projects / cvc5.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Merge branch 'master' into uniq-ptr
[cvc5.git] / src / theory / strings / theory_strings.cpp
1 /********************* */
2 /*! \file theory_strings.cpp
3 ** \verbatim
4 ** Top contributors (to current version):
5 ** Andrew Reynolds, Tianyi Liang, Tim King
6 ** This file is part of the CVC4 project.
7 ** Copyright (c) 2009-2016 by the authors listed in the file AUTHORS
8 ** in the top-level source directory) and their institutional affiliations.
9 ** All rights reserved. See the file COPYING in the top-level source
10 ** directory for licensing information.\endverbatim
11 **
12 ** \brief Implementation of the theory of strings.
13 **
14 ** Implementation of the theory of strings.
15 **/
16
17 #include "theory/strings/theory_strings.h"
18
19 #include <cmath>
20
21 #include "expr/kind.h"
22 #include "options/strings_options.h"
23 #include "smt/logic_exception.h"
24 #include "smt/smt_statistics_registry.h"
25 #include "smt/command.h"
26 #include "theory/rewriter.h"
27 #include "theory/strings/theory_strings_rewriter.h"
28 #include "theory/strings/type_enumerator.h"
29 #include "theory/theory_model.h"
30 #include "theory/valuation.h"
31 #include "theory/quantifiers/term_database.h"
32
33 using namespace std;
34 using namespace CVC4::context;
35
36 namespace CVC4 {
37 namespace theory {
38 namespace strings {
39
40 Node TheoryStrings::TermIndex::add( TNode n, unsigned index, TheoryStrings* t, Node er, std::vector< Node >& c ) {
41 if( index==n.getNumChildren() ){
42 if( d_data.isNull() ){
43 d_data = n;
44 }
45 return d_data;
46 }else{
47 Assert( index<n.getNumChildren() );
48 TNode nir = t->getRepresentative( n[index] );
49 //if it is empty, and doing CONCAT, ignore
50 if( nir==er && n.getKind()==kind::STRING_CONCAT ){
51 return add( n, index+1, t, er, c );
52 }else{
53 c.push_back( nir );
54 return d_children[nir].add( n, index+1, t, er, c );
55 }
56 }
57 }
58
59
60 TheoryStrings::TheoryStrings(context::Context* c, context::UserContext* u,
61 OutputChannel& out, Valuation valuation,
62 const LogicInfo& logicInfo)
63 : Theory(THEORY_STRINGS, c, u, out, valuation, logicInfo),
64 RMAXINT(LONG_MAX),
65 d_notify( *this ),
66 d_equalityEngine(d_notify, c, "theory::strings::TheoryStrings", true),
67 d_conflict(c, false),
68 d_infer(c),
69 d_infer_exp(c),
70 d_nf_pairs(c),
71 d_loop_antec(u),
72 d_pregistered_terms_cache(u),
73 d_registered_terms_cache(u),
74 d_length_lemma_terms_cache(u),
75 d_skolem_ne_reg_cache(u),
76 d_preproc(u),
77 d_preproc_cache(u),
78 d_extf_infer_cache(c),
79 d_extf_infer_cache_u(u),
80 d_ee_disequalities(c),
81 d_congruent(c),
82 d_proxy_var(u),
83 d_proxy_var_to_length(u),
84 d_functionsTerms(c),
85 d_has_extf(c, false ),
86 d_regexp_memberships(c),
87 d_regexp_ucached(u),
88 d_regexp_ccached(c),
89 d_pos_memberships(c),
90 d_neg_memberships(c),
91 d_inter_cache(c),
92 d_inter_index(c),
93 d_processed_memberships(c),
94 d_regexp_ant(c),
95 d_input_vars(u),
96 d_input_var_lsum(u),
97 d_cardinality_lits(u),
98 d_curr_cardinality(c, 0)
99 {
100 d_extt = new ExtTheory( this );
101 d_extt->addFunctionKind( kind::STRING_SUBSTR );
102 d_extt->addFunctionKind( kind::STRING_STRIDOF );
103 d_extt->addFunctionKind( kind::STRING_ITOS );
104 d_extt->addFunctionKind( kind::STRING_U16TOS );
105 d_extt->addFunctionKind( kind::STRING_U32TOS );
106 d_extt->addFunctionKind( kind::STRING_STOI );
107 d_extt->addFunctionKind( kind::STRING_STOU16 );
108 d_extt->addFunctionKind( kind::STRING_STOU32 );
