pol)

nf being a number field in nfinit format considered as base field, and pol a polynomial defining a relative extension over nf, this computes data to work in the relative extension. The main variable of pol must be of higher priority (see Label se:priority) than that of nf, and the coefficients of pol must be in nf.

The result is a row vector, whose components are technical. In the following description, we let K be the base field defined by nf and L/K the large field associated to the rnf. Furthermore, we let m = [K:Q] the degree of the base field, n = [L:K] the relative degree, r_1 and r_2 the number of real and complex places of K. Access to this information via member functions is preferred since the specific data organization specified below will change in the future.

rnf[1](rnf.pol) contains the relative polynomial pol.

rnf[2] contains the integer basis [A,d] of K, as (integral) elements of L/Q. More precisely, A is a vector of polynomial with integer coefficients, d is a denominator, and the integer basis is given by A/d.

rnf[3] (rnf.disc) is a two-component row vector [d(L/K),s] where d(L/K) is the relative ideal discriminant of L/K and s is the discriminant of L/K viewed as an element of K^*/(K^*)^2, in other words it is the output of rnfdisc.

rnf[4](rnf.index) is the ideal index f, i.e. such that d(pol)Z_K = f^2d(L/K).

rnf[5] is currently unused.

rnf[6] is currently unused.

rnf[7] (rnf.zk) is the pseudo-basis (A,I) for the maximal order Z_L as a Z_K-module: A is the relative integral pseudo basis expressed as polynomials (in the variable of pol) with polmod coefficients in nf, and the second component I is the ideal list of the pseudobasis in HNF.

rnf[8] is the inverse matrix of the integral basis matrix, with coefficients polmods in nf.

rnf[9] is currently unused.

rnf[10] (rnf.nf) is nf.

rnf[11] is the output of rnfequation(K, pol, 1). Namely, a vector [P, a, k] describing the absolute extension L/Q: P is an absolute equation, more conveniently obtained as rnf.polabs; a expresses the generator alpha = y mod K.pol of the number field K as an element of L, i.e. a polynomial modulo the absolute equation P;

k is a small integer such that, if beta is an abstract root of pol and alpha the generator of K given above, then P(beta + kalpha) = 0.

@3Caveat.. Be careful if k != 0 when dealing simultaneously with absolute and relative quantities since L = Q(beta + kalpha) = K(alpha), and the generator chosen for the absolute extension is not the same as for the relative one. If this happens, one can of course go on working, but we advise to change the relative polynomial so that its root becomes beta + k alpha. Typical GP instructions would be

    [P,a,k] = rnfequation(K, pol, 1);
    if (k, pol = subst(pol, x, x - k*Mod(y, K.pol)));
    L = rnfinit(K, pol);

rnf[12] is by default unused and set equal to 0. This field is used to store further information about the field as it becomes available (which is rarely needed, hence would be too expensive to compute during the initial rnfinit call).

The library syntax is GEN rnfinit(GEN nf, GEN pol).

rnfisabelian(nf,T)

T being a relative polynomial with coefficients in nf, return 1 if it defines an abelian extension, and 0 otherwise.

  ? K = nfinit(y^2 + 23);
  ? rnfisabelian(K, x^3 - 3*x - y)
  %2 = 1

The library syntax is long rnfisabelian(GEN nf, GEN T).

rnfisfree(bnf,x)

Given bnf as output by bnfinit, and either a polynomial x with coefficients in bnf defining a relative extension L of bnf, or a pseudo-basis x of such an extension, returns true (1) if L/bnf is free, false (0) if not.

The library syntax is long rnfisfree(GEN bnf, GEN x).

rnfisnorm(T,a,{flag = 0})

Similar to bnfisnorm but in the relative case. T is as output by rnfisnorminit applied to the extension L/K. This tries to decide whether the element a in K is the norm of some x in the extension L/K.

The output is a vector [x,q], where a = Norm (x)*q. The algorithm looks for a solution x which is an S-integer, with S a list of places of K containing at least the ramified primes, the generators of the class group of L, as well as those primes dividing a. If L/K is Galois, then this is enough; otherwise, flag is used to add more primes to S: all the places above the primes p <= flag (resp. p|flag) if flag > 0 (resp. flag < 0).

The answer is guaranteed (i.e. a is a norm iff q = 1) if the field is Galois, or, under GRH, if S contains all primes less than 12 log ^2| disc (M)|, where M is the normal closure of L/K.

If rnfisnorminit has determined (or was told) that L/K is Galois, and flag != 0, a Warning is issued (so that you can set flag = 1 to check whether L/K is known to be Galois, according to T). Example:

  bnf = bnfinit(y^3 + y^2 - 2*y - 1);
  p = x^2 + Mod(y^2 + 2*y + 1, bnf.pol);
  T = rnfisnorminit(bnf, p);
  rnfisnorm(T, 17)

checks whether 17 is a norm in the Galois extension Q(beta) / Q(alpha), where alpha^3 + alpha^2 - 2alpha - 1 = 0 and beta^2 + alpha^2 + 2alpha + 1 = 0 (it is).

The library syntax is GEN rnfisnorm(GEN T, GEN a, long flag).

rnfisnorminit(pol,polrel,{flag = 2})

Let K be defined by a root of pol, and L/K the extension defined by the polynomial polrel. As usual, pol can in fact be an nf, or bnf, etc; if pol has degree 1 (the base field is Q), polrel is also allowed to be an nf, etc. Computes technical data needed by rnfisnorm to solve norm equations Nx = a, for x in L, and a in K.

If flag = 0, do not care whether L/K is Galois or not.

If flag = 1, L/K is assumed to be Galois (unchecked), which speeds up rnfisnorm.

If flag = 2, let the routine determine whether L/K is Galois.

The library syntax is GEN rnfisnorminit(GEN pol, GEN polrel, long flag).

rnfkummer(bnr,{subgp},{d = 0})

bnr being as output by bnrinit, finds a relative equation for the class field corresponding to the module in bnr and the given congruence subgroup (the full ray class field if subgp is omitted). If d is positive, outputs the list of all relative equations of degree d contained in the ray class field defined by bnr, with the same conductor as (bnr, subgp).

@3Warning. This routine only works for subgroups of prime index. It uses Kummer theory, adjoining necessary roots of unity (it needs to compute a tough bnfinit here), and finds a generator via Hecke's characterization of ramification in Kummer extensions of prime degree. If your extension does not have prime degree, for the time being, you have to split it by hand as a tower / compositum of such extensions.

The library syntax is GEN rnfkummer(GEN bnr, GEN subgp = NULL, long d, long prec).

rnflllgram(nf,pol,order)

Given a polynomial pol with coefficients in nf defining a relative extension L and a suborder order of L (of maximal rank), as output by rnfpseudobasis(nf,pol) or similar, gives [[neworder],U], where neworder is a reduced order and U is the unimodular transformation matrix.

The library syntax is GEN rnflllgram(GEN nf, GEN pol, GEN order, long prec).

rnfnormgroup(bnr,pol)

bnr being a big ray class field as output by bnrinit and pol a relative polynomial defining an Abelian extension, computes the norm group (alias Artin or Takagi group) corresponding to the Abelian extension of bnf = bnr.bnf defined by pol, where the module corresponding to bnr is assumed to be a multiple of the conductor (i.e. pol defines a subextension of bnr). The result is the HNF defining the norm group on the given generators of bnr.gen. Note that neither the fact that pol defines an Abelian extension nor the fact that the module is a multiple of the conductor is checked. The result is undefined if the assumption is not correct.

The library syntax is GEN rnfnormgroup(GEN bnr, GEN pol).

rnfpolred(nf,pol)

THIS FUNCTION IS OBSOLETE: use rnfpolredbest instead. Relative version of polred. Given a monic polynomial pol with coefficients in nf, finds a list of relative polynomials defining some subfields, hopefully simpler and containing the original field. In the present version 2.7.4, this is slower and less efficient than rnfpolredbest.

@3Remark. this function is based on an incomplete reduction theory of lattices over number fields, implemented by rnflllgram, which deserves to be improved.

The library syntax is GEN rnfpolred(GEN nf, GEN pol, long prec).

rnfpolredabs(nf,pol,{flag = 0})

THIS FUNCTION IS OBSOLETE: use rnfpolredbest instead. Relative version of polredabs. Given a monic polynomial pol with coefficients in nf, finds a simpler relative polynomial defining the same field. The binary digits of flag mean

The binary digits of flag correspond to 1: add information to convert elements to the new representation, 2: absolute polynomial, instead of relative, 16: possibly use a suborder of the maximal order. More precisely:

0: default, return P

1: returns [P,a] where P is the default output and a, a t_POLMOD modulo P, is a root of pol.

2: returns Pabs, an absolute, instead of a relative, polynomial. Same as but faster than

    rnfequation(nf, rnfpolredabs(nf,pol))

3: returns [Pabs,a,b], where Pabs is an absolute polynomial as above, a, b are t_POLMOD modulo Pabs, roots of nf.pol and pol respectively.

16: possibly use a suborder of the maximal order. This is slower than the default when the relative discriminant is smooth, and much faster otherwise. See Label se:polredabs.

@3Warning. In the present implementation, rnfpolredabs produces smaller polynomials than rnfpolred and is usually faster, but its complexity is still exponential in the absolute degree. The function rnfpolredbest runs in polynomial time, and tends to return polynomials with smaller discriminants.

The library syntax is GEN rnfpolredabs(GEN nf, GEN pol, long flag).

rnfpolredbest(nf,pol,{flag = 0})

Relative version of polredbest. Given a monic polynomial pol with coefficients in nf, finds a simpler relative polynomial P defining the same field. As opposed to rnfpolredabs this function does not return a smallest (canonical) polynomial with respect to some measure, but it does run in polynomial time.

The binary digits of flag correspond to 1: add information to convert elements to the new representation, 2: absolute polynomial, instead of relative. More precisely:

0: default, return P

1: returns [P,a] where P is the default output and a, a t_POLMOD modulo P, is a root of pol.

2: returns Pabs, an absolute, instead of a relative, polynomial. Same as but faster than

    rnfequation(nf, rnfpolredbest(nf,pol))

3: returns [Pabs,a,b], where Pabs is an absolute polynomial as above, a, b are t_POLMOD modulo Pabs, roots of nf.pol and pol respectively.

  ? K = nfinit(y^3-2); pol = x^2 +x*y + y^2;
  ? [P, a] = rnfpolredbest(K,pol,1);
  ? P
  %3 = x^2 - x + Mod(y - 1, y^3 - 2)
  ? a
  %4 = Mod(Mod(2*y^2+3*y+4,y^3-2)*x + Mod(-y^2-2*y-2,y^3-2),
           x^2 - x + Mod(y-1,y^3-2))
  ? subst(K.pol,y,a)
  %5 = 0
  ? [Pabs, a, b] = rnfpolredbest(K,pol,3);
  ? Pabs
  %7 = x^6 - 3*x^5 + 5*x^3 - 3*x + 1
  ? a
  %8 = Mod(-x^2+x+1, x^6-3*x^5+5*x^3-3*x+1)
  ? b
  %9 = Mod(2*x^5-5*x^4-3*x^3+10*x^2+5*x-5, x^6-3*x^5+5*x^3-3*x+1)
  ? subst(K.pol,y,a)
  %10 = 0
  ? substvec(pol,[x,y],[a,b])
  %11 = 0

The library syntax is GEN rnfpolredbest(GEN nf, GEN pol, long flag).

rnfpseudobasis(nf,pol)

Given a number field nf as output by nfinit and a polynomial pol with coefficients in nf defining a relative extension L of nf, computes a pseudo-basis (A,I) for the maximal order Z_L viewed as a Z_K-module, and the relative discriminant of L. This is output as a four-element row vector [A,I,D,d], where D is the relative ideal discriminant and d is the relative discriminant considered as an element of nf^*/{nf^*}^2.

The library syntax is GEN rnfpseudobasis(GEN nf, GEN pol).

rnfsteinitz(nf,x)

Given a number field nf as output by nfinit and either a polynomial x with coefficients in nf defining a relative extension L of nf, or a pseudo-basis x of such an extension as output for example by rnfpseudobasis, computes another pseudo-basis (A,I) (not in HNF in general) such that all the ideals of I except perhaps the last one are equal to the ring of integers of nf, and outputs the four-component row vector [A,I,D,d] as in rnfpseudobasis. The name of this function comes from the fact that the ideal class of the last ideal of I, which is well defined, is the Steinitz class of the Z_K-module Z_L (its image in SK_0(Z_K)).

The library syntax is GEN rnfsteinitz(GEN nf, GEN x).

subgrouplist(bnr,{bound},{flag = 0})

bnr being as output by bnrinit or a list of cyclic components of a finite Abelian group G, outputs the list of subgroups of G. Subgroups are given as HNF left divisors of the SNF matrix corresponding to G.

If flag = 0 (default) and bnr is as output by bnrinit, gives only the subgroups whose modulus is the conductor. Otherwise, the modulus is not taken into account.

If bound is present, and is a positive integer, restrict the output to subgroups of index less than bound. If bound is a vector containing a single positive integer B, then only subgroups of index exactly equal to B are computed. For instance

  ? subgrouplist([6,2])
  %1 = [[6, 0; 0, 2], [2, 0; 0, 2], [6, 3; 0, 1], [2, 1; 0, 1], [3, 0; 0, 2],
  [1, 0; 0, 2], [6, 0; 0, 1], [2, 0; 0, 1], [3, 0; 0, 1], [1, 0; 0, 1]]
  ? subgrouplist([6,2],3)    \\ index less than 3
  %2 = [[2, 1; 0, 1], [1, 0; 0, 2], [2, 0; 0, 1], [3, 0; 0, 1], [1, 0; 0, 1]]
  ? subgrouplist([6,2],[3])  \\ index 3
  %3 = [[3, 0; 0, 1]]
  ? bnr = bnrinit(bnfinit(x), [120,[1]], 1);
  ? L = subgrouplist(bnr, [8]);

In the last example, L corresponds to the 24 subfields of Q(zeta_{120}), of degree 8 and conductor 120 oo (by setting flag, we see there are a total of 43 subgroups of degree 8).

  ? vector(#L, i, galoissubcyclo(bnr, L[i]))

will produce their equations. (For a general base field, you would have to rely on bnrstark, or rnfkummer.)

The library syntax is GEN subgrouplist0(GEN bnr, GEN bound = NULL, long flag).

zetak(nfz,x,{flag = 0})

znf being a number field initialized by zetakinit (not by nfinit), computes the value of the Dedekind zeta function of the number field at the complex number x. If flag = 1 computes Dedekind Lambda function instead (i.e. the product of the Dedekind zeta function by its gamma and exponential factors).

@3CAVEAT. This implementation is not satisfactory and must be rewritten. In particular

@3* The accuracy of the result depends in an essential way on the accuracy of both the zetakinit program and the current accuracy. Be wary in particular that x of large imaginary part or, on the contrary, very close to an ordinary integer will suffer from precision loss, yielding fewer significant digits than expected. Computing with 28 digits of relative accuracy, we have

  ? zeta(3)
  %1 = 1.202056903159594285399738161
  ? zeta(3-1e-20)
  %2 = 1.202056903159594285401719424
  ? zetak(zetakinit(x), 3-1e-20)
  %3 = 1.2020569031595952919  \\ 5 digits are wrong
  ? zetak(zetakinit(x), 3-1e-28)
  %4 = -25.33411749           \\ junk

@3* As the precision increases, results become unexpectedly completely wrong:

  ? \p100
  ? zetak(zetakinit(x^2-5), -1) - 1/30
  %1 = 7.26691813 E-108    \\ perfect
  ? \p150
  ? zetak(zetakinit(x^2-5), -1) - 1/30
  %2 = -2.486113578 E-156  \\ perfect
  ? \p200
  ? zetak(zetakinit(x^2-5), -1) - 1/30
  %3 = 4.47... E-75        \\ more than half of the digits are wrong
  ? \p250
  ? zetak(zetakinit(x^2-5), -1) - 1/30
  %4 = 1.6 E43             \\ junk

The library syntax is GEN gzetakall(GEN nfz, GEN x, long flag, long prec). See also GEN glambdak(GEN znf, GEN x, long prec) or GEN gzetak(GEN znf, GEN x, long prec).

zetakinit(bnf)

Computes a number of initialization data concerning the number field associated to bnf so as to be able to compute the Dedekind zeta and lambda functions, respectively zetak(x) and zetak(x,1), at the current real precision. If you do not need the bnfinit data somewhere else, you may call it with an irreducible polynomial instead of a bnf: it will call bnfinit itself.

The result is a 9-component vector v whose components are very technical and cannot really be used except through the zetak function.

This function is very inefficient and should be rewritten. It needs to computes millions of coefficients of the corresponding Dirichlet series if the precision is big. Unless the discriminant is small it will not be able to handle more than 9 digits of relative precision. For instance, zetakinit(x^8 - 2) needs 440MB of memory at default precision.

This function will fail with the message

   *** bnrL1: overflow in zeta_get_N0 [need too many primes].

@3if the approximate functional equation requires us to sum too many terms (if the discriminant of the number field is too large).

The library syntax is GEN initzeta(GEN bnf, long prec).


Polynomials and power series

We group here all functions which are specific to polynomials or power series. Many other functions which can be applied on these objects are described in the other sections. Also, some of the functions described here can be applied to other types.

O(p^e)

If p is an integer greater than 2, returns a p-adic 0 of precision e. In all other cases, returns a power series zero with precision given by e v, where v is the X-adic valuation of p with respect to its main variable.

The library syntax is GEN ggrando(). GEN zeropadic(GEN p, long e) for a p-adic and GEN zeroser(long v, long e) for a power series zero in variable v.

bezoutres(A,B,{v})

Deprecated alias for polresultantext

The library syntax is GEN polresultantext0(GEN A, GEN B, long v = -1), where v is a variable number.

deriv(x,{v})

Derivative of x with respect to the main variable if v is omitted, and with respect to v otherwise. The derivative of a scalar type is zero, and the derivative of a vector or matrix is done componentwise. One can use x' as a shortcut if the derivative is with respect to the main variable of x.

By definition, the main variable of a t_POLMOD is the main variable among the coefficients from its two polynomial components (representative and modulus); in other words, assuming a polmod represents an element of R[X]/(T(X)), the variable X is a mute variable and the derivative is taken with respect to the main variable used in the base ring R.

The library syntax is GEN deriv(GEN x, long v = -1), where v is a variable number.

diffop(x,v,d,{n = 1})

Let v be a vector of variables, and d a vector of the same length, return the image of x by the n-power (1 if n is not given) of the differential operator D that assumes the value d[i] on the variable v[i]. The value of D on a scalar type is zero, and D applies componentwise to a vector or matrix. When applied to a t_POLMOD, if no value is provided for the variable of the modulus, such value is derived using the implicit function theorem.

Some examples: This function can be used to differentiate formal expressions: If E = exp (X^2) then we have E' = 2*X*E. We can derivate X*exp(X^2) as follow:

  ? diffop(E*X,[X,E],[1,2*X*E])
  %1 = (2*X^2 + 1)*E

Let Sin and Cos be two function such that Sin^2+Cos^2 = 1 and Cos' = -Sin. We can differentiate Sin/Cos as follow, PARI inferring the value of Sin' from the equation:

  ? diffop(Mod('Sin/'Cos,'Sin^2+'Cos^2-1),['Cos],[-'Sin])
  %1 = Mod(1/Cos^2, Sin^2 + (Cos^2 - 1))

Compute the Bell polynomials (both complete and partial) via the Faa di Bruno formula:

  Bell(k,n=-1)=
  {
    my(var(i)=eval(Str("X",i)));
    my(x,v,dv);
    v=vector(k,i,if(i==1,'E,var(i-1)));
    dv=vector(k,i,if(i==1,'X*var(1)*'E,var(i)));
    x=diffop('E,v,dv,k)/'E;
    if(n<0,subst(x,'X,1),polcoeff(x,n,'X))
  }

The library syntax is GEN diffop0(GEN x, GEN v, GEN d, long n).

For n = 1, the function GEN diffop(GEN x, GEN v, GEN d) is also available.

eval(x)

Replaces in x the formal variables by the values that have been assigned to them after the creation of x. This is mainly useful in GP, and not in library mode. Do not confuse this with substitution (see subst).

If x is a character string, eval(x) executes x as a GP command, as if directly input from the keyboard, and returns its output.

  ? x1 = "one"; x2 = "two";
  ? n = 1; eval(Str("x", n))
  %2 = "one"
  ? f = "exp"; v = 1;
  ? eval(Str(f, "(", v, ")"))
  %4 = 2.7182818284590452353602874713526624978

@3Note that the first construct could be implemented in a simpler way by using a vector x = ["one","two"]; x[n], and the second by using a closure f = exp; f(v). The final example is more interesting:

  ? genmat(u,v) = matrix(u,v,i,j, eval( Str("x",i,j) ));
  ? genmat(2,3)   \\ generic 2 x 3 matrix
  %2 =
  [x11 x12 x13]
  [x21 x22 x23]

A syntax error in the evaluation expression raises an e_SYNTAX exception, which can be trapped as usual:

  ? 1a
   ***   unused characters: 1a
   ***                       ^-
  ? E(expr) =
    {
      iferr(eval(expr),
            e, print("syntax error"),
            errname(e) == "e_SYNTAX");
    }
  ? E("1+1")
  %1 = 2
  ? E("1a")
  syntax error

The library syntax is geval(GEN x).

factorpadic(pol,p,r)

p-adic factorization of the polynomial pol to precision r, the result being a two-column matrix as in factor. Note that this is not the same as a factorization over Z/p^rZ (polynomials over that ring do not form a unique factorization domain, anyway), but approximations in Q/p^rZ of the true factorization in Q_p[X].

  ? factorpadic(x^2 + 9, 3,5)
  %1 =
  [(1 + O(3^5))*x^2 + O(3^5)*x + (3^2 + O(3^5)) 1]
  ? factorpadic(x^2 + 1, 5,3)
  %2 =
  [  (1 + O(5^3))*x + (2 + 5 + 2*5^2 + O(5^3)) 1]
  [(1 + O(5^3))*x + (3 + 3*5 + 2*5^2 + O(5^3)) 1]

@3 The factors are normalized so that their leading coefficient is a power of p. The method used is a modified version of the round 4 algorithm of Zassenhaus.

If pol has inexact t_PADIC coefficients, this is not always well-defined; in this case, the polynomial is first made integral by dividing out the p-adic content, then lifted to Z using truncate coefficientwise. Hence we actually factor exactly a polynomial which is only p-adically close to the input. To avoid pitfalls, we advise to only factor polynomials with exact rational coefficients.

The library syntax is factorpadic(GEN f,GEN p, long r) . The function factorpadic0 is deprecated, provided for backward compatibility.

intformal(x,{v})

formal integration of x with respect to the variable v (wrt. the main variable if v is omitted). Since PARI cannot represent logarithmic or arctangent terms, any such term in the result will yield an error:

   ? intformal(x^2)
   %1 = 1/3*x^3
   ? intformal(x^2, y)
   %2 = y*x^2
   ? intformal(1/x)
     ***   at top-level: intformal(1/x)
     ***                 ^--------------
     *** intformal: domain error in intformal: residue(series, pole) != 0

The argument x can be of any type. When x is a rational function, we assume that the base ring is an integral domain of characteristic zero.

By definition, the main variable of a t_POLMOD is the main variable among the coefficients from its two polynomial components (representative and modulus); in other words, assuming a polmod represents an element of R[X]/(T(X)), the variable X is a mute variable and the integral is taken with respect to the main variable used in the base ring R. In particular, it is meaningless to integrate with respect to the main variable of x.mod:

  ? intformal(Mod(1,x^2+1), 'x)
  *** intformal: incorrect priority in intformal: variable x = x

The library syntax is GEN integ(GEN x, long v = -1), where v is a variable number.

padicappr(pol,a)

Vector of p-adic roots of the polynomial pol congruent to the p-adic number a modulo p, and with the same p-adic precision as a. The number a can be an ordinary p-adic number (type t_PADIC, i.e. an element of Z_p) or can be an integral element of a finite extension of Q_p, given as a t_POLMOD at least one of whose coefficients is a t_PADIC. In this case, the result is the vector of roots belonging to the same extension of Q_p as a.

The library syntax is GEN padicappr(GEN pol, GEN a). Also available is GEN Zp_appr(GEN f, GEN a) when a is a t_PADIC.

padicfields(p, N, {flag = 0})

Returns a vector of polynomials generating all the extensions of degree N of the field Q_p of p-adic rational numbers; N is allowed to be a 2-component vector [n,d], in which case we return the extensions of degree n and discriminant p^d.

The list is minimal in the sense that two different polynomials generate non-isomorphic extensions; in particular, the number of polynomials is the number of classes of non-isomorphic extensions. If P is a polynomial in this list, alpha is any root of P and K = Q_p(alpha), then alpha is the sum of a uniformizer and a (lift of a) generator of the residue field of K; in particular, the powers of alpha generate the ring of p-adic integers of K.

If flag = 1, replace each polynomial P by a vector [P, e, f, d, c] where e is the ramification index, f the residual degree, d the valuation of the discriminant, and c the number of conjugate fields. If flag = 2, only return the number of extensions in a fixed algebraic closure (Krasner's formula), which is much faster.

The library syntax is GEN padicfields0(GEN p, GEN N, long flag). Also available is GEN padicfields(GEN p, long n, long d, long flag), which computes extensions of Q_p of degree n and discriminant p^d.

polchebyshev(n,{flag = 1},{a = 'x})

Returns the n-th Chebyshev polynomial of the first kind T_n (flag = 1) or the second kind U_n (flag = 2), evaluated at a ('x by default). Both series of polynomials satisfy the 3-term relation

   P_{n+1} = 2xP_n - P_{n-1},

and are determined by the initial conditions U_0 = T_0 = 1, T_1 = x, U_1 = 2x. In fact T_n' = n U_{n-1} and, for all complex numbers z, we have T_n( cos z) = cos (nz) and U_{n-1}( cos z) = sin (nz)/ sin z. If n >= 0, then these polynomials have degree n. For n < 0, T_n is equal to T_{-n} and U_n is equal to -U_{-2-n}. In particular, U_{-1} = 0.

The library syntax is GEN polchebyshev_eval(long n, long flag, GEN a = NULL). Also available are GEN polchebyshev(long n, long flag, long v), GEN polchebyshev1(long n, long v) and GEN polchebyshev2(long n, long v) for T_n and U_n respectively.

polcoeff(x,n,{v})

Coefficient of degree n of the polynomial x, with respect to the main variable if v is omitted, with respect to v otherwise. If n is greater than the degree, the result is zero.

Naturally applies to scalars (polynomial of degree 0), as well as to rational functions whose denominator is a monomial. It also applies to power series: if n is less than the valuation, the result is zero. If it is greater than the largest significant degree, then an error message is issued.

For greater flexibility, x can be a vector or matrix type and the function then returns component(x,n).

The library syntax is GEN polcoeff0(GEN x, long n, long v = -1), where v is a variable number.

polcyclo(n,{a = 'x})

n-th cyclotomic polynomial, evaluated at a ('x by default). The integer n must be positive.

Algorithm used: reduce to the case where n is squarefree; to compute the cyclotomic polynomial, use Phi_{np}(x) = Phi_n(x^p)/Phi(x); to compute it evaluated, use Phi_n(x) = prod_{d | n} (x^d-1)^{mu(n/d)}. In the evaluated case, the algorithm assumes that a^d - 1 is either 0 or invertible, for all d | n. If this is not the case (the base ring has zero divisors), use subst(polcyclo(n),x,a).

The library syntax is GEN polcyclo_eval(long n, GEN a = NULL). The variant GEN polcyclo(long n, long v) returns the n-th cyclotomic polynomial in variable v.

polcyclofactors(f)

Returns a vector of polynomials, whose product is the product of distinct cyclotomic polynomials dividing f.

  ? f = x^10+5*x^8-x^7+8*x^6-4*x^5+8*x^4-3*x^3+7*x^2+3;
  ? v = polcyclofactors(f)
  %2 = [x^2 + 1, x^2 + x + 1, x^4 - x^3 + x^2 - x + 1]
  ? apply(poliscycloprod, v)
  %3 = [1, 1, 1]
  ? apply(poliscyclo, v)
  %4 = [4, 3, 10]

@3In general, the polynomials are products of cyclotomic polynomials and not themselves irreducible:

  ? g = x^8+2*x^7+6*x^6+9*x^5+12*x^4+11*x^3+10*x^2+6*x+3;
  ? polcyclofactors(g)
  %2 = [x^6 + 2*x^5 + 3*x^4 + 3*x^3 + 3*x^2 + 2*x + 1]
  ? factor(%[1])
  %3 =
  [            x^2 + x + 1 1]
  [x^4 + x^3 + x^2 + x + 1 1]

The library syntax is GEN polcyclofactors(GEN f).

poldegree(x,{v})

Degree of the polynomial x in the main variable if v is omitted, in the variable v otherwise.

The degree of 0 is a fixed negative number, whose exact value should not be used. The degree of a non-zero scalar is 0. Finally, when x is a non-zero polynomial or rational function, returns the ordinary degree of x. Raise an error otherwise.

The library syntax is long poldegree(GEN x, long v = -1), where v is a variable number.

poldisc(pol,{v})

Discriminant of the polynomial pol in the main variable if v is omitted, in v otherwise. The algorithm used is the subresultant algorithm.

The library syntax is GEN poldisc0(GEN pol, long v = -1), where v is a variable number.

poldiscreduced(f)

Reduced discriminant vector of the (integral, monic) polynomial f. This is the vector of elementary divisors of Z[alpha]/f'(alpha)Z[alpha], where alpha is a root of the polynomial f. The components of the result are all positive, and their product is equal to the absolute value of the discriminant of f.

The library syntax is GEN reduceddiscsmith(GEN f).

polgraeffe(f)

Returns the Graeffe transform g of f, such that g(x^2) = f(x) f(-x).

The library syntax is GEN polgraeffe(GEN f).

polhensellift(A, B, p, e)

Given a prime p, an integral polynomial A whose leading coefficient is a p-unit, a vector B of integral polynomials that are monic and pairwise relatively prime modulo p, and whose product is congruent to A/{lc}(A) modulo p, lift the elements of B to polynomials whose product is congruent to A modulo p^e.

