path: root/src/algebra/intalg.spad.pamphlet

\documentclass{article}
\usepackage{axiom}
\begin{document}
\title{\$SPAD/src/algebra intalg.spad}
\author{Manuel Bronstein}
\maketitle
\begin{abstract}
\end{abstract}
\eject
\tableofcontents
\eject
\section{package DBLRESP DoubleResultantPackage}
<<package DBLRESP DoubleResultantPackage>>=
)abbrev package DBLRESP DoubleResultantPackage
++ Residue resultant
++ Author: Manuel Bronstein
++ Date Created: 1987
++ Date Last Updated: 12 July 1990
++ Description:
++ This package provides functions for computing the residues
++ of a function on an algebraic curve.
DoubleResultantPackage(F, UP, UPUP, R): Exports == Implementation where
  F   : Field
  UP  : UnivariatePolynomialCategory F
  UPUP: UnivariatePolynomialCategory Fraction UP
  R   : FunctionFieldCategory(F, UP, UPUP)

  RF  ==> Fraction UP
  UP2 ==> SparseUnivariatePolynomial UP
  UP3 ==> SparseUnivariatePolynomial UP2

  Exports ==> with
    doubleResultant: (R, UP -> UP) -> UP
      ++ doubleResultant(f, ') returns p(x) whose roots are
      ++ rational multiples of the residues of f at all its
      ++ finite poles. Argument ' is the derivation to use.

  Implementation ==> add
    import CommuteUnivariatePolynomialCategory(F, UP, UP2)
    import UnivariatePolynomialCommonDenominator(UP, RF, UPUP)

    UP22   : UP   -> UP2
    UP23   : UPUP -> UP3
    remove0: UP   -> UP             -- removes the power of x dividing p

    remove0 p ==
      primitivePart((p exquo monomial(1, minimumDegree p))::UP)

    UP22 p ==
      map(#1::UP, p)$UnivariatePolynomialCategoryFunctions2(F,UP,UP,UP2)

    UP23 p ==
      map(UP22(retract(#1)@UP),
          p)$UnivariatePolynomialCategoryFunctions2(RF, UPUP, UP2, UP3)

    doubleResultant(h, derivation) ==
      cd := splitDenominator lift h
      d  := (cd.den exquo (g := gcd(cd.den, derivation(cd.den))))::UP
      r  := swap primitivePart swap resultant(UP23(cd.num)
          - ((monomial(1, 1)$UP :: UP2) * UP22(g * derivation d))::UP3,
                                              UP23 definingPolynomial())
      remove0 resultant(r, UP22 d)

@
\section{package INTHERAL AlgebraicHermiteIntegration}
<<package INTHERAL AlgebraicHermiteIntegration>>=
)abbrev package INTHERAL AlgebraicHermiteIntegration
++ Hermite integration, algebraic case
++ Author: Manuel Bronstein
++ Date Created: 1987
++ Date Last Updated: 25 July 1990
++ Description: algebraic Hermite redution.
AlgebraicHermiteIntegration(F,UP,UPUP,R):Exports == Implementation where
  F   : Field
  UP  : UnivariatePolynomialCategory F
  UPUP: UnivariatePolynomialCategory Fraction UP
  R   : FunctionFieldCategory(F, UP, UPUP)

  N   ==> NonNegativeInteger
  RF  ==> Fraction UP

  Exports ==> with
    HermiteIntegrate: (R, UP -> UP) -> Record(answer:R, logpart:R)
      ++ HermiteIntegrate(f, ') returns \spad{[g,h]} such that
      ++ \spad{f = g' + h} and h has a only simple finite normal poles.

  Implementation ==> add
    localsolve: (Matrix UP, Vector UP, UP) -> Vector UP

-- the denominator of f should have no prime factor P s.t. P | P'
-- (which happens only for P = t in the exponential case)
    HermiteIntegrate(f, derivation) ==
      ratform:R := 0
      n    := rank()
      m    := transpose((mat:= integralDerivationMatrix derivation).num)
      inum := (cform := integralCoordinates f).num
      if ((iden := cform.den) exquo (e := mat.den)) case "failed" then
        iden := (coef := (e exquo gcd(e, iden))::UP) * iden
        inum := coef * inum
      for trm in factors squareFree iden | (j:= trm.exponent) > 1 repeat
        u':=(u:=(iden exquo (v:=trm.factor)**(j::N))::UP) * derivation v
        sys := ((u * v) exquo e)::UP * m
        nn := minRowIndex sys - minIndex inum
        while j > 1 repeat
          j := j - 1
          p := - j * u'
          sol := localsolve(sys + scalarMatrix(n, p), inum, v)
          ratform := ratform + integralRepresents(sol, v ** (j::N))
          inum    := [((qelt(inum, i) - p * qelt(sol, i) -
                        dot(row(sys, i - nn), sol))
                          exquo v)::UP - u * derivation qelt(sol, i)
                             for i in minIndex inum .. maxIndex inum]
        iden := u * v
      [ratform, integralRepresents(inum, iden)]