109 d_extt->addFunctionKind( kind::STRING_STRREPL );
110 d_extt->addFunctionKind( kind::STRING_STRCTN );
111 d_extt->addFunctionKind( kind::STRING_IN_REGEXP );
112
113 // The kinds we are treating as function application in congruence
114 d_equalityEngine.addFunctionKind(kind::STRING_IN_REGEXP);
115 d_equalityEngine.addFunctionKind(kind::STRING_LENGTH);
116 d_equalityEngine.addFunctionKind(kind::STRING_CONCAT);
117 if( options::stringLazyPreproc() ){
118 d_equalityEngine.addFunctionKind(kind::STRING_STRCTN);
119 d_equalityEngine.addFunctionKind(kind::STRING_SUBSTR);
120 d_equalityEngine.addFunctionKind(kind::STRING_ITOS);
121 d_equalityEngine.addFunctionKind(kind::STRING_STOI);
122 d_equalityEngine.addFunctionKind(kind::STRING_U16TOS);
123 d_equalityEngine.addFunctionKind(kind::STRING_STOU16);
124 d_equalityEngine.addFunctionKind(kind::STRING_U32TOS);
125 d_equalityEngine.addFunctionKind(kind::STRING_STOU32);
126 d_equalityEngine.addFunctionKind(kind::STRING_STRIDOF);
127 d_equalityEngine.addFunctionKind(kind::STRING_STRREPL);
128 }
129
130 d_zero = NodeManager::currentNM()->mkConst( Rational( 0 ) );
131 d_one = NodeManager::currentNM()->mkConst( Rational( 1 ) );
132 d_emptyString = NodeManager::currentNM()->mkConst( ::CVC4::String("") );
133 std::vector< Node > nvec;
134 d_emptyRegexp = NodeManager::currentNM()->mkNode( kind::REGEXP_EMPTY, nvec );
135 d_true = NodeManager::currentNM()->mkConst( true );
136 d_false = NodeManager::currentNM()->mkConst( false );
137
138 d_card_size = 128;
139 }
140
141 TheoryStrings::~TheoryStrings() {
142 for( std::map< Node, EqcInfo* >::iterator it = d_eqc_info.begin(); it != d_eqc_info.end(); ++it ){
143 delete it->second;
144 }
145 delete d_extt;
146 }
147
148 Node TheoryStrings::getRepresentative( Node t ) {
149 if( d_equalityEngine.hasTerm( t ) ){
150 return d_equalityEngine.getRepresentative( t );
151 }else{
152 return t;
153 }
154 }
155
156 bool TheoryStrings::hasTerm( Node a ){
157 return d_equalityEngine.hasTerm( a );
158 }
159
160 bool TheoryStrings::areEqual( Node a, Node b ){
161 if( a==b ){
162 return true;
163 }else if( hasTerm( a ) && hasTerm( b ) ){
164 return d_equalityEngine.areEqual( a, b );
165 }else{
166 return false;
167 }
168 }
169
170 bool TheoryStrings::areDisequal( Node a, Node b ){
171 if( a==b ){
172 return false;
173 }else{
174 if( hasTerm( a ) && hasTerm( b ) ) {
175 Node ar = d_equalityEngine.getRepresentative( a );
176 Node br = d_equalityEngine.getRepresentative( b );
177 return ( ar!=br && ar.isConst() && br.isConst() ) || d_equalityEngine.areDisequal( ar, br, false );
178 }else{
179 Node ar = getRepresentative( a );
180 Node br = getRepresentative( b );
181 return ar!=br && ar.isConst() && br.isConst();
182 }
183 }
184 }
185
186 Node TheoryStrings::getLengthExp( Node t, std::vector< Node >& exp, Node te ){
187 Assert( areEqual( t, te ) );
188 Node lt = mkLength( te );
189 if( hasTerm( lt ) ){
190 // use own length if it exists, leads to shorter explanation
191 return lt;
192 }else{
193 EqcInfo * ei = getOrMakeEqcInfo( t, false );
194 Node length_term = ei ? ei->d_length_term : Node::null();
195 if( length_term.isNull() ){
196 //typically shouldnt be necessary
197 length_term = t;
198 }
199 Debug("strings") << "TheoryStrings::getLengthTerm " << t << " is " << length_term << std::endl;
200 addToExplanation( length_term, te, exp );
201 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, length_term ) );
202 }
203 }
204
205 Node TheoryStrings::getLength( Node t, std::vector< Node >& exp ) {
206 return getLengthExp( t, exp, t );