More generally, if T is an integral polynomial irreducible mod p, and B is a factorization of A over the finite field F_p[t]/(T), you can lift it to Z_p[t]/(T, p^e) by replacing the p argument with [p,T]:

  ? { T = t^3 - 2; p = 7; A = x^2 + t + 1;
      B = [x + (3*t^2 + t + 1), x + (4*t^2 + 6*t + 6)];
      r = polhensellift(A, B, [p, T], 6) }
  %1 = [x + (20191*t^2 + 50604*t + 75783), x + (97458*t^2 + 67045*t + 41866)]
  ? liftall( r[1] * r[2] * Mod(Mod(1,p^6),T) )
  %2 = x^2 + (t + 1)

The library syntax is GEN polhensellift(GEN A, GEN B, GEN p, long e).

polhermite(n,{a = 'x})

n-th Hermite polynomial H_n evaluated at a ('x by default), i.e.

   H_n(x) = (-1)^n e^{x^2} (d^n)/(dx^n)e^{-x^2}.

The library syntax is GEN polhermite_eval(long n, GEN a = NULL). The variant GEN polhermite(long n, long v) returns the n-th Hermite polynomial in variable v.

polinterpolate(X,{Y},{x},{&e})

Given the data vectors X and Y of the same length n (X containing the x-coordinates, and Y the corresponding y-coordinates), this function finds the interpolating polynomial passing through these points and evaluates it at x. If Y is omitted, return the polynomial interpolating the (i,X[i]). If present, e will contain an error estimate on the returned value.

The library syntax is GEN polint(GEN X, GEN Y = NULL, GEN x = NULL, GEN *e = NULL).

poliscyclo(f)

Returns 0 if f is not a cyclotomic polynomial, and n > 0 if f = Phi_n, the n-th cyclotomic polynomial.

  ? poliscyclo(x^4-x^2+1)
  %1 = 12
  ? polcyclo(12)
  %2 = x^4 - x^2 + 1
  ? poliscyclo(x^4-x^2-1)
  %3 = 0

The library syntax is long poliscyclo(GEN f).

poliscycloprod(f)

Returns 1 if f is a product of cyclotomic polynomial, and 0 otherwise.

  ? f = x^6+x^5-x^3+x+1;
  ? poliscycloprod(f)
  %2 = 1
  ? factor(f)
  %3 =
  [  x^2 + x + 1 1]
  [x^4 - x^2 + 1 1]
  ? [ poliscyclo(T) | T <- %[,1] ]
  %4 = [3, 12]
  ? polcyclo(3) * polcyclo(12)
  %5 = x^6 + x^5 - x^3 + x + 1

The library syntax is long poliscycloprod(GEN f).

polisirreducible(pol)

pol being a polynomial (univariate in the present version 2.7.4), returns 1 if pol is non-constant and irreducible, 0 otherwise. Irreducibility is checked over the smallest base field over which pol seems to be defined.

The library syntax is long isirreducible(GEN pol).

pollead(x,{v})

Leading coefficient of the polynomial or power series x. This is computed with respect to the main variable of x if v is omitted, with respect to the variable v otherwise.

The library syntax is GEN pollead(GEN x, long v = -1), where v is a variable number.

pollegendre(n,{a = 'x})

n-th Legendre polynomial evaluated at a ('x by default).

The library syntax is GEN pollegendre_eval(long n, GEN a = NULL). To obtain the n-th Legendre polynomial in variable v, use GEN pollegendre(long n, long v).

polrecip(pol)

Reciprocal polynomial of pol, i.e. the coefficients are in reverse order. pol must be a polynomial.

The library syntax is GEN polrecip(GEN pol).

polresultant(x,y,{v},{flag = 0})

Resultant of the two polynomials x and y with exact entries, with respect to the main variables of x and y if v is omitted, with respect to the variable v otherwise. The algorithm assumes the base ring is a domain. If you also need the u and v such that x*u + y*v = {Res}(x,y), use the polresultantext function.

If flag = 0 (default), uses the the algorithm best suited to the inputs, either the subresultant algorithm (Lazard/Ducos variant, generic case), a modular algorithm (inputs in Q[X]) or Sylvester's matrix (inexact inputs).

If flag = 1, uses the determinant of Sylvester's matrix instead; this should always be slower than the default.

The library syntax is GEN polresultant0(GEN x, GEN y, long v = -1, long flag), where v is a variable number.

polresultantext(A,B,{v})

Finds polynomials U and V such that A*U + B*V = R, where R is the resultant of U and V with respect to the main variables of A and B if v is omitted, and with respect to v otherwise. Returns the row vector [U,V,R]. The algorithm used (subresultant) assumes that the base ring is a domain.

  ? A = x*y; B = (x+y)^2;
  ? [U,V,R] = polresultantext(A, B)
  %2 = [-y*x - 2*y^2, y^2, y^4]
  ? A*U + B*V
  %3 = y^4
  ? [U,V,R] = polresultantext(A, B, y)
  %4 = [-2*x^2 - y*x, x^2, x^4]
  ? A*U+B*V
  %5 = x^4

The library syntax is GEN polresultantext0(GEN A, GEN B, long v = -1), where v is a variable number. Also available is GEN polresultantext(GEN x, GEN y).

polroots(x)

Complex roots of the polynomial x, given as a column vector where each root is repeated according to its multiplicity. The precision is given as for transcendental functions: in GP it is kept in the variable realprecision and is transparent to the user, but it must be explicitly given as a second argument in library mode.

The algorithm used is a modification of A. Schönhage's root-finding algorithm, due to and originally implemented by X. Gourdon. Barring bugs, it is guaranteed to converge and to give the roots to the required accuracy.

The library syntax is GEN roots(GEN x, long prec).

polrootsmod(pol,p,{flag = 0})

Row vector of roots modulo p of the polynomial pol. Multiple roots are not repeated.

  ? polrootsmod(x^2-1,2)
  %1 = [Mod(1, 2)]~

If p is very small, you may set flag = 1, which uses a naive search.

The library syntax is GEN rootmod0(GEN pol, GEN p, long flag).

polrootspadic(x,p,r)

Vector of p-adic roots of the polynomial pol, given to p-adic precision r p is assumed to be a prime. Multiple roots are not repeated. Note that this is not the same as the roots in Z/p^rZ, rather it gives approximations in Z/p^rZ of the true roots living in Q_p.

  ? polrootspadic(x^3 - x^2 + 64, 2, 5)
  %1 = [2^3 + O(2^5), 2^3 + 2^4 + O(2^5), 1 + O(2^5)]~

If pol has inexact t_PADIC coefficients, this is not always well-defined; in this case, the polynomial is first made integral by dividing out the p-adic content, then lifted to Z using truncate coefficientwise. Hence the roots given are approximations of the roots of an exact polynomial which is p-adically close to the input. To avoid pitfalls, we advise to only factor polynomials with eact rational coefficients.

The library syntax is GEN rootpadic(GEN x, GEN p, long r).

polsturm(pol,{a},{b})

Number of real roots of the real squarefree polynomial pol in the interval ]a,b], using Sturm's algorithm. a (resp. b) is taken to be - oo (resp. + oo ) if omitted.

The library syntax is long sturmpart(GEN pol, GEN a = NULL, GEN b = NULL). Also available is long sturm(GEN pol) (total number of real roots).

polsubcyclo(n,d,{v = 'x})

Gives polynomials (in variable v) defining the sub-Abelian extensions of degree d of the cyclotomic field Q(zeta_n), where d | phi(n).

If there is exactly one such extension the output is a polynomial, else it is a vector of polynomials, possibly empty. To get a vector in all cases, use concat([], polsubcyclo(n,d)).

The function galoissubcyclo allows to specify exactly which sub-Abelian extension should be computed.

The library syntax is GEN polsubcyclo(long n, long d, long v = -1), where v is a variable number.

polsylvestermatrix(x,y)

Forms the Sylvester matrix corresponding to the two polynomials x and y, where the coefficients of the polynomials are put in the columns of the matrix (which is the natural direction for solving equations afterwards). The use of this matrix can be essential when dealing with polynomials with inexact entries, since polynomial Euclidean division doesn't make much sense in this case.

The library syntax is GEN sylvestermatrix(GEN x, GEN y).

polsym(x,n)

Creates the column vector of the symmetric powers of the roots of the polynomial x up to power n, using Newton's formula.

The library syntax is GEN polsym(GEN x, long n).

poltchebi(n,{v = 'x})

Deprecated alias for polchebyshev

The library syntax is GEN polchebyshev1(long n, long v = -1), where v is a variable number.

polzagier(n,m)

Creates Zagier's polynomial P_n^{(m)} used in the functions sumalt and sumpos (with flag = 1). One must have m <= n. The exact definition can be found in ``Convergence acceleration of alternating series'', Cohen et al., Experiment. Math., vol. 9, 2000, pp. 3--12.

The library syntax is GEN polzag(long n, long m).

serconvol(x,y)

Convolution (or Hadamard product) of the two power series x and y; in other words if x = sum a_k*X^k and y = sum b_k*X^k then serconvol(x,y) = sum a_k*b_k*X^k.

The library syntax is GEN convol(GEN x, GEN y).

serlaplace(x)

x must be a power series with non-negative exponents. If x = sum (a_k/k!)*X^k then the result is sum a_k*X^k.

The library syntax is GEN laplace(GEN x).

serreverse(s)

Reverse power series of s, i.e. the series t such that t(s) = x; s must be a power series whose valuation is exactly equal to one.

  ? \ps 8
  ? t = serreverse(tan(x))
  %2 = x - 1/3*x^3 + 1/5*x^5 - 1/7*x^7 + O(x^8)
  ? tan(t)
  %3 = x + O(x^8)

The library syntax is GEN serreverse(GEN s).

subst(x,y,z)

Replace the simple variable y by the argument z in the ``polynomial'' expression x. Every type is allowed for x, but if it is not a genuine polynomial (or power series, or rational function), the substitution will be done as if the scalar components were polynomials of degree zero. In particular, beware that:

  ? subst(1, x, [1,2; 3,4])
  %1 =
  [1 0]
  [0 1]
  ? subst(1, x, Mat([0,1]))
    ***   at top-level: subst(1,x,Mat([0,1])
    ***                 ^--------------------
    *** subst: forbidden substitution by a non square matrix.

@3 If x is a power series, z must be either a polynomial, a power series, or a rational function. Finally, if x is a vector, matrix or list, the substitution is applied to each individual entry.

Use the function substvec to replace several variables at once, or the function substpol to replace a polynomial expression.

The library syntax is GEN gsubst(GEN x, long y, GEN z), where y is a variable number.

substpol(x,y,z)

Replace the ``variable'' y by the argument z in the ``polynomial'' expression x. Every type is allowed for x, but the same behavior as subst above apply.

The difference with subst is that y is allowed to be any polynomial here. The substitution is done moding out all components of x (recursively) by y - t, where t is a new free variable of lowest priority. Then substituting t by z in the resulting expression. For instance

  ? substpol(x^4 + x^2 + 1, x^2, y)
  %1 = y^2 + y + 1
  ? substpol(x^4 + x^2 + 1, x^3, y)
  %2 = x^2 + y*x + 1
  ? substpol(x^4 + x^2 + 1, (x+1)^2, y)
  %3 = (-4*y - 6)*x + (y^2 + 3*y - 3)

The library syntax is GEN gsubstpol(GEN x, GEN y, GEN z). Further, GEN gdeflate(GEN T, long v, long d) attempts to write T(x) in the form t(x^d), where x = pol_x(v), and returns NULL if the substitution fails (for instance in the example %2 above).

substvec(x,v,w)

v being a vector of monomials of degree 1 (variables), w a vector of expressions of the same length, replace in the expression x all occurrences of v_i by w_i. The substitutions are done simultaneously; more precisely, the v_i are first replaced by new variables in x, then these are replaced by the w_i:

  ? substvec([x,y], [x,y], [y,x])
  %1 = [y, x]
  ? substvec([x,y], [x,y], [y,x+y])
  %2 = [y, x + y]     \\ not [y, 2*y]

The library syntax is GEN gsubstvec(GEN x, GEN v, GEN w).

sumformal(f,{v})

formal sum of the polynomial expression f with respect to the main variable if v is omitted, with respect to the variable v otherwise; it is assumed that the base ring has characteristic zero. In other words, considering f as a polynomial function in the variable v, returns F, a polynomial in v vanishing at 0, such that F(b) - F(a) = sum_{v = a+1}^b f(v):

  ? sumformal(n)  \\ 1 + ... + n
  %1 = 1/2*n^2 + 1/2*n
  ? f(n) = n^3+n^2+1;
  ? F = sumformal(f(n))  \\ f(1) + ... + f(n)
  %3 = 1/4*n^4 + 5/6*n^3 + 3/4*n^2 + 7/6*n
  ? sum(n = 1, 2000, f(n)) == subst(F, n, 2000)
  %4 = 1
  ? sum(n = 1001, 2000, f(n)) == subst(F, n, 2000) - subst(F, n, 1000)
  %5 = 1
  ? sumformal(x^2 + x*y + y^2, y)
  %6 = y*x^2 + (1/2*y^2 + 1/2*y)*x + (1/3*y^3 + 1/2*y^2 + 1/6*y)
  ? x^2 * y + x * sumformal(y) + sumformal(y^2) == %
  %7 = 1

The library syntax is GEN sumformal(GEN f, long v = -1), where v is a variable number.

taylor(x,t,{d = seriesprecision})

Taylor expansion around 0 of x with respect to the simple variable t. x can be of any reasonable type, for example a rational function. Contrary to Ser, which takes the valuation into account, this function adds O(t^d) to all components of x.

  ? taylor(x/(1+y), y, 5)
  %1 = (y^4 - y^3 + y^2 - y + 1)*x + O(y^5)
  ? Ser(x/(1+y), y, 5)
   ***   at top-level: Ser(x/(1+y),y,5)
   ***                 ^----------------
   *** Ser: main variable must have higher priority in gtoser.

The library syntax is GEN tayl(GEN x, long t, long precdl), where t is a variable number.

thue(tnf,a,{sol})

Returns all solutions of the equation P(x,y) = a in integers x and y, where tnf was created with thueinit(P). If present, sol must contain the solutions of Norm (x) = a modulo units of positive norm in the number field defined by P (as computed by bnfisintnorm). If there are infinitely many solutions, an error will be issued.

It is allowed to input directly the polynomial P instead of a tnf, in which case, the function first performs thueinit(P,0). This is very wasteful if more than one value of a is required.

If tnf was computed without assuming GRH (flag 1 in thueinit), then the result is unconditional. Otherwise, it depends in principle of the truth of the GRH, but may still be unconditionally correct in some favorable cases. The result is conditional on the GRH if a != +- 1 and, P has a single irreducible rational factor, whose associated tentative class number h and regulator R (as computed assuming the GRH) satisfy

@3* h > 1,

@3* R/0.2 > 1.5.

Here's how to solve the Thue equation x^{13} - 5y^{13} = - 4:

  ? tnf = thueinit(x^13 - 5);
  ? thue(tnf, -4)
  %1 = [[1, 1]]

@3In this case, one checks that bnfinit(x^13 -5).no is 1. Hence, the only solution is (x,y) = (1,1), and the result is unconditional. On the other hand:

  ? P = x^3-2*x^2+3*x-17; tnf = thueinit(P);
  ? thue(tnf, -15)
  %2 = [[1, 1]]  \\ a priori conditional on the GRH.
  ? K = bnfinit(P); K.no
  %3 = 3
  ? K.reg
  %4 = 2.8682185139262873674706034475498755834

This time the result is conditional. All results computed using this particular tnf are likewise conditional, except for a right-hand side of +- 1. The above result is in fact correct, so we did not just disprove the GRH:

  ? tnf = thueinit(x^3-2*x^2+3*x-17, 1 /*unconditional*/);
  ? thue(tnf, -15)
  %4 = [[1, 1]]

Note that reducible or non-monic polynomials are allowed:

  ? tnf = thueinit((2*x+1)^5 * (4*x^3-2*x^2+3*x-17), 1);
  ? thue(tnf, 128)
  %2 = [[-1, 0], [1, 0]]

@3Reducible polynomials are in fact much easier to handle.

The library syntax is GEN thue(GEN tnf, GEN a, GEN sol = NULL).

thueinit(P,{flag = 0})

Initializes the tnf corresponding to P, a univariate polynomial with integer coefficients. The result is meant to be used in conjunction with thue to solve Thue equations P(X / Y)Y^{ deg P} = a, where a is an integer.

If flag is non-zero, certify results unconditionally. Otherwise, assume GRH, this being much faster of course. In the latter case, the result may still be unconditionally correct, see thue. For instance in most cases where P is reducible (not a pure power of an irreducible), or conditional computed class groups are trivial or the right hand side is +-1, then results are always unconditional.

The library syntax is GEN thueinit(GEN P, long flag, long prec).


Vectors, matrices, linear algebra and sets

Note that most linear algebra functions operating on subspaces defined by generating sets (such as mathnf, qflll, etc.) take matrices as arguments. As usual, the generating vectors are taken to be the columns of the given matrix.

Since PARI does not have a strong typing system, scalars live in unspecified commutative base rings. It is very difficult to write robust linear algebra routines in such a general setting. We thus assume that the base ring is a domain and work over its field of fractions. If the base ring is not a domain, one gets an error as soon as a non-zero pivot turns out to be non-invertible. Some functions, e.g. mathnf or mathnfmod, specifically assume that the base ring is Z.

algdep(x,k,{flag = 0})

x being real/complex, or p-adic, finds a polynomial of degree at most k with integer coefficients having x as approximate root. Note that the polynomial which is obtained is not necessarily the ``correct'' one. In fact it is not even guaranteed to be irreducible. One can check the closeness either by a polynomial evaluation (use subst), or by computing the roots of the polynomial given by algdep (use polroots).

Internally, lindep([1,x,...,x^k], flag) is used. A non-zero value of flag may improve on the default behavior if the input number is known to a huge accuracy, and you suspect the last bits are incorrect (this truncates the number, throwing away the least significant bits), but default values are usually sufficient:

  ? \p200
  ? algdep(2^(1/6)+3^(1/5), 30);      \\ wrong in 0.8s
  ? algdep(2^(1/6)+3^(1/5), 30, 100); \\ wrong in 0.4s
  ? algdep(2^(1/6)+3^(1/5), 30, 170); \\ right in 0.8s
  ? algdep(2^(1/6)+3^(1/5), 30, 200); \\ wrong in 1.0s
  ? \p250
  ? algdep(2^(1/6)+3^(1/5), 30);      \\ right in 1.0s
  ? algdep(2^(1/6)+3^(1/5), 30, 200); \\ right in 1.0s
  ? \p500
  ? algdep(2^(1/6)+3^(1/5), 30);      \\ right in 2.9s
  ? \p1000
  ? algdep(2^(1/6)+3^(1/5), 30);      \\ right in 10.6s

The changes in defaultprecision only affect the quality of the initial approximation to 2^{1/6} + 3^{1/5}, algdep itself uses exact operations (the size of its operands depend on the accuracy of the input of course: more accurate input means slower operations).

Proceeding by increments of 5 digits of accuracy, algdep with default flag produces its first correct result at 205 digits, and from then on a steady stream of correct results.

The above example is the test case studied in a 2000 paper by Borwein and Lisonek: Applications of integer relation algorithms, Discrete Math., 217, p. 65--82. The version of PARI tested there was 1.39, which succeeded reliably from precision 265 on, in about 200 as much time as the current version.

The library syntax is GEN algdep0(GEN x, long k, long flag). Also available is GEN algdep(GEN x, long k) (flag = 0).

charpoly(A,{v = 'x},{flag = 5})

characteristic polynomial of A with respect to the variable v, i.e. determinant of v*I-A if A is a square matrix.

  ? charpoly([1,2;3,4]);
  %1 = x^2 - 5*x - 2
  ? charpoly([1,2;3,4],, 't)
  %2 = t^2 - 5*t - 2

If A is not a square matrix, the function returns the characteristic polynomial of the map ``multiplication by A'' if A is a scalar:

  ? charpoly(Mod(x+2, x^3-2))
  %1 = x^3 - 6*x^2 + 12*x - 10
  ? charpoly(I)
  %2 = x^2 + 1
  ? charpoly(quadgen(5))
  %3 = x^2 - x - 1
  ? charpoly(ffgen(ffinit(2,4)))
  %4 = Mod(1, 2)*x^4 + Mod(1, 2)*x^3 + Mod(1, 2)*x^2 + Mod(1, 2)*x + Mod(1, 2)

The value of flag is only significant for matrices, and we advise to stick to the default value. Let n be the dimension of A.

If flag = 0, same method (Le Verrier's) as for computing the adjoint matrix, i.e. using the traces of the powers of A. Assumes that n! is invertible; uses O(n^4) scalar operations.

If flag = 1, uses Lagrange interpolation which is usually the slowest method. Assumes that n! is invertible; uses O(n^4) scalar operations.

If flag = 2, uses the Hessenberg form. Assumes that the base ring is a field. Uses O(n^3) scalar operations, but suffers from coefficient explosion unless the base field is finite or R.

If flag = 3, uses Berkowitz's division free algorithm, valid over any ring (commutative, with unit). Uses O(n^4) scalar operations.

If flag = 4, x must be integral. Uses a modular algorithm: Hessenberg form for various small primes, then Chinese remainders.

If flag = 5 (default), uses the ``best'' method given x. This means we use Berkowitz unless the base ring is Z (use flag = 4) or a field where coefficient explosion does not occur, e.g. a finite field or the reals (use flag = 2).

The library syntax is GEN charpoly0(GEN A, long v = -1, long flag), where v is a variable number. Also available are GEN charpoly(GEN x, long v) (flag = 5), GEN caract(GEN A, long v) (flag = 1), GEN carhess(GEN A, long v) (flag = 2), GEN carberkowitz(GEN A, long v) (flag = 3) and GEN caradj(GEN A, long v, GEN *pt). In this last case, if pt is not NULL, *pt receives the address of the adjoint matrix of A (see matadjoint), so both can be obtained at once.

concat(x,{y})

Concatenation of x and y. If x or y is not a vector or matrix, it is considered as a one-dimensional vector. All types are allowed for x and y, but the sizes must be compatible. Note that matrices are concatenated horizontally, i.e. the number of rows stays the same. Using transpositions, one can concatenate them vertically, but it is often simpler to use matconcat.

  ? x = matid(2); y = 2*matid(2);
  ? concat(x,y)
  %2 =
  [1 0 2 0]
  [0 1 0 2]
  ? concat(x~,y~)~
  %3 =
  [1 0]
  [0 1]
  [2 0]
  [0 2]
  ? matconcat([x;y])
  %4 =
  [1 0]
  [0 1]
  [2 0]
  [0 2]

@3 To concatenate vectors sideways (i.e. to obtain a two-row or two-column matrix), use Mat instead, or matconcat:

  ? x = [1,2];
  ? y = [3,4];
  ? concat(x,y)
  %3 = [1, 2, 3, 4]
  ? Mat([x,y]~)
  %4 =
  [1 2]
  [3 4]
  ? matconcat([x;y])
  %5 =
  [1 2]
  [3 4]

Concatenating a row vector to a matrix having the same number of columns will add the row to the matrix (top row if the vector is x, i.e. comes first, and bottom row otherwise).

The empty matrix [;] is considered to have a number of rows compatible with any operation, in particular concatenation. (Note that this is not the case for empty vectors [ ] or [ ]~.)

If y is omitted, x has to be a row vector or a list, in which case its elements are concatenated, from left to right, using the above rules.

  ? concat([1,2], [3,4])
  %1 = [1, 2, 3, 4]
  ? a = [[1,2]~, [3,4]~]; concat(a)
  %2 =
  [1 3]
  [2 4]
  ? concat([1,2; 3,4], [5,6]~)
  %3 =
  [1 2 5]
  [3 4 6]
  ? concat([%, [7,8]~, [1,2,3,4]])
  %5 =
  [1 2 5 7]
  [3 4 6 8]
  [1 2 3 4]

The library syntax is GEN concat(GEN x, GEN y = NULL). GEN concat1(GEN x) is a shortcut for concat(x,NULL).

forqfvec(v,q,b,expr)

q being a square and symmetric matrix representing a positive definite quadratic form, evaluate expr for all vector v such that q(v) <= b. The formal variable v runs through all such vectors in turn.

  ? forqfvec(v, [3,2;2,3], 3, print(v))
  [0, 1]~
  [1, 0]~
  [-1, 1]~

The library syntax is void forqfvec0(GEN v, GEN q = NULL, GEN b). The following function is also available: void forqfvec(void *E, long (*fun)(void *, GEN, double), GEN q, GEN b): Evaluate fun(E,v,m) on all v such that q(v) < b, where v is a t_VECSMALL and m = q(v) is a C double. The function fun must return 0, unless forqfvec should stop, in which case, it should return 1.

lindep(v,{flag = 0})

finds a small non-trivial integral linear combination between components of v. If none can be found return an empty vector.

If v is a vector with real/complex entries we use a floating point (variable precision) LLL algorithm. If flag = 0 the accuracy is chosen internally using a crude heuristic. If flag > 0 the computation is done with an accuracy of flag decimal digits. To get meaningful results in the latter case, the parameter flag should be smaller than the number of correct decimal digits in the input.

  ? lindep([sqrt(2), sqrt(3), sqrt(2)+sqrt(3)])
  %1 = [-1, -1, 1]~

If v is p-adic, flag is ignored and the algorithm LLL-reduces a suitable (dual) lattice.

  ? lindep([1, 2 + 3 + 3^2 + 3^3 + 3^4 + O(3^5)])
  %2 = [1, -2]~

If v is a matrix, flag is ignored and the function returns a non trivial kernel vector (combination of the columns).

  ? lindep([1,2,3;4,5,6;7,8,9])
  %3 = [1, -2, 1]~

If v contains polynomials or power series over some base field, finds a linear relation with coefficients in the field.

  ? lindep([x*y, x^2 + y, x^2*y + x*y^2, 1])
  %4 = [y, y, -1, -y^2]~

@3For better control, it is preferable to use t_POL rather than t_SER in the input, otherwise one gets a linear combination which is t-adically small, but not necessarily 0. Indeed, power series are first converted to the minimal absolute accuracy occurring among the entries of v (which can cause some coefficients to be ignored), then truncated to polynomials:

  ? v = [t^2+O(t^4), 1+O(t^2)]; L=lindep(v)
  %1 = [1, 0]~
  ? v*L
  %2 = t^2+O(t^4)  \\ small but not 0

The library syntax is GEN lindep0(GEN v, long flag). Also available are GEN lindep(GEN v) (real/complex entries, flag = 0), GEN lindep2(GEN v, long flag) (real/complex entries) GEN padic_lindep(GEN v) (p-adic entries) and GEN Xadic_lindep(GEN v) (polynomial entries). Finally GEN deplin(GEN v) returns a non-zero kernel vector for a t_MAT input.

listcreate()

Creates an empty list. This routine used to have a mandatory argument, which is now ignored (for backward compatibility). In fact, this function has become redundant and obsolete; it will disappear in future versions of PARI: just use List()

listinsert(L,x,n)

Inserts the object x at position n in L (which must be of type t_LIST). This has complexity O(#L - n + 1): all the remaining elements of list (from position n+1 onwards) are shifted to the right.

The library syntax is GEN listinsert(GEN L, GEN x, long n).

listkill(L)

Obsolete, retained for backward compatibility. Just use L = List() instead of listkill(L). In most cases, you won't even need that, e.g. local variables are automatically cleared when a user function returns.

The library syntax is void listkill(GEN L).

listpop(list,{n})

Removes the n-th element of the list list (which must be of type t_LIST). If n is omitted, or greater than the list current length, removes the last element. If the list is already empty, do nothing. This runs in time O(#L - n + 1).

The library syntax is void listpop(GEN list, long n).

listput(list,x,{n})

Sets the n-th element of the list list (which must be of type t_LIST) equal to x. If n is omitted, or greater than the list length, appends x. You may put an element into an occupied cell (not changing the list length), but it is easier to use the standard list[n] = x construct. This runs in time O(#L) in the worst case (when the list must be reallocated), but in time O(1) on average: any number of successive listputs run in time O(#L), where #L denotes the list final length.

The library syntax is GEN listput(GEN list, GEN x, long n).

listsort(L,{flag = 0})

Sorts the t_LIST list in place, with respect to the (somewhat arbitrary) universal comparison function cmp. In particular, the ordering is the same as for sets and setsearch can be used on a sorted list.

  ? L = List([1,2,4,1,3,-1]); listsort(L); L
  %1 = List([-1, 1, 1, 2, 3, 4])
  ? setsearch(L, 4)
  %2 = 6
  ? setsearch(L, -2)
  %3 = 0

@3This is faster than the vecsort command since the list is sorted in place: no copy is made. No value returned.

If flag is non-zero, suppresses all repeated coefficients.

The library syntax is void listsort(GEN L, long flag).

matadjoint(M,{flag = 0})

adjoint matrix of M, i.e. a matrix N of cofactors of M, satisfying M*N = det (M)* Id . M must be a (non-necessarily invertible) square matrix of dimension n. If flag is 0 or omitted, we try to use Leverrier-Faddeev's algorithm, which assumes that n! invertible. If it fails or flag = 1, compute T = charpoly(M) independently first and return (-1)^{n-1} (T(x)-T(0))/x evaluated at M.

  ? a = [1,2,3;3,4,5;6,7,8] * Mod(1,4);
  %2 =
  [Mod(1, 4) Mod(2, 4) Mod(3, 4)]
  [Mod(3, 4) Mod(0, 4) Mod(1, 4)]
  [Mod(2, 4) Mod(3, 4) Mod(0, 4)]

@3 Both algorithms use O(n^4) operations in the base ring, and are usually slower than computing the characteristic polynomial or the inverse of M directly.

The library syntax is GEN matadjoint0(GEN M, long flag). Also available are GEN adj(GEN x) (flag = 0) and GEN adjsafe(GEN x) (flag = 1).

matcompanion(x)

The left companion matrix to the non-zero polynomial x.

The library syntax is GEN matcompanion(GEN x).

matconcat(v)

Returns a t_MAT built from the entries of v, which may be a t_VEC (concatenate horizontally), a t_COL (concatenate vertically), or a t_MAT (concatenate vertically each column, and concatenate vertically the resulting matrices). The entries of v are always considered as matrices: they can themselves be t_VEC (seen as a row matrix), a t_COL seen as a column matrix), a t_MAT, or a scalar (seen as an 1 x 1 matrix).