    localsolve(mat, vec, modulus) ==
      ans:Vector(UP) := new(nrows mat, 0)
      diagonal? mat =>
        for i in minIndex ans .. maxIndex ans
          for j in minRowIndex mat .. maxRowIndex mat
            for k in minColIndex mat .. maxColIndex mat repeat
              (bc := extendedEuclidean(qelt(mat, j, k), modulus,
                qelt(vec, i))) case "failed" => return new(0, 0)
              qsetelt_!(ans, i, bc.coef1)
        ans
      sol := particularSolution(map(#1::RF, mat)$MatrixCategoryFunctions2(UP,
                         Vector UP, Vector UP, Matrix UP, RF,
                           Vector RF, Vector RF, Matrix RF),
                             map(#1::RF, vec)$VectorFunctions2(UP,
                               RF))$LinearSystemMatrixPackage(RF,
                                        Vector RF, Vector RF, Matrix RF)
      sol case "failed" => new(0, 0)
      for i in minIndex ans .. maxIndex ans repeat
        (bc := extendedEuclidean(denom qelt(sol, i), modulus, 1))
          case "failed" => return new(0, 0)
        qsetelt_!(ans, i, (numer qelt(sol, i) * bc.coef1) rem modulus)
      ans

@
\section{package INTALG AlgebraicIntegrate}
<<package INTALG AlgebraicIntegrate>>=
)abbrev package INTALG AlgebraicIntegrate
++ Integration of an algebraic function
++ Author: Manuel Bronstein
++ Date Created: 1987
++ Date Last Updated: 19 May 1993
++ Description:
++ This package provides functions for integrating a function
++ on an algebraic curve.
AlgebraicIntegrate(R0, F, UP, UPUP, R): Exports == Implementation where
  R0   : Join(IntegralDomain, RetractableTo Integer)
  F    : Join(AlgebraicallyClosedField, FunctionSpace R0)
  UP   : UnivariatePolynomialCategory F
  UPUP : UnivariatePolynomialCategory Fraction UP
  R    : FunctionFieldCategory(F, UP, UPUP)

  SE  ==> Symbol
  Z   ==> Integer
  Q   ==> Fraction Z
  SUP ==> SparseUnivariatePolynomial F
  QF  ==> Fraction UP
  GP  ==> LaurentPolynomial(F, UP)
  K   ==> Kernel F
  IR  ==> IntegrationResult R
  UPQ ==> SparseUnivariatePolynomial Q
  UPR ==> SparseUnivariatePolynomial R
  FRQ ==> Factored UPQ
  FD  ==> FiniteDivisor(F, UP, UPUP, R)
  FAC ==> Record(factor:UPQ, exponent:Z)
  LOG ==> Record(scalar:Q, coeff:UPR, logand:UPR)
  DIV ==> Record(num:R, den:UP, derivden:UP, gd:UP)
  FAIL0 ==> error "integrate: implementation incomplete (constant residues)"
  FAIL1==> error "integrate: implementation incomplete (non-algebraic residues)"
  FAIL2 ==> error "integrate: implementation incomplete (residue poly has multiple non-linear factors)"
  FAIL3 ==> error "integrate: implementation incomplete (has polynomial part)"
  NOTI  ==> error "Not integrable (provided residues have no relations)"

  Exports ==> with
    algintegrate  : (R, UP -> UP) -> IR
      ++ algintegrate(f, d) integrates f with respect to the derivation d.
    palgintegrate : (R, UP -> UP) -> IR
      ++ palgintegrate(f, d) integrates f with respect to the derivation d.
      ++ Argument f must be a pure algebraic function.
    palginfieldint: (R, UP -> UP) -> Union(R, "failed")
      ++ palginfieldint(f, d) returns an algebraic function g
      ++ such that \spad{dg = f} if such a g exists, "failed" otherwise.
      ++ Argument f must be a pure algebraic function.