207 }
208
209 void TheoryStrings::setMasterEqualityEngine(eq::EqualityEngine* eq) {
210 d_equalityEngine.setMasterEqualityEngine(eq);
211 }
212
213 void TheoryStrings::addSharedTerm(TNode t) {
214 Debug("strings") << "TheoryStrings::addSharedTerm(): "
215 << t << " " << t.getType().isBoolean() << endl;
216 d_equalityEngine.addTriggerTerm(t, THEORY_STRINGS);
217 Debug("strings") << "TheoryStrings::addSharedTerm() finished" << std::endl;
218 }
219
220 EqualityStatus TheoryStrings::getEqualityStatus(TNode a, TNode b) {
221 if( d_equalityEngine.hasTerm(a) && d_equalityEngine.hasTerm(b) ){
222 if (d_equalityEngine.areEqual(a, b)) {
223 // The terms are implied to be equal
224 return EQUALITY_TRUE;
225 }
226 if (d_equalityEngine.areDisequal(a, b, false)) {
227 // The terms are implied to be dis-equal
228 return EQUALITY_FALSE;
229 }
230 }
231 return EQUALITY_UNKNOWN;
232 }
233
234 void TheoryStrings::propagate(Effort e) {
235 // direct propagation now
236 }
237
238 bool TheoryStrings::propagate(TNode literal) {
239 Debug("strings-propagate") << "TheoryStrings::propagate(" << literal << ")" << std::endl;
240 // If already in conflict, no more propagation
241 if (d_conflict) {
242 Debug("strings-propagate") << "TheoryStrings::propagate(" << literal << "): already in conflict" << std::endl;
243 return false;
244 }
245 // Propagate out
246 bool ok = d_out->propagate(literal);
247 if (!ok) {
248 d_conflict = true;
249 }
250 return ok;
251 }
252
253 /** explain */
254 void TheoryStrings::explain(TNode literal, std::vector<TNode>& assumptions) {
255 Debug("strings-explain") << "Explain " << literal << " " << d_conflict << std::endl;
256 bool polarity = literal.getKind() != kind::NOT;
257 TNode atom = polarity ? literal : literal[0];
258 unsigned ps = assumptions.size();
259 std::vector< TNode > tassumptions;
260 if (atom.getKind() == kind::EQUAL || atom.getKind() == kind::IFF) {
261 if( atom[0]!=atom[1] ){
262 d_equalityEngine.explainEquality(atom[0], atom[1], polarity, tassumptions);
263 }
264 } else {
265 d_equalityEngine.explainPredicate(atom, polarity, tassumptions);
266 }
267 for( unsigned i=0; i<tassumptions.size(); i++ ){
268 if( std::find( assumptions.begin(), assumptions.end(), tassumptions[i] )==assumptions.end() ){
269 assumptions.push_back( tassumptions[i] );
270 }
271 }
272 Debug("strings-explain-debug") << "Explanation for " << literal << " was " << std::endl;
273 for( unsigned i=ps; i<assumptions.size(); i++ ){
274 Debug("strings-explain-debug") << " " << assumptions[i] << std::endl;
275 }
276 }
277
278 Node TheoryStrings::explain( TNode literal ){
279 std::vector< TNode > assumptions;
280 explain( literal, assumptions );
281 if( assumptions.empty() ){
282 return d_true;
283 }else if( assumptions.size()==1 ){
284 return assumptions[0];
285 }else{
286 return NodeManager::currentNM()->mkNode( kind::AND, assumptions );
287 }
288 }
289
290 bool TheoryStrings::getCurrentSubstitution( int effort, std::vector< Node >& vars,
291 std::vector< Node >& subs, std::map< Node, std::vector< Node > >& exp ) {
292 Trace("strings-subs") << "getCurrentSubstitution, effort = " << effort << std::endl;
293 for( unsigned i=0; i<vars.size(); i++ ){
294 Node n = vars[i];
295 Trace("strings-subs") << " get subs for " << n << "..." << std::endl;
296 if( effort>=3 ){
297 //model values
298 Node mv = d_valuation.getModel()->getRepresentative( n );
299 Trace("strings-subs") << " model val : " << mv << std::endl;
300 subs.push_back( mv );
301 }else{
302 Node nr = getRepresentative( n );