  ? A=[1,2;3,4]; B=[5,6]~; C=[7,8]; D=9;
  ? matconcat([A, B]) \\ horizontal
  %1 =
  [1 2 5]
  [3 4 6]
  ? matconcat([A, C]~) \\ vertical
  %2 =
  [1 2]
  [3 4]
  [7 8]
  ? matconcat([A, B; C, D]) \\ block matrix
  %3 =
  [1 2 5]
  [3 4 6]
  [7 8 9]

@3 If the dimensions of the entries to concatenate do not match up, the above rules are extended as follows:

@3* each entry v_{i,j} of v has a natural length and height: 1 x 1 for a scalar, 1 x n for a t_VEC of length n, n x 1 for a t_COL, m x n for an m x n t_MAT

@3* let H_i be the maximum over j of the lengths of the v_{i,j}, let L_j be the maximum over i of the heights of the v_{i,j}. The dimensions of the (i,j)-th block in the concatenated matrix are H_i x L_j.

@3* a scalar s = v_{i,j} is considered as s times an identity matrix of the block dimension \min (H_i,L_j)

@3* blocks are extended by 0 columns on the right and 0 rows at the bottom, as needed.

  ? matconcat([1, [2,3]~, [4,5,6]~]) \\ horizontal
  %4 =
  [1 2 4]
  [0 3 5]
  [0 0 6]
  ? matconcat([1, [2,3], [4,5,6]]~) \\ vertical
  %5 =
  [1 0 0]
  [2 3 0]
  [4 5 6]
  ? matconcat([B, C; A, D]) \\ block matrix
  %6 =
  [5 0 7 8]
  [6 0 0 0]
  [1 2 9 0]
  [3 4 0 9]
  ? U=[1,2;3,4]; V=[1,2,3;4,5,6;7,8,9];
  ? matconcat(matdiagonal([U, V])) \\ block diagonal
  %7 =
  [1 2 0 0 0]
  [3 4 0 0 0]
  [0 0 1 2 3]
  [0 0 4 5 6]
  [0 0 7 8 9]

The library syntax is GEN matconcat(GEN v).

matdet(x,{flag = 0})

Determinant of the square matrix x.

If flag = 0, uses an appropriate algorithm depending on the coefficients:

@3* integer entries: modular method due to Dixon, Pernet and Stein.

@3* real or p-adic entries: classical Gaussian elimination using maximal pivot.

@3* intmod entries: classical Gaussian elimination using first non-zero pivot.

@3* other cases: Gauss-Bareiss.

If flag = 1, uses classical Gaussian elimination with appropriate pivoting strategy (maximal pivot for real or p-adic coefficients). This is usually worse than the default.

The library syntax is GEN det0(GEN x, long flag). Also available are GEN det(GEN x) (flag = 0), GEN det2(GEN x) (flag = 1) and GEN ZM_det(GEN x) for integer entries.

matdetint(B)

Let B be an m x n matrix with integer coefficients. The determinant D of the lattice generated by the columns of B is the square root of det (B^T B) if B has maximal rank m, and 0 otherwise.

This function uses the Gauss-Bareiss algorithm to compute a positive multiple of D. When B is square, the function actually returns D = | det B|.

This function is useful in conjunction with mathnfmod, which needs to know such a multiple. If the rank is maximal and the matrix non-square, you can obtain D exactly using

    matdet( mathnfmod(B, matdetint(B)) )

Note that as soon as one of the dimensions gets large (m or n is larger than 20, say), it will often be much faster to use mathnf(B, 1) or mathnf(B, 4) directly.

The library syntax is GEN detint(GEN B).

matdiagonal(x)

x being a vector, creates the diagonal matrix whose diagonal entries are those of x.

  ? matdiagonal([1,2,3]);
  %1 =
  [1 0 0]
  [0 2 0]
  [0 0 3]

@3Block diagonal matrices are easily created using matconcat:

  ? U=[1,2;3,4]; V=[1,2,3;4,5,6;7,8,9];
  ? matconcat(matdiagonal([U, V]))
  %1 =
  [1 2 0 0 0]
  [3 4 0 0 0]
  [0 0 1 2 3]
  [0 0 4 5 6]
  [0 0 7 8 9]

The library syntax is GEN diagonal(GEN x).

mateigen(x,{flag = 0})

Returns the (complex) eigenvectors of x as columns of a matrix. If flag = 1, return [L,H], where L contains the eigenvalues and H the corresponding eigenvectors; multiple eigenvalues are repeated according to the eigenspace dimension (which may be less than the eigenvalue multiplicity in the characteristic polynomial).

This function first computes the characteristic polynomial of x and approximates its complex roots (lambda_i), then tries to compute the eigenspaces as kernels of the x - lambda_i. This algorithm is ill-conditioned and is likely to miss kernel vectors if some roots of the characteristic polynomial are close, in particular if it has multiple roots.

  ? A = [13,2; 10,14]; mateigen(A)
  %1 =
  [-1/2 2/5]
  [   1   1]
  ? [L,H] = mateigen(A, 1);
  ? L
  %3 = [9, 18]
  ? H
  %4 =
  [-1/2 2/5]
  [   1   1]

@3 For symmetric matrices, use qfjacobi instead; for Hermitian matrices, compute

   A = real(x);
   B = imag(x);
   y = matconcat([A, -B; B, A]);

@3and apply qfjacobi to y.

The library syntax is GEN mateigen(GEN x, long flag, long prec). Also available is GEN eigen(GEN x, long prec) (flag = 0)

matfrobenius(M,{flag},{v = 'x})

Returns the Frobenius form of the square matrix M. If flag = 1, returns only the elementary divisors as a vector of polynomials in the variable v. If flag = 2, returns a two-components vector [F,B] where F is the Frobenius form and B is the basis change so that M = B^{-1}FB.

The library syntax is GEN matfrobenius(GEN M, long flag, long v = -1), where v is a variable number.

mathess(x)

Returns a matrix similar to the square matrix x, which is in upper Hessenberg form (zero entries below the first subdiagonal).

The library syntax is GEN hess(GEN x).

mathilbert(n)

x being a long, creates the Hilbert matrixof order x, i.e. the matrix whose coefficient (i,j) is 1/ (i+j-1).

The library syntax is GEN mathilbert(long n).

mathnf(M,{flag = 0})

Let R be a Euclidean ring, equal to Z or to K[X] for some field K. If M is a (not necessarily square) matrix with entries in R, this routine finds the upper triangular Hermite normal form of M. If the rank of M is equal to its number of rows, this is a square matrix. In general, the columns of the result form a basis of the R-module spanned by the columns of M.

The values 0,1,2,3 of flag have a binary meaning, analogous to the one in matsnf; in this case, binary digits of flag mean:

@3* 1 (complete output): if set, outputs [H,U], where H is the Hermite normal form of M, and U is a transformation matrix such that MU = [0|H]. The matrix U belongs to {GL}(R). When M has a large kernel, the entries of U are in general huge.

@3* 2 (generic input): Deprecated. If set, assume that R = K[X] is a polynomial ring; otherwise, assume that R = Z. This flag is now useless since the routine always checks whether the matrix has integral entries.

@3For these 4 values, we use a naive algorithm, which behaves well in small dimension only. Larger values correspond to different algorithms, are restricted to integer matrices, and all output the unimodular matrix U. From now on all matrices have integral entries.

@3* flag = 4, returns [H,U] as in ``complete output'' above, using a variant of LLL reduction along the way. The matrix U is provably small in the L_2 sense, and in general close to optimal; but the reduction is in general slow, although provably polynomial-time.

If flag = 5, uses Batut's algorithm and output [H,U,P], such that H and U are as before and P is a permutation of the rows such that P applied to MU gives H. This is in general faster than flag = 4 but the matrix U is usually worse; it is heuristically smaller than with the default algorithm.

When the matrix is dense and the dimension is large (bigger than 100, say), flag = 4 will be fastest. When M has maximal rank, then

    H = mathnfmod(M, matdetint(M))

@3will be even faster. You can then recover U as M^{-1}H.

  ? M = matrix(3,4,i,j,random([-5,5]))
  %1 =
  [ 0 2  3  0]
  [-5 3 -5 -5]
  [ 4 3 -5  4]
  ? [H,U] = mathnf(M, 1);
  ? U
  %3 =
  [-1 0 -1 0]
  [ 0 5  3 2]
  [ 0 3  1 1]
  [ 1 0  0 0]
  ? H
  %5 =
  [19 9 7]
  [ 0 9 1]
  [ 0 0 1]
  ? M*U
  %6 =
  [0 19 9 7]
  [0  0 9 1]
  [0  0 0 1]

For convenience, M is allowed to be a t_VEC, which is then automatically converted to a t_MAT, as per the Mat function. For instance to solve the generalized extended gcd problem, one may use

  ? v = [116085838, 181081878, 314252913,10346840];
  ? [H,U] = mathnf(v, 1);
  ? U
  %2 =
  [ 103 -603    15  -88]
  [-146   13 -1208  352]
  [  58  220   678 -167]
  [-362 -144   381 -101]
  ? v*U
  %3 = [0, 0, 0, 1]

@3This also allows to input a matrix as a t_VEC of t_COLs of the same length (which Mat would concatenate to the t_MAT having those columns):

  ? v = [[1,0,4]~, [3,3,4]~, [0,-4,-5]~]; mathnf(v)
  %1 =
  [47 32 12]
  [ 0  1  0]
  [ 0  0  1]

The library syntax is GEN mathnf0(GEN M, long flag). Also available are GEN hnf(GEN M) (flag = 0) and GEN hnfall(GEN M) (flag = 1). To reduce huge relation matrices (sparse with small entries, say dimension 400 or more), you can use the pair hnfspec / hnfadd. Since this is quite technical and the calling interface may change, they are not documented yet. Look at the code in basemath/hnf_snf.c.

mathnfmod(x,d)

If x is a (not necessarily square) matrix of maximal rank with integer entries, and d is a multiple of the (non-zero) determinant of the lattice spanned by the columns of x, finds the upper triangular Hermite normal form of x.

If the rank of x is equal to its number of rows, the result is a square matrix. In general, the columns of the result form a basis of the lattice spanned by the columns of x. Even when d is known, this is in general slower than mathnf but uses much less memory.

The library syntax is GEN hnfmod(GEN x, GEN d).

mathnfmodid(x,d)

Outputs the (upper triangular) Hermite normal form of x concatenated with the diagonal matrix with diagonal d. Assumes that x has integer entries. Variant: if d is an integer instead of a vector, concatenate d times the identity matrix.

  ? m=[0,7;-1,0;-1,-1]
  %1 =
  [ 0  7]
  [-1  0]
  [-1 -1]
  ? mathnfmodid(m, [6,2,2])
  %2 =
  [2 1 1]
  [0 1 0]
  [0 0 1]
  ? mathnfmodid(m, 10)
  %3 =
  [10 7 3]
  [ 0 1 0]
  [ 0 0 1]

The library syntax is GEN hnfmodid(GEN x, GEN d).

mathouseholder(Q,v)

applies a sequence Q of Householder transforms, as returned by matqr(M,1) to the vector or matrix v.

The library syntax is GEN mathouseholder(GEN Q, GEN v).

matid(n)

Creates the n x n identity matrix.

The library syntax is GEN matid(long n).

matimage(x,{flag = 0})

Gives a basis for the image of the matrix x as columns of a matrix. A priori the matrix can have entries of any type. If flag = 0, use standard Gauss pivot. If flag = 1, use matsupplement (much slower: keep the default flag!).

The library syntax is GEN matimage0(GEN x, long flag). Also available is GEN image(GEN x) (flag = 0).

matimagecompl(x)

Gives the vector of the column indices which are not extracted by the function matimage, as a permutation (t_VECSMALL). Hence the number of components of matimagecompl(x) plus the number of columns of matimage(x) is equal to the number of columns of the matrix x.

The library syntax is GEN imagecompl(GEN x).

matindexrank(x)

x being a matrix of rank r, returns a vector with two t_VECSMALL components y and z of length r giving a list of rows and columns respectively (starting from 1) such that the extracted matrix obtained from these two vectors using vecextract(x,y,z) is invertible.

The library syntax is GEN indexrank(GEN x).

matintersect(x,y)

x and y being two matrices with the same number of rows each of whose columns are independent, finds a basis of the Q-vector space equal to the intersection of the spaces spanned by the columns of x and y respectively. The faster function idealintersect can be used to intersect fractional ideals (projective Z_K modules of rank 1); the slower but much more general function nfhnf can be used to intersect general Z_K-modules.

The library syntax is GEN intersect(GEN x, GEN y).

matinverseimage(x,y)

Given a matrix x and a column vector or matrix y, returns a preimage z of y by x if one exists (i.e such that x z = y), an empty vector or matrix otherwise. The complete inverse image is z + {Ker} x, where a basis of the kernel of x may be obtained by matker.

  ? M = [1,2;2,4];
  ? matinverseimage(M, [1,2]~)
  %2 = [1, 0]~
  ? matinverseimage(M, [3,4]~)
  %3 = []~    \\ no solution
  ? matinverseimage(M, [1,3,6;2,6,12])
  %4 =
  [1 3 6]
  [0 0 0]
  ? matinverseimage(M, [1,2;3,4])
  %5 = [;]    \\ no solution
  ? K = matker(M)
  %6 =
  [-2]
  [1]

The library syntax is GEN inverseimage(GEN x, GEN y).

matisdiagonal(x)

Returns true (1) if x is a diagonal matrix, false (0) if not.

The library syntax is GEN isdiagonal(GEN x).

matker(x,{flag = 0})

Gives a basis for the kernel of the matrix x as columns of a matrix. The matrix can have entries of any type, provided they are compatible with the generic arithmetic operations (+, x and /).

If x is known to have integral entries, set flag = 1.

The library syntax is GEN matker0(GEN x, long flag). Also available are GEN ker(GEN x) (flag = 0), GEN keri(GEN x) (flag = 1).

matkerint(x,{flag = 0})

Gives an LLL-reduced Z-basis for the lattice equal to the kernel of the matrix x as columns of the matrix x with integer entries (rational entries are not permitted).

If flag = 0, uses an integer LLL algorithm.

If flag = 1, uses matrixqz(x,-2). Many orders of magnitude slower than the default: never use this.

The library syntax is GEN matkerint0(GEN x, long flag). See also GEN kerint(GEN x) (flag = 0), which is a trivial wrapper around

  ZM_lll(ZM_lll(x, 0.99, LLL_KER), 0.99, LLL_INPLACE);

@3Remove the outermost ZM_lll if LLL-reduction is not desired (saves time).

matmuldiagonal(x,d)

Product of the matrix x by the diagonal matrix whose diagonal entries are those of the vector d. Equivalent to, but much faster than x*matdiagonal(d).

The library syntax is GEN matmuldiagonal(GEN x, GEN d).

matmultodiagonal(x,y)

Product of the matrices x and y assuming that the result is a diagonal matrix. Much faster than x*y in that case. The result is undefined if x*y is not diagonal.

The library syntax is GEN matmultodiagonal(GEN x, GEN y).

matpascal(n,{q})

Creates as a matrix the lower triangular Pascal triangle of order x+1 (i.e. with binomial coefficients up to x). If q is given, compute the q-Pascal triangle (i.e. using q-binomial coefficients).

The library syntax is GEN matqpascal(long n, GEN q = NULL). Also available is GEN matpascal(GEN x).

matqr(M,{flag = 0})

Returns [Q,R], the QR-decomposition of the square invertible matrix M with real entries: Q is orthogonal and R upper triangular. If flag = 1, the orthogonal matrix is returned as a sequence of Householder transforms: applying such a sequence is stabler and faster than multiplication by the corresponding Q matrix. More precisely, if

    [Q,R] = matqr(M);
    [q,r] = matqr(M, 1);

@3then r = R and mathouseholder(q, M) is R; furthermore

    mathouseholder(q, matid(#M)) == Q~

@3the inverse of Q. This function raises an error if the precision is too low or x is singular.

The library syntax is GEN matqr(GEN M, long flag, long prec).

matrank(x)

Rank of the matrix x.

The library syntax is long rank(GEN x).

matrix(m,n,{X},{Y},{expr = 0})

Creation of the m x n matrix whose coefficients are given by the expression expr. There are two formal parameters in expr, the first one (X) corresponding to the rows, the second (Y) to the columns, and X goes from 1 to m, Y goes from 1 to n. If one of the last 3 parameters is omitted, fill the matrix with zeroes.

matrixqz(A,{p = 0})

A being an m x n matrix in M_{m,n}(Q), let {Im}_Q A (resp. {Im}_Z A) the Q-vector space (resp. the Z-module) spanned by the columns of A. This function has varying behavior depending on the sign of p:

If p >= 0, A is assumed to have maximal rank n <= m. The function returns a matrix B belongs to M_{m,n}(Z), with {Im}_Q B = {Im}_Q A, such that the GCD of all its n x n minors is coprime to p; in particular, if p = 0 (default), this GCD is 1.

  ? minors(x) = vector(#x[,1], i, matdet(x[^i,]));
  ? A = [3,1/7; 5,3/7; 7,5/7]; minors(A)
  %1 = [4/7, 8/7, 4/7]   \\ determinants of all 2x2 minors
  ? B = matrixqz(A)
  %2 =
  [3 1]
  [5 2]
  [7 3]
  ? minors(%)
  %3 = [1, 2, 1]   \\ B integral with coprime minors

If p = -1, returns the HNF basis of the lattice Z^n cap {Im}_Z A.

If p = -2, returns the HNF basis of the lattice Z^n cap {Im}_Q A.

  ? matrixqz(A,-1)
  %4 =
  [8 5]
  [4 3]
  [0 1]
  ? matrixqz(A,-2)
  %5 =
  [2 -1]
  [1 0]
  [0 1]

The library syntax is GEN matrixqz0(GEN A, GEN p = NULL).

matsize(x)

x being a vector or matrix, returns a row vector with two components, the first being the number of rows (1 for a row vector), the second the number of columns (1 for a column vector).

The library syntax is GEN matsize(GEN x).

matsnf(X,{flag = 0})

If X is a (singular or non-singular) matrix outputs the vector of elementary divisors of X, i.e. the diagonal of the Smith normal form of X, normalized so that d_n | d_{n-1} | ... | d_1.

The binary digits of flag mean:

1 (complete output): if set, outputs [U,V,D], where U and V are two unimodular matrices such that UXV is the diagonal matrix D. Otherwise output only the diagonal of D. If X is not a square matrix, then D will be a square diagonal matrix padded with zeros on the left or the top.

2 (generic input): if set, allows polynomial entries, in which case the input matrix must be square. Otherwise, assume that X has integer coefficients with arbitrary shape.

4 (cleanup): if set, cleans up the output. This means that elementary divisors equal to 1 will be deleted, i.e. outputs a shortened vector D' instead of D. If complete output was required, returns [U',V',D'] so that U'XV' = D' holds. If this flag is set, X is allowed to be of the form `vector of elementary divisors' or [U,V,D] as would normally be output with the cleanup flag unset.

The library syntax is GEN matsnf0(GEN X, long flag).

matsolve(M,B)

M being an invertible matrix and B a column vector, finds the solution X of MX = B, using Dixon p-adic lifting method if M and B are integral and Gaussian elimination otherwise. This has the same effect as, but is faster, than M^{-1}*B.

The library syntax is GEN gauss(GEN M, GEN B). For integral input, the function GEN ZM_gauss(GEN M,GEN B) is also available.

matsolvemod(M,D,B,{flag = 0})

M being any integral matrix, D a column vector of non-negative integer moduli, and B an integral column vector, gives a small integer solution to the system of congruences sum_i m_{i,j}x_j = b_i (mod d_i) if one exists, otherwise returns zero. Shorthand notation: B (resp. D) can be given as a single integer, in which case all the b_i (resp. d_i) above are taken to be equal to B (resp. D).

  ? M = [1,2;3,4];
  ? matsolvemod(M, [3,4]~, [1,2]~)
  %2 = [-2, 0]~
  ? matsolvemod(M, 3, 1) \\ M X = [1,1]~ over F_3
  %3 = [-1, 1]~
  ? matsolvemod(M, [3,0]~, [1,2]~) \\ x + 2y = 1 (mod 3), 3x + 4y = 2 (in Z)
  %4 = [6, -4]~

If flag = 1, all solutions are returned in the form of a two-component row vector [x,u], where x is a small integer solution to the system of congruences and u is a matrix whose columns give a basis of the homogeneous system (so that all solutions can be obtained by adding x to any linear combination of columns of u). If no solution exists, returns zero.

The library syntax is GEN matsolvemod0(GEN M, GEN D, GEN B, long flag). Also available are GEN gaussmodulo(GEN M, GEN D, GEN B) (flag = 0) and GEN gaussmodulo2(GEN M, GEN D, GEN B) (flag = 1).

matsupplement(x)

Assuming that the columns of the matrix x are linearly independent (if they are not, an error message is issued), finds a square invertible matrix whose first columns are the columns of x, i.e. supplement the columns of x to a basis of the whole space.

  ? matsupplement([1;2])
  %1 =
  [1 0]
  [2 1]

Raises an error if x has 0 columns, since (due to a long standing design bug), the dimension of the ambient space (the number of rows) is unknown in this case:

  ? matsupplement(matrix(2,0))
    ***   at top-level: matsupplement(matrix
    ***                 ^--------------------
    *** matsupplement: sorry, suppl [empty matrix] is not yet implemented.

The library syntax is GEN suppl(GEN x).

mattranspose(x)

Transpose of x (also x~). This has an effect only on vectors and matrices.

The library syntax is GEN gtrans(GEN x).

minpoly(A,{v = 'x})

minimal polynomial of A with respect to the variable v., i.e. the monic polynomial P of minimal degree (in the variable v) such that P(A) = 0.

The library syntax is GEN minpoly(GEN A, long v = -1), where v is a variable number.

norml2(x)

Square of the L^2-norm of x. More precisely, if x is a scalar, norml2(x) is defined to be the square of the complex modulus of x (real t_QUADs are not supported). If x is a polynomial, a (row or column) vector or a matrix, norml2(x) is defined recursively as sum_i norml2(x_i), where (x_i) run through the components of x. In particular, this yields the usual sum |x_i|^2 (resp. sum |x_{i,j}|^2) if x is a polynomial or vector (resp. matrix) with complex components.

  ? norml2( [ 1, 2, 3 ] )      \\ vector
  %1 = 14
  ? norml2( [ 1, 2; 3, 4] )   \\ matrix
  %2 = 30
  ? norml2( 2*I + x )
  %3 = 5
  ? norml2( [ [1,2], [3,4], 5, 6 ] )   \\ recursively defined
  %4 = 91

The library syntax is GEN gnorml2(GEN x).

normlp(x,{p})

L^p-norm of x; sup norm if p is omitted. More precisely, if x is a scalar, normlp(x, p) is defined to be abs(x). If x is a polynomial, a (row or column) vector or a matrix:

@3* if p is omitted, normlp(x) is defined recursively as \max_i normlp(x_i)), where (x_i) run through the components of x. In particular, this yields the usual sup norm if x is a polynomial or vector with complex components.

@3* otherwise, normlp(x, p) is defined recursively as (sum_i normlp^p(x_i,p))^{1/p}. In particular, this yields the usual (sum |x_i|^p)^{1/p} if x is a polynomial or vector with complex components.

  ? v = [1,-2,3]; normlp(v)      \\ vector
  %1 = 3
  ? M = [1,-2;-3,4]; normlp(M)   \\ matrix
  %2 = 4
  ? T = (1+I) + I*x^2; normlp(T)
  %3 = 1.4142135623730950488016887242096980786
  ? normlp([[1,2], [3,4], 5, 6])   \\ recursively defined
  %4 = 6
  ? normlp(v, 1)
  %5 = 6
  ? normlp(M, 1)
  %6 = 10
  ? normlp(T, 1)
  %7 = 2.4142135623730950488016887242096980786

The library syntax is GEN gnormlp(GEN x, GEN p = NULL, long prec).

qfauto(G,{fl})

G being a square and symmetric matrix with integer entries representing a positive definite quadratic form, outputs the automorphism group of the associate lattice. Since this requires computing the minimal vectors, the computations can become very lengthy as the dimension grows. G can also be given by an qfisominit structure. See qfisominit for the meaning of fl.

The output is a two-components vector [o,g] where o is the group order and g is the list of generators (as a vector). For each generator H, the equality G = {^t}H G H holds.

The interface of this function is experimental and will likely change in the future.

This function implements an algorithm of Plesken and Souvignier, following Souvignier's implementation.

The library syntax is GEN qfauto0(GEN G, GEN fl = NULL). Also available is GEN qfauto(GEN G, GEN fl) where G is a vector of zm.

qfautoexport(qfa,{flag})

qfa being an automorphism group as output by qfauto, export the underlying matrix group as a string suitable for (no flags or flag = 0) GAP or (flag = 1) Magma. The following example computes the size of the matrix group using GAP:

  ? G = qfauto([2,1;1,2])
  %1 = [12, [[-1, 0; 0, -1], [0, -1; 1, 1], [1, 1; 0, -1]]]
  ? s = qfautoexport(G)
  %2 = "Group([[-1, 0], [0, -1]], [[0, -1], [1, 1]], [[1, 1], [0, -1]])"
  ? extern("echo \"Order("s");\" | gap -q")
  %3 = 12

The library syntax is GEN qfautoexport(GEN qfa, long flag).

qfbil(x,y,{q})

Evaluate the bilinear form q (symmetric matrix) at the vectors (x,y); if q omitted, use the standard Euclidean scalar product, corresponding to the identity matrix.

Roughly equivalent to x~ * q * y, but a little faster and more convenient (does not distinguish between column and row vectors):

  ? x = [1,2,3]~; y = [-1,0,1]~; qfbil(x,y)
  %1 = 2
  ? q = [1,2,3;2,2,-1;3,-1,0]; qfbil(x,y, q)
  %2 = -13
  ? for(i=1,10^6, qfbil(x,y,q))
  %3 = 568ms
  ? for(i=1,10^6, x~*q*y)
  %4 = 717ms

@3The associated quadratic form is also available, as qfnorm, slightly faster:

  ? for(i=1,10^6, qfnorm(x,q))
  time = 444ms
  ? for(i=1,10^6, qfnorm(x))
  time = 176 ms.
  ? for(i=1,10^6, qfbil(x,y))
  time = 208 ms.

The library syntax is GEN qfbil(GEN x, GEN y, GEN q = NULL).

qfgaussred(q)

decomposition into squares of the quadratic form represented by the symmetric matrix q. The result is a matrix whose diagonal entries are the coefficients of the squares, and the off-diagonal entries on each line represent the bilinear forms. More precisely, if (a_{ij}) denotes the output, one has

   q(x) = sum_i a_{ii} (x_i + sum_{j != i} a_{ij} x_j)^2

? qfgaussred([0,1;1,0]) %1 = [1/2 1]

  [-1 -1/2]

@3This means that 2xy = (1/2)(x+y)^2 - (1/2)(x-y)^2.

The library syntax is GEN qfgaussred(GEN q). GEN qfgaussred_positive(GEN q) assumes that q is positive definite and is a little faster; returns NULL if a vector with negative norm occurs (non positive matrix or too many rounding errors).

qfisom(G,H,{fl})

G, H being square and symmetric matrices with integer entries representing positive definite quadratic forms, return an invertible matrix S such that G = {^t}S H S. This defines a isomorphism between the corresponding lattices. Since this requires computing the minimal vectors, the computations can become very lengthy as the dimension grows. See qfisominit for the meaning of fl.

G can also be given by an qfisominit structure which is preferable if several forms H need to be compared to G.

This function implements an algorithm of Plesken and Souvignier, following Souvignier's implementation.

The library syntax is GEN qfisom0(GEN G, GEN H, GEN fl = NULL). Also available is GEN qfisom(GEN G, GEN H, GEN fl) where G is a vector of zm, and H is a zm.

qfisominit(G,{fl})

G being a square and symmetric matrix with integer entries representing a positive definite quadratic form, return an isom structure allowing to compute isomorphisms between G and other quadratic forms faster.

The interface of this function is experimental and will likely change in future release.

If present, the optional parameter fl must be a t_VEC with two components. It allows to specify the invariants used, which can make the computation faster or slower. The components are

@3* fl[1] Depth of scalar product combination to use.

@3* fl[2] Maximum level of Bacher polynomials to use.

Since this function computes the minimal vectors, it can become very lengthy as the dimension of G grows.

The library syntax is GEN qfisominit0(GEN G, GEN fl = NULL). Also available is GEN qfisominit(GEN F, GEN fl) where F is a vector of zm.

qfjacobi(A)

Apply Jacobi's eigenvalue algorithm to the real symmetric matrix A. This returns [L, V], where

@3* L is the vector of (real) eigenvalues of A, sorted in increasing order,

@3* V is the corresponding orthogonal matrix of eigenvectors of A.

  ? \p19
  ? A = [1,2;2,1]; mateigen(A)
  %1 =
  [-1 1]
  [ 1 1]
  ? [L, H] = qfjacobi(A);
  ? L
  %3 = [-1.000000000000000000, 3.000000000000000000]~
  ? H
  %4 =
  [ 0.7071067811865475245 0.7071067811865475244]
  [-0.7071067811865475244 0.7071067811865475245]
  ? norml2( (A-L[1])*H[,1] )       \\ approximate eigenvector
  %5 = 9.403954806578300064 E-38
  ? norml2(H*H~ - 1)
  %6 = 2.350988701644575016 E-38   \\ close to orthogonal

The library syntax is GEN jacobi(GEN A, long prec).

qflll(x,{flag = 0})

LLL algorithm applied to the columns of the matrix x. The columns of x may be linearly dependent. The result is a unimodular transformation matrix T such that x .T is an LLL-reduced basis of the lattice generated by the column vectors of x. Note that if x is not of maximal rank T will not be square. The LLL parameters are (0.51,0.99), meaning that the Gram-Schmidt coefficients for the final basis satisfy mu_{i,j} <= |0.51|, and the Lov\'{a}sz's constant is 0.99.

If flag = 0 (default), assume that x has either exact (integral or rational) or real floating point entries. The matrix is rescaled, converted to integers and the behavior is then as in flag = 1.

If flag = 1, assume that x is integral. Computations involving Gram-Schmidt vectors are approximate, with precision varying as needed (Lehmer's trick, as generalized by Schnorr). Adapted from Nguyen and Stehlé's algorithm and Stehl\'e's code (fplll-1.3).