  Implementation ==> add
    import FD
    import DoubleResultantPackage(F, UP, UPUP, R)
    import PointsOfFiniteOrder(R0, F, UP, UPUP, R)
    import AlgebraicHermiteIntegration(F, UP, UPUP, R)
    import InnerCommonDenominator(Z, Q, List Z, List Q)
    import FunctionSpaceUnivariatePolynomialFactor(R0, F, UP)
    import PolynomialCategoryQuotientFunctions(IndexedExponents K,
                         K, R0, SparseMultivariatePolynomial(R0, K), F)

    F2R        : F  -> R
    F2UPR      : F  -> UPR
    UP2SUP     : UP -> SUP
    SUP2UP     : SUP -> UP
    UPQ2F      : UPQ -> UP
    univ       : (F, K) -> QF
    pLogDeriv  : (LOG, R -> R) -> R
    nonLinear  : List FAC -> Union(FAC, "failed")
    mkLog      : (UP, Q, R, F) -> List LOG
    R2UP       : (R, K) -> UPR
    alglogint  : (R, UP -> UP) -> Union(List LOG, "failed")
    palglogint : (R, UP -> UP) -> Union(List LOG, "failed")
    trace00    : (DIV, UP, List LOG) -> Union(List LOG,"failed")
    trace0     : (DIV, UP, Q, FD)    -> Union(List LOG, "failed")
    trace1     : (DIV, UP, List Q, List FD, Q) -> Union(List LOG, "failed")
    nonQ       : (DIV, UP)           -> Union(List LOG, "failed")
    rlift      : (F, K, K) -> R
    varRoot?   : (UP, F -> F) -> Boolean
    algintexp  : (R, UP -> UP) -> IR
    algintprim : (R, UP -> UP) -> IR

    dummy:R := 0

    dumx  := kernel(new()$SE)$K
    dumy  := kernel(new()$SE)$K

    F2UPR f == F2R(f)::UPR
    F2R f   == f::UP::QF::R

    algintexp(f, derivation) ==
      d := (c := integralCoordinates f).den
      v := c.num
      vp:Vector(GP) := new(n := #v, 0)
      vf:Vector(QF) := new(n, 0)
      for i in minIndex v .. maxIndex v repeat
        r := separate(qelt(v, i) / d)$GP
        qsetelt_!(vf, i, r.fracPart)
        qsetelt_!(vp, i, r.polyPart)
      ff := represents(vf, w := integralBasis())
      h := HermiteIntegrate(ff, derivation)
      p := represents(map(convert(#1)@QF, vp)$VectorFunctions2(GP, QF), w)
      zero?(h.logpart) and zero? p => h.answer::IR
      (u := alglogint(h.logpart, derivation)) case "failed" =>
                       mkAnswer(h.answer, empty(), [[p + h.logpart, dummy]])
      zero? p => mkAnswer(h.answer, u::List(LOG), empty())
      FAIL3

    algintprim(f, derivation) ==
      h := HermiteIntegrate(f, derivation)
      zero?(h.logpart) => h.answer::IR
      (u := alglogint(h.logpart, derivation)) case "failed" =>
                       mkAnswer(h.answer, empty(), [[h.logpart, dummy]])
      mkAnswer(h.answer, u::List(LOG), empty())

    -- checks whether f = +/[ci (ui)'/(ui)]
    -- f dx must have no pole at infinity
    palglogint(f, derivation) ==
      rec := algSplitSimple(f, derivation)
      ground?(r := doubleResultant(f, derivation)) => "failed"
-- r(z) has roots which are the residues of f at all its poles
      (u  := qfactor r) case "failed" => nonQ(rec, r)
      (fc := nonLinear(lf := factors(u::FRQ))) case "failed" => FAIL2
-- at this point r(z) = fc(z) (z - b1)^e1 .. (z - bk)^ek
-- where the ri's are rational numbers, and fc(z) is arbitrary
-- (fc can be linear too)
-- la = [b1....,bk]  (all rational residues)
      la := [- coefficient(q.factor, 0) for q in remove_!(fc::FAC, lf)]
-- ld = [D1,...,Dk] where Di is the sum of places where f has residue bi
      ld  := [divisor(rec.num, rec.den, rec.derivden, rec.gd, b::F) for b in la]
      pp  := UPQ2F(fc.factor)
-- bb = - sum of all the roots of fc (i.e. the other residues)
      zero?(bb := coefficient(fc.factor,
           (degree(fc.factor) - 1)::NonNegativeInteger)) =>
              -- cd = [[a1,...,ak], d]  such that bi = ai/d
              cd  := splitDenominator la
              -- g = gcd(a1,...,ak), so bi = (g/d) ci  with ci = bi / g
              -- so [g/d] is a basis for [a1,...,ak] over the integers
              g   := gcd(cd.num)
              -- dv0 is the divisor +/[ci Di] corresponding to all the residues
              -- of f except the ones which are root of fc(z)
              dv0 := +/[(a quo g) * dv for a in cd.num for dv in ld]
              trace0(rec, pp, g / cd.den, dv0)
      trace1(rec, pp, la, ld, bb)