303 std::map< Node, Node >::iterator itc = d_eqc_to_const.find( nr );
304 if( itc!=d_eqc_to_const.end() ){
305 //constant equivalence classes
306 Trace("strings-subs") << " constant eqc : " << d_eqc_to_const_exp[nr] << " " << d_eqc_to_const_base[nr] << " " << nr << std::endl;
307 subs.push_back( itc->second );
308 if( !d_eqc_to_const_exp[nr].isNull() ){
309 exp[n].push_back( d_eqc_to_const_exp[nr] );
310 }
311 if( !d_eqc_to_const_base[nr].isNull() ){
312 addToExplanation( n, d_eqc_to_const_base[nr], exp[n] );
313 }
314 }else if( effort>=1 && effort<3 && n.getType().isString() ){
315 //normal forms
316 Node ns = getNormalString( d_normal_forms_base[nr], exp[n] );
317 subs.push_back( ns );
318 Trace("strings-subs") << " normal eqc : " << ns << " " << d_normal_forms_base[nr] << " " << nr << std::endl;
319 if( !d_normal_forms_base[nr].isNull() ) {
320 addToExplanation( n, d_normal_forms_base[nr], exp[n] );
321 }
322 }else{
323 //representative?
324 //Trace("strings-subs") << " representative : " << nr << std::endl;
325 //addToExplanation( n, nr, exp[n] );
326 //subs.push_back( nr );
327 subs.push_back( n );
328 }
329 }
330 }
331 return true;
332 }
333
334 int TheoryStrings::getReduction( int effort, Node n, Node& nr ) {
335 //determine the effort level to process the extf at
336 // 0 - at assertion time, 1+ - after no other reduction is applicable
337 Assert( d_extf_info_tmp.find( n )!=d_extf_info_tmp.end() );
338 if( d_extf_info_tmp[n].d_model_active ){
339 int r_effort = -1;
340 int pol = d_extf_info_tmp[n].d_pol;
341 if( n.getKind()==kind::STRING_STRCTN ){
342 if( pol==1 ){
343 r_effort = 1;
344 }else if( pol==-1 ){
345 if( effort==2 ){
346 Node x = n[0];
347 Node s = n[1];
348 std::vector< Node > lexp;
349 Node lenx = getLength( x, lexp );
350 Node lens = getLength( s, lexp );
351 if( areEqual( lenx, lens ) ){
352 Trace("strings-extf-debug") << " resolve extf : " << n << " based on equal lengths disequality." << std::endl;
353 //we can reduce to disequality when lengths are equal
354 if( !areDisequal( x, s ) ){
355 lexp.push_back( lenx.eqNode(lens) );
356 lexp.push_back( n.negate() );
357 Node xneqs = x.eqNode(s).negate();
358 sendInference( lexp, xneqs, "NEG-CTN-EQL", true );
359 }
360 return 1;
361 }else if( !areDisequal( lenx, lens ) ){
362 //split on their lenths
363 sendSplit( lenx, lens, "NEG-CTN-SP" );
364 }else{
365 r_effort = 2;
366 }
367 }
368 }
369 }else{
370 if( options::stringLazyPreproc() ){
371 if( n.getKind()==kind::STRING_SUBSTR ){
372 r_effort = 1;
373 }else if( n.getKind()!=kind::STRING_IN_REGEXP ){
374 r_effort = 2;
375 }
376 }
377 }
378 if( effort==r_effort ){
379 Node c_n = pol==-1 ? n.negate() : n;
380 if( d_preproc_cache.find( c_n )==d_preproc_cache.end() ){
381 d_preproc_cache[ c_n ] = true;
382 Trace("strings-process-debug") << "Process reduction for " << n << ", pol = " << pol << std::endl;
383 if( n.getKind()==kind::STRING_STRCTN && pol==1 ){
384 Node x = n[0];
385 Node s = n[1];
386 //positive contains reduces to a equality
387 Node sk1 = mkSkolemCached( x, s, sk_id_ctn_pre, "sc1" );
388 Node sk2 = mkSkolemCached( x, s, sk_id_ctn_post, "sc2" );
389 Node eq = Rewriter::rewrite( x.eqNode( mkConcat( sk1, s, sk2 ) ) );
390 std::vector< Node > exp_vec;
391 exp_vec.push_back( n );
392 sendInference( d_empty_vec, exp_vec, eq, "POS-CTN", true );
393 //we've reduced this n
394 Trace("strings-extf-debug") << " resolve extf : " << n << " based on positive contain reduction." << std::endl;
395 return 1;
396 }else{
397 // for STRING_SUBSTR, STRING_STRCTN with pol=-1,