If flag = 2, x should be an integer matrix whose columns are linearly independent. Returns a partially reduced basis for x, using an unpublished algorithm by Peter Montgomery: a basis is said to be partially reduced if |v_i +- v_j| >= |v_i| for any two distinct basis vectors v_i, v_j.

This is faster than flag = 1, esp. when one row is huge compared to the other rows (knapsack-style), and should quickly produce relatively short vectors. The resulting basis is not LLL-reduced in general. If LLL reduction is eventually desired, avoid this partial reduction: applying LLL to the partially reduced matrix is significantly slower than starting from a knapsack-type lattice.

If flag = 4, as flag = 1, returning a vector [K, T] of matrices: the columns of K represent a basis of the integer kernel of x (not LLL-reduced in general) and T is the transformation matrix such that x.T is an LLL-reduced Z-basis of the image of the matrix x.

If flag = 5, case as case 4, but x may have polynomial coefficients.

If flag = 8, same as case 0, but x may have polynomial coefficients.

The library syntax is GEN qflll0(GEN x, long flag). Also available are GEN lll(GEN x) (flag = 0), GEN lllint(GEN x) (flag = 1), and GEN lllkerim(GEN x) (flag = 4).

qflllgram(G,{flag = 0})

Same as qflll, except that the matrix G = x~ * x is the Gram matrix of some lattice vectors x, and not the coordinates of the vectors themselves. In particular, G must now be a square symmetric real matrix, corresponding to a positive quadratic form (not necessarily definite: x needs not have maximal rank). The result is a unimodular transformation matrix T such that x.T is an LLL-reduced basis of the lattice generated by the column vectors of x. See qflll for further details about the LLL implementation.

If flag = 0 (default), assume that G has either exact (integral or rational) or real floating point entries. The matrix is rescaled, converted to integers and the behavior is then as in flag = 1.

If flag = 1, assume that G is integral. Computations involving Gram-Schmidt vectors are approximate, with precision varying as needed (Lehmer's trick, as generalized by Schnorr). Adapted from Nguyen and Stehlé's algorithm and Stehl\'e's code (fplll-1.3).

flag = 4: G has integer entries, gives the kernel and reduced image of x.

flag = 5: same as 4, but G may have polynomial coefficients.

The library syntax is GEN qflllgram0(GEN G, long flag). Also available are GEN lllgram(GEN G) (flag = 0), GEN lllgramint(GEN G) (flag = 1), and GEN lllgramkerim(GEN G) (flag = 4).

qfminim(x,{b},{m},{flag = 0})

x being a square and symmetric matrix representing a positive definite quadratic form, this function deals with the vectors of x whose norm is less than or equal to b, enumerated using the Fincke-Pohst algorithm, storing at most m vectors (no limit if m is omitted). The function searches for the minimal non-zero vectors if b is omitted. The behavior is undefined if x is not positive definite (a ``precision too low'' error is most likely, although more precise error messages are possible). The precise behavior depends on flag.

If flag = 0 (default), seeks at most 2m vectors. The result is a three-component vector, the first component being the number of vectors found, the second being the maximum norm found, and the last vector is a matrix whose columns are the vectors found, only one being given for each pair +- v (at most m such pairs, unless m was omitted). The vectors are returned in no particular order.

If flag = 1, ignores m and returns [N,v], where v is a non-zero vector of length N <= b, or [] if no non-zero vector has length <= b. If no explicit b is provided, return a vector of smallish norm (smallest vector in an LLL-reduced basis).

In these two cases, x must have integral entries. The implementation uses low precision floating point computations for maximal speed, which gives incorrect result when x has large entries. (The condition is checked in the code and the routine raises an error if large rounding errors occur.) A more robust, but much slower, implementation is chosen if the following flag is used:

If flag = 2, x can have non integral real entries. In this case, if b is omitted, the ``minimal'' vectors only have approximately the same norm. If b is omitted, m is an upper bound for the number of vectors that will be stored and returned, but all minimal vectors are nevertheless enumerated. If m is omitted, all vectors found are stored and returned; note that this may be a huge vector!

  ? x = matid(2);
  ? qfminim(x)  \\ 4 minimal vectors of norm 1: F<+->[0,1], F<+->[1,0]
  %2 = [4, 1, [0, 1; 1, 0]]
  ? { x =
  [4, 2, 0, 0, 0,-2, 0, 0, 0, 0, 0, 0, 1,-1, 0, 0, 0, 1, 0,-1, 0, 0, 0,-2;
   2, 4,-2,-2, 0,-2, 0, 0, 0, 0, 0, 0, 0,-1, 0, 0, 0, 0, 0,-1, 0, 1,-1,-1;
   0,-2, 4, 0,-2, 0, 0, 0, 0, 0, 0, 0,-1, 1, 0, 0, 1, 0, 0, 1,-1,-1, 0, 0;
   0,-2, 0, 4, 0, 0, 0, 0, 0, 0, 0, 0, 1,-1, 0, 0, 0, 1,-1, 0, 1,-1, 1, 0;
   0, 0,-2, 0, 4, 0, 0, 0, 1,-1, 0, 0, 1, 0, 0, 0,-2, 0, 0,-1, 1, 1, 0, 0;
  -2, -2,0, 0, 0, 4,-2, 0,-1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0,-1, 1, 1;
   0, 0, 0, 0, 0,-2, 4,-2, 0, 0, 0, 0, 0, 1, 0, 0, 0,-1, 0, 0, 0, 1,-1, 0;
   0, 0, 0, 0, 0, 0,-2, 4, 0, 0, 0, 0,-1, 0, 0, 0, 0, 0,-1,-1,-1, 0, 1, 0;
   0, 0, 0, 0, 1,-1, 0, 0, 4, 0,-2, 0, 1, 1, 0,-1, 0, 1, 0, 0, 0, 0, 0, 0;
   0, 0, 0, 0,-1, 0, 0, 0, 0, 4, 0, 0, 1, 1,-1, 1, 0, 0, 0, 1, 0, 0, 1, 0;
   0, 0, 0, 0, 0, 0, 0, 0,-2, 0, 4,-2, 0,-1, 0, 0, 0,-1, 0,-1, 0, 0, 0, 0;
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0,-2, 4,-1, 1, 0, 0,-1, 1, 0, 1, 1, 1,-1, 0;
   1, 0,-1, 1, 1, 0, 0,-1, 1, 1, 0,-1, 4, 0, 0, 1, 0, 1, 1, 0, 1, 0, 1,-1;
  -1,-1, 1,-1, 0, 0, 1, 0, 1, 1,-1, 1, 0, 4, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1;
   0, 0, 0, 0, 0, 0, 0, 0, 0,-1, 0, 0, 0, 1, 4, 0, 0, 0, 1, 0, 0, 0, 0, 0;
   0, 0, 0, 0, 0, 0, 0, 0,-1, 1, 0, 0, 1, 1, 0, 4, 0, 0, 0, 0, 1, 1, 0, 0;
   0, 0, 1, 0,-2, 0, 0, 0, 0, 0, 0,-1, 0, 0, 0, 0, 4, 1, 1, 1, 0, 0, 1, 1;
   1, 0, 0, 1, 0, 0,-1, 0, 1, 0,-1, 1, 1, 0, 0, 0, 1, 4, 0, 1, 1, 0, 1, 0;
   0, 0, 0,-1, 0, 1, 0,-1, 0, 0, 0, 0, 1, 1, 1, 0, 1, 0, 4, 0, 1, 1, 0, 1;
  -1, -1,1, 0,-1, 1, 0,-1, 0, 1,-1, 1, 0, 1, 0, 0, 1, 1, 0, 4, 0, 0, 1, 1;
   0, 0,-1, 1, 1, 0, 0,-1, 0, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0, 4, 1, 0, 1;
   0, 1,-1,-1, 1,-1, 1, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 0, 1, 0, 1, 4, 0, 1;
   0,-1, 0, 1, 0, 1,-1, 1, 0, 1, 0,-1, 1, 0, 0, 0, 1, 1, 0, 1, 0, 0, 4, 1;
  -2,-1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0,-1, 1, 0, 0, 1, 0, 1, 1, 1, 1, 1, 4]; }
  ? qfminim(x,,0)  \\ the Leech lattice has 196560 minimal vectors of norm 4
  time = 648 ms.
  %4 = [196560, 4, [;]]
  ? qfminim(x,,0,2); \\ safe algorithm. Slower and unnecessary here.
  time = 18,161 ms.
  %5 = [196560, 4.000061035156250000, [;]]

@3 In the last example, we store 0 vectors to limit memory use. All minimal vectors are nevertheless enumerated. Provided parisize is about 50MB, qfminim(x) succeeds in 2.5 seconds.

The library syntax is GEN qfminim0(GEN x, GEN b = NULL, GEN m = NULL, long flag, long prec). Also available are GEN minim(GEN x, GEN b = NULL, GEN m = NULL) (flag = 0), GEN minim2(GEN x, GEN b = NULL, GEN m = NULL) (flag = 1). GEN minim_raw(GEN x, GEN b = NULL, GEN m = NULL) (do not perform LLL reduction on x).

qfnorm(x,{q})

Evaluate the binary quadratic form q (symmetric matrix) at the vector x. If q omitted, use the standard Euclidean form, corresponding to the identity matrix.

Equivalent to x~ * q * x, but about twice faster and more convenient (does not distinguish between column and row vectors):

  ? x = [1,2,3]~; qfnorm(x)
  %1 = 14
  ? q = [1,2,3;2,2,-1;3,-1,0]; qfnorm(x, q)
  %2 = 23
  ? for(i=1,10^6, qfnorm(x,q))
  time = 384ms.
  ? for(i=1,10^6, x~*q*x)
  time = 729ms.

@3We also allow t_MATs of compatible dimensions for x, and return x~ * q * x in this case as well:

  ? M = [1,2,3;4,5,6;7,8,9]; qfnorm(M) \\ Gram matrix
  %5 =
  [66  78  90]
  [78  93 108]
  [90 108 126]
  ? for(i=1,10^6, qfnorm(M,q))
  time = 2,144 ms.
  ? for(i=1,10^6, M~*q*M)
  time = 2,793 ms.

@3The polar form is also available, as qfbil.

The library syntax is GEN qfnorm(GEN x, GEN q = NULL).

qfperfection(G)

G being a square and symmetric matrix with integer entries representing a positive definite quadratic form, outputs the perfection rank of the form. That is, gives the rank of the family of the s symmetric matrices v_iv_i^t, where s is half the number of minimal vectors and the v_i (1 <= i <= s) are the minimal vectors.

Since this requires computing the minimal vectors, the computations can become very lengthy as the dimension of x grows.

The library syntax is GEN perf(GEN G).

qfrep(q,B,{flag = 0})

q being a square and symmetric matrix with integer entries representing a positive definite quadratic form, count the vectors representing successive integers.

@3* If flag = 0, count all vectors. Outputs the vector whose i-th entry, 1 <= i <= B is half the number of vectors v such that q(v) = i.

@3* If flag = 1, count vectors of even norm. Outputs the vector whose i-th entry, 1 <= i <= B is half the number of vectors such that q(v) = 2i.

  ? q = [2, 1; 1, 3];
  ? qfrep(q, 5)
  %2 = Vecsmall([0, 1, 2, 0, 0]) \\ 1 vector of norm 2, 2 of norm 3, etc.
  ? qfrep(q, 5, 1)
  %3 = Vecsmall([1, 0, 0, 1, 0]) \\ 1 vector of norm 2, 0 of norm 4, etc.

This routine uses a naive algorithm based on qfminim, and will fail if any entry becomes larger than 2^{31} (or 2^{63}).

The library syntax is GEN qfrep0(GEN q, GEN B, long flag).

qfsign(x)

Returns [p,m] the signature of the quadratic form represented by the symmetric matrix x. Namely, p (resp. m) is the number of positive (resp. negative) eigenvalues of x.The result is computed using Gaussian reduction.

The library syntax is GEN qfsign(GEN x).

seralgdep(s,p,r)

finds a linear relation between powers (1,s, ..., s^p) of the series s, with polynomial coefficients of degree <= r. In case no relation is found, return 0.

  ? s = 1 + 10*y - 46*y^2 + 460*y^3 - 5658*y^4 + 77740*y^5 + O(y^6);
  ? seralgdep(s, 2, 2)
  %2 = -x^2 + (8*y^2 + 20*y + 1)
  ? subst(%, x, s)
  %3 = O(y^6)
  ? seralgdep(s, 1, 3)
  %4 = (-77*y^2 - 20*y - 1)*x + (310*y^3 + 231*y^2 + 30*y + 1)
  ? seralgdep(s, 1, 2)
  %5 = 0

@3The series main variable must not be x, so as to be able to express the result as a polynomial in x.

The library syntax is GEN seralgdep(GEN s, long p, long r).

setbinop(f,X,{Y})

The set whose elements are the f(x,y), where x,y run through X,Y. respectively. If Y is omitted, assume that X = Y and that f is symmetric: f(x,y) = f(y,x) for all x,y in X.

  ? X = [1,2,3]; Y = [2,3,4];
  ? setbinop((x,y)->x+y, X,Y) \\ set X + Y
  %2 = [3, 4, 5, 6, 7]
  ? setbinop((x,y)->x-y, X,Y) \\ set X - Y
  %3 = [-3, -2, -1, 0, 1]
  ? setbinop((x,y)->x+y, X)   \\ set 2X = X + X
  %2 = [2, 3, 4, 5, 6]

The library syntax is GEN setbinop(GEN f, GEN X, GEN Y = NULL).

setintersect(x,y)

Intersection of the two sets x and y (see setisset). If x or y is not a set, the result is undefined.

The library syntax is GEN setintersect(GEN x, GEN y).

setisset(x)

Returns true (1) if x is a set, false (0) if not. In PARI, a set is a row vector whose entries are strictly increasing with respect to a (somewhat arbitrary) universal comparison function. To convert any object into a set (this is most useful for vectors, of course), use the function Set.

  ? a = [3, 1, 1, 2];
  ? setisset(a)
  %2 = 0
  ? Set(a)
  %3 = [1, 2, 3]

The library syntax is long setisset(GEN x).

setminus(x,y)

Difference of the two sets x and y (see setisset), i.e. set of elements of x which do not belong to y. If x or y is not a set, the result is undefined.

The library syntax is GEN setminus(GEN x, GEN y).

setsearch(S,x,{flag = 0})

Determines whether x belongs to the set S (see setisset).

We first describe the default behaviour, when flag is zero or omitted. If x belongs to the set S, returns the index j such that S[j] = x, otherwise returns 0.

  ? T = [7,2,3,5]; S = Set(T);
  ? setsearch(S, 2)
  %2 = 1
  ? setsearch(S, 4)      \\ not found
  %3 = 0
  ? setsearch(T, 7)      \\ search in a randomly sorted vector
  %4 = 0 \\ WRONG !

If S is not a set, we also allow sorted lists with respect to the cmp sorting function, without repeated entries, as per listsort(L,1); otherwise the result is undefined.

  ? L = List([1,4,2,3,2]); setsearch(L, 4)
  %1 = 0 \\ WRONG !
  ? listsort(L, 1); L    \\ sort L first
  %2 = List([1, 2, 3, 4])
  ? setsearch(L, 4)
  %3 = 4                 \\ now correct

If flag is non-zero, this function returns the index j where x should be inserted, and 0 if it already belongs to S. This is meant to be used for dynamically growing (sorted) lists, in conjunction with listinsert.

  ? L = List([1,5,2,3,2]); listsort(L,1); L
  %1 = List([1,2,3,5])
  ? j = setsearch(L, 4, 1)  \\ 4 should have been inserted at index j
  %2 = 4
  ? listinsert(L, 4, j); L
  %3 = List([1, 2, 3, 4, 5])

The library syntax is long setsearch(GEN S, GEN x, long flag).

setunion(x,y)

Union of the two sets x and y (see setisset). If x or y is not a set, the result is undefined.

The library syntax is GEN setunion(GEN x, GEN y).

trace(x)

This applies to quite general x. If x is not a matrix, it is equal to the sum of x and its conjugate, except for polmods where it is the trace as an algebraic number.

For x a square matrix, it is the ordinary trace. If x is a non-square matrix (but not a vector), an error occurs.

The library syntax is GEN gtrace(GEN x).

vecextract(x,y,{z})

Extraction of components of the vector or matrix x according to y. In case x is a matrix, its components are the columns of x. The parameter y is a component specifier, which is either an integer, a string describing a range, or a vector.

If y is an integer, it is considered as a mask: the binary bits of y are read from right to left, but correspond to taking the components from left to right. For example, if y = 13 = (1101)_2 then the components 1,3 and 4 are extracted.

If y is a vector (t_VEC, t_COL or t_VECSMALL), which must have integer entries, these entries correspond to the component numbers to be extracted, in the order specified.

If y is a string, it can be

@3* a single (non-zero) index giving a component number (a negative index means we start counting from the end).

@3* a range of the form "a..b", where a and b are indexes as above. Any of a and b can be omitted; in this case, we take as default values a = 1 and b = -1, i.e. the first and last components respectively. We then extract all components in the interval [a,b], in reverse order if b < a.

In addition, if the first character in the string is ^, the complement of the given set of indices is taken.

If z is not omitted, x must be a matrix. y is then the row specifier, and z the column specifier, where the component specifier is as explained above.

  ? v = [a, b, c, d, e];
  ? vecextract(v, 5)         \\ mask
  %1 = [a, c]
  ? vecextract(v, [4, 2, 1]) \\ component list
  %2 = [d, b, a]
  ? vecextract(v, "2..4")    \\ interval
  %3 = [b, c, d]
  ? vecextract(v, "-1..-3")  \\ interval + reverse order
  %4 = [e, d, c]
  ? vecextract(v, "^2")      \\ complement
  %5 = [a, c, d, e]
  ? vecextract(matid(3), "2..", "..")
  %6 =
  [0 1 0]
  [0 0 1]

The range notations v[i..j] and v[^i] (for t_VEC or t_COL) and M[i..j, k..l] and friends (for t_MAT) implement a subset of the above, in a simpler and faster way, hence should be preferred in most common situations. The following features are not implemented in the range notation:

@3* reverse order,

@3* omitting either a or b in a..b.

The library syntax is GEN extract0(GEN x, GEN y, GEN z = NULL).

vecsearch(v,x,{cmpf})

Determines whether x belongs to the sorted vector or list v: return the (positive) index where x was found, or 0 if it does not belong to v.

If the comparison function cmpf is omitted, we assume that v is sorted in increasing order, according to the standard comparison function < , thereby restricting the possible types for x and the elements of v (integers, fractions or reals).

If cmpf is present, it is understood as a comparison function and we assume that v is sorted according to it, see vecsort for how to encode comparison functions.

  ? v = [1,3,4,5,7];
  ? vecsearch(v, 3)
  %2 = 2
  ? vecsearch(v, 6)
  %3 = 0 \\ not in the list
  ? vecsearch([7,6,5], 5) \\ unsorted vector: result undefined
  %4 = 0

By abuse of notation, x is also allowed to be a matrix, seen as a vector of its columns; again by abuse of notation, a t_VEC is considered as part of the matrix, if its transpose is one of the matrix columns.

  ? v = vecsort([3,0,2; 1,0,2]) \\ sort matrix columns according to lex order
  %1 =
  [0 2 3]
  [0 2 1]
  ? vecsearch(v, [3,1]~)
  %2 = 3
  ? vecsearch(v, [3,1])  \\ can search for x or x~
  %3 = 3
  ? vecsearch(v, [1,2])
  %4 = 0 \\ not in the list

@3

The library syntax is long vecsearch(GEN v, GEN x, GEN cmpf = NULL).

vecsort(x,{cmpf},{flag = 0})

Sorts the vector x in ascending order, using a mergesort method. x must be a list, vector or matrix (seen as a vector of its columns). Note that mergesort is stable, hence the initial ordering of ``equal'' entries (with respect to the sorting criterion) is not changed.

If cmpf is omitted, we use the standard comparison function lex, thereby restricting the possible types for the elements of x (integers, fractions or reals and vectors of those). If cmpf is present, it is understood as a comparison function and we sort according to it. The following possibilities exist:

@3* an integer k: sort according to the value of the k-th subcomponents of the components of x.

@3* a vector: sort lexicographically according to the components listed in the vector. For example, if cmpf = [2,1,3], sort with respect to the second component, and when these are equal, with respect to the first, and when these are equal, with respect to the third.

@3* a comparison function (t_CLOSURE), with two arguments x and y, and returning an integer which is < 0, > 0 or = 0 if x < y, x > y or x = y respectively. The sign function is very useful in this context:

  ? vecsort([3,0,2; 1,0,2]) \\ sort columns according to lex order
  %1 =
  [0 2 3]
  [0 2 1]
  ? vecsort(v, (x,y)->sign(y-x))            \\ reverse sort
  ? vecsort(v, (x,y)->sign(abs(x)-abs(y)))  \\ sort by increasing absolute value
  ? cmpf(x,y) = my(dx = poldisc(x), dy = poldisc(y)); sign(abs(dx) - abs(dy))
  ? vecsort([x^2+1, x^3-2, x^4+5*x+1], cmpf)

@3 The last example used the named cmpf instead of an anonymous function, and sorts polynomials with respect to the absolute value of their discriminant. A more efficient approach would use precomputations to ensure a given discriminant is computed only once:

  ? DISC = vector(#v, i, abs(poldisc(v[i])));
  ? perm = vecsort(vector(#v,i,i), (x,y)->sign(DISC[x]-DISC[y]))
  ? vecextract(v, perm)

@3Similar ideas apply whenever we sort according to the values of a function which is expensive to compute.

@3The binary digits of flag mean:

@3* 1: indirect sorting of the vector x, i.e. if x is an n-component vector, returns a permutation of [1,2,...,n] which applied to the components of x sorts x in increasing order. For example, vecextract(x, vecsort(x,,1)) is equivalent to vecsort(x).

@3* 4: use descending instead of ascending order.

@3* 8: remove ``duplicate'' entries with respect to the sorting function (keep the first occurring entry). For example:

    ? vecsort([Pi,Mod(1,2),z], (x,y)->0, 8)   \\ make everything compare equal
    %1 = [3.141592653589793238462643383]
    ? vecsort([[2,3],[0,1],[0,3]], 2, 8)
    %2 = [[0, 1], [2, 3]]

The library syntax is GEN vecsort0(GEN x, GEN cmpf = NULL, long flag).

vecsum(v)

Return the sum of the component of the vector v

The library syntax is GEN vecsum(GEN v).

vector(n,{X},{expr = 0})

Creates a row vector (type t_VEC) with n components whose components are the expression expr evaluated at the integer points between 1 and n. If one of the last two arguments is omitted, fill the vector with zeroes.

Avoid modifying X within expr; if you do, the formal variable still runs from 1 to n. In particular, vector(n,i,expr) is not equivalent to

  v = vector(n)
  for (i = 1, n, v[i] = expr)

as the following example shows:

  n = 3
  v = vector(n); vector(n, i, i++)            ----> [2, 3, 4]
  v = vector(n); for (i = 1, n, v[i] = i++)   ----> [2, 0, 4]

vectorsmall(n,{X},{expr = 0})

Creates a row vector of small integers (type t_VECSMALL) with n components whose components are the expression expr evaluated at the integer points between 1 and n. If one of the last two arguments is omitted, fill the vector with zeroes.

vectorv(n,{X},{expr = 0})

As vector, but returns a column vector (type t_COL).


Sums, products, integrals and similar functions

Although the gp calculator is programmable, it is useful to have a number of preprogrammed loops, including sums, products, and a certain number of recursions. Also, a number of functions from numerical analysis like numerical integration and summation of series will be described here.

One of the parameters in these loops must be the control variable, hence a simple variable name. In the descriptions, the letter X will always denote any simple variable name, and represents the formal parameter used in the function. The expression to be summed, integrated, etc. is any legal PARI expression, including of course expressions using loops.

@3Library mode. Since it is easier to program directly the loops in library mode, these functions are mainly useful for GP programming. On the other hand, numerical routines code a function (to be integrated, summed, etc.) with two parameters named

    GEN (*eval)(void*,GEN)
    void *E;  \\ context: eval(E, x) must evaluate your function at x.

see the Libpari manual for details.

@3Numerical integration. Starting with version 2.2.9 the ``double exponential'' univariate integration method is implemented in intnum and its variants. Romberg integration is still available under the name intnumromb, but superseded. It is possible to compute numerically integrals to thousands of decimal places in reasonable time, as long as the integrand is regular. It is also reasonable to compute numerically integrals in several variables, although more than two becomes lengthy. The integration domain may be non-compact, and the integrand may have reasonable singularities at endpoints. To use intnum, you must split the integral into a sum of subintegrals where the function has no singularities except at the endpoints. Polynomials in logarithms are not considered singular, and neglecting these logs, singularities are assumed to be algebraic (asymptotic to C(x-a)^{-alpha} for some alpha > -1 when x is close to a), or to correspond to simple discontinuities of some (higher) derivative of the function. For instance, the point 0 is a singularity of {abs}(x).

See also the discrete summation methods below, sharing the prefix sum.

derivnum(X = a,expr)

Numerical derivation of expr with respect to X at X = a.

  ? derivnum(x=0,sin(exp(x))) - cos(1)
  %1 = -1.262177448 E-29

A clumsier approach, which would not work in library mode, is

  ? f(x) = sin(exp(x))
  ? f'(0) - cos(1)
  %1 = -1.262177448 E-29

When a is a power series, compute derivnum(t = a,f) as f'(a) = (f(a))'/a'.

The library syntax is derivnum(void *E, GEN (*eval)(void*,GEN), GEN a, long prec). Also available is GEN derivfun(void *E, GEN (*eval)(void *, GEN), GEN a, long prec), which also allows power series for a.

intcirc(X = a,R,expr,{tab})

Numerical integration of (2iPi)^{-1}expr with respect to X on the circle |X-a |= R. In other words, when expr is a meromorphic function, sum of the residues in the corresponding disk. tab is as in intnum, except that if computed with intnuminit it should be with the endpoints [-1, 1].

  ? \p105
  ? intcirc(s=1, 0.5, zeta(s)) - 1
  %1 = -2.398082982 E-104 - 7.94487211 E-107*I

The library syntax is intcirc(void *E, GEN (*eval)(void*,GEN), GEN a,GEN R,GEN tab, long prec).

intfouriercos(X = a,b,z,expr,{tab})

Numerical integration of expr(X) cos (2Pi zX) from a to b, in other words Fourier cosine transform (from a to b) of the function represented by expr. Endpoints a and b are coded as in intnum, and are not necessarily at infinity, but if they are, oscillations (i.e. [[+-1],alpha I]) are forbidden.

The library syntax is intfouriercos(void *E, GEN (*eval)(void*,GEN), GEN a, GEN b, GEN z, GEN tab, long prec).

intfourierexp(X = a,b,z,expr,{tab})

Numerical integration of expr(X) exp (-2iPi zX) from a to b, in other words Fourier transform (from a to b) of the function represented by expr. Note the minus sign. Endpoints a and b are coded as in intnum, and are not necessarily at infinity but if they are, oscillations (i.e. [[+-1],alpha I]) are forbidden.

The library syntax is intfourierexp(void *E, GEN (*eval)(void*,GEN), GEN a, GEN b, GEN z, GEN tab, long prec).

intfouriersin(X = a,b,z,expr,{tab})

Numerical integration of expr(X) sin (2Pi zX) from a to b, in other words Fourier sine transform (from a to b) of the function represented by expr. Endpoints a and b are coded as in intnum, and are not necessarily at infinity but if they are, oscillations (i.e. [[+-1],alpha I]) are forbidden.

The library syntax is intfouriersin(void *E, GEN (*eval)(void*,GEN), GEN a, GEN b, GEN z, GEN tab, long prec).

intfuncinit(X = a,b,expr,{flag = 0},{m = 0})

Initialize tables for use with integral transforms such as intmellininv, etc., where a and b are coded as in intnum, expr is the function s(X) to which the integral transform is to be applied (which will multiply the weights of integration) and m is as in intnuminit. If flag is nonzero, assumes that s(-X) = \overline{s(X)}, which makes the computation twice as fast. See intmellininvshort for examples of the use of this function, which is particularly useful when the function s(X) is lengthy to compute, such as a gamma product.

The library syntax is intfuncinit(void *E, GEN (*eval)(void*,GEN), GEN a,GEN b,long m, long flag, long prec). Note that the order of m and flag are reversed compared to the GP syntax.

intlaplaceinv(X = sig,z,expr,{tab})

Numerical integration of (2iPi)^{-1}expr(X)e^{Xz} with respect to X on the line Re (X) = sig. In other words, inverse Laplace transform of the function corresponding to expr at the value z.

sig is coded as follows. Either it is a real number sigma, equal to the abscissa of integration, and then the integrand is assumed to be slowly decreasing when the imaginary part of the variable tends to +- oo . Or it is a two component vector [sigma,alpha], where sigma is as before, and either alpha = 0 for slowly decreasing functions, or alpha > 0 for functions decreasing like exp (-alpha t). Note that it is not necessary to choose the exact value of alpha. tab is as in intnum.

It is often a good idea to use this function with a value of m one or two higher than the one chosen by default (which can be viewed thanks to the function intnumstep), or to increase the abscissa of integration sigma. For example:

  ? \p 105
  ? intlaplaceinv(x=2, 1, 1/x) - 1
  time = 350 ms.
  %1 = 7.37... E-55 + 1.72... E-54*I \\ not so good
  ? m = intnumstep()
  %2 = 7
  ? intlaplaceinv(x=2, 1, 1/x, m+1) - 1
  time = 700 ms.
  %3 = 3.95... E-97 + 4.76... E-98*I \\ better
  ? intlaplaceinv(x=2, 1, 1/x, m+2) - 1
  time = 1400 ms.
  %4 = 0.E-105 + 0.E-106*I \\ perfect but slow.
  ? intlaplaceinv(x=5, 1, 1/x) - 1
  time = 340 ms.
  %5 = -5.98... E-85 + 8.08... E-85*I \\ better than %1
  ? intlaplaceinv(x=5, 1, 1/x, m+1) - 1
  time = 680 ms.
  %6 = -1.09... E-106 + 0.E-104*I \\ perfect, fast.
  ? intlaplaceinv(x=10, 1, 1/x) - 1
  time = 340 ms.
  %7 = -4.36... E-106 + 0.E-102*I \\ perfect, fastest, but why sig = 10?
  ? intlaplaceinv(x=100, 1, 1/x) - 1
  time = 330 ms.
  %7 = 1.07... E-72 + 3.2... E-72*I \\ too far now...