    UPQ2F p ==
      map(#1::F, p)$UnivariatePolynomialCategoryFunctions2(Q,UPQ,F,UP)

    UP2SUP p ==
       map(#1, p)$UnivariatePolynomialCategoryFunctions2(F, UP, F, SUP)

    SUP2UP p ==
       map(#1, p)$UnivariatePolynomialCategoryFunctions2(F, SUP, F, UP)

    varRoot?(p, derivation) ==
      for c in coefficients primitivePart p repeat
        derivation(c) ~= 0 => return true
      false

    pLogDeriv(log, derivation) ==
      map(derivation, log.coeff) ~= 0 =>
                 error "can only handle logs with constant coefficients"
      one?(n := degree(log.coeff)) =>
        c := - (leadingCoefficient reductum log.coeff)
             / (leadingCoefficient log.coeff)
        ans := (log.logand) c
        (log.scalar)::R * c * derivation(ans) / ans
      numlog := map(derivation, log.logand)
      (diflog := extendedEuclidean(log.logand, log.coeff, numlog)) case
          "failed" => error "this shouldn't happen"
      algans := diflog.coef1
      ans:R := 0
      for i in 0..n-1 repeat
        algans := (algans * monomial(1, 1)) rem log.coeff
        ans    := ans + coefficient(algans, i)
      (log.scalar)::R * ans

    R2UP(f, k) ==
      x := dumx :: F
      g := (map(#1 x, lift f)$UnivariatePolynomialCategoryFunctions2(QF,
                              UPUP, F, UP)) (y := dumy::F)
      map(rlift(#1, dumx, dumy), univariate(g, k,
         minPoly k))$UnivariatePolynomialCategoryFunctions2(F,SUP,R,UPR)

    univ(f, k) ==
      g := univariate(f, k)
      (SUP2UP numer g) / (SUP2UP denom g)

    rlift(f, kx, ky) ==
      reduce map(univ(#1, kx), retract(univariate(f,
         ky))@SUP)$UnivariatePolynomialCategoryFunctions2(F,SUP,QF,UPUP)

    nonQ(rec, p) ==
      empty? rest(lf := factors ffactor primitivePart p) =>
                       trace00(rec, first(lf).factor, empty()$List(LOG))
      FAIL1

-- case when the irreducible factor p has roots which sum to 0
-- p is assumed doubly transitive for now
    trace0(rec, q, r, dv0) ==
      lg:List(LOG) :=
        zero? dv0 => empty()
        (rc0 := torsionIfCan dv0) case "failed" => NOTI
        mkLog(1, r / (rc0.order::Q), rc0.function, 1)
      trace00(rec, q, lg)

    trace00(rec, pp, lg) ==
      p0 := divisor(rec.num, rec.den, rec.derivden, rec.gd,
                    alpha0 := zeroOf UP2SUP pp)
      q  := (pp exquo (monomial(1, 1)$UP - alpha0::UP))::UP
      alpha := rootOf UP2SUP q
      dvr := divisor(rec.num, rec.den, rec.derivden, rec.gd, alpha) - p0
      (rc := torsionIfCan dvr) case "failed" =>
        degree(pp) <= 2 => "failed"
        NOTI
      concat(lg, mkLog(q, inv(rc.order::Q), rc.function, alpha))