398 // STRING_STRIDOF, STRING_ITOS, STRING_U16TOS, STRING_U32TOS, STRING_STOI, STRING_STOU16, STRING_STOU32, STRING_STRREPL
399 std::vector< Node > new_nodes;
400 Node res = d_preproc.simplify( n, new_nodes );
401 Assert( res!=n );
402 new_nodes.push_back( NodeManager::currentNM()->mkNode( res.getType().isBoolean() ? kind::IFF : kind::EQUAL, res, n ) );
403 Node nnlem = new_nodes.size()==1 ? new_nodes[0] : NodeManager::currentNM()->mkNode( kind::AND, new_nodes );
404 nnlem = Rewriter::rewrite( nnlem );
405 Trace("strings-red-lemma") << "Reduction_" << effort << " lemma : " << nnlem << std::endl;
406 Trace("strings-red-lemma") << "...from " << n << std::endl;
407 sendInference( d_empty_vec, nnlem, "Reduction", true );
408 //we've reduced this n
409 Trace("strings-extf-debug") << " resolve extf : " << n << " based on reduction." << std::endl;
410 return 1;
411 }
412 }else{
413 return 1;
414 }
415 }
416 }
417 return 0;
418 }
419
420 /////////////////////////////////////////////////////////////////////////////
421 // NOTIFICATIONS
422 /////////////////////////////////////////////////////////////////////////////
423
424
425 void TheoryStrings::presolve() {
426 Debug("strings-presolve") << "TheoryStrings::Presolving : get fmf options " << (options::stringFMF() ? "true" : "false") << std::endl;
427
428 if(!options::stdASCII()) {
429 d_card_size = 256;
430 }
431 }
432
433
434 /////////////////////////////////////////////////////////////////////////////
435 // MODEL GENERATION
436 /////////////////////////////////////////////////////////////////////////////
437
438
439 void TheoryStrings::collectModelInfo( TheoryModel* m, bool fullModel ) {
440 Trace("strings-model") << "TheoryStrings : Collect model info, fullModel = " << fullModel << std::endl;
441 Trace("strings-model") << "TheoryStrings : assertEqualityEngine." << std::endl;
442 m->assertEqualityEngine( &d_equalityEngine );
443 // Generate model
444 std::vector< Node > nodes;
445 getEquivalenceClasses( nodes );
446 std::map< Node, Node > processed;
447 std::vector< std::vector< Node > > col;
448 std::vector< Node > lts;
449 separateByLength( nodes, col, lts );
450 //step 1 : get all values for known lengths
451 std::vector< Node > lts_values;
452 std::map< unsigned, bool > values_used;
453 for( unsigned i=0; i<col.size(); i++ ) {
454 Trace("strings-model") << "Checking length for {";
455 for( unsigned j=0; j<col[i].size(); j++ ) {
456 if( j>0 ) {
457 Trace("strings-model") << ", ";
458 }
459 Trace("strings-model") << col[i][j];
460 }
461 Trace("strings-model") << " } (length is " << lts[i] << ")" << std::endl;
462 if( lts[i].isConst() ) {
463 lts_values.push_back( lts[i] );
464 Assert(lts[i].getConst<Rational>() <= RMAXINT, "Exceeded LONG_MAX in string model");
465 unsigned lvalue = lts[i].getConst<Rational>().getNumerator().toUnsignedInt();
466 values_used[ lvalue ] = true;
467 }else{
468 //get value for lts[i];
469 if( !lts[i].isNull() ){
470 Node v = d_valuation.getModelValue(lts[i]);
471 Trace("strings-model") << "Model value for " << lts[i] << " is " << v << std::endl;
472 lts_values.push_back( v );
473 Assert(v.getConst<Rational>() <= RMAXINT, "Exceeded LONG_MAX in string model");
474 unsigned lvalue = v.getConst<Rational>().getNumerator().toUnsignedInt();
475 values_used[ lvalue ] = true;
476 }else{
477 //Trace("strings-model-warn") << "No length for eqc " << col[i][0] << std::endl;
478 //Assert( false );
479 lts_values.push_back( Node::null() );
480 }
481 }
482 }
483 ////step 2 : assign arbitrary values for unknown lengths?