The library syntax is intlaplaceinv(void *E, GEN (*eval)(void*,GEN), GEN sig,GEN z, GEN tab, long prec).

intmellininv(X = sig,z,expr,{tab})

Numerical integration of (2iPi)^{-1}expr(X)z^{-X} with respect to X on the line Re (X) = sig, in other words, inverse Mellin transform of the function corresponding to expr at the value z.

sig is coded as follows. Either it is a real number sigma, equal to the abscissa of integration, and then the integrated is assumed to decrease exponentially fast, of the order of exp (-t) when the imaginary part of the variable tends to +- oo . Or it is a two component vector [sigma,alpha], where sigma is as before, and either alpha = 0 for slowly decreasing functions, or alpha > 0 for functions decreasing like exp (-alpha t), such as gamma products. Note that it is not necessary to choose the exact value of alpha, and that alpha = 1 (equivalent to sig alone) is usually sufficient. tab is as in intnum.

As all similar functions, this function is provided for the convenience of the user, who could use intnum directly. However it is in general better to use intmellininvshort.

  ? \p 105
  ? intmellininv(s=2,4, gamma(s)^3);
  time = 1,190 ms. \\ reasonable.
  ? \p 308
  ? intmellininv(s=2,4, gamma(s)^3);
  time = 51,300 ms. \\ slow because of F<Gamma>(s)^3.

The library syntax is intmellininv(void *E, GEN (*eval)(void*,GEN), GEN sig, GEN z, GEN tab, long prec).

intmellininvshort(sig,z,tab)

Numerical integration of (2iPi)^{-1}s(X)z^{-X} with respect to X on the line Re (X) = sig. In other words, inverse Mellin transform of s(X) at the value z. Here s(X) is implicitly contained in tab in intfuncinit format, typically

  tab = intfuncinit(T = [-1], [1], s(sig + I*T))

or similar commands. Take the example of the inverse Mellin transform of Gamma(s)^3 given in intmellininv:

  ? \p 105
  ? oo = [1]; \\ for clarity
  ? A = intmellininv(s=2,4, gamma(s)^3);
  time = 2,500 ms. \\ not too fast because of F<Gamma>(s)^3.
  \\  function of real type, decreasing as  F<exp> (-3F<Pi>/2.|t|)
  ? tab = intfuncinit(t=[-oo, 3*Pi/2],[oo, 3*Pi/2], gamma(2+I*t)^3, 1);
  time = 1,370 ms.
  ? intmellininvshort(2,4, tab) - A
  time = 50 ms.
  %4 = -1.26... - 3.25...E-109*I \\ 50 times faster than C<A> and perfect.
  ? tab2 = intfuncinit(t=-oo, oo, gamma(2+I*t)^3, 1);
  ? intmellininvshort(2,4, tab2)
  %6 = -1.2...E-42 - 3.2...E-109*I  \\ 63 digits lost

In the computation of tab, it was not essential to include the exact exponential decrease of Gamma(2+it)^3. But as the last example shows, a rough indication must be given, otherwise slow decrease is assumed, resulting in catastrophic loss of accuracy.

The library syntax is GEN intmellininvshort(GEN sig, GEN z, GEN tab, long prec).

intnum(X = a,b,expr,{tab})

Numerical integration of expr on ]a,b[ with respect to X. The integrand may have values belonging to a vector space over the real numbers; in particular, it can be complex-valued or vector-valued. But it is assumed that the function is regular on ]a,b[. If the endpoints a and b are finite and the function is regular there, the situation is simple:

  ? intnum(x = 0,1, x^2)
  %1 = 0.3333333333333333333333333333
  ? intnum(x = 0,Pi/2, [cos(x), sin(x)])
  %2 = [1.000000000000000000000000000, 1.000000000000000000000000000]

An endpoint equal to +- oo is coded as the single-component vector [+-1]. You are welcome to set, e.g oo = [1] or INFINITY = [1], then using +oo, -oo, -INFINITY, etc. will have the expected behavior.

  ? oo = [1];  \\ for clarity
  ? intnum(x = 1,+oo, 1/x^2)
  %2 = 1.000000000000000000000000000

In basic usage, it is assumed that the function does not decrease exponentially fast at infinity:

  ? intnum(x=0,+oo, exp(-x))
    ***   at top-level: intnum(x=0,+oo,exp(-
    ***                 ^--------------------
    *** exp: exponent (expo) overflow

We shall see in a moment how to avoid the last problem, after describing the last argument tab, which is both optional and technical. The routine uses weights, which are mostly independent of the function being integrated, evaluated at many sampling points. If tab is

@3* a positive integer m, we use 2^m sampling points, hopefully increasing accuracy. But note that the running time is roughly proportional to 2^m. One may try consecutive values of m until they give the same value up to an accepted error. If tab is omitted, the algorithm guesses a reasonable value for m depending on the current precision only, which should be sufficient for regular functions. That value may be obtained from intnumstep, and increased in case of difficulties.

@3* a set of integration tables as output by intnuminit, they are used directly. This is useful if several integrations of the same type are performed (on the same kind of interval and functions, for a given accuracy), in particular for multivariate integrals, since we then skip expensive precomputations.

@3Specifying the behavior at endpoints. This is done as follows. An endpoint a is either given as such (a scalar, real or complex, or [+-1] for +- oo ), or as a two component vector [a,alpha], to indicate the behavior of the integrand in a neighborhood of a.

If a is finite, the code [a,alpha] means the function has a singularity of the form (x-a)^{alpha}, up to logarithms. (If alpha >= 0, we only assume the function is regular, which is the default assumption.) If a wrong singularity exponent is used, the result will lose a catastrophic number of decimals:

  ? intnum(x=0, 1, x^(-1/2))         \\ assume x^{-1/2} is regular at 0
  %1 = 1.999999999999999999990291881
  ? intnum(x=[0,-1/2], 1, x^(-1/2))  \\ no, it's not
  %2 = 2.000000000000000000000000000
  ? intnum(x=[0,-1/10], 1, x^(-1/2))
  %3 = 1.999999999999999999999946438 \\ using a wrong exponent is bad

If a is +- oo , which is coded as [+- 1], the situation is more complicated, and [[+-1],alpha] means:

@3* alpha = 0 (or no alpha at all, i.e. simply [+-1]) assumes that the integrand tends to zero, but not exponentially fast, and not oscillating such as sin (x)/x.

@3* alpha > 0 assumes that the function tends to zero exponentially fast approximately as exp (-alpha x). This includes oscillating but quickly decreasing functions such as exp (-x) sin (x).

  ? oo = [1];
  ? intnum(x=0, +oo, exp(-2*x))
    ***   at top-level: intnum(x=0,+oo,exp(-
    ***                 ^--------------------
    *** exp: exponent (expo) overflow
  ? intnum(x=0, [+oo, 2], exp(-2*x))
  %1 = 0.5000000000000000000000000000 \\ OK!
  ? intnum(x=0, [+oo, 4], exp(-2*x))
  %2 = 0.4999999999999999999961990984 \\ wrong exponent  ==E<gt>  imprecise result
  ? intnum(x=0, [+oo, 20], exp(-2*x))
  %2 = 0.4999524997739071283804510227 \\ disaster

@3* alpha < -1 assumes that the function tends to 0 slowly, like x^{alpha}. Here it is essential to give the correct alpha, if possible, but on the other hand alpha <= -2 is equivalent to alpha = 0, in other words to no alpha at all.

The last two codes are reserved for oscillating functions. Let k > 0 real, and g(x) a non-oscillating function tending slowly to 0 (e.g. like a negative power of x), then

@3* alpha = k * I assumes that the function behaves like cos (kx)g(x).

@3* alpha = -k* I assumes that the function behaves like sin (kx)g(x).

@3Here it is critical to give the exact value of k. If the oscillating part is not a pure sine or cosine, one must expand it into a Fourier series, use the above codings, and sum the resulting contributions. Otherwise you will get nonsense. Note that cos (kx), and similarly sin (kx), means that very function, and not a translated version such as cos (kx+a).

@3Note. If f(x) = cos (kx)g(x) where g(x) tends to zero exponentially fast as exp (-alpha x), it is up to the user to choose between [[+-1],alpha] and [[+-1],k* I], but a good rule of thumb is that if the oscillations are much weaker than the exponential decrease, choose [[+-1],alpha], otherwise choose [[+-1],k* I], although the latter can reasonably be used in all cases, while the former cannot. To take a specific example, in the inverse Mellin transform, the integrand is almost always a product of an exponentially decreasing and an oscillating factor. If we choose the oscillating type of integral we perhaps obtain the best results, at the expense of having to recompute our functions for a different value of the variable z giving the transform, preventing us to use a function such as intmellininvshort. On the other hand using the exponential type of integral, we obtain less accurate results, but we skip expensive recomputations. See intmellininvshort and intfuncinit for more explanations.

We shall now see many examples to get a feeling for what the various parameters achieve. All examples below assume precision is set to 105 decimal digits. We first type

  ? \p 105
  ? oo = [1]  \\ for clarity

@3Apparent singularities. Even if the function f(x) represented by expr has no singularities, it may be important to define the function differently near special points. For instance, if f(x) = 1 /( exp (x)-1) - exp (-x)/x, then int_0^ oo f(x)dx = gamma, Euler's constant Euler. But

  ? f(x) = 1/(exp(x)-1) - exp(-x)/x
  ? intnum(x = 0, [oo,1],  f(x)) - Euler
  %1 = 6.00... E-67

thus only correct to 67 decimal digits. This is because close to 0 the function f is computed with an enormous loss of accuracy. A better solution is

  ? f(x) = 1/(exp(x)-1)-exp(-x)/x
  ? F = truncate( f(t + O(t^7)) ); \\ expansion around t = 0
  ? g(x) = if (x > 1e-18, f(x), subst(F,t,x))  \\ note that 6.18 > 105
  ? intnum(x = 0, [oo,1],  g(x)) - Euler
  %2 = 0.E-106 \\ perfect

It is up to the user to determine constants such as the 10^{-18} and 7 used above.

@3True singularities. With true singularities the result is worse. For instance

  ? intnum(x = 0, 1,  1/sqrt(x)) - 2
  %1 = -1.92... E-59 \\ only 59 correct decimals
  ? intnum(x = [0,-1/2], 1,  1/sqrt(x)) - 2
  %2 = 0.E-105 \\ better

@3Oscillating functions.

  ? intnum(x = 0, oo, sin(x) / x) - Pi/2
  %1 = 20.78.. \\ nonsense
  ? intnum(x = 0, [oo,1], sin(x)/x) - Pi/2
  %2 = 0.004.. \\ bad
  ? intnum(x = 0, [oo,-I], sin(x)/x) - Pi/2
  %3 = 0.E-105 \\ perfect
  ? intnum(x = 0, [oo,-I], sin(2*x)/x) - Pi/2  \\ oops, wrong k
  %4 = 0.07...
  ? intnum(x = 0, [oo,-2*I], sin(2*x)/x) - Pi/2
  %5 = 0.E-105 \\ perfect
  ? intnum(x = 0, [oo,-I], sin(x)^3/x) - Pi/4
  %6 = 0.0092... \\ bad
  ? sin(x)^3 - (3*sin(x)-sin(3*x))/4
  %7 = O(x^17)

@3 We may use the above linearization and compute two oscillating integrals with ``infinite endpoints'' [oo, -I] and [oo, -3*I] respectively, or notice the obvious change of variable, and reduce to the single integral (1/2)int_0^ oo sin (x)/xdx. We finish with some more complicated examples:

  ? intnum(x = 0, [oo,-I], (1-cos(x))/x^2) - Pi/2
  %1 = -0.0004... \\ bad
  ? intnum(x = 0, 1, (1-cos(x))/x^2) \
  + intnum(x = 1, oo, 1/x^2) - intnum(x = 1, [oo,I], cos(x)/x^2) - Pi/2
  %2 = -2.18... E-106 \\ OK
  ? intnum(x = 0, [oo, 1], sin(x)^3*exp(-x)) - 0.3
  %3 = 5.45... E-107 \\ OK
  ? intnum(x = 0, [oo,-I], sin(x)^3*exp(-x)) - 0.3
  %4 = -1.33... E-89 \\ lost 16 decimals. Try higher m:
  ? m = intnumstep()
  %5 = 7 \\ the value of m actually used above.
  ? tab = intnuminit(0,[oo,-I], m+1); \\ try m one higher.
  ? intnum(x = 0, oo, sin(x)^3*exp(-x), tab) - 0.3
  %6 = 5.45... E-107 \\ OK this time.

@3Warning. Like sumalt, intnum often assigns a reasonable value to diverging integrals. Use these values at your own risk! For example:

  ? intnum(x = 0, [oo, -I], x^2*sin(x))
  %1 = -2.0000000000...

Note the formula

   int_0^ oo sin (x)/x^sdx = cos (Pi s/2) Gamma(1-s) ,

a priori valid only for 0 < Re (s) < 2, but the right hand side provides an analytic continuation which may be evaluated at s = -2...

@3Multivariate integration. Using successive univariate integration with respect to different formal parameters, it is immediate to do naive multivariate integration. But it is important to use a suitable intnuminit to precompute data for the internal integrations at least!

For example, to compute the double integral on the unit disc x^2+y^2 <= 1 of the function x^2+y^2, we can write

  ? tab = intnuminit(-1,1);
  ? intnum(x=-1,1, intnum(y=-sqrt(1-x^2),sqrt(1-x^2), x^2+y^2, tab), tab)

The first tab is essential, the second optional. Compare:

  ? tab = intnuminit(-1,1);
  time = 30 ms.
  ? intnum(x=-1,1, intnum(y=-sqrt(1-x^2),sqrt(1-x^2), x^2+y^2));
  time = 54,410 ms. \\ slow
  ? intnum(x=-1,1, intnum(y=-sqrt(1-x^2),sqrt(1-x^2), x^2+y^2, tab), tab);
  time = 7,210 ms.  \\ faster

However, the intnuminit program is usually pessimistic when it comes to choosing the integration step 2^{-m}. It is often possible to improve the speed by trial and error. Continuing the above example:

  ? test(M) =
  {
  tab = intnuminit(-1,1, M);
  intnum(x=-1,1, intnum(y=-sqrt(1-x^2),sqrt(1-x^2), x^2+y^2,tab), tab) - Pi/2
  }
  ? m = intnumstep() \\ what value of m did it take?
  %1 = 7
  ? test(m - 1)
  time = 1,790 ms.
  %2 = -2.05... E-104 \\ 4 = 2^2 times faster and still OK.
  ? test(m - 2)
  time = 430 ms.
  %3 = -1.11... E-104 \\ 16 = 2^4 times faster and still OK.
  ? test(m - 3)
  time = 120 ms.
  %3 = -7.23... E-60 \\ 64 = 2^6 times faster, lost 45 decimals.

The library syntax is intnum(void *E, GEN (*eval)(void*,GEN), GEN a,GEN b,GEN tab, long prec), where an omitted tab is coded as NULL.

intnuminit(a,b,{m = 0})

Initialize tables for integration from a to b, where a and b are coded as in intnum. Only the compactness, the possible existence of singularities, the speed of decrease or the oscillations at infinity are taken into account, and not the values. For instance intnuminit(-1,1) is equivalent to intnuminit(0,Pi), and intnuminit([0,-1/2],[1]) is equivalent to intnuminit([-1],[-1,-1/2]). If m is not given, it is computed according to the current precision. Otherwise the integration step is 1/2^m. Reasonable values of m are m = 6 or m = 7 for 100 decimal digits, and m = 9 for 1000 decimal digits.

The result is technical, but in some cases it is useful to know the output. Let x = phi(t) be the change of variable which is used. tab[1] contains the integer m as above, either given by the user or computed from the default precision, and can be recomputed directly using the function intnumstep. tab[2] and tab[3] contain respectively the abscissa and weight corresponding to t = 0 (phi(0) and phi'(0)). tab[4] and tab[5] contain the abscissas and weights corresponding to positive t = nh for 1 <= n <= N and h = 1/2^m (phi(nh) and phi'(nh)). Finally tab[6] and tab[7] contain either the abscissas and weights corresponding to negative t = nh for -N <= n <= -1, or may be empty (but not always) if phi(t) is an odd function (implicitly we would have tab[6] = -tab[4] and tab[7] = tab[5]).

The library syntax is GEN intnuminit(GEN a, GEN b, long m, long prec).

intnuminitgen(t,a,b,ph,{m = 0},{flag = 0})

Initialize tables for integrations from a to b using abscissas ph(t) and weights ph'(t). Note that there is no equal sign after the variable name t since t always goes from - oo to + oo , but it is ph(t) which goes from a to b, and this is not checked. If flag = 1 or 2, multiply the reserved table length by 4^{flag}, to avoid corresponding error.

The library syntax is intnuminitgen(void *E, GEN (*eval)(void*,GEN), GEN a, GEN b, long m, long flag, long prec)

intnumromb(X = a,b,expr,{flag = 0})

Numerical integration of expr (smooth in ]a,b[), with respect to X. Suitable for low accuracy; if expr is very regular (e.g. analytic in a large region) and high accuracy is desired, try intnum first.

Set flag = 0 (or omit it altogether) when a and b are not too large, the function is smooth, and can be evaluated exactly everywhere on the interval [a,b].

If flag = 1, uses a general driver routine for doing numerical integration, making no particular assumption (slow).

flag = 2 is tailored for being used when a or b are infinite. One must have ab > 0, and in fact if for example b = + oo , then it is preferable to have a as large as possible, at least a >= 1.

If flag = 3, the function is allowed to be undefined (but continuous) at a or b, for example the function sin (x)/x at x = 0.

The user should not require too much accuracy: 18 or 28 decimal digits is OK, but not much more. In addition, analytical cleanup of the integral must have been done: there must be no singularities in the interval or at the boundaries. In practice this can be accomplished with a simple change of variable. Furthermore, for improper integrals, where one or both of the limits of integration are plus or minus infinity, the function must decrease sufficiently rapidly at infinity. This can often be accomplished through integration by parts. Finally, the function to be integrated should not be very small (compared to the current precision) on the entire interval. This can of course be accomplished by just multiplying by an appropriate constant.

Note that infinity can be represented with essentially no loss of accuracy by an appropriate huge number. However beware of real underflow when dealing with rapidly decreasing functions. For example, in order to compute the int_0^ oo e^{-x^2}dx to 28 decimal digits, then one can set infinity equal to 10 for example, and certainly not to 1e1000.

The library syntax is intnumromb(void *E, GEN (*eval)(void*,GEN), GEN a, GEN b, long flag, long prec), where eval(x, E) returns the value of the function at x. You may store any additional information required by eval in E, or set it to NULL.

intnumstep()

Give the value of m used in all the intnum and sumnum programs, hence such that the integration step is equal to 1/2^m.

The library syntax is long intnumstep(long prec).

prod(X = a,b,expr,{x = 1})

Product of expression expr, initialized at x, the formal parameter X going from a to b. As for sum, the main purpose of the initialization parameter x is to force the type of the operations being performed. For example if it is set equal to the integer 1, operations will start being done exactly. If it is set equal to the real 1., they will be done using real numbers having the default precision. If it is set equal to the power series 1+O(X^k) for a certain k, they will be done using power series of precision at most k. These are the three most common initializations.

@3As an extreme example, compare

  ? prod(i=1, 100, 1 - X^i);  \\ this has degree 5050 !!
  time = 128 ms.
  ? prod(i=1, 100, 1 - X^i, 1 + O(X^101))
  time = 8 ms.
  %2 = 1 - X - X^2 + X^5 + X^7 - X^12 - X^15 + X^22 + X^26 - X^35 - X^40 + \
  X^51 + X^57 - X^70 - X^77 + X^92 + X^100 + O(X^101)

Of course, in this specific case, it is faster to use eta, which is computed using Euler's formula.

  ? prod(i=1, 1000, 1 - X^i, 1 + O(X^1001));
  time = 589 ms.
  ? \ps1000
  seriesprecision = 1000 significant terms
  ? eta(X) - %
  time = 8ms.
  %4 = O(X^1001)

The library syntax is produit(GEN a, GEN b, char *expr, GEN x).

prodeuler(X = a,b,expr)

Product of expression expr, initialized at 1. (i.e. to a real number equal to 1 to the current realprecision), the formal parameter X ranging over the prime numbers between a and b.

The library syntax is prodeuler(void *E, GEN (*eval)(void*,GEN), GEN a,GEN b, long prec).

prodinf(X = a,expr,{flag = 0})

infinite product of expression expr, the formal parameter X starting at a. The evaluation stops when the relative error of the expression minus 1 is less than the default precision. In particular, non-convergent products result in infinite loops. The expressions must always evaluate to an element of C.

If flag = 1, do the product of the (1+expr) instead.

The library syntax is prodinf(void *E, GEN (*eval)(void*,GEN), GEN a, long prec) (flag = 0), or prodinf1 with the same arguments (flag = 1).

solve(X = a,b,expr)

Find a real root of expression expr between a and b, under the condition expr(X = a) * expr(X = b) <= 0. (You will get an error message roots must be bracketed in solve if this does not hold.) This routine uses Brent's method and can fail miserably if expr is not defined in the whole of [a,b] (try solve(x = 1, 2, tan(x))).

The library syntax is zbrent(void *E,GEN (*eval)(void*,GEN),GEN a,GEN b,long prec).

sum(X = a,b,expr,{x = 0})

Sum of expression expr, initialized at x, the formal parameter going from a to b. As for prod, the initialization parameter x may be given to force the type of the operations being performed.

@3As an extreme example, compare

  ? sum(i=1, 10^4, 1/i); \\ rational number: denominator has 4345 digits.
  time = 236 ms.
  ? sum(i=1, 5000, 1/i, 0.)
  time = 8 ms.
  %2 = 9.787606036044382264178477904

The library syntax is somme(GEN a, GEN b, char *expr, GEN x).

sumalt(X = a,expr,{flag = 0})

Numerical summation of the series expr, which should be an alternating series, the formal variable X starting at a. Use an algorithm of Cohen, Villegas and Zagier (Experiment. Math. 9 (2000), no. 1, 3--12).

If flag = 1, use a variant with slightly different polynomials. Sometimes faster.

The routine is heuristic and a rigorous proof assumes that the values of expr are the moments of a positive measure on [0,1]. Divergent alternating series can sometimes be summed by this method, as well as series which are not exactly alternating (see for example Label se:user_defined). It should be used to try and guess the value of an infinite sum. (However, see the example at the end of Label se:userfundef.)

If the series already converges geometrically, suminf is often a better choice:

  ? \p28
  ? sumalt(i = 1, -(-1)^i / i)  - log(2)
  time = 0 ms.
  %1 = -2.524354897 E-29
  ? suminf(i = 1, -(-1)^i / i)   \\ Had to hit < C-C > 
    ***   at top-level: suminf(i=1,-(-1)^i/i)
    ***                                ^------
    *** suminf: user interrupt after 10min, 20,100 ms.
  ? \p1000
  ? sumalt(i = 1, -(-1)^i / i)  - log(2)
  time = 90 ms.
  %2 = 4.459597722 E-1002
  ? sumalt(i = 0, (-1)^i / i!) - exp(-1)
  time = 670 ms.
  %3 = -4.03698781490633483156497361352190615794353338591897830587 E-944
  ? suminf(i = 0, (-1)^i / i!) - exp(-1)
  time = 110 ms.
  %4 = -8.39147638 E-1000   \\  faster and more accurate

The library syntax is sumalt(void *E, GEN (*eval)(void*,GEN),GEN a,long prec). Also available is sumalt2 with the same arguments (flag = 1).

sumdiv(n,X,expr)

Sum of expression expr over the positive divisors of n. This function is a trivial wrapper essentially equivalent to

    D = divisors(n);
    for (i = 1, #D, X = D[i]; eval(expr))

@3(except that X is lexically scoped to the sumdiv loop). If expr is a multiplicative function, use sumdivmult.

sumdivmult(n,d,expr)

Sum of multiplicative expression expr over the positive divisors d of n. Assume that expr evaluates to f(d) where f is multiplicative: f(1) = 1 and f(ab) = f(a)f(b) for coprime a and b.

suminf(X = a,expr)

infinite sum of expression expr, the formal parameter X starting at a. The evaluation stops when the relative error of the expression is less than the default precision for 3 consecutive evaluations. The expressions must always evaluate to a complex number.

If the series converges slowly, make sure realprecision is low (even 28 digits may be too much). In this case, if the series is alternating or the terms have a constant sign, sumalt and sumpos should be used instead.

  ? \p28
  ? suminf(i = 1, -(-1)^i / i)   \\ Had to hit < C-C > 
    ***   at top-level: suminf(i=1,-(-1)^i/i)
    ***                                ^------
    *** suminf: user interrupt after 10min, 20,100 ms.
  ? sumalt(i = 1, -(-1)^i / i) - log(2)
  time = 0 ms.
  %1 = -2.524354897 E-29

The library syntax is suminf(void *E, GEN (*eval)(void*,GEN), GEN a, long prec).

sumnum(X = a,sig,expr,{tab},{flag = 0})

Numerical summation of expr, the variable X taking integer values from ceiling of a to + oo , where expr is assumed to be a holomorphic function f(X) for Re (X) >= sigma.

The parameter sigma belongs to R is coded in the argument sig as follows: it is either

@3* a real number sigma. Then the function f is assumed to decrease at least as 1/X^2 at infinity, but not exponentially;

@3* a two-component vector [sigma,alpha], where sigma is as before, alpha < -1. The function f is assumed to decrease like X^{alpha}. In particular, alpha <= -2 is equivalent to no alpha at all.

@3* a two-component vector [sigma,alpha], where sigma is as before, alpha > 0. The function f is assumed to decrease like exp (-alpha X). In this case it is essential that alpha be exactly the rate of exponential decrease, and it is usually a good idea to increase the default value of m used for the integration step. In practice, if the function is exponentially decreasing sumnum is slower and less accurate than sumpos or suminf, so should not be used.

The function uses the intnum routines and integration on the line Re (s) = sigma. The optional argument tab is as in intnum, except it must be initialized with sumnuminit instead of intnuminit.

When tab is not precomputed, sumnum can be slower than sumpos, when the latter is applicable. It is in general faster for slowly decreasing functions.

Finally, if flag is nonzero, we assume that the function f to be summed is of real type, i.e. satisfies \overline{f(z)} = f(\overline{z}), which speeds up the computation.

  ? \p 308
  ? a = sumpos(n=1, 1/(n^3+n+1));
  time = 1,410 ms.
  ? tab = sumnuminit(2);
  time = 1,620 ms. \\ slower but done once and for all.
  ? b = sumnum(n=1, 2, 1/(n^3+n+1), tab);
  time = 460 ms. \\ 3 times as fast as C<sumpos>
  ? a - b
  %4 = -1.0... E-306 + 0.E-320*I \\ perfect.
  ? sumnum(n=1, 2, 1/(n^3+n+1), tab, 1) - a; \\ function of real type
  time = 240 ms.
  %2 = -1.0... E-306 \\ twice as fast, no imaginary part.
  ? c = sumnum(n=1, 2, 1/(n^2+1), tab, 1);
  time = 170 ms. \\ fast
  ? d = sumpos(n=1, 1 / (n^2+1));
  time = 2,700 ms. \\ slow.
  ? d - c
  time = 0 ms.
  %5 = 1.97... E-306 \\ perfect.

For slowly decreasing function, we must indicate singularities:

  ? \p 308
  ? a = sumnum(n=1, 2, n^(-4/3));
  time = 9,930 ms. \\ slow because of the computation of n^{-4/3}.
  ? a - zeta(4/3)
  time = 110 ms.
  %1 = -2.42... E-107 \\ lost 200 decimals because of singularity at  oo 
  ? b = sumnum(n=1, [2,-4/3], n^(-4/3), /*omitted*/, 1); \\ of real type
  time = 12,210 ms.
  ? b - zeta(4/3)
  %3 = 1.05... E-300 \\ better

Since the complex values of the function are used, beware of determination problems. For instance:

  ? \p 308
  ? tab = sumnuminit([2,-3/2]);
  time = 1,870 ms.
  ? sumnum(n=1,[2,-3/2], 1/(n*sqrt(n)), tab,1) - zeta(3/2)
  time = 690 ms.
  %1 = -1.19... E-305 \\ fast and correct
  ? sumnum(n=1,[2,-3/2], 1/sqrt(n^3), tab,1) - zeta(3/2)
  time = 730 ms.
  %2 = -1.55... \\ nonsense. However
  ? sumnum(n=1,[2,-3/2], 1/n^(3/2), tab,1) - zeta(3/2)
  time = 8,990 ms.
  %3 = -1.19... E-305 \\ perfect, as 1/(n* F<sqrt> {n}) above but much slower

For exponentially decreasing functions, sumnum is given for completeness, but one of suminf or sumpos should always be preferred. If you experiment with such functions and sumnum anyway, indicate the exact rate of decrease and increase m by 1 or 2:

  ? suminf(n=1, 2^(-n)) - 1
  time = 10 ms.
  %1 = -1.11... E-308 \\ fast and perfect
  ? sumpos(n=1, 2^(-n)) - 1
  time = 10 ms.
  %2 = -2.78... E-308 \\ also fast and perfect
  ? sumnum(n=1,2, 2^(-n)) - 1
  %3 = -1.321115060 E320 + 0.E311*I \\ nonsense
  ? sumnum(n=1, [2,log(2)], 2^(-n), /*omitted*/, 1) - 1 \\ of real type
  time = 5,860 ms.
  %4 = -1.5... E-236 \\ slow and lost 70 decimals
  ? m = intnumstep()
  %5 = 9
  ? sumnum(n=1,[2,log(2)], 2^(-n), m+1, 1) - 1
  time = 11,770 ms.
  %6 = -1.9... E-305 \\ now perfect, but slow.

The library syntax is sumnum(void *E, GEN (*eval)(void*,GEN), GEN a,GEN sig,GEN tab,long flag, long prec).

sumnumalt(X = a,sig,expr,{tab},{flag = 0})

Numerical summation of (-1)^Xexpr(X), the variable X taking integer values from ceiling of a to + oo , where expr is assumed to be a holomorphic function for Re (X) >= sig (or sig[1]).

