X-Git-Url: https://www.ginac.de/ginac.git//ginac.git?p=ginac.git;a=blobdiff_plain;f=ginac%2Findexed.cpp;h=bd41f44b0b38a271d243564c945d674bea9acaa3;hp=34fbc65f93122eeee823b6cb3e0da8138719b1f6;hb=ef2ed9a711c416722c34548015b4ae3621948c1c;hpb=695f6ae955ec530cded8f21efd5569df39447f76 diff --git a/ginac/indexed.cpp b/ginac/indexed.cpp index 34fbc65f..bd41f44b 100644 --- a/ginac/indexed.cpp +++ b/ginac/indexed.cpp @@ -3,7 +3,7 @@ * Implementation of GiNaC's indexed expressions. */ /* - * GiNaC Copyright (C) 1999-2005 Johannes Gutenberg University Mainz, Germany + * GiNaC Copyright (C) 1999-2006 Johannes Gutenberg University Mainz, Germany * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -17,7 +17,7 @@ * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #include @@ -35,8 +35,11 @@ #include "operators.h" #include "lst.h" #include "archive.h" +#include "symbol.h" #include "utils.h" #include "integral.h" +#include "matrix.h" +#include "inifcns.h" namespace GiNaC { @@ -51,7 +54,7 @@ GINAC_IMPLEMENT_REGISTERED_CLASS_OPT(indexed, exprseq, indexed::indexed() : symtree(not_symmetric()) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; } ////////// @@ -60,79 +63,79 @@ indexed::indexed() : symtree(not_symmetric()) indexed::indexed(const ex & b) : inherited(b), symtree(not_symmetric()) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const ex & b, const ex & i1) : inherited(b, i1), symtree(not_symmetric()) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const ex & b, const ex & i1, const ex & i2) : inherited(b, i1, i2), symtree(not_symmetric()) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const ex & b, const ex & i1, const ex & i2, const ex & i3) : inherited(b, i1, i2, i3), symtree(not_symmetric()) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const ex & b, const ex & i1, const ex & i2, const ex & i3, const ex & i4) : inherited(b, i1, i2, i3, i4), symtree(not_symmetric()) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const ex & b, const symmetry & symm, const ex & i1, const ex & i2) : inherited(b, i1, i2), symtree(symm) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const ex & b, const symmetry & symm, const ex & i1, const ex & i2, const ex & i3) : inherited(b, i1, i2, i3), symtree(symm) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const ex & b, const symmetry & symm, const ex & i1, const ex & i2, const ex & i3, const ex & i4) : inherited(b, i1, i2, i3, i4), symtree(symm) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const ex & b, const exvector & v) : inherited(b), symtree(not_symmetric()) { seq.insert(seq.end(), v.begin(), v.end()); - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const ex & b, const symmetry & symm, const exvector & v) : inherited(b), symtree(symm) { seq.insert(seq.end(), v.begin(), v.end()); - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; validate(); } indexed::indexed(const symmetry & symm, const exprseq & es) : inherited(es), symtree(symm) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; } indexed::indexed(const symmetry & symm, const exvector & v, bool discardable) : inherited(v, discardable), symtree(symm) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; } indexed::indexed(const symmetry & symm, std::auto_ptr vp) : inherited(vp), symtree(symm) { - tinfo_key = TINFO_indexed; + tinfo_key = &indexed::tinfo_static; } ////////// @@ -295,6 +298,9 @@ ex indexed::eval(int level) const return f * thiscontainer(v); } + if(this->tinfo()==&indexed::tinfo_static && seq.size()==1) + return base; + // Canonicalize indices according to the symmetry properties if (seq.size() > 2) { exvector v = seq; @@ -312,6 +318,20 @@ ex indexed::eval(int level) const return ex_to(base).eval_indexed(*this); } +ex indexed::real_part() const +{ + if(op(0).info(info_flags::real)) + return *this; + return real_part_function(*this).hold(); +} + +ex indexed::imag_part() const +{ + if(op(0).info(info_flags::real)) + return 0; + return imag_part_function(*this).hold(); +} + ex indexed::thiscontainer(const exvector & v) const { return indexed(ex_to(symtree), v); @@ -322,6 +342,14 @@ ex indexed::thiscontainer(std::auto_ptr vp) const return indexed(ex_to(symtree), vp); } +unsigned indexed::return_type() const +{ + if(is_a(op(0))) + return return_types::commutative; + else + return op(0).return_type(); +} + ex indexed::expand(unsigned options) const { GINAC_ASSERT(seq.size() > 0); @@ -507,22 +535,6 @@ struct is_summation_idx : public std::unary_function { } }; -exvector power::get_free_indices() const -{ - // Get free indices of basis - exvector basis_indices = basis.get_free_indices(); - - if (exponent.info(info_flags::even)) { - // If the exponent is an even number, then any "free" index that - // forms a dummy pair with itself is actually a summation index - exvector really_free; - std::remove_copy_if(basis_indices.begin(), basis_indices.end(), - std::back_inserter(really_free), is_summation_idx()); - return really_free; - } else - return basis_indices; -} - exvector integral::get_free_indices() const { if (a.get_free_indices().size() || b.get_free_indices().size()) @@ -530,6 +542,15 @@ exvector integral::get_free_indices() const return f.get_free_indices(); } +template size_t number_of_type(const exvector&v) +{ + size_t number = 0; + for(exvector::const_iterator i=v.begin(); i!=v.end(); ++i) + if(is_exactly_a(*i)) + ++number; + return number; +} + /** Rename dummy indices in an expression. * * @param e Expression to work on @@ -538,10 +559,10 @@ exvector integral::get_free_indices() const * @param global_dummy_indices The set of dummy indices that have appeared * before and which we would like to use in "e", too. This gets updated * by the function */ -static ex rename_dummy_indices(const ex & e, exvector & global_dummy_indices, exvector & local_dummy_indices) +template static ex rename_dummy_indices(const ex & e, exvector & global_dummy_indices, exvector & local_dummy_indices) { - size_t global_size = global_dummy_indices.size(), - local_size = local_dummy_indices.size(); + size_t global_size = number_of_type(global_dummy_indices), + local_size = number_of_type(local_dummy_indices); // Any local dummy indices at all? if (local_size == 0) @@ -555,7 +576,7 @@ static ex rename_dummy_indices(const ex & e, exvector & global_dummy_indices, ex int remaining = local_size - global_size; exvector::const_iterator it = local_dummy_indices.begin(), itend = local_dummy_indices.end(); while (it != itend && remaining > 0) { - if (find_if(global_dummy_indices.begin(), global_dummy_indices.end(), bind2nd(op0_is_equal(), *it)) == global_dummy_indices.end()) { + if (is_exactly_a(*it) && find_if(global_dummy_indices.begin(), global_dummy_indices.end(), bind2nd(idx_is_equal_ignore_dim(), *it)) == global_dummy_indices.end()) { global_dummy_indices.push_back(*it); global_size++; remaining--; @@ -573,11 +594,13 @@ static ex rename_dummy_indices(const ex & e, exvector & global_dummy_indices, ex exvector local_syms, global_syms; local_syms.reserve(local_size); global_syms.reserve(local_size); - for (size_t i=0; i(local_dummy_indices[i])) + local_syms.push_back(local_dummy_indices[i].op(0)); shaker_sort(local_syms.begin(), local_syms.end(), ex_is_less(), ex_swap()); - for (size_t i=0; i(global_dummy_indices[i])) + global_syms.push_back(global_dummy_indices[i].op(0)); shaker_sort(global_syms.begin(), global_syms.end(), ex_is_less(), ex_swap()); // Remove common indices @@ -669,16 +692,15 @@ struct ex_base_is_less : public std::binary_function { } }; -/** Simplify product