484 // confirmed by calculus invariant, see paper
485 Trace("strings-model") << "Assign to equivalence classes..." << std::endl;
486 //step 3 : assign values to equivalence classes that are pure variables
487 for( unsigned i=0; i<col.size(); i++ ){
488 std::vector< Node > pure_eq;
489 Trace("strings-model") << "The equivalence classes ";
490 for( unsigned j=0; j<col[i].size(); j++ ) {
491 Trace("strings-model") << col[i][j] << " ";
492 //check if col[i][j] has only variables
493 if( !col[i][j].isConst() ){
494 Assert( d_normal_forms.find( col[i][j] )!=d_normal_forms.end() );
495 if( d_normal_forms[col[i][j]].size()==1 ){//&& d_normal_forms[col[i][j]][0]==col[i][j] ){
496 pure_eq.push_back( col[i][j] );
497 }
498 }else{
499 processed[col[i][j]] = col[i][j];
500 }
501 }
502 Trace("strings-model") << "have length " << lts_values[i] << std::endl;
503
504 //assign a new length if necessary
505 if( !pure_eq.empty() ){
506 if( lts_values[i].isNull() ){
507 unsigned lvalue = 0;
508 while( values_used.find( lvalue )!=values_used.end() ){
509 lvalue++;
510 }
511 Trace("strings-model") << "*** Decide to make length of " << lvalue << std::endl;
512 lts_values[i] = NodeManager::currentNM()->mkConst( Rational( lvalue ) );
513 values_used[ lvalue ] = true;
514 }
515 Trace("strings-model") << "Need to assign values of length " << lts_values[i] << " to equivalence classes ";
516 for( unsigned j=0; j<pure_eq.size(); j++ ){
517 Trace("strings-model") << pure_eq[j] << " ";
518 }
519 Trace("strings-model") << std::endl;
520
521
522 //use type enumerator
523 Assert(lts_values[i].getConst<Rational>() <= RMAXINT, "Exceeded LONG_MAX in string model");
524 StringEnumeratorLength sel(lts_values[i].getConst<Rational>().getNumerator().toUnsignedInt());
525 for( unsigned j=0; j<pure_eq.size(); j++ ){
526 Assert( !sel.isFinished() );
527 Node c = *sel;
528 while( d_equalityEngine.hasTerm( c ) ){
529 ++sel;
530 Assert( !sel.isFinished() );
531 c = *sel;
532 }
533 ++sel;
534 Trace("strings-model") << "*** Assigned constant " << c << " for " << pure_eq[j] << std::endl;
535 processed[pure_eq[j]] = c;
536 m->assertEquality( pure_eq[j], c, true );
537 }
538 }
539 }
540 Trace("strings-model") << "String Model : Pure Assigned." << std::endl;
541 //step 4 : assign constants to all other equivalence classes
542 for( unsigned i=0; i<nodes.size(); i++ ){
543 if( processed.find( nodes[i] )==processed.end() ){
544 Assert( d_normal_forms.find( nodes[i] )!=d_normal_forms.end() );
545 Trace("strings-model") << "Construct model for " << nodes[i] << " based on normal form ";
546 for( unsigned j=0; j<d_normal_forms[nodes[i]].size(); j++ ) {
547 if( j>0 ) Trace("strings-model") << " ++ ";
548 Trace("strings-model") << d_normal_forms[nodes[i]][j];
549 Node r = getRepresentative( d_normal_forms[nodes[i]][j] );
550 if( !r.isConst() && processed.find( r )==processed.end() ){
551 Trace("strings-model") << "(UNPROCESSED)";
552 }
553 }
554 Trace("strings-model") << std::endl;
555 std::vector< Node > nc;
556 for( unsigned j=0; j<d_normal_forms[nodes[i]].size(); j++ ) {
557 Node r = getRepresentative( d_normal_forms[nodes[i]][j] );
558 Assert( r.isConst() || processed.find( r )!=processed.end() );
559 nc.push_back(r.isConst() ? r : processed[r]);
560 }
561 Node cc = mkConcat( nc );
562 Assert( cc.getKind()==kind::CONST_STRING );
563 Trace("strings-model") << "*** Determined constant " << cc << " for " << nodes[i] << std::endl;
564 processed[nodes[i]] = cc;