@3Warning. This function uses the intnum routines and is orders of magnitude slower than sumalt. It is only given for completeness and should not be used in practice.

@3Warning 2. The expression expr must not include the (-1)^X coefficient. Thus sumalt(n = a,(-1)^nf(n)) is (approximately) equal to sumnumalt(n = a,sig,f(n)).

sig is coded as in sumnum. However for slowly decreasing functions (where sig is coded as [sigma,alpha] with alpha < -1), it is not really important to indicate alpha. In fact, as for sumalt, the program will often give meaningful results (usually analytic continuations) even for divergent series. On the other hand the exponential decrease must be indicated.

tab is as in intnum, but if used must be initialized with sumnuminit. If flag is nonzero, assumes that the function f to be summed is of real type, i.e. satisfies \overline{f(z)} = f(\overline{z}), and then twice faster when tab is precomputed.

  ? \p 308
  ? tab = sumnuminit(2, /*omitted*/, -1); \\ abscissa F<sigma> = 2, alternating sums.
  time = 1,620 ms. \\ slow, but done once and for all.
  ? a = sumnumalt(n=1, 2, 1/(n^3+n+1), tab, 1);
  time = 230 ms. \\ similar speed to C<sumnum>
  ? b = sumalt(n=1, (-1)^n/(n^3+n+1));
  time = 0 ms. \\ infinitely faster!
  ? a - b
  time = 0 ms.
  %1 = -1.66... E-308 \\ perfect

The library syntax is sumnumalt(void *E, GEN (*eval)(void*,GEN), GEN a, GEN sig, GEN tab, long flag, long prec).

sumnuminit(sig, {m = 0}, {sgn = 1})

Initialize tables for numerical summation using sumnum (with sgn = 1) or sumnumalt (with sgn = -1), sig is the abscissa of integration coded as in sumnum, and m is as in intnuminit.

The library syntax is GEN sumnuminit(GEN sig, long m, long sgn, long prec).

sumpos(X = a,expr,{flag = 0})

Numerical summation of the series expr, which must be a series of terms having the same sign, the formal variable X starting at a. The algorithm used is Van Wijngaarden's trick for converting such a series into an alternating one, then we use sumalt. For regular functions, the function sumnum is in general much faster once the initializations have been made using sumnuminit.

The routine is heuristic and assumes that expr is more or less a decreasing function of X. In particular, the result will be completely wrong if expr is 0 too often. We do not check either that all terms have the same sign. As sumalt, this function should be used to try and guess the value of an infinite sum.

If flag = 1, use slightly different polynomials. Sometimes faster.

The library syntax is sumpos(void *E, GEN (*eval)(void*,GEN),GEN a,long prec). Also available is sumpos2 with the same arguments (flag = 1).


Plotting functions

Although plotting is not even a side purpose of PARI, a number of plotting functions are provided. Moreover, a lot of people FOOTNOTE<<< Among these, special thanks go to Klaus-Peter Nischke who suggested the recursive plotting and forking/resizing stuff the graphical window, and Ilya Zakharevich who rewrote the graphic code from scratch implementing many new primitives (splines, clipping). Nils Skoruppa and Bill Allombert wrote the Qt and fltk graphic drivers respectively. >>> suggested ideas or submitted patches for this section of the code. There are three types of graphic functions.

High-level plotting functions

(all the functions starting with ploth) in which the user has little to do but explain what type of plot he wants, and whose syntax is similar to the one used in the preceding section.

Low-level plotting functions

(called rectplot functions, sharing the prefix plot), where every drawing primitive (point, line, box, etc.) is specified by the user. These low-level functions work as follows. You have at your disposal 16 virtual windows which are filled independently, and can then be physically ORed on a single window at user-defined positions. These windows are numbered from 0 to 15, and must be initialized before being used by the function plotinit, which specifies the height and width of the virtual window (called a rectwindow in the sequel). At all times, a virtual cursor (initialized at [0,0]) is associated to the window, and its current value can be obtained using the function plotcursor.

A number of primitive graphic objects (called rect objects) can then be drawn in these windows, using a default color associated to that window (which can be changed using the plotcolor function) and only the part of the object which is inside the window will be drawn, with the exception of polygons and strings which are drawn entirely. The ones sharing the prefix plotr draw relatively to the current position of the virtual cursor, the others use absolute coordinates. Those having the prefix plotrecth put in the rectwindow a large batch of rect objects corresponding to the output of the related ploth function.

Finally, the actual physical drawing is done using plotdraw. The rectwindows are preserved so that further drawings using the same windows at different positions or different windows can be done without extra work. To erase a window, use plotkill. It is not possible to partially erase a window: erase it completely, initialize it again, then fill it with the graphic objects that you want to keep.

In addition to initializing the window, you may use a scaled window to avoid unnecessary conversions. For this, use plotscale. As long as this function is not called, the scaling is simply the number of pixels, the origin being at the upper left and the y-coordinates going downwards.

Plotting functions are platform independent, but a number of graphical drivers are available for screen output: X11-windows (hence also for GUI's based on X11 such as Openwindows and Motif), and the Qt and FLTK graphical libraries. The physical window opened by plotdraw or any of the ploth* functions is completely separated from gp (technically, a fork is done, and the non-graphical memory is immediately freed in the child process), which means you can go on working in the current gp session, without having to kill the window first. This window can be closed, enlarged or reduced using the standard window manager functions. No zooming procedure is implemented though (yet).

Functions for PostScript output

in the same way that printtex allows you to have a TeX output corresponding to printed results, the functions starting with ps allow you to have PostScript output of the plots. This will not be identical with the screen output, but sufficiently close. Note that you can use PostScript output even if you do not have the plotting routines enabled. The PostScript output is written in a file whose name is derived from the psfile default (./pari.ps if you did not tamper with it). Each time a new PostScript output is asked for, the PostScript output is appended to that file. Hence you probably want to remove this file, or change the value of psfile, in between plots. On the other hand, in this manner, as many plots as desired can be kept in a single file.

Library mode

None of the graphic functions are available within the PARI library, you must be under gp to use them. The reason for that is that you really should not use PARI for heavy-duty graphical work, there are better specialized alternatives around. This whole set of routines was only meant as a convenient, but simple-minded, visual aid. If you really insist on using these in your program (we warned you), the source (plot*.c) should be readable enough for you to achieve something.

plot(X = a,b,expr,{Ymin},{Ymax})

Crude ASCII plot of the function represented by expression expr from a to b, with Y ranging from Ymin to Ymax. If Ymin (resp. Ymax) is not given, the minimum (resp. the maximum) of the computed values of the expression is used instead.

plotbox(w,x2,y2)

Let (x1,y1) be the current position of the virtual cursor. Draw in the rectwindow w the outline of the rectangle which is such that the points (x1,y1) and (x2,y2) are opposite corners. Only the part of the rectangle which is in w is drawn. The virtual cursor does not move.

plotclip(w)

`clips' the content of rectwindow w, i.e remove all parts of the drawing that would not be visible on the screen. Together with plotcopy this function enables you to draw on a scratchpad before committing the part you're interested in to the final picture.

plotcolor(w,c)

Set default color to c in rectwindow w. This is only implemented for the X-windows, fltk and Qt graphing engines. Possible values for c are given by the graphcolormap default, factory setting are

1 = black, 2 = blue, 3 = violetred, 4 = red, 5 = green, 6 = grey, 7 = gainsborough.

but this can be considerably extended.

plotcopy(sourcew,destw,dx,dy,{flag = 0})

Copy the contents of rectwindow sourcew to rectwindow destw with offset (dx,dy). If flag's bit 1 is set, dx and dy express fractions of the size of the current output device, otherwise dx and dy are in pixels. dx and dy are relative positions of northwest corners if other bits of flag vanish, otherwise of: 2: southwest, 4: southeast, 6: northeast corners

plotcursor(w)

Give as a 2-component vector the current (scaled) position of the virtual cursor corresponding to the rectwindow w.

plotdraw(list, {flag = 0})

Physically draw the rectwindows given in list which must be a vector whose number of components is divisible by 3. If list = [w1,x1,y1,w2,x2,y2,...], the windows w1, w2, etc. are physically placed with their upper left corner at physical position (x1,y1), (x2,y2),...respectively, and are then drawn together. Overlapping regions will thus be drawn twice, and the windows are considered transparent. Then display the whole drawing in a special window on your screen. If flag != 0, x1, y1 etc. express fractions of the size of the current output device

ploth(X = a,b,expr,{flags = 0},{n = 0})

High precision plot of the function y = f(x) represented by the expression expr, x going from a to b. This opens a specific window (which is killed whenever you click on it), and returns a four-component vector giving the coordinates of the bounding box in the form [xmin,xmax,ymin,ymax].

@3Important note. ploth may evaluate expr thousands of times; given the relatively low resolution of plotting devices, few significant digits of the result will be meaningful. Hence you should keep the current precision to a minimum (e.g. 9) before calling this function.

n specifies the number of reference point on the graph, where a value of 0 means we use the hardwired default values (1000 for general plot, 1500 for parametric plot, and 8 for recursive plot).

If no flag is given, expr is either a scalar expression f(X), in which case the plane curve y = f(X) will be drawn, or a vector [f_1(X),...,f_k(X)], and then all the curves y = f_i(X) will be drawn in the same window.

@3The binary digits of flag mean:

@3* 1 = Parametric: parametric plot. Here expr must be a vector with an even number of components. Successive pairs are then understood as the parametric coordinates of a plane curve. Each of these are then drawn.

For instance:

  ploth(X=0,2*Pi,[sin(X),cos(X)], "Parametric")
  ploth(X=0,2*Pi,[sin(X),cos(X)])
  ploth(X=0,2*Pi,[X,X,sin(X),cos(X)], "Parametric")

@3draw successively a circle, two entwined sinusoidal curves and a circle cut by the line y = x.

@3* 2 = Recursive: recursive plot. If this flag is set, only one curve can be drawn at a time, i.e. expr must be either a two-component vector (for a single parametric curve, and the parametric flag has to be set), or a scalar function. The idea is to choose pairs of successive reference points, and if their middle point is not too far away from the segment joining them, draw this as a local approximation to the curve. Otherwise, add the middle point to the reference points. This is fast, and usually more precise than usual plot. Compare the results of

  ploth(X=-1,1, sin(1/X), "Recursive")
  ploth(X=-1,1, sin(1/X))

for instance. But beware that if you are extremely unlucky, or choose too few reference points, you may draw some nice polygon bearing little resemblance to the original curve. For instance you should never plot recursively an odd function in a symmetric interval around 0. Try

  ploth(x = -20, 20, sin(x), "Recursive")

to see why. Hence, it's usually a good idea to try and plot the same curve with slightly different parameters.

The other values toggle various display options:

@3* 4 = no_Rescale: do not rescale plot according to the computed extrema. This is used in conjunction with plotscale when graphing multiple functions on a rectwindow (as a plotrecth call):

    s = plothsizes();
    plotinit(0, s[2]-1, s[2]-1);
    plotscale(0, -1,1, -1,1);
    plotrecth(0, t=0,2*Pi, [cos(t),sin(t)], "Parametric|no_Rescale")
    plotdraw([0, -1,1]);

This way we get a proper circle instead of the distorted ellipse produced by

    ploth(t=0,2*Pi, [cos(t),sin(t)], "Parametric")

@3* 8 = no_X_axis: do not print the x-axis.

@3* 16 = no_Y_axis: do not print the y-axis.

@3* 32 = no_Frame: do not print frame.

@3* 64 = no_Lines: only plot reference points, do not join them.

@3* 128 = Points_too: plot both lines and points.

@3* 256 = Splines: use splines to interpolate the points.

@3* 512 = no_X_ticks: plot no x-ticks.

@3* 1024 = no_Y_ticks: plot no y-ticks.

@3* 2048 = Same_ticks: plot all ticks with the same length.

@3* 4096 = Complex: is a parametric plot but where each member of expr is considered a complex number encoding the two coordinates of a point. For instance:

  ploth(X=0,2*Pi,exp(I*X), "Complex")
  ploth(X=0,2*Pi,[(1+I)*X,exp(I*X)], "Complex")

@3will draw respectively a circle and a circle cut by the line y = x.

plothraw(listx,listy,{flag = 0})

Given listx and listy two vectors of equal length, plots (in high precision) the points whose (x,y)-coordinates are given in listx and listy. Automatic positioning and scaling is done, but with the same scaling factor on x and y. If flag is 1, join points, other non-0 flags toggle display options and should be combinations of bits 2^k, k >= 3 as in ploth.

plothsizes({flag = 0})

Return data corresponding to the output window in the form of a 6-component vector: window width and height, sizes for ticks in horizontal and vertical directions (this is intended for the gnuplot interface and is currently not significant), width and height of characters.

If flag = 0, sizes of ticks and characters are in pixels, otherwise are fractions of the screen size

plotinit(w,{x},{y},{flag = 0})

Initialize the rectwindow w, destroying any rect objects you may have already drawn in w. The virtual cursor is set to (0,0). The rectwindow size is set to width x and height y; omitting either x or y means we use the full size of the device in that direction. If flag = 0, x and y represent pixel units. Otherwise, x and y are understood as fractions of the size of the current output device (hence must be between 0 and 1) and internally converted to pixels.

The plotting device imposes an upper bound for x and y, for instance the number of pixels for screen output. These bounds are available through the plothsizes function. The following sequence initializes in a portable way (i.e independent of the output device) a window of maximal size, accessed through coordinates in the [0,1000] x [0,1000] range:

  s = plothsizes();
  plotinit(0, s[1]-1, s[2]-1);
  plotscale(0, 0,1000, 0,1000);

plotkill(w)

Erase rectwindow w and free the corresponding memory. Note that if you want to use the rectwindow w again, you have to use plotinit first to specify the new size. So it's better in this case to use plotinit directly as this throws away any previous work in the given rectwindow.

plotlines(w,X,Y,{flag = 0})

Draw on the rectwindow w the polygon such that the (x,y)-coordinates of the vertices are in the vectors of equal length X and Y. For simplicity, the whole polygon is drawn, not only the part of the polygon which is inside the rectwindow. If flag is non-zero, close the polygon. In any case, the virtual cursor does not move.

X and Y are allowed to be scalars (in this case, both have to). There, a single segment will be drawn, between the virtual cursor current position and the point (X,Y). And only the part thereof which actually lies within the boundary of w. Then move the virtual cursor to (X,Y), even if it is outside the window. If you want to draw a line from (x1,y1) to (x2,y2) where (x1,y1) is not necessarily the position of the virtual cursor, use plotmove(w,x1,y1) before using this function.

plotlinetype(w,type)

Change the type of lines subsequently plotted in rectwindow w. type -2 corresponds to frames, -1 to axes, larger values may correspond to something else. w = -1 changes highlevel plotting. This is only taken into account by the gnuplot interface.

plotmove(w,x,y)

Move the virtual cursor of the rectwindow w to position (x,y).

plotpoints(w,X,Y)

Draw on the rectwindow w the points whose (x,y)-coordinates are in the vectors of equal length X and Y and which are inside w. The virtual cursor does not move. This is basically the same function as plothraw, but either with no scaling factor or with a scale chosen using the function plotscale.

As was the case with the plotlines function, X and Y are allowed to be (simultaneously) scalar. In this case, draw the single point (X,Y) on the rectwindow w (if it is actually inside w), and in any case move the virtual cursor to position (x,y).

plotpointsize(w,size)

Changes the ``size'' of following points in rectwindow w. If w = -1, change it in all rectwindows. This only works in the gnuplot interface.

plotpointtype(w,type)

Change the type of points subsequently plotted in rectwindow w. type = -1 corresponds to a dot, larger values may correspond to something else. w = -1 changes highlevel plotting. This is only taken into account by the gnuplot interface.

plotrbox(w,dx,dy)

Draw in the rectwindow w the outline of the rectangle which is such that the points (x1,y1) and (x1+dx,y1+dy) are opposite corners, where (x1,y1) is the current position of the cursor. Only the part of the rectangle which is in w is drawn. The virtual cursor does not move.

plotrecth(w,X = a,b,expr,{flag = 0},{n = 0})

Writes to rectwindow w the curve output of ploth(w,X = a,b,expr,flag,n). Returns a vector for the bounding box.

plotrecthraw(w,data,{flags = 0})

Plot graph(s) for data in rectwindow w. flag has the same significance here as in ploth, though recursive plot is no more significant.

data is a vector of vectors, each corresponding to a list a coordinates. If parametric plot is set, there must be an even number of vectors, each successive pair corresponding to a curve. Otherwise, the first one contains the x coordinates, and the other ones contain the y-coordinates of curves to plot.

plotrline(w,dx,dy)

Draw in the rectwindow w the part of the segment (x1,y1)-(x1+dx,y1+dy) which is inside w, where (x1,y1) is the current position of the virtual cursor, and move the virtual cursor to (x1+dx,y1+dy) (even if it is outside the window).

plotrmove(w,dx,dy)

Move the virtual cursor of the rectwindow w to position (x1+dx,y1+dy), where (x1,y1) is the initial position of the cursor (i.e. to position (dx,dy) relative to the initial cursor).

plotrpoint(w,dx,dy)

Draw the point (x1+dx,y1+dy) on the rectwindow w (if it is inside w), where (x1,y1) is the current position of the cursor, and in any case move the virtual cursor to position (x1+dx,y1+dy).

plotscale(w,x1,x2,y1,y2)

Scale the local coordinates of the rectwindow w so that x goes from x1 to x2 and y goes from y1 to y2 (x2 < x1 and y2 < y1 being allowed). Initially, after the initialization of the rectwindow w using the function plotinit, the default scaling is the graphic pixel count, and in particular the y axis is oriented downwards since the origin is at the upper left. The function plotscale allows to change all these defaults and should be used whenever functions are graphed.

plotstring(w,x,{flags = 0})

Draw on the rectwindow w the String x (see Label se:strings), at the current position of the cursor.

flag is used for justification: bits 1 and 2 regulate horizontal alignment: left if 0, right if 2, center if 1. Bits 4 and 8 regulate vertical alignment: bottom if 0, top if 8, v-center if 4. Can insert additional small gap between point and string: horizontal if bit 16 is set, vertical if bit 32 is set (see the tutorial for an example).

psdraw(list, {flag = 0})

Same as plotdraw, except that the output is a PostScript program appended to the psfile, and flag != 0 scales the plot from size of the current output device to the standard PostScript plotting size

psploth(X = a,b,expr,{flags = 0},{n = 0})

Same as ploth, except that the output is a PostScript program appended to the psfile.

psplothraw(listx,listy,{flag = 0})

Same as plothraw, except that the output is a PostScript program appended to the psfile.


Programming in GP: control statements

A number of control statements are available in GP. They are simpler and have a syntax slightly different from their C counterparts, but are quite powerful enough to write any kind of program. Some of them are specific to GP, since they are made for number theorists. As usual, X will denote any simple variable name, and seq will always denote a sequence of expressions, including the empty sequence.

@3Caveat. In constructs like

      for (X = a,b, seq)

the variable X is lexically scoped to the loop, leading to possibly unexpected behavior:

      n = 5;
      for (n = 1, 10,
        if (something_nice(), break);
      );
      \\  at this point C<n> is 5 !

If the sequence seq modifies the loop index, then the loop is modified accordingly:

      ? for (n = 1, 10, n += 2; print(n))
      3
      6
      9
      12

break({n = 1})

Interrupts execution of current seq, and immediately exits from the n innermost enclosing loops, within the current function call (or the top level loop); the integer n must be positive. If n is greater than the number of enclosing loops, all enclosing loops are exited.

breakpoint()

Interrupt the program and enter the breakloop. The program continues when the breakloop is exited.

  ? f(N,x)=my(z=x^2+1);breakpoint();gcd(N,z^2+1-z);
  ? f(221,3)
    ***   at top-level: f(221,3)
    ***                 ^--------
    ***   in function f: my(z=x^2+1);breakpoint();gcd(N,z
    ***                              ^--------------------
    ***   Break loop: type <Return> to continue; 'break' to go back to GP
  break> z
  10
  break>
  %2 = 13

dbg_down({n = 1})

(In the break loop) go down n frames. This allows to cancel a previous call to dbg_up.

dbg_err()

In the break loop, return the error data of the current error, if any. See iferr for details about error data. Compare:

  ? iferr(1/(Mod(2,12019)^(6!)-1),E,Vec(E))
  %1 = ["e_INV", "Fp_inv", Mod(119, 12019)]
  ? 1/(Mod(2,12019)^(6!)-1)
    ***   at top-level: 1/(Mod(2,12019)^(6!)-
    ***                  ^--------------------
    *** _/_: impossible inverse in Fp_inv: Mod(119, 12019).
    ***   Break loop: type 'break' to go back to GP prompt
  break> Vec(dbg_err())
  ["e_INV", "Fp_inv", Mod(119, 12019)]

dbg_up({n = 1})

(In the break loop) go up n frames. This allows to inspect data of the parent function. To cancel a dbg_up call, use dbg_down

dbg_x(A{,n})

Print the inner structure of A, complete if n is omitted, up to level n otherwise. This is useful for debugging. This is similar to \x but does not require A to be an history entry. In particular, it can be used in the break loop.

for(X = a,b,seq)

Evaluates seq, where the formal variable X goes from a to b. Nothing is done if a > b. a and b must be in R.

forcomposite(n = a,{b},seq)

Evaluates seq, where the formal variable n ranges over the composite numbers between the non-negative real numbers a to b, including a and b if they are composite. Nothing is done if a > b.

  ? forcomposite(n = 0, 10, print(n))
  4
  6
  8
  9
  10

@3Omitting b means we will run through all composites >= a, starting an infinite loop; it is expected that the user will break out of the loop himself at some point, using break or return.

Note that the value of n cannot be modified within seq:

  ? forcomposite(n = 2, 10, n = [])
   ***   at top-level: forcomposite(n=2,10,n=[])
   ***                                      ^---
   ***   index read-only: was changed to [].

fordiv(n,X,seq)

Evaluates seq, where the formal variable X ranges through the divisors of n (see divisors, which is used as a subroutine). It is assumed that factor can handle n, without negative exponents. Instead of n, it is possible to input a factorization matrix, i.e. the output of factor(n).

This routine uses divisors as a subroutine, then loops over the divisors. In particular, if n is an integer, divisors are sorted by increasing size.

To avoid storing all divisors, possibly using a lot of memory, the following (much slower) routine loops over the divisors using essentially constant space:

  FORDIV(N)=
  { my(P, E);
    P = factor(N); E = P[,2]; P = P[,1];
    forvec( v = vector(#E, i, [0,E[i]]),
    X = factorback(P, v)
    \\ ...
  );
  }
  ? for(i=1,10^5, FORDIV(i))
  time = 3,445 ms.
  ? for(i=1,10^5, fordiv(i, d, ))
  time = 490 ms.

forell(E,a,b,seq)

Evaluates seq, where the formal variable E = [name, M, G] ranges through all elliptic curves of conductors from a to b. In this notation name is the curve name in Cremona's elliptic curve database, M is the minimal model, G is a Z-basis of the free part of the Mordell-Weil group E(Q).

  ? forell(E, 1, 500, my([name,M,G] = E); \
      if (#G > 1, print(name)))
  389a1
  433a1
  446d1

The elldata database must be installed and contain data for the specified conductors.

The library syntax is forell(void *data, long (*call)(void*,GEN), long a, long b).

forpart(X = k,seq,{a = k},{n = k})

Evaluate seq over the partitions X = [x_1,...x_n] of the integer k, i.e. increasing sequences x_1 <= x_2... <= x_n of sum x_1+...+ x_n = k. By convention, 0 admits only the empty partition and negative numbers have no partitions. A partition is given by a t_VECSMALL, where parts are sorted in nondecreasing order:

  ? forpart(X=3, print(X))
  Vecsmall([3])
  Vecsmall([1, 2])
  Vecsmall([1, 1, 1])

@3Optional parameters n and a are as follows:

@3* n = nmax (resp. n = [nmin,nmax]) restricts partitions to length less than nmax (resp. length between nmin and nmax), where the length is the number of nonzero entries.

@3* a = amax (resp. a = [amin,amax]) restricts the parts to integers less than amax (resp. between amin and amax).

By default, parts are positive and we remove zero entries unless amin <= 0, in which case X is of constant length nmax.

  \\ at most 3 non-zero parts, all <= 4
  ? forpart(v=5,print(Vec(v)),4,3)
  [1, 4]
  [2, 3]
  [1, 1, 3]
  [1, 2, 2]
  \\ between 2 and 4 parts less than 5, fill with zeros
  ? forpart(v=5,print(Vec(v)),[0,5],[2,4])
  [0, 0, 1, 4]
  [0, 0, 2, 3]
  [0, 1, 1, 3]
  [0, 1, 2, 2]
  [1, 1, 1, 2]

@3 The behavior is unspecified if X is modified inside the loop.

The library syntax is forpart(void *data, long (*call)(void*,GEN), long k, GEN a, GEN n).

forprime(p = a,{b},seq)

Evaluates seq, where the formal variable p ranges over the prime numbers between the real numbers a to b, including a and b if they are prime. More precisely, the value of p is incremented to nextprime(p + 1), the smallest prime strictly larger than p, at the end of each iteration. Nothing is done if a > b.

  ? forprime(p = 4, 10, print(p))
  5
  7

@3Omitting b means we will run through all primes >= a, starting an infinite loop; it is expected that the user will break out of the loop himself at some point, using break or return.

Note that the value of p cannot be modified within seq:

  ? forprime(p = 2, 10, p = [])
   ***   at top-level: forprime(p=2,10,p=[])
   ***                                   ^---
   ***   prime index read-only: was changed to [].

forstep(X = a,b,s,seq)

Evaluates seq, where the formal variable X goes from a to b, in increments of s. Nothing is done if s > 0 and a > b or if s < 0 and a < b. s must be in R^* or a vector of steps [s_1,...,s_n]. In the latter case, the successive steps are used in the order they appear in s.

  ? forstep(x=5, 20, [2,4], print(x))
  5
  7
  11
  13
  17
  19

forsubgroup(H = G,{bound},seq)

Evaluates seq for each subgroup H of the abelian group G (given in SNF form or as a vector of elementary divisors).

If bound is present, and is a positive integer, restrict the output to subgroups of index less than bound. If bound is a vector containing a single positive integer B, then only subgroups of index exactly equal to B are computed

The subgroups are not ordered in any obvious way, unless G is a p-group in which case Birkhoff's algorithm produces them by decreasing index. A subgroup is given as a matrix whose columns give its generators on the implicit generators of G. For example, the following prints all subgroups of index less than 2 in G = Z/2Z g_1 x Z/2Z g_2:

  ? G = [2,2]; forsubgroup(H=G, 2, print(H))
  [1; 1]
  [1; 2]
  [2; 1]
  [1, 0; 1, 1]

The last one, for instance is generated by (g_1, g_1 + g_2). This routine is intended to treat huge groups, when subgrouplist is not an option due to the sheer size of the output.

For maximal speed the subgroups have been left as produced by the algorithm. To print them in canonical form (as left divisors of G in HNF form), one can for instance use

  ? G = matdiagonal([2,2]); forsubgroup(H=G, 2, print(mathnf(concat(G,H))))
  [2, 1; 0, 1]
  [1, 0; 0, 2]
  [2, 0; 0, 1]
  [1, 0; 0, 1]

Note that in this last representation, the index [G:H] is given by the determinant. See galoissubcyclo and galoisfixedfield for applications to Galois theory.

The library syntax is forsubgroup(void *data, long (*call)(void*,GEN), GEN G, GEN bound).

forvec(X = v,seq,{flag = 0})

Let v be an n-component vector (where n is arbitrary) of two-component vectors [a_i,b_i] for 1 <= i <= n. This routine evaluates seq, where the formal variables X[1],..., X[n] go from a_1 to b_1,..., from a_n to b_n, i.e. X goes from [a_1,...,a_n] to [b_1,...,b_n] with respect to the lexicographic ordering. (The formal variable with the highest index moves the fastest.) If flag = 1, generate only nondecreasing vectors X, and if flag = 2, generate only strictly increasing vectors X.

The type of X is the same as the type of v: t_VEC or t_COL.

if(a,{seq1},{seq2})

Evaluates the expression sequence seq1 if a is non-zero, otherwise the expression seq2. Of course, seq1 or seq2 may be empty:

if (a,seq) evaluates seq if a is not equal to zero (you don't have to write the second comma), and does nothing otherwise,

if (a,,seq) evaluates seq if a is equal to zero, and does nothing otherwise. You could get the same result using the ! (not) operator: if (!a,seq).

The value of an if statement is the value of the branch that gets evaluated: for instance

  x = if(n % 4 == 1, y, z);

@3sets x to y if n is 1 modulo 4, and to z otherwise.

Successive 'else' blocks can be abbreviated in a single compound if as follows:

  if (test1, seq1,
      test2, seq2,
      ...
      testn, seqn,
      seqdefault);

@3is equivalent to

  if (test1, seq1
           , if (test2, seq2
                      , ...
                        if (testn, seqn, seqdefault)...));

For instance, this allows to write traditional switch / case constructions:

  if (x == 0, do0(),
      x == 1, do1(),
      x == 2, do2(),
      dodefault());

@3Remark. The boolean operators && and || are evaluated according to operator precedence as explained in Label se:operators, but, contrary to other operators, the evaluation of the arguments is stopped as soon as the final truth value has been determined. For instance

  if (x != 0 && f(1/x), ...)

@3is a perfectly safe statement.

@3Remark. Functions such as break and next operate on loops, such as forxxx, while, until. The if statement is not a loop. (Obviously!)

iferr(seq1,E,seq2{,pred})

Evaluates the expression sequence seq1. If an error occurs, set the formal parameter E set to the error data. If pred is not present or evaluates to true, catch the error and evaluate seq2. Both pred and seq2 can reference E. The error type is given by errname(E), and other data can be accessed using the component function. The code seq2 should check whether the error is the one expected. In the negative the error can be rethrown using error(E) (and possibly caught by an higher iferr instance). The following uses iferr to implement Lenstra's ECM factoring method

  ? ecm(N, B = 1000!, nb = 100)=
    {
      for(a = 1, nb,
        iferr(ellmul(ellinit([a,1]*Mod(1,N)), [0,1]*Mod(1,N), B),
          E, return(gcd(lift(component(E,2)),N)),
          errname(E)=="e_INV" && type(component(E,2)) == "t_INTMOD"))
    }
  ? ecm(2^101-1)
  %2 = 7432339208719

The return value of iferr itself is the value of seq2 if an error occurs, and the value of seq1 otherwise. We now describe the list of valid error types, and the associated error data E; in each case, we list in order the components of E, accessed via component(E,1), component(E,2), etc.