of indexed expressions (commutative, noncommutative and - * simple squares), return list of free indices. */ -ex simplify_indexed_product(const ex & e, exvector & free_indices, exvector & dummy_indices, const scalar_products & sp) +/* An auxiliary function used by simplify_indexed() and expand_dummy_sum() + * It returns an exvector of factors from the supplied product */ +static void product_to_exvector(const ex & e, exvector & v, bool & non_commutative) { // Remember whether the product was commutative or noncommutative // (because we chop it into factors and need to reassemble later) - bool non_commutative = is_exactly_a(e); + non_commutative = is_exactly_a(e); // Collect factors in an exvector, store squares twice - exvector v; v.reserve(e.nops() * 2); if (is_exactly_a(e)) { @@ -691,7 +713,7 @@ ex simplify_indexed_product(const ex & e, exvector & free_indices, exvector & du ex f = e.op(i); if (is_exactly_a(f) && f.op(1).is_equal(_ex2)) { v.push_back(f.op(0)); - v.push_back(f.op(0)); + v.push_back(f.op(0)); } else if (is_exactly_a(f)) { // Noncommutative factor found, split it as well non_commutative = true; // everything becomes noncommutative, ncmul will sort out the commutative factors later @@ -701,6 +723,34 @@ ex simplify_indexed_product(const ex & e, exvector & free_indices, exvector & du v.push_back(f); } } +} + +template ex idx_symmetrization(const ex& r,const exvector& local_dummy_indices) +{ exvector dummy_syms; + dummy_syms.reserve(r.nops()); + for (exvector::const_iterator it = local_dummy_indices.begin(); it != local_dummy_indices.end(); ++it) + if(is_exactly_a(*it)) + dummy_syms.push_back(it->op(0)); + if(dummy_syms.size() < 2) + return r; + ex q=symmetrize(r, dummy_syms); + return q; +} + +// Forward declaration needed in absence of friend injection, C.f. [namespace.memdef]: +ex simplify_indexed(const ex & e, exvector & free_indices, exvector & dummy_indices, const scalar_products & sp); + +/** Simplify product of indexed expressions (commutative, noncommutative and + * simple squares), return list of free indices. */ +ex simplify_indexed_product(const ex & e, exvector & free_indices, exvector & dummy_indices, const scalar_products & sp) +{ + // Collect factors in an exvector + exvector v; + + // Remember whether the product was commutative or noncommutative + // (because we chop it into factors and need to reassemble later) + bool non_commutative; + product_to_exvector(e, v, non_commutative); // Perform contractions bool something_changed = false; @@ -742,20 +792,12 @@ try_again: // At least one dummy index, is it a defined scalar product? bool contracted = false; - if (free.empty()) { - - // Find minimal dimension of all indices of both factors - exvector::const_iterator dit = ex_to(*it1).seq.begin() + 1, ditend = ex_to(*it1).seq.end(); - ex dim = ex_to(*dit).get_dim(); - ++dit; - for (; dit != ditend; ++dit) { - dim = minimal_dim(dim, ex_to(*dit).get_dim()); - } - dit = ex_to(*it2).seq.begin() + 1; - ditend = ex_to(*it2).seq.end(); - for (; dit != ditend; ++dit) { - dim = minimal_dim(dim, ex_to(*dit).get_dim()); - } + if (free.empty() && it1->nops()==2 && it2->nops()==2) { + + ex dim = minimal_dim( + ex_to(it1->op(1)).get_dim(), + ex_to(it2->op(1)).get_dim() + ); // User-defined scalar product? if (sp.is_defined(*it1, *it2, dim)) { @@ -850,19 +892,26 @@ contraction_done: // The result should be symmetric with respect to exchange of dummy // indices, so if the symmetrization vanishes, the whole expression is // zero. This detects things like eps.i.j.k * p.j * p.k = 0. - if (local_dummy_indices.size() >= 2) { - exvector dummy_syms; - dummy_syms.reserve(local_dummy_indices.size()); - for (exvector::const_iterator it = local_dummy_indices.begin(); it != local_dummy_indices.end(); ++it) - dummy_syms.push_back(it->op(0)); - if (symmetrize(r, dummy_syms).is_zero()) { - free_indices.clear(); - return _ex0; - } + ex q = idx_symmetrization(r, local_dummy_indices); + if (q.is_zero()) { + free_indices.clear(); + return _ex0; + } + q = idx_symmetrization(q, local_dummy_indices); + if (q.is_zero()) { + free_indices.clear(); + return _ex0; + } + q = idx_symmetrization(q, local_dummy_indices); + if (q.is_zero()) { + free_indices.clear(); + return _ex0; } // Dummy index renaming - r = rename_dummy_indices(r, dummy_indices, local_dummy_indices); + r = rename_dummy_indices(r, dummy_indices, local_dummy_indices); + r = rename_dummy_indices(r, dummy_indices, local_dummy_indices); + r = rename_dummy_indices(r, dummy_indices, local_dummy_indices); // Product of indexed object with a scalar? if (is_exactly_a(r) && r.nops() == 2 @@ -929,6 +978,17 @@ public: } }; +bool hasindex(const ex &x, const ex &sym) +{ + if(is_a(x) && x.op(0)==sym) + return true; + else + for(size_t i=0; i(e_expanded, dummy_indices, local_dummy_indices); + e_expanded = rename_dummy_indices(e_expanded, dummy_indices, local_dummy_indices); + e_expanded = rename_dummy_indices(e_expanded, dummy_indices, local_dummy_indices); + return e_expanded; } // Simplification of sum = sum of simplifications, check consistency of @@ -1001,18 +1064,19 @@ ex simplify_indexed(const ex & e, exvector & free_indices, exvector & dummy_indi if (num_terms_orig < 2 || dummy_indices.size() < 2) return sum; - // Yes, construct vector of all dummy index symbols - exvector dummy_syms; - dummy_syms.reserve(dummy_indices.size()); - for (exvector::const_iterator it = dummy_indices.begin(); it != dummy_indices.end(); ++it) - dummy_syms.push_back(it->op(0)); - // Chop the sum into terms and symmetrize each one over the dummy // indices std::vector terms; for (size_t i=0; iop(0))) + dummy_indices_of_term.push_back(*i); + ex term_symm = idx_symmetrization(term, dummy_indices_of_term); + term_symm = idx_symmetrization(term_symm, dummy_indices_of_term); + term_symm = idx_symmetrization(term_symm, dummy_indices_of_term); if (term_symm.is_zero()) continue; terms.push_back(terminfo(term, term_symm)); @@ -1278,4 +1342,195 @@ void scalar_products::debugprint() const } } +exvector get_all_dummy_indices_safely(const ex & e) +{ + if (is_a(e)) + return ex_to(e).get_dummy_indices(); + else if (is_a(e) && e.op(1)==2) { + return e.op(0).get_free_indices(); + } + else if (is_a(e) || is_a(e)) { + exvector dummies; + exvector free_indices; + for (int i=0; i(e)) { + exvector result; + for(int i=0; i(new_vec), + ex_is_less()); + result.swap(new_vec); + } + return result; + } + return exvector(); +} + +/** Returns all dummy indices from the exvector */ +exvector get_all_dummy_indices(const ex & e) +{ + exvector p; + bool nc; + product_to_exvector(e, p, nc); + exvector::const_iterator ip = p.begin(), ipend = p.end(); + exvector v, v1; + while (ip != ipend) { + if (is_a(*ip)) { + v1 = ex_to(*ip).get_dummy_indices(); + v.insert(v.end(), v1.begin(), v1.end()); + exvector::const_iterator ip1 = ip+1; + while (ip1 != ipend) { + if (is_a(*ip1)) { + v1 = ex_to(*ip).get_dummy_indices(ex_to(*ip1)); + v.insert(v.end(), v1.begin(), v1.end()); + } + ++ip1; + } + } + ++ip; + } + return v; +} + +lst rename_dummy_indices_uniquely(const exvector & va, const exvector & vb) +{ + exvector common_indices; + set_intersection(va.begin(), va.end(), vb.begin(), vb.end(), std::back_insert_iterator(common_indices), ex_is_less()); + if (common_indices.empty()) { + return lst(lst(), lst()); + } else { + exvector new_indices, old_indices; + old_indices.reserve(2*common_indices.size()); + new_indices.reserve(2*common_indices.size()); + exvector::const_iterator ip = common_indices.begin(), ipend = common_indices.end(); + while (ip != ipend) { + ex newsym=(new symbol)->setflag(status_flags::dynallocated); + ex newidx; + if(is_exactly_a(*ip)) + newidx = (new spinidx(newsym, ex_to(*ip).get_dim(), + ex_to(*ip).is_covariant(), + ex_to(*ip).is_dotted())) + -> setflag(status_flags::dynallocated); + else if (is_exactly_a(*ip)) + newidx = (new varidx(newsym, ex_to(*ip).get_dim(), + ex_to(*ip).is_covariant())) + -> setflag(status_flags::dynallocated); + else + newidx = (new idx(newsym, ex_to(*ip).get_dim())) + -> setflag(status_flags::dynallocated); + old_indices.push_back(*ip); + new_indices.push_back(newidx); + if(is_a(*ip)) { + old_indices.push_back(ex_to(*ip).toggle_variance()); + new_indices.push_back(ex_to(newidx).toggle_variance()); + } + ++ip; + } + return lst(lst(old_indices.begin(), old_indices.end()), lst(new_indices.begin(), new_indices.end())); + } +} + +ex rename_dummy_indices_uniquely(const exvector & va, const exvector & vb, const ex & b) +{ + lst indices_subs = rename_dummy_indices_uniquely(va, vb); + return (indices_subs.op(0).nops()>0 ? b.subs(ex_to(indices_subs.op(0)), ex_to(indices_subs.op(1)), subs_options::no_pattern|subs_options::no_index_renaming) : b); +} + +ex rename_dummy_indices_uniquely(const ex & a, const ex & b) +{ + exvector va = get_all_dummy_indices_safely(a); + if (va.size() > 0) { + exvector vb = get_all_dummy_indices_safely(b); + if (vb.size() > 0) { + sort(va.begin(), va.end(), ex_is_less()); + sort(vb.begin(), vb.end(), ex_is_less()); + lst indices_subs = rename_dummy_indices_uniquely(va, vb); + if (indices_subs.op(0).nops() > 0) + return b.subs(ex_to(indices_subs.op(0)), ex_to(indices_subs.op(1)), subs_options::no_pattern|subs_options::no_index_renaming); + } + } + return b; +} + +ex rename_dummy_indices_uniquely(exvector & va, const ex & b, bool modify_va) +{ + if (va.size() > 0) { + exvector vb = get_all_dummy_indices_safely(b); + if (vb.size() > 0) { + sort(vb.begin(), vb.end(), ex_is_less()); + lst indices_subs = rename_dummy_indices_uniquely(va, vb); + if (indices_subs.op(0).nops() > 0) { + if (modify_va) { + for (lst::const_iterator i = ex_to(indices_subs.op(1)).begin(); i != ex_to(indices_subs.op(1)).end(); ++i) + va.push_back(*i); + exvector uncommon_indices; + set_difference(vb.begin(), vb.end(), indices_subs.op(0).begin(), indices_subs.op(0).end(), std::back_insert_iterator(uncommon_indices), ex_is_less()); + exvector::const_iterator ip = uncommon_indices.begin(), ipend = uncommon_indices.end(); + while (ip != ipend) { + va.push_back(*ip); + ++ip; + } + sort(va.begin(), va.end(), ex_is_less()); + } + return b.subs(ex_to(indices_subs.op(0)), ex_to(indices_subs.op(1)), subs_options::no_pattern|subs_options::no_index_renaming); + } + } + } + return b; +} + +ex expand_dummy_sum(const ex & e, bool subs_idx) +{ + ex e_expanded = e.expand(); + pointer_to_map_function_1arg fcn(expand_dummy_sum, subs_idx); + if (is_a(e_expanded) || is_a(e_expanded) || is_a(e_expanded)) { + return e_expanded.map(fcn); + } else if (is_a(e_expanded) || is_a(e_expanded) || is_a(e_expanded) || is_a(e_expanded)) { + exvector v; + if (is_a(e_expanded)) + v = ex_to(e_expanded).get_dummy_indices(); + else + v = get_all_dummy_indices(e_expanded); + ex result = e_expanded; + for(exvector::const_iterator it=v.begin(); it!=v.end(); ++it) { + ex nu = *it; + if (ex_to(nu).get_dim().info(info_flags::nonnegint)) { + int idim = ex_to(ex_to(nu).get_dim()).to_int(); + ex en = 0; + for (int i=0; i < idim; i++) { + if (subs_idx && is_a(nu)) { + ex other = ex_to(nu).toggle_variance(); + en += result.subs(lst( + nu == idx(i, idim), + other == idx(i, idim) + )); + } else { + en += result.subs( nu.op(0) == i ); + } + } + result = en; + } + } + return result; + } else { + return e; + } +} + } // namespace GiNaC