565 m->assertEquality( nodes[i], cc, true );
566 }
567 }
568 //Trace("strings-model") << "String Model : Assigned." << std::endl;
569 Trace("strings-model") << "String Model : Finished." << std::endl;
570 }
571
572 /////////////////////////////////////////////////////////////////////////////
573 // MAIN SOLVER
574 /////////////////////////////////////////////////////////////////////////////
575
576
577 void TheoryStrings::preRegisterTerm(TNode n) {
578 if( d_pregistered_terms_cache.find(n) == d_pregistered_terms_cache.end() ) {
579 d_pregistered_terms_cache.insert(n);
580 //check for logic exceptions
581 if( !options::stringExp() ){
582 if( n.getKind()==kind::STRING_STRIDOF ||
583 n.getKind() == kind::STRING_ITOS || n.getKind() == kind::STRING_U16TOS || n.getKind() == kind::STRING_U32TOS ||
584 n.getKind() == kind::STRING_STOI || n.getKind() == kind::STRING_STOU16 || n.getKind() == kind::STRING_STOU32 ||
585 n.getKind() == kind::STRING_STRREPL || n.getKind() == kind::STRING_STRCTN ){
586 std::stringstream ss;
587 ss << "Term of kind " << n.getKind() << " not supported in default mode, try --strings-exp";
588 throw LogicException(ss.str());
589 }
590 }
591 switch( n.getKind() ) {
592 case kind::EQUAL: {
593 d_equalityEngine.addTriggerEquality(n);
594 break;
595 }
596 case kind::STRING_IN_REGEXP: {
597 d_out->requirePhase(n, true);
598 d_equalityEngine.addTriggerPredicate(n);
599 d_equalityEngine.addTerm(n[0]);
600 d_equalityEngine.addTerm(n[1]);
601 break;
602 }
603 default: {
604 TypeNode tn = n.getType();
605 if( tn.isString() ) {
606 registerTerm( n, 0 );
607 // FMF
608 if( n.getKind() == kind::VARIABLE && options::stringFMF() ){
609 d_input_vars.insert(n);
610 }
611 d_equalityEngine.addTerm(n);
612 } else if (tn.isBoolean()) {
613 // Get triggered for both equal and dis-equal
614 d_equalityEngine.addTriggerPredicate(n);
615 } else {
616 // Function applications/predicates
617 d_equalityEngine.addTerm(n);
618 if( options::stringExp() ){
619 //collect extended functions here: some may not be asserted to strings (such as those with return type Int),
620 // but we need to record them so they are treated properly
621 d_extt->registerTermRec( n );
622 }
623 }
624 //concat terms do not contribute to theory combination? TODO: verify
625 if( n.hasOperator() && kindToTheoryId( n.getKind() )==THEORY_STRINGS && n.getKind()!=kind::STRING_CONCAT ){
626 d_functionsTerms.push_back( n );
627 }
628 }
629 }
630 }
631 }
632
633 Node TheoryStrings::expandDefinition(LogicRequest &logicRequest, Node node) {
634 Trace("strings-exp-def") << "TheoryStrings::expandDefinition : " << node << std::endl;
635 return node;
636 }
637
638
639 void TheoryStrings::check(Effort e) {
640 if (done() && e<EFFORT_FULL) {
641 return;
642 }
643
644 TimerStat::CodeTimer checkTimer(d_checkTime);
645
646 bool polarity;
647 TNode atom;
648
649 /*if(getLogicInfo().hasEverything()) {
650 WarningOnce() << "WARNING: strings not supported in default configuration (ALL_SUPPORTED).\n"
651 << "To suppress this warning in the future use proper logic symbol, e.g. (set-logic QF_S)." << std::endl;
652 }
653 }*/
654
655 if( !done() && !hasTerm( d_emptyString ) ) {
656 preRegisterTerm( d_emptyString );
657 }
658
659 // Trace("strings-process") << "Theory of strings, check : " << e << std::endl;
660 Trace("strings-check") << "Theory of strings, check : " << e << std::endl;
661 while ( !done() && !d_conflict ) {
662 // Get all the assertions