@3Internal errors, ``system'' errors.

@3* "e_ARCH". A requested feature s is not available on this architecture or operating system. E has one component (t_STR): the missing feature name s.

@3* "e_BUG". A bug in the PARI library, in function s. E has one component (t_STR): the function name s.

@3* "e_FILE". Error while trying to open a file. E has two components, 1 (t_STR): the file type (input, output, etc.), 2 (t_STR): the file name.

@3* "e_IMPL". A requested feature s is not implemented. E has one component, 1 (t_STR): the feature name s.

@3* "e_PACKAGE". Missing optional package s. E has one component, 1 (t_STR): the package name s.

@3Syntax errors, type errors.

@3* "e_DIM". The dimensions of arguments x and y submitted to function s does not match up. E.g., multiplying matrices of inconsistent dimension, adding vectors of different lengths,... E has three component, 1 (t_STR): the function name s, 2: the argument x, 3: the argument y.

@3* "e_FLAG". A flag argument is out of bounds in function s. E has one component, 1 (t_STR): the function name s.

@3* "e_NOTFUNC". Generated by the PARI evaluator; tried to use a GEN x which is not a t_CLOSURE in a function call syntax (as in f = 1; f(2);). E has one component, 1: the offending GEN x.

@3* "e_OP". Impossible operation between two objects than cannot be typecast to a sensible common domain for deeper reasons than a type mismatch, usually for arithmetic reasons. As in O(2) + O(3): it is valid to add two t_PADICs, provided the underlying prime is the same; so the addition is not forbidden a priori for type reasons, it only becomes so when inspecting the objects and trying to perform the operation. E has three components, 1 (t_STR): the operator name op, 2: first argument, 3: second argument.

@3* "e_TYPE". An argument x of function s had an unexpected type. (As in factor("blah").) E has two components, 1 (t_STR): the function name s, 2: the offending argument x.

@3* "e_TYPE2". Forbidden operation between two objects than cannot be typecast to a sensible common domain, because their types do not match up. (As in Mod(1,2) + Pi.) E has three components, 1 (t_STR): the operator name op, 2: first argument, 3: second argument.

@3* "e_PRIORITY". Object o in function s contains variables whose priority is incompatible with the expected operation. E.g. Pol([x,1], 'y): this raises an error because it's not possible to create a polynomial whose coefficients involve variables with higher priority than the main variable. E has four components: 1 (t_STR): the function name s, 2: the offending argument o, 3 (t_STR): an operator op describing the priority error, 4 (t_POL): the variable v describing the priority error. The argument satisfies variable(x) op variable(v).

@3* "e_VAR". The variables of arguments x and y submitted to function s does not match up. E.g., considering the algebraic number Mod(t,t^2+1) in nfinit(x^2+1). E has three component, 1 (t_STR): the function name s, 2 (t_POL): the argument x, 3 (t_POL): the argument y.

@3Overflows.

@3* "e_COMPONENT". Trying to access an inexistent component in a vector/matrix/list in a function: the index is less than 1 or greater than the allowed length. E has four components, 1 (t_STR): the function name 2 (t_STR): an operator op ( < or > ), 2 (t_GEN): a numerical limit l bounding the allowed range, 3 (GEN): the index x. It satisfies x op l.

@3* "e_DOMAIN". An argument is not in the function's domain. E has five components, 1 (t_STR): the function name, 2 (t_STR): the mathematical name of the out-of-domain argument 3 (t_STR): an operator op describing the domain error, 4 (t_GEN): the numerical limit l describing the domain error, 5 (GEN): the out-of-domain argument x. The argument satisfies x op l, which prevents it from belonging to the function's domain.

@3* "e_MAXPRIME". A function using the precomputed list of prime numbers ran out of primes. E has one component, 1 (t_INT): the requested prime bound, which overflowed primelimit or 0 (bound is unknown).

@3* "e_MEM". A call to pari_malloc or pari_realloc failed. E has no component.

@3* "e_OVERFLOW". An object in function s becomes too large to be represented within PARI's hardcoded limits. (As in 2^2^2^10 or exp(1e100), which overflow in lg and expo.) E has one component, 1 (t_STR): the function name s.

@3* "e_PREC". Function s fails because input accuracy is too low. (As in floor(1e100) at default accuracy.) E has one component, 1 (t_STR): the function name s.

@3* "e_STACK". The PARI stack overflows. E has no component.

@3Errors triggered intentionally.

@3* "e_ALARM". A timeout, generated by the alarm function. E has one component (t_STR): the error message to print.

@3* "e_USER". A user error, as triggered by error(g_1,...,g_n). E has one component, 1 (t_VEC): the vector of n arguments given to error.

@3Mathematical errors.

@3* "e_CONSTPOL". An argument of function s is a constant polynomial, which does not make sense. (As in galoisinit(Pol(1)).) E has one component, 1 (t_STR): the function name s.

@3* "e_COPRIME". Function s expected coprime arguments, and did receive x,y, which were not. E has three component, 1 (t_STR): the function name s, 2: the argument x, 3: the argument y.

@3* "e_INV". Tried to invert a non-invertible object x in function s. E has two components, 1 (t_STR): the function name s, 2: the non-invertible x. If x = Mod(a,b) is a t_INTMOD and a is not 0 mod b, this allows to factor the modulus, as gcd(a,b) is a non-trivial divisor of b.

@3* "e_IRREDPOL". Function s expected an irreducible polynomial, and did receive T, which was not. (As in nfinit(x^2-1).) E has two component, 1 (t_STR): the function name s, 2 (t_POL): the polynomial x.

@3* "e_MISC". Generic uncategorized error. E has one component (t_STR): the error message to print.

@3* "e_MODULUS". moduli x and y submitted to function s are inconsistent. As in

     nfalgtobasis(nfinit(t^3-2), Mod(t,t^2+1)

E has three component, 1 (t_STR): the function s, 2: the argument x, 3: the argument x.

@3* "e_NEGVAL". An argument of function s is a power series with negative valuation, which does not make sense. (As in cos(1/x).) E has one component, 1 (t_STR): the function name s.

@3* "e_PRIME". Function s expected a prime number, and did receive p, which was not. (As in idealprimedec(nf, 4).) E has two component, 1 (t_STR): the function name s, 2: the argument p.

@3* "e_ROOTS0". An argument of function s is a zero polynomial, and we need to consider its roots. (As in polroots(0).) E has one component, 1 (t_STR): the function name s.

@3* "e_SQRTN". Trying to compute an n-th root of x, which does not exist, in function s. (As in sqrt(Mod(-1,3)).) E has two components, 1 (t_STR): the function name s, 2: the argument x.

next({n = 1})

Interrupts execution of current seq, resume the next iteration of the innermost enclosing loop, within the current function call (or top level loop). If n is specified, resume at the n-th enclosing loop. If n is bigger than the number of enclosing loops, all enclosing loops are exited.

return({x = 0})

Returns from current subroutine, with result x. If x is omitted, return the (void) value (return no result, like print).

until(a,seq)

Evaluates seq until a is not equal to 0 (i.e. until a is true). If a is initially not equal to 0, seq is evaluated once (more generally, the condition on a is tested after execution of the seq, not before as in while).

while(a,seq)

While a is non-zero, evaluates the expression sequence seq. The test is made before evaluating the seq, hence in particular if a is initially equal to zero the seq will not be evaluated at all.


Programming in GP: other specific functions

In addition to the general PARI functions, it is necessary to have some functions which will be of use specifically for gp, though a few of these can be accessed under library mode. Before we start describing these, we recall the difference between strings and keywords (see Label se:strings): the latter don't get expanded at all, and you can type them without any enclosing quotes. The former are dynamic objects, where everything outside quotes gets immediately expanded.

Strprintf(fmt,{x}*)

Returns a string built from the remaining arguments according to the format fmt. The format consists of ordinary characters (not %), printed unchanged, and conversions specifications. See printf.

addhelp(sym,str)

Changes the help message for the symbol sym. The string str is expanded on the spot and stored as the online help for sym. It is recommended to document global variables and user functions in this way, although gp will not protest if you don't.

You can attach a help text to an alias, but it will never be shown: aliases are expanded by the ? help operator and we get the help of the symbol the alias points to. Nothing prevents you from modifying the help of built-in PARI functions. But if you do, we would like to hear why you needed it!

Without addhelp, the standard help for user functions consists of its name and definition.

  gp> f(x) = x^2;
  gp> ?f
  f =
    (x)->x^2

@3Once addhelp is applied to f, the function code is no longer included. It can still be consulted by typing the function name:

  gp> addhelp(f, "Square")
  gp> ?f
  Square
  gp> f
  %2 = (x)->x^2

The library syntax is void addhelp(const char *sym, const char *str).

alarm({s = 0},{code})

If code is omitted, trigger an e_ALARM exception after s seconds, cancelling any previously set alarm; stop a pending alarm if s = 0 or is omitted.

Otherwise, if s is positive, the function evaluates code, aborting after s seconds. The return value is the value of code if it ran to completion before the alarm timeout, and a t_ERROR object otherwise.

    ? p = nextprime(10^25); q = nextprime(10^26); N = p*q;
    ? E = alarm(1, factor(N));
    ? type(E)
    %3 = "t_ERROR"
    ? print(E)
    %4 = error("alarm interrupt after 964 ms.")
    ? alarm(10, factor(N));   \\ enough time
    %5 =
    [ 10000000000000000000000013 1]
    [100000000000000000000000067 1]

@3Here is a more involved example: the function timefact(N,sec) below tries to factor N and gives up after sec seconds, returning a partial factorisation.

  \\ Time-bounded partial factorization
  default(factor_add_primes,1);
  timefact(N,sec)=
  {
    F = alarm(sec, factor(N));
    if (type(F) == "t_ERROR", factor(N, 2^24), F);
  }

@3We either return the factorization directly, or replace the t_ERROR result by a simple bounded factorization factor(N, 2^24). Note the factor_add_primes trick: any prime larger than 2^{24} discovered while attempting the initial factorization is stored and remembered. When the alarm rings, the subsequent bounded factorization finds it right away.

@3Caveat. It is not possible to set a new alarm within another alarm code: the new timer erases the parent one.

alias(newsym,sym)

Defines the symbol newsym as an alias for the the symbol sym:

  ? alias("det", "matdet");
  ? det([1,2;3,4])
  %1 = -2

You are not restricted to ordinary functions, as in the above example: to alias (from/to) member functions, prefix them with `_.'; to alias operators, use their internal name, obtained by writing _ in lieu of the operators argument: for instance, _! and !_ are the internal names of the factorial and the logical negation, respectively.

  ? alias("mod", "_.mod");
  ? alias("add", "_+_");
  ? alias("_.sin", "sin");
  ? mod(Mod(x,x^4+1))
  %2 = x^4 + 1
  ? add(4,6)
  %3 = 10
  ? Pi.sin
  %4 = 0.E-37

Alias expansion is performed directly by the internal GP compiler. Note that since alias is performed at compilation-time, it does not require any run-time processing, however it only affects GP code compiled after the alias command is evaluated. A slower but more flexible alternative is to use variables. Compare

  ? fun = sin;
  ? g(a,b) = intnum(t=a,b,fun(t));
  ? g(0, Pi)
  %3 = 2.0000000000000000000000000000000000000
  ? fun = cos;
  ? g(0, Pi)
  %5 = 1.8830410776607851098 E-39

with

  ? alias(fun, sin);
  ? g(a,b) = intnum(t=a,b,fun(t));
  ? g(0,Pi)
  %2 = 2.0000000000000000000000000000000000000
  ? alias(fun, cos);  \\ Oops. Does not affect *previous* definition!
  ? g(0,Pi)
  %3 = 2.0000000000000000000000000000000000000
  ? g(a,b) = intnum(t=a,b,fun(t)); \\ Redefine, taking new alias into account
  ? g(0,Pi)
  %5 = 1.8830410776607851098 E-39

A sample alias file misc/gpalias is provided with the standard distribution.

The library syntax is void alias0(const char *newsym, const char *sym).

allocatemem({s = 0})

This special operation changes the stack size after initialization. x must be a non-negative integer. If x > 0, a new stack of at least x bytes is allocated. We may allocate more than x bytes if x is way too small, or for alignment reasons: the current formula is \max(16*ceil{x/16}, 500032) bytes.

If x = 0, the size of the new stack is twice the size of the old one. The old stack is discarded.

@3Warning. This function should be typed at the gp prompt in interactive usage, or left by itself at the start of batch files. It cannot be used meaningfully in loop-like constructs, or as part of a larger expression sequence, e.g

     allocatemem(); x = 1;   \\ This will not set C<x>!

In fact, all loops are immediately exited, user functions terminated, and the rest of the sequence following allocatemem() is silently discarded, as well as all pending sequences of instructions. We just go on reading the next instruction sequence from the file we're in (or from the user). In particular, we have the following possibly unexpected behavior: in

     read("file.gp"); x = 1

@3were file.gp contains an allocatemem statement, the x = 1 is never executed, since all pending instructions in the current sequence are discarded.

The technical reason is that this routine moves the stack, so temporary objects created during the current expression evaluation are not correct anymore. (In particular byte-compiled expressions, which are allocated on the stack.) To avoid accessing obsolete pointers to the old stack, this routine ends by a longjmp.

@3Remark. If the operating system cannot allocate the desired x bytes, a loop halves the allocation size until it succeeds:

  ? allocatemem(5*10^10)
   ***   Warning: not enough memory, new stack 50000000000.
   ***   Warning: not enough memory, new stack 25000000000.
   ***   Warning: not enough memory, new stack 12500000000.
   ***   Warning: new stack size = 6250000000 (5960.464 Mbytes).

apply(f, A)

Apply the t_CLOSURE f to the entries of A. If A is a scalar, return f(A). If A is a polynomial or power series, apply f on all coefficients. If A is a vector or list, return the elements f(x) where x runs through A. If A is a matrix, return the matrix whose entries are the f(A[i,j]).

  ? apply(x->x^2, [1,2,3,4])
  %1 = [1, 4, 9, 16]
  ? apply(x->x^2, [1,2;3,4])
  %2 =
  [1 4]
  [9 16]
  ? apply(x->x^2, 4*x^2 + 3*x+ 2)
  %3 = 16*x^2 + 9*x + 4

@3Note that many functions already act componentwise on vectors or matrices, but they almost never act on lists; in this case, apply is a good solution:

  ? L = List([Mod(1,3), Mod(2,4)]);
  ? lift(L)
    ***   at top-level: lift(L)
    ***                 ^-------
    *** lift: incorrect type in lift.
  ? apply(lift, L);
  %2 = List([1, 2])

@3Remark. For v a t_VEC, t_COL, t_LIST or t_MAT, the alternative set-notations

  [g(x) | x <- v, f(x)]
  [x | x <- v, f(x)]
  [g(x) | x <- v]

@3 are available as shortcuts for

  apply(g, select(f, Vec(v)))
  select(f, Vec(v))
  apply(g, Vec(v))

@3respectively:

  ? L = List([Mod(1,3), Mod(2,4)]);
  ? [ lift(x) | x<-L ]
  %2 = [1, 2]

The library syntax is genapply(void *E, GEN (*fun)(void*,GEN), GEN a).

default({key},{val})

Returns the default corresponding to keyword key. If val is present, sets the default to val first (which is subject to string expansion first). Typing default() (or \d) yields the complete default list as well as their current values. See Label se:defaults for an introduction to GP defaults, Label se:gp_defaults for a list of available defaults, and Label se:meta for some shortcut alternatives. Note that the shortcuts are meant for interactive use and usually display more information than default.

The library syntax is GEN default0(const char *key = NULL, const char *val = NULL).

errname(E)

Returns the type of the error message E as a string.

The library syntax is GEN errname(GEN E).

error({str}*)

Outputs its argument list (each of them interpreted as a string), then interrupts the running gp program, returning to the input prompt. For instance

  error("n = ", n, " is not squarefree!")

extern(str)

The string str is the name of an external command (i.e. one you would type from your UNIX shell prompt). This command is immediately run and its output fed into gp, just as if read from a file.

externstr(str)

The string str is the name of an external command (i.e. one you would type from your UNIX shell prompt). This command is immediately run and its output is returned as a vector of GP strings, one component per output line.

getabstime()

Returns the time (in milliseconds) elapsed since gp startup. This provides a reentrant version of gettime:

  my (t = getabstime());
  ...
  print("Time: ", getabstime() - t);

The library syntax is long getabstime().

getenv(s)

Return the value of the environment variable s if it is defined, otherwise return 0.

The library syntax is GEN gp_getenv(const char *s).

getheap()

Returns a two-component row vector giving the number of objects on the heap and the amount of memory they occupy in long words. Useful mainly for debugging purposes.

The library syntax is GEN getheap().

getrand()

Returns the current value of the seed used by the pseudo-random number generator random. Useful mainly for debugging purposes, to reproduce a specific chain of computations. The returned value is technical (reproduces an internal state array), and can only be used as an argument to setrand.

The library syntax is GEN getrand().

getstack()

Returns the current value of top-avma, i.e. the number of bytes used up to now on the stack. Useful mainly for debugging purposes.

The library syntax is long getstack().

gettime()

Returns the time (in milliseconds) elapsed since either the last call to gettime, or to the beginning of the containing GP instruction (if inside gp), whichever came last.

For a reentrant version, see getabstime.

The library syntax is long gettime().

global(list of variables)

Obsolete. Scheduled for deletion.

inline(x,...,z)

(Experimental) declare x,..., z as inline variables. Such variables behave like lexically scoped variable (see my()) but with unlimited scope. It is however possible to exit the scope by using uninline(). When used in a GP script, it is recommended to call uninline() before the script's end to avoid inline variables leaking outside the script.

input()

Reads a string, interpreted as a GP expression, from the input file, usually standard input (i.e. the keyboard). If a sequence of expressions is given, the result is the result of the last expression of the sequence. When using this instruction, it is useful to prompt for the string by using the print1 function. Note that in the present version 2.19 of pari.el, when using gp under GNU Emacs (see Label se:emacs) one must prompt for the string, with a string which ends with the same prompt as any of the previous ones (a "? " will do for instance).

install(name,code,{gpname},{lib})

Loads from dynamic library lib the function name. Assigns to it the name gpname in this gp session, with prototype code (see below). If gpname is omitted, uses name. If lib is omitted, all symbols known to gp are available: this includes the whole of libpari.so and possibly others (such as libc.so).

Most importantly, install gives you access to all non-static functions defined in the PARI library. For instance, the function GEN addii(GEN x, GEN y) adds two PARI integers, and is not directly accessible under gp (it is eventually called by the + operator of course):

  ? install("addii", "GG")
  ? addii(1, 2)
  %1 = 3

It also allows to add external functions to the gp interpreter. For instance, it makes the function system obsolete:

  ? install(system, vs, sys,/*omitted*/)
  ? sys("ls gp*")
  gp.c            gp.h            gp_rl.c

@3This works because system is part of libc.so, which is linked to gp. It is also possible to compile a shared library yourself and provide it to gp in this way: use gp2c, or do it manually (see the modules_build variable in pari.cfg for hints).

Re-installing a function will print a warning and update the prototype code if needed. However, it will not reload a symbol from the library, even if the latter has been recompiled.

@3Prototype. We only give a simplified description here, covering most functions, but there are many more possibilities. The full documentation is available in libpari.dvi, see

    ??prototype

@3* First character i, l, v : return type int / long / void. (Default: GEN)

@3* One letter for each mandatory argument, in the same order as they appear in the argument list: G (GEN), & (GEN*), L (long), s (char *), n (variable).

@3* p to supply realprecision (usually long prec in the argument list), P to supply seriesprecision (usually long precdl).

@3We also have special constructs for optional arguments and default values:

@3* DG (optional GEN, NULL if omitted),

@3* D& (optional GEN*, NULL if omitted),

@3* Dn (optional variable, -1 if omitted),

For instance the prototype corresponding to

    long issquareall(GEN x, GEN *n = NULL)

@3is lGD&.

@3Caution. This function may not work on all systems, especially when gp has been compiled statically. In that case, the first use of an installed function will provoke a Segmentation Fault (this should never happen with a dynamically linked executable). If you intend to use this function, please check first on some harmless example such as the one above that it works properly on your machine.

The library syntax is void gpinstall(const char *name, const char *code, const char *gpname, const char *lib).

kill(sym)

Restores the symbol sym to its ``undefined'' status, and deletes any help messages associated to sym using addhelp. Variable names remain known to the interpreter and keep their former priority: you cannot make a variable ``less important" by killing it!

  ? z = y = 1; y
  %1 = 1
  ? kill(y)
  ? y            \\ restored to ``undefined'' status
  %2 = y
  ? variable()
  %3 = [x, y, z] \\ but the variable name y is still known, with y > z !

For the same reason, killing a user function (which is an ordinary variable holding a t_CLOSURE) does not remove its name from the list of variable names.

If the symbol is associated to a variable --- user functions being an important special case ---, one may use the quote operator a = 'a to reset variables to their starting values. However, this will not delete a help message associated to a, and is also slightly slower than kill(a).

  ? x = 1; addhelp(x, "foo"); x
  %1 = 1
  ? x = 'x; x   \\ same as 'kill', except we don't delete help.
  %2 = x
  ? ?x
  foo

On the other hand, kill is the only way to remove aliases and installed functions.

  ? alias(fun, sin);
  ? kill(fun);
  ? install(addii, GG);
  ? kill(addii);

The library syntax is void kill0(const char *sym).

print({str}*)

Outputs its (string) arguments in raw format, ending with a newline.

print1({str}*)

Outputs its (string) arguments in raw format, without ending with a newline. Note that you can still embed newlines within your strings, using the \n notation !

printf(fmt,{x}*)

This function is based on the C library command of the same name. It prints its arguments according to the format fmt, which specifies how subsequent arguments are converted for output. The format is a character string composed of zero or more directives:

@3* ordinary characters (not %), printed unchanged,

@3* conversions specifications (% followed by some characters) which fetch one argument from the list and prints it according to the specification.

More precisely, a conversion specification consists in a %, one or more optional flags (among #, 0, -, +, ` '), an optional decimal digit string specifying a minimal field width, an optional precision in the form of a period (`.') followed by a decimal digit string, and the conversion specifier (among d,i, o, u, x,X, p, e,E, f, g,G, s).

@3The flag characters. The character % is followed by zero or more of the following flags:

@3* #: The value is converted to an ``alternate form''. For o conversion (octal), a 0 is prefixed to the string. For x and X conversions (hexa), respectively 0x and 0X are prepended. For other conversions, the flag is ignored.

@3* 0: The value should be zero padded. For d, i, o, u, x, X e, E, f, F, g, and G conversions, the value is padded on the left with zeros rather than blanks. (If the 0 and - flags both appear, the 0 flag is ignored.)

@3* -: The value is left adjusted on the field boundary. (The default is right justification.) The value is padded on the right with blanks, rather than on the left with blanks or zeros. A - overrides a 0 if both are given.

@3* ` ' (a space): A blank is left before a positive number produced by a signed conversion.

@3* +: A sign (+ or -) is placed before a number produced by a signed conversion. A + overrides a space if both are used.

@3The field width. An optional decimal digit string (whose first digit is non-zero) specifying a minimum field width. If the value has fewer characters than the field width, it is padded with spaces on the left (or right, if the left-adjustment flag has been given). In no case does a small field width cause truncation of a field; if the value is wider than the field width, the field is expanded to contain the conversion result. Instead of a decimal digit string, one may write * to specify that the field width is given in the next argument.

@3The precision. An optional precision in the form of a period (`.') followed by a decimal digit string. This gives the number of digits to appear after the radix character for e, E, f, and F conversions, the maximum number of significant digits for g and G conversions, and the maximum number of characters to be printed from an s conversion. Instead of a decimal digit string, one may write * to specify that the field width is given in the next argument.

@3The length modifier. This is ignored under gp, but necessary for libpari programming. Description given here for completeness:

@3* l: argument is a long integer.

@3* P: argument is a GEN.

@3The conversion specifier. A character that specifies the type of conversion to be applied.

@3* d, i: A signed integer.

@3* o, u, x, X: An unsigned integer, converted to unsigned octal (o), decimal (u) or hexadecimal (x or X) notation. The letters abcdef are used for x conversions; the letters ABCDEF are used for X conversions.

@3* e, E: The (real) argument is converted in the style [ -]d.ddd e[ -]dd, where there is one digit before the decimal point, and the number of digits after it is equal to the precision; if the precision is missing, use the current realprecision for the total number of printed digits. If the precision is explicitly 0, no decimal-point character appears. An E conversion uses the letter E rather than e to introduce the exponent.

@3* f, F: The (real) argument is converted in the style [ -]ddd.ddd, where the number of digits after the decimal point is equal to the precision; if the precision is missing, use the current realprecision for the total number of printed digits. If the precision is explicitly 0, no decimal-point character appears. If a decimal point appears, at least one digit appears before it.

@3* g, G: The (real) argument is converted in style e or f (or E or F for G conversions) [ -]ddd.ddd, where the total number of digits printed is equal to the precision; if the precision is missing, use the current realprecision. If the precision is explicitly 0, it is treated as 1. Style e is used when the decimal exponent is < -4, to print 0., or when the integer part cannot be decided given the known significant digits, and the f format otherwise.

@3* c: The integer argument is converted to an unsigned char, and the resulting character is written.

@3* s: Convert to a character string. If a precision is given, no more than the specified number of characters are written.

@3* p: Print the address of the argument in hexadecimal (as if by %#x).

@3* %: A % is written. No argument is converted. The complete conversion specification is %%.

@3Examples:

  ? printf("floor: %d, field width 3: %3d, with sign: %+3d\n", Pi, 1, 2);
  floor: 3, field width 3:   1, with sign:  +2
  ? printf("%.5g %.5g %.5g\n",123,123/456,123456789);
  123.00 0.26974 1.2346 e8
  ? printf("%-2.5s:%2.5s:%2.5s\n", "P", "PARI", "PARIGP");
  P :PARI:PARIG
  \\ min field width and precision given by arguments
  ? x = 23; y=-1/x; printf("x=%+06.2f y=%+0*.*f\n", x, 6, 2, y);
  x=+23.00 y=-00.04
  \\ minimum fields width 5, pad left with zeroes
  ? for (i = 2, 5, printf("%05d\n", 10^i))
  00100
  01000
  10000
  100000  \\ don't truncate fields whose length is larger than the minimum width
  ? printf("%.2f  |%06.2f|", Pi,Pi)
  3.14  |  3.14|

@3All numerical conversions apply recursively to the entries of vectors and matrices:

  ? printf("%4d", [1,2,3]);
  [   1,   2,   3]
  ? printf("%5.2f", mathilbert(3));
  [ 1.00  0.50  0.33]
  [ 0.50  0.33  0.25]
  [ 0.33  0.25  0.20]

@3Technical note. Our implementation of printf deviates from the C89 and C99 standards in a few places:

@3* whenever a precision is missing, the current realprecision is used to determine the number of printed digits (C89: use 6 decimals after the radix character).

@3* in conversion style e, we do not impose that the exponent has at least two digits; we never write a + sign in the exponent; 0 is printed in a special way, always as 0.Eexp.

@3* in conversion style f, we switch to style e if the exponent is greater or equal to the precision.

@3* in conversion g and G, we do not remove trailing zeros from the fractional part of the result; nor a trailing decimal point; 0 is printed in a special way, always as 0.Eexp.

printsep(sep,{str}*)

Outputs its (string) arguments in raw format, ending with a newline. Successive entries are separated by sep:

  ? printsep(":", 1,2,3,4)
  1:2:3:4

printsep1(sep,{str}*)

Outputs its (string) arguments in raw format, without ending with a newline. Successive entries are separated by sep:

  ? printsep1(":", 1,2,3,4);print("|")
  1:2:3:4

printtex({str}*)

Outputs its (string) arguments in TeX format. This output can then be used in a TeX manuscript. The printing is done on the standard output. If you want to print it to a file you should use writetex (see there).

Another possibility is to enable the log default (see Label se:defaults). You could for instance do:

  default(logfile, "new.tex");
  default(log, 1);
  printtex(result);

quit({status = 0})

Exits gp and return to the system with exit status status, a small integer. A non-zero exit status normally indicates abnormal termination. (Note: the system actually sees only status mod 256, see your man pages for exit(3) or wait(2)).

read({filename})

Reads in the file filename (subject to string expansion). If filename is omitted, re-reads the last file that was fed into gp. The return value is the result of the last expression evaluated.

If a GP binary file is read using this command (see Label se:writebin), the file is loaded and the last object in the file is returned.

In case the file you read in contains an allocatemem statement (to be generally avoided), you should leave read instructions by themselves, and not part of larger instruction sequences.

readstr({filename})

Reads in the file filename and return a vector of GP strings, each component containing one line from the file. If filename is omitted, re-reads the last file that was fed into gp.

readvec({filename})

Reads in the file filename (subject to string expansion). If filename is omitted, re-reads the last file that was fed into gp. The return value is a vector whose components are the evaluation of all sequences of instructions contained in the file. For instance, if file contains

  1
  2
  3

then we will get:

  ? \r a
  %1 = 1
  %2 = 2
  %3 = 3
  ? read(a)
  %4 = 3
  ? readvec(a)
  %5 = [1, 2, 3]

In general a sequence is just a single line, but as usual braces and \ may be used to enter multiline sequences.

The library syntax is GEN gp_readvec_file(const char *filename). The underlying library function GEN gp_readvec_stream(FILE *f) is usually more flexible.

select(f, A, {flag = 0})

We first describe the default behavior, when flag is 0 or omitted. Given a vector or list A and a t_CLOSURE f, select returns the elements x of A such that f(x) is non-zero. In other words, f is seen as a selection function returning a boolean value.

  ? select(x->isprime(x), vector(50,i,i^2+1))
  %1 = [2, 5, 17, 37, 101, 197, 257, 401, 577, 677, 1297, 1601]
  ? select(x->(x<100), %)
  %2 = [2, 5, 17, 37]

@3returns the primes of the form i^2+1 for some i <= 50, then the elements less than 100 in the preceding result. The select function also applies to a matrix A, seen as a vector of columns, i.e. it selects columns instead of entries, and returns the matrix whose columns are the selected ones.