-- case when the irreducible factor p has roots which sum <> 0
-- the residues of f are of the form [a1,...,ak] rational numbers
-- plus all the roots of q(z), which is squarefree
-- la is the list of residues la := [a1,...,ak]
-- ld is the list of divisors [D1,...Dk] where Di is the sum of all the
-- places where f has residue ai
-- q(z) is assumed doubly transitive for now.
-- let [alpha_1,...,alpha_m] be the roots of q(z)
-- in this function, b = - alpha_1 - ... - alpha_m is <> 0
-- which implies only one generic log term
    trace1(rec, q, la, ld, b) ==
-- cd = [[b1,...,bk], d]  such that ai / b = bi / d
      cd  := splitDenominator [a / b for a in la]
-- then, a basis for all the residues of f over the integers is
-- [beta_1 = - alpha_1 / d,..., beta_m = - alpha_m / d], since:
--      alpha_i = - d beta_i
--      ai = (ai / b) * b = (bi / d) * b = b1 * beta_1 + ... + bm * beta_m
-- linear independence is a consequence of the doubly transitive assumption
-- v0 is the divisor +/[bi Di] corresponding to the residues [a1,...,ak]
      v0 := +/[a * dv for a in cd.num for dv in ld]
-- alpha is a generic root of q(z)
      alpha := rootOf UP2SUP q
-- v is the divisor corresponding to all the residues
      v := v0 - cd.den * divisor(rec.num, rec.den, rec.derivden, rec.gd, alpha)
      (rc := torsionIfCan v) case "failed" =>   -- non-torsion case
        degree(q) <= 2 => "failed"       -- guaranteed doubly-transitive
        NOTI                             -- maybe doubly-transitive
      mkLog(q, inv((- rc.order * cd.den)::Q), rc.function, alpha)

    mkLog(q, scalr, lgd, alpha) ==
      degree(q) <= 1 =>
        [[scalr, monomial(1, 1)$UPR - F2UPR alpha, lgd::UPR]]
      [[scalr,
         map(F2R, q)$UnivariatePolynomialCategoryFunctions2(F,UP,R,UPR),
                                           R2UP(lgd, retract(alpha)@K)]]

-- return the non-linear factor, if unique
-- or any linear factor if they are all linear
    nonLinear l ==
      found:Boolean := false
      ans := first l
      for q in l repeat
        if degree(q.factor) > 1 then
          found => return "failed"
          found := true
          ans   := q
      ans

-- f dx must be locally integral at infinity
    palginfieldint(f, derivation) ==
      h := HermiteIntegrate(f, derivation)
      zero?(h.logpart) => h.answer
      "failed"

-- f dx must be locally integral at infinity
    palgintegrate(f, derivation) ==
      h := HermiteIntegrate(f, derivation)
      zero?(h.logpart) => h.answer::IR
      (not integralAtInfinity?(h.logpart)) or
        ((u := palglogint(h.logpart, derivation)) case "failed") =>
                      mkAnswer(h.answer, empty(), [[h.logpart, dummy]])
      zero?(difFirstKind := h.logpart - +/[pLogDeriv(lg,
            differentiate(#1, derivation)) for lg in u::List(LOG)]) =>
                mkAnswer(h.answer, u::List(LOG), empty())
      mkAnswer(h.answer, u::List(LOG), [[difFirstKind, dummy]])

-- for mixed functions. f dx not assumed locally integral at infinity
    algintegrate(f, derivation) ==
      zero? degree(x' := derivation(x := monomial(1, 1)$UP)) =>
         algintprim(f, derivation)
      ((xx := x' exquo x) case UP) and
        (retractIfCan(xx::UP)@Union(F, "failed") case F) =>
          algintexp(f, derivation)
      error "should not happen"

    alglogint(f, derivation) ==
      varRoot?(doubleResultant(f, derivation),
                         retract(derivation(#1::UP))@F) => "failed"
      FAIL0

@
\section{License}
<<license>>=
--Copyright (c) 1991-2002, The Numerical ALgorithms Group Ltd.
--All rights reserved.
--
--Redistribution and use in source and binary forms, with or without
--modification, are permitted provided that the following conditions are
--met:
--
--    - Redistributions of source code must retain the above copyright
--      notice, this list of conditions and the following disclaimer.
--
--    - Redistributions in binary form must reproduce the above copyright
--      notice, this list of conditions and the following disclaimer in
--      the documentation and/or other materials provided with the
--      distribution.
--
--    - Neither the name of The Numerical ALgorithms Group Ltd. nor the
--      names of its contributors may be used to endorse or promote products
--      derived from this software without specific prior written permission.
--
--THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
--IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
--TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
--PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
--OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
--EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
--PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
--PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
--LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
--NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
--SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
@
<<*>>=
<<license>>

-- SPAD files for the integration world should be compiled in the
-- following order:
--
--   intaux  rderf  intrf  curve  curvepkg  divisor  pfo
--   INTALG  intaf  efstruc  rdeef  intef  irexpand  integrat

<<package DBLRESP DoubleResultantPackage>>
<<package INTHERAL AlgebraicHermiteIntegration>>
<<package INTALG AlgebraicIntegrate>>
@
\eject
\begin{thebibliography}{99}
\bibitem{1} nothing
\end{thebibliography}
\end{document}