@3Remark. For v a t_VEC, t_COL, t_LIST or t_MAT, the alternative set-notations

  [g(x) | x <- v, f(x)]
  [x | x <- v, f(x)]
  [g(x) | x <- v]

are available as shortcuts for

  apply(g, select(f, Vec(v)))
  select(f, Vec(v))
  apply(g, Vec(v))

@3respectively:

  ? [ x | x <- vector(50,i,i^2+1), isprime(x) ]
  %1 = [2, 5, 17, 37, 101, 197, 257, 401, 577, 677, 1297, 1601]

@3If flag = 1, this function returns instead the indices of the selected elements, and not the elements themselves (indirect selection):

  ? V = vector(50,i,i^2+1);
  ? select(x->isprime(x), V, 1)
  %2 = Vecsmall([1, 2, 4, 6, 10, 14, 16, 20, 24, 26, 36, 40])
  ? vecextract(V, %)
  %3 = [2, 5, 17, 37, 101, 197, 257, 401, 577, 677, 1297, 1601]

The following function lists the elements in (Z/NZ)^*:

  ? invertibles(N) = select(x->gcd(x,N) == 1, [1..N])

@3Finally

  ? select(x->x, M)

@3selects the non-0 entries in M. If the latter is a t_MAT, we extract the matrix of non-0 columns. Note that removing entries instead of selecting them just involves replacing the selection function f with its negation:

  ? select(x->!isprime(x), vector(50,i,i^2+1))

The library syntax is genselect(void *E, long (*fun)(void*,GEN), GEN a). Also available is GEN genindexselect(void *E, long (*fun)(void*, GEN), GEN a), corresponding to flag = 1.

setrand(n)

Reseeds the random number generator using the seed n. No value is returned. The seed is either a technical array output by getrand, or a small positive integer, used to generate deterministically a suitable state array. For instance, running a randomized computation starting by setrand(1) twice will generate the exact same output.

The library syntax is void setrand(GEN n).

system(str)

str is a string representing a system command. This command is executed, its output written to the standard output (this won't get into your logfile), and control returns to the PARI system. This simply calls the C system command.

trap({e}, {rec}, seq)

THIS FUNCTION IS OBSOLETE: use iferr, which has a nicer and much more powerful interface. For compatibility's sake we now describe the obsolete function trap.

This function tries to evaluate seq, trapping runtime error e, that is effectively preventing it from aborting computations in the usual way; the recovery sequence rec is executed if the error occurs and the evaluation of rec becomes the result of the command. If e is omitted, all exceptions are trapped. See Label se:errorrec for an introduction to error recovery under gp.

  ? \\ trap division by 0
  ? inv(x) = trap (e_INV, INFINITY, 1/x)
  ? inv(2)
  %1 = 1/2
  ? inv(0)
  %2 = INFINITY

Note that seq is effectively evaluated up to the point that produced the error, and the recovery sequence is evaluated starting from that same context, it does not "undo" whatever happened in the other branch (restore the evaluation context):

  ? x = 1; trap (, /* recover: */ x, /* try: */ x = 0; 1/x)
  %1 = 0

@3Note. The interface is currently not adequate for trapping individual exceptions. In the current version 2.7.4, the following keywords are recognized, but the name list will be expanded and changed in the future (all library mode errors can be trapped: it's a matter of defining the keywords to gp):

e_ALARM: alarm time-out

e_ARCH: not available on this architecture or operating system

e_STACK: the PARI stack overflows

e_INV: impossible inverse

e_IMPL: not yet implemented

e_OVERFLOW: all forms of arithmetic overflow, including length or exponent overflow (when a larger value is supplied than the implementation can handle).

e_SYNTAX: syntax error

e_MISC: miscellaneous error

e_TYPE: wrong type

e_USER: user error (from the error function)

The library syntax is GEN trap0(const char *e = NULL, GEN rec = NULL, GEN seq = NULL).

type(x)

This is useful only under gp. Returns the internal type name of the PARI object x as a string. Check out existing type names with the metacommand \t. For example type(1) will return "t_INT".

The library syntax is GEN type0(GEN x). The macro typ is usually simpler to use since it returns a long that can easily be matched with the symbols t_*. The name type was avoided since it is a reserved identifier for some compilers.

uninline()

(Experimental) Exit the scope of all current inline variables.

version()

Returns the current version number as a t_VEC with three integer components (major version number, minor version number and patchlevel); if your sources were obtained through our version control system, this will be followed by further more precise arguments, including e.g. a git commit hash.

This function is present in all versions of PARI following releases 2.3.4 (stable) and 2.4.3 (testing).

Unless you are working with multiple development versions, you probably only care about the 3 first numeric components. In any case, the lex function offers a clever way to check against a particular version number, since it will compare each successive vector entry, numerically or as strings, and will not mind if the vectors it compares have different lengths:

     if (lex(version(), [2,3,5]) >= 0,
       \\ code to be executed if we are running 2.3.5 or more recent.
     ,
       \\ compatibility code
     );

@3On a number of different machines, version() could return either of

   %1 = [2, 3, 4]    \\ released version, stable branch
   %1 = [2, 4, 3]    \\ released version, testing branch
   %1 = [2, 6, 1, 15174, ""505ab9b"] \\ development

In particular, if you are only working with released versions, the first line of the gp introductory message can be emulated by

     [M,m,p] = version();
     printf("GP/PARI CALCULATOR Version %s.%s.%s", M,m,p);

@3If you are working with many development versions of PARI/GP, the 4th and/or 5th components can be profitably included in the name of your logfiles, for instance.

@3Technical note. For development versions obtained via git, the 4th and 5th components are liable to change eventually, but we document their current meaning for completeness. The 4th component counts the number of reachable commits in the branch (analogous to svn's revision number), and the 5th is the git commit hash. In particular, lex comparison still orders correctly development versions with respect to each others or to released versions (provided we stay within a given branch, e.g. master)!

The library syntax is GEN pari_version().

warning({str}*)

Outputs the message ``user warning'' and the argument list (each of them interpreted as a string). If colors are enabled, this warning will be in a different color, making it easy to distinguish.

  warning(n, " is very large, this might take a while.")

whatnow(key)

If keyword key is the name of a function that was present in GP version 1.39.15 or lower, outputs the new function name and syntax, if it changed at all (387 out of 560 did).

write(filename,{str}*)

Writes (appends) to filename the remaining arguments, and appends a newline (same output as print).

write1(filename,{str}*)

Writes (appends) to filename the remaining arguments without a trailing newline (same output as print1).

writebin(filename,{x})

Writes (appends) to filename the object x in binary format. This format is not human readable, but contains the exact internal structure of x, and is much faster to save/load than a string expression, as would be produced by write. The binary file format includes a magic number, so that such a file can be recognized and correctly input by the regular read or \r function. If saved objects refer to (polynomial) variables that are not defined in the new session, they will be displayed in a funny way (see Label se:kill). Installed functions and history objects can not be saved via this function.

If x is omitted, saves all user variables from the session, together with their names. Reading such a ``named object'' back in a gp session will set the corresponding user variable to the saved value. E.g after

  x = 1; writebin("log")

reading log into a clean session will set x to 1. The relative variables priorities (see Label se:priority) of new variables set in this way remain the same (preset variables retain their former priority, but are set to the new value). In particular, reading such a session log into a clean session will restore all variables exactly as they were in the original one.

Just as a regular input file, a binary file can be compressed using gzip, provided the file name has the standard .gz extension.

In the present implementation, the binary files are architecture dependent and compatibility with future versions of gp is not guaranteed. Hence binary files should not be used for long term storage (also, they are larger and harder to compress than text files).

The library syntax is void gpwritebin(const char *filename, GEN x = NULL).

writetex(filename,{str}*)

As write, in TeX format.


Parallel programming

These function are only available if PARI was configured using Configure --mt = .... Two multithread interfaces are supported:

@3* POSIX threads

@3* Message passing interface (MPI)

As a rule, POSIX threads are well-suited for single systems, while MPI is used by most clusters. However the parallel GP interface does not depend on the chosen multithread interface: a properly written GP program will work identically with both.

parapply(f, x)

Parallel evaluation of f on the elements of x. The function f must not access global variables or variables declared with local(), and must be free of side effects.

  parapply(factor,[2^256 + 1, 2^193 - 1])

factors 2^{256} + 1 and 2^{193} - 1 in parallel.

  {
    my(E = ellinit([1,3]), V = vector(12,i,randomprime(2^200)));
    parapply(p->ellcard(E,p), V)
  }

computes the order of E(F_p) for 12 random primes of 200 bits.

The library syntax is GEN parapply(GEN f, GEN x).

pareval(x)

Parallel evaluation of the elements of x, where x is a vector of closures. The closures must be of arity 0, must not access global variables or variables declared with local and must be free of side effects.

The library syntax is GEN pareval(GEN x).

parfor(i = a,{b},expr1,{j},{expr2})

Evaluates the sequence expr2 (dependent on i and j) for i between a and b, in random order, computed in parallel; in this sequence expr2, substitute the variable j by the value of expr1 (dependent on i). If b is omitted, the loop will not stop.

It is allowed for expr2 to exit the loop using break/next/return; however in that case, expr2 will still be evaluated for all remaining value of i less than the current one, unless a subsequent break/next/return happens.

parforprime(p = a,{b},expr1,{j},{expr2})

Evaluates the sequence expr2 (dependent on p and j) for p prime between a and b, in random order, computed in parallel. Substitute for j the value of expr1 (dependent on p). If b is omitted, the loop will not stop.

It is allowed fo expr2 to exit the loop using break/next/return, however in that case, expr2 will still be evaluated for all remaining value of p less than the current one, unless a subsequent break/next/return happens.

parselect(f, A, {flag = 0})

Selects elements of A according to the selection function f, done in parallel. If flag is 1, return the indices of those elements (indirect selection) The function f must not access global variables or variables declared with local(), and must be free of side effects.

The library syntax is GEN parselect(GEN f, GEN A, long flag ).

parsum(i = a,b,expr,{x})

Sum of expression expr, initialized at x, the formal parameter going from a to b, evaluated in parallel in random order. The expression expr must not access global variables or variables declared with local(), and must be free of side effects.

  parsum(i=1,1000,ispseudoprime(2^prime(i)-1))

returns the numbers of prime numbers among the first 1000 Mersenne numbers.

parvector(N,i,expr)

As vector(N,i,expr) but the evaluations of expr are done in parallel. The expression expr must not access global variables or variables declared with local(), and must be free of side effects.

  parvector(10,i,quadclassunit(2^(100+i)+1).no)

computes the class numbers in parallel.


GP defaults

This section documents the GP defaults

TeXstyle\label{se:def,TeXstyle}

The bits of this default allow gp to use less rigid TeX formatting commands in the logfile. This default is only taken into account when log = 3. The bits of TeXstyle have the following meaning

2: insert \right / \left pairs where appropriate.

4: insert discretionary breaks in polynomials, to enhance the probability of a good line break.

The default value is 0.

breakloop\label{se:def,breakloop}

If true, enables the ``break loop'' debugging mode, see Label se:break_loop.

The default value is 1 if we are running an interactive gp session, and 0 otherwise.

colors\label{se:def,colors}

This default is only usable if gp is running within certain color-capable terminals. For instance rxvt, color_xterm and modern versions of xterm under X Windows, or standard Linux/DOS text consoles. It causes gp to use a small palette of colors for its output. With xterms, the colormap used corresponds to the resources Xterm*colorn where n ranges from 0 to 15 (see the file misc/color.dft for an example). Accepted values for this default are strings "a_1,...,a_k" where k <= 7 and each a_i is either

@3* the keyword no (use the default color, usually black on transparent background)

@3* an integer between 0 and 15 corresponding to the aforementioned colormap

@3* a triple [c_0,c_1,c_2] where c_0 stands for foreground color, c_1 for background color, and c_2 for attributes (0 is default, 1 is bold, 4 is underline).

The output objects thus affected are respectively error messages, history numbers, prompt, input line, output, help messages, timer (that's seven of them). If k < 7, the remaining a_i are assumed to be no. For instance

  default(colors, "9, 5, no, no, 4")

typesets error messages in color 9, history numbers in color 5, output in color 4, and does not affect the rest.

A set of default colors for dark (reverse video or PC console) and light backgrounds respectively is activated when colors is set to darkbg, resp. lightbg (or any proper prefix: d is recognized as an abbreviation for darkbg). A bold variant of darkbg, called boldfg, is provided if you find the former too pale.

 In the present version, this default is incompatible with PariEmacs.
Changing it will just fail silently (the alternative would be to display
escape sequences as is, since Emacs will refuse to interpret them).
You must customize color highlighting from the PariEmacs side, see its
documentation.

The default value is "" (no colors).

compatible\label{se:def,compatible}

The GP function names and syntax have changed tremendously between versions 1.xx and 2.00. To help you cope with this we provide some kind of backward compatibility, depending on the value of this default:

  compatible = 0: no backward compatibility. In this mode, a very handy function, to be described in Label se:whatnow, is whatnow, which tells you what has become of your favorite functions, which gp suddenly can't seem to remember.

  compatible = 1: warn when using obsolete functions, but otherwise accept them. The output uses the new conventions though, and there may be subtle incompatibilities between the behavior of former and current functions, even when they share the same name (the current function is used in such cases, of course!). We thought of this one as a transitory help for gp old-timers. Thus, to encourage switching to compatible = 0, it is not possible to disable the warning.

  compatible = 2: use only the old function naming scheme (as used up to version 1.39.15), but taking case into account. Thus I ( = sqrt {-1}) is not the same as i (user variable, unbound by default), and you won't get an error message using i as a loop index as used to be the case.

  compatible = 3: try to mimic exactly the former behavior. This is not always possible when functions have changed in a fundamental way. But these differences are usually for the better (they were meant to, anyway), and will probably not be discovered by the casual user.

One adverse side effect is that any user functions and aliases that have been defined before changing compatible will get erased if this change modifies the function list, i.e. if you move between groups {0,1} and {2,3} (variables are unaffected). We of course strongly encourage you to try and get used to the setting compatible = 0.

Note that the default new_galois_format is another compatibility setting, which is completely independent of compatible.

The default value is 0.

datadir\label{se:def,datadir}

The name of directory containing the optional data files. For now, this includes the elldata, galdata, galpol, seadata packages.

The default value is \datadir (the location of installed precomputed data, can be specified via Configure --datadir = ).

debug\label{se:def,debug}

Debugging level. If it is non-zero, some extra messages may be printed, according to what is going on (see \g).

The default value is 0 (no debugging messages).

debugfiles\label{se:def,debugfiles}

File usage debugging level. If it is non-zero, gp will print information on file descriptors in use, from PARI's point of view (see \gf).

The default value is 0 (no debugging messages).

debugmem\label{se:def,debugmem}

Memory debugging level. If it is non-zero, gp will regularly print information on memory usage. If it's greater than 2, it will indicate any important garbage collecting and the function it is taking place in (see \gm).

@3Important Note: As it noticeably slows down the performance, the first functionality (memory usage) is disabled if you're not running a version compiled for debugging (see Appendix A).

The default value is 0 (no debugging messages).

echo\label{se:def,echo}

This toggle is either 1 (on) or 0 (off). When echo mode is on, each command is reprinted before being executed. This can be useful when reading a file with the \r or read commands. For example, it is turned on at the beginning of the test files used to check whether gp has been built correctly (see \e).

The default value is 0 (no echo).

factor_add_primes\label{se:def,factor_add_primes}

This toggle is either 1 (on) or 0 (off). If on, the integer factorization machinery calls addprimes on primes factor that were difficult to find (larger than 2^24), so they are automatically tried first in other factorizations. If a routine is performing (or has performed) a factorization and is interrupted by an error or via Control-C, this lets you recover the prime factors already found. The downside is that a huge addprimes table unrelated to the current computations will slow down arithmetic functions relying on integer factorization; one should then empty the table using removeprimes.

The default value is 0.

factor_proven\label{se:def,factor_proven}

This toggle is either 1 (on) or 0 (off). By default, the factors output by the integer factorization machinery are only pseudo-primes, not proven primes. If this toggle is set, a primality proof is done for each factor and all results depending on integer factorization are fully proven. This flag does not affect partial factorization when it is explicitly requested. It also does not affect the private table managed by addprimes: its entries are included as is in factorizations, without being tested for primality.

The default value is 0.

format\label{se:def,format}

Of the form x.n, where x (conversion style) is a letter in {e,f,g}, and n (precision) is an integer; this affects the way real numbers are printed:

@3* If the conversion style is e, real numbers are printed in scientific format, always with an explicit exponent, e.g. 3.3 E-5.

@3* In style f, real numbers are generally printed in fixed floating point format without exponent, e.g. 0.000033. A large real number, whose integer part is not well defined (not enough significant digits), is printed in style e. For instance 10.^100 known to ten significant digits is always printed in style e.

@3* In style g, non-zero real numbers are printed in f format, except when their decimal exponent is < -4, in which case they are printed in e format. Real zeroes (of arbitrary exponent) are printed in e format.

The precision n is the number of significant digits printed for real numbers, except if n < 0 where all the significant digits will be printed (initial default 28, or 38 for 64-bit machines). For more powerful formatting possibilities, see printf and Strprintf.

The default value is "g.28" and "g.38" on 32-bit and 64-bit machines, respectively.

graphcolormap\label{se:def,graphcolormap}

A vector of colors, to be used by hi-res graphing routines. Its length is arbitrary, but it must contain at least 3 entries: the first 3 colors are used for background, frame/ticks and axes respectively. All colors in the colormap may be freely used in plotcolor calls.

A color is either given as in the default by character strings or by an RGB code. For valid character strings, see the standard rgb.txt file in X11 distributions, where we restrict to lowercase letters and remove all whitespace from color names. An RGB code is a vector with 3 integer entries between 0 and 255. For instance [250, 235, 215] and "antiquewhite" represent the same color. RGB codes are cryptic but often easier to generate.

The default value is ["white", "black", "blue", "violetred", "red", "green", "grey", "gainsboro"].

graphcolors\label{se:def,graphcolors}

Entries in the graphcolormap that will be used to plot multi-curves. The successive curves are drawn in colors

graphcolormap[graphcolors[1]], graphcolormap[graphcolors[2]], ...

cycling when the graphcolors list is exhausted.

The default value is [4,5].

help\label{se:def,help}

Name of the external help program to use from within gp when extended help is invoked, usually through a ?? or ??? request (see Label se:exthelp), or M-H under readline (see Label se:readline).

The default value is the path to the gphelp script we install.

histfile\label{se:def,histfile}

Name of a file where gp will keep a history of all input commands (results are omitted). If this file exists when the value of histfile changes, it is read in and becomes part of the session history. Thus, setting this default in your gprc saves your readline history between sessions. Setting this default to the empty string "" changes it to < undefined >

The default value is < undefined > (no history file).

histsize\label{se:def,histsize}

gp keeps a history of the last histsize results computed so far, which you can recover using the % notation (see Label se:history). When this number is exceeded, the oldest values are erased. Tampering with this default is the only way to get rid of the ones you do not need anymore.

The default value is 5000.

lines\label{se:def,lines}

If set to a positive value, gp prints at most that many lines from each result, terminating the last line shown with [+++] if further material has been suppressed. The various print commands (see Label se:gp_program) are unaffected, so you can always type print(%) or \a to view the full result. If the actual screen width cannot be determined, a ``line'' is assumed to be 80 characters long.

The default value is 0.

linewrap\label{se:def,linewrap}

If set to a positive value, gp wraps every single line after printing that many characters.

The default value is 0 (unset).

log\label{se:def,log}

This can be either 0 (off) or 1, 2, 3 (on, see below for the various modes). When logging mode is turned on, gp opens a log file, whose exact name is determined by the logfile default. Subsequently, all the commands and results will be written to that file (see \l). In case a file with this precise name already existed, it will not be erased: your data will be appended at the end.

The specific positive values of log have the following meaning

1: plain logfile

2: emit color codes to the logfile (if colors is set).

3: write LaTeX output to the logfile (can be further customized using TeXstyle).

The default value is 0.

logfile\label{se:def,logfile}

Name of the log file to be used when the log toggle is on. Environment and time expansion are performed.

The default value is "pari.log".

nbthreads\label{se:def,nbthreads}

Number of threads to use for parallel computing. The exact meaning an default depend on the mt engine used:

@3* single: not used (always one thread).

@3* pthread: number of threads (unlimited, default: number of core)

@3* mpi: number of MPI process to use (limited to the number allocated by mpirun, default: use all allocated process).

new_galois_format\label{se:def,new_galois_format}

This toggle is either 1 (on) or 0 (off). If on, the polgalois command will use a different, more consistent, naming scheme for Galois groups. This default is provided to ensure that scripts can control this behavior and do not break unexpectedly.

The default value is 0. This value will change to 1 (set) in the next major version.

output\label{se:def,output}

There are three possible values: 0 ( =  raw), 1 ( =  prettymatrix), or 3 ( =  external prettyprint). This means that, independently of the default format for reals which we explained above, you can print results in three ways:

@3* raw format, i.e. a format which is equivalent to what you input, including explicit multiplication signs, and everything typed on a line instead of two dimensional boxes. This can have several advantages, for instance it allows you to pick the result with a mouse or an editor, and to paste it somewhere else.

@3* prettymatrix format: this is identical to raw format, except that matrices are printed as boxes instead of horizontally. This is prettier, but takes more space and cannot be used for input. Column vectors are still printed horizontally.

@3* external prettyprint: pipes all gp output in TeX format to an external prettyprinter, according to the value of prettyprinter. The default script (tex2mail) converts its input to readable two-dimensional text.

Independently of the setting of this default, an object can be printed in any of the three formats at any time using the commands \a and \m and \B respectively.

The default value is 1 (prettymatrix).

parisize\label{se:def,parisize}

gp, and in fact any program using the PARI library, needs a stack in which to do its computations. parisize is the stack size, in bytes. It is strongly recommended you increase this default (using the -s command-line switch, or a gprc) if you can afford it. Don't increase it beyond the actual amount of RAM installed on your computer or gp will spend most of its time paging.

In case of emergency, you can use the allocatemem function to increase parisize, once the session is started.

The default value is 4M, resp. 8M on a 32-bit, resp. 64-bit machine.

path\label{se:def,path}

This is a list of directories, separated by colons ':' (semicolons ';' in the DOS world, since colons are preempted for drive names). When asked to read a file whose name is not given by an absolute path (does not start with /, ./ or ../), gp will look for it in these directories, in the order they were written in path. Here, as usual, . means the current directory, and .. its immediate parent. Environment expansion is performed.

The default value is ".:~:~/gp" on UNIX systems, ".;C:\;C:\GP" on DOS, OS/2 and Windows, and "." otherwise.

prettyprinter\label{se:def,prettyprinter}

The name of an external prettyprinter to use when output is 3 (alternate prettyprinter). Note that the default tex2mail looks much nicer than the built-in ``beautified format'' (output = 2).

The default value is "tex2mail -TeX -noindent -ragged -by_par".

primelimit\label{se:def,primelimit}

gp precomputes a list of all primes less than primelimit at initialization time, and can build fast sieves on demand to quickly iterate over primes up to the square of primelimit. These are used by many arithmetic functions, usually for trial division purposes. The maximal value is 2^{32} - 2049 (resp 2^{64} - 2049) on a 32-bit (resp. 64-bit) machine, but values beyond 10^8, allowing to iterate over primes up to 10^{16}, do not seem useful.

Since almost all arithmetic functions eventually require some table of prime numbers, PARI guarantees that the first 6547 primes, up to and including 65557, are precomputed, even if primelimit is 1.

This default is only used on startup: changing it will not recompute a new table.

@3Deprecated feature. primelimit was used in some situations by algebraic number theory functions using the nf_PARTIALFACT flag (nfbasis, nfdisc, nfinit,...): this assumes that all primes p > primelimit have a certain property (the equation order is p-maximal). This is never done by default, and must be explicitly set by the user of such functions. Nevertheless, these functions now provide a more flexible interface, and their use of the global default primelimit is deprecated.

@3Deprecated feature. factor(N, 0) was used to partially factor integers by removing all prime factors <= primelimit. Don't use this, supply an explicit bound: factor(N, bound), which avoids relying on an unpredictable global variable.

The default value is 500k.

prompt\label{se:def,prompt}

A string that will be printed as prompt. Note that most usual escape sequences are available there: \e for Esc, \n for Newline,..., \\ for \. Time expansion is performed.

This string is sent through the library function strftime (on a Unix system, you can try man strftime at your shell prompt). This means that % constructs have a special meaning, usually related to the time and date. For instance, %H = hour (24-hour clock) and %M = minute [00,59] (use %% to get a real %).

If you use readline, escape sequences in your prompt will result in display bugs. If you have a relatively recent readline (see the comment at the end of \secref{se:def,colors}), you can brace them with special sequences (\[ and \]), and you will be safe. If these just result in extra spaces in your prompt, then you'll have to get a more recent readline. See the file misc/gprc.dft for an example.

S< >Caution: PariEmacs needs to know about the prompt pattern to separate your input from previous gp results, without ambiguity. It is not a trivial problem to adapt automatically this regular expression to an arbitrary prompt (which can be self-modifying!). See PariEmacs's documentation.

The default value is "? ".

prompt_cont\label{se:def,prompt_cont}

A string that will be printed to prompt for continuation lines (e.g. in between braces, or after a line-terminating backslash). Everything that applies to prompt applies to prompt_cont as well.

The default value is "".

psfile\label{se:def,psfile}

Name of the default file where gp is to dump its PostScript drawings (these are appended, so that no previous data are lost). Environment and time expansion are performed.

The default value is "pari.ps".

readline\label{se:def,readline}

Switches readline line-editing facilities on and off. This may be useful if you are running gp in a Sun cmdtool, which interacts badly with readline. Of course, until readline is switched on again, advanced editing features like automatic completion and editing history are not available.

The default value is 1.

realprecision\label{se:def,realprecision}

The number of significant digits used to convert exact inputs given to transcendental functions (see Label se:trans), or to create absolute floating point constants (input as 1.0 or Pi for instance). Unless you tamper with the format default, this is also the number of significant digits used to print a t_REAL number; format will override this latter behaviour, and allow you to have a large internal precision while outputting few digits for instance.

Note that PARI's internal precision works on a word basis (by increments of 32 or 64 bits), hence may be a little larger than the number of decimal digits you expected. For instance to get 2 decimal digits you need one word of precision which, on a 64-bit machine, actually gives you 19 digits (19 < log _{10}(2^{64}) < 20). The value returned when typing default(realprecision) is the internal number of significant digits, not the number of printed digits:

  ? default(realprecision, 2)
        realprecision = 19 significant digits (2 digits displayed)
  ? default(realprecision)
  %1 = 19

The default value is 38, resp. 28, on a 64-bit, resp .32-bit, machine.

recover\label{se:def,recover}

This toggle is either 1 (on) or 0 (off). If you change this to 0, any error becomes fatal and causes the gp interpreter to exit immediately. Can be useful in batch job scripts.

The default value is 1.

secure\label{se:def,secure}

This toggle is either 1 (on) or 0 (off). If on, the system and extern command are disabled. These two commands are potentially dangerous when you execute foreign scripts since they let gp execute arbitrary UNIX commands. gp will ask for confirmation before letting you (or a script) unset this toggle.

The default value is 0.

seriesprecision\label{se:def,seriesprecision}

Number of significant terms when converting a polynomial or rational function to a power series (see \ps).

The default value is 16.

simplify\label{se:def,simplify}

This toggle is either 1 (on) or 0 (off). When the PARI library computes something, the type of the result is not always the simplest possible. The only type conversions which the PARI library does automatically are rational numbers to integers (when they are of type t_FRAC and equal to integers), and similarly rational functions to polynomials (when they are of type t_RFRAC and equal to polynomials). This feature is useful in many cases, and saves time, but can be annoying at times. Hence you can disable this and, whenever you feel like it, use the function simplify (see Chapter 3) which allows you to simplify objects to the simplest possible types recursively (see \y).

The default value is 1.

sopath\label{se:def,sopath}

This is a list of directories, separated by colons ':' (semicolons ';' in the DOS world, since colons are preempted for drive names). When asked to install an external symbol from a shared library whose name is not given by an absolute path (does not start with /, ./ or ../), gp will look for it in these directories, in the order they were written in sopath. Here, as usual, . means the current directory, and .. its immediate parent. Environment expansion is performed.

The default value is "", corresponding to an empty list of directories: install will use the library name as input (and look in the current directory if the name is not an absolute path).

strictargs\label{se:def,strictargs}

This toggle is either 1 (on) or 0 (off). If on, all arguments to new user functions are mandatory unless the function supplies an explicit default value. Otherwise arguments have the default value 0.

In this example,

    fun(a,b=2)=a+b

a is mandatory, while b is optional. If strictargs is on:

  ? fun()
   ***   at top-level: fun()
   ***                 ^-----
   ***   in function fun: a,b=2
   ***                    ^-----
   ***   missing mandatory argument 'a' in user function.

This applies to functions defined while strictargs is on. Changing strictargs does not affect the behavior of previously defined functions.

The default value is 0.

strictmatch\label{se:def,strictmatch}

This toggle is either 1 (on) or 0 (off). If on, unused characters after a sequence has been processed will produce an error. Otherwise just a warning is printed. This can be useful when you are unsure how many parentheses you have to close after complicated nested loops. Please do not use this; find a decent text-editor instead.

The default value is 1.

threadsize\label{se:def,threadsize}

In parallel mode, each thread needs its own private stack in which to do its computations, see parisize. This value determines the size in bytes of the stacks of each thread, so the total memory allocated will be parisize+nbthreads x threadsize.

If set to 0, the value used is the same as parisize.

The default value is 0.

timer\label{se:def,timer}

This toggle is either 1 (on) or 0 (off). Every instruction sequence in the gp calculator (anything ended by a newline in your input) is timed, to some accuracy depending on the hardware and operating system. When timer is on, each such timing is printed immediately before the output as follows:

  ? factor(2^2^7+1)
  time = 108 ms.     \\ this line omitted if 'timer' is 0
  %1 =
  [     59649589127497217 1]
  [5704689200685129054721 1]

@3(See also # and ##.)

The time measured is the user CPU time, not including the time for printing the results. If the time is negligible ( < 1 ms.), nothing is printed: in particular, no timing should be printed when defining a user function or an alias, or installing a symbol from the library.

The default value is 0 (off).