663 Assertion assertion = get();
664 TNode fact = assertion.assertion;
665
666 Trace("strings-assertion") << "get assertion: " << fact << endl;
667 polarity = fact.getKind() != kind::NOT;
668 atom = polarity ? fact : fact[0];
669
670 //assert pending fact
671 assertPendingFact( atom, polarity, fact );
672 }
673 doPendingFacts();
674
675 if( !d_conflict && ( ( e == EFFORT_FULL && !d_valuation.needCheck() ) || ( e==EFFORT_STANDARD && options::stringEager() ) ) ) {
676 Trace("strings-check") << "Theory of strings full effort check " << std::endl;
677
678 if(Trace.isOn("strings-eqc")) {
679 for( unsigned t=0; t<2; t++ ) {
680 eq::EqClassesIterator eqcs2_i = eq::EqClassesIterator( &d_equalityEngine );
681 Trace("strings-eqc") << (t==0 ? "STRINGS:" : "OTHER:") << std::endl;
682 while( !eqcs2_i.isFinished() ){
683 Node eqc = (*eqcs2_i);
684 bool print = (t==0 && eqc.getType().isString() ) || (t==1 && !eqc.getType().isString() );
685 if (print) {
686 eq::EqClassIterator eqc2_i = eq::EqClassIterator( eqc, &d_equalityEngine );
687 Trace("strings-eqc") << "Eqc( " << eqc << " ) : { ";
688 while( !eqc2_i.isFinished() ) {
689 if( (*eqc2_i)!=eqc && (*eqc2_i).getKind()!=kind::EQUAL ){
690 Trace("strings-eqc") << (*eqc2_i) << " ";
691 }
692 ++eqc2_i;
693 }
694 Trace("strings-eqc") << " } " << std::endl;
695 EqcInfo * ei = getOrMakeEqcInfo( eqc, false );
696 if( ei ){
697 Trace("strings-eqc-debug") << "* Length term : " << ei->d_length_term.get() << std::endl;
698 Trace("strings-eqc-debug") << "* Cardinality lemma k : " << ei->d_cardinality_lem_k.get() << std::endl;
699 Trace("strings-eqc-debug") << "* Normalization length lemma : " << ei->d_normalized_length.get() << std::endl;
700 }
701 }
702 ++eqcs2_i;
703 }
704 Trace("strings-eqc") << std::endl;
705 }
706 Trace("strings-eqc") << std::endl;
707 }
708
709 bool addedLemma = false;
710 bool addedFact;
711 do{
712 Trace("strings-process") << "----check, next round---" << std::endl;
713 checkInit();
714 Trace("strings-process") << "Done check init, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
715 if( !hasProcessed() ){
716 checkExtfEval();
717 Trace("strings-process") << "Done check extended functions eval, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
718 if( !hasProcessed() ){
719 checkFlatForms();
720 Trace("strings-process") << "Done check flat forms, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
721 if( !hasProcessed() && e==EFFORT_FULL ){
722 checkNormalForms();
723 Trace("strings-process") << "Done check normal forms, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
724 if( !hasProcessed() ){
725 if( options::stringEagerLen() ){
726 checkLengthsEqc();
727 Trace("strings-process") << "Done check lengths, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
728 }
729 if( !hasProcessed() ){
730 if( options::stringExp() && !options::stringGuessModel() ){
731 checkExtfReductions( 2 );
732 Trace("strings-process") << "Done check extended functions reduction 2, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
733 }
734 if( !hasProcessed() ){
735 checkMemberships();
736 Trace("strings-process") << "Done check memberships, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
737 if( !hasProcessed() ){
738 checkCardinality();
739 Trace("strings-process") << "Done check cardinality, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
740 }
741 }
742 }
743 }
744 }
745 }
746 }
747 //flush the facts
748 addedFact = !d_pending.empty();
749 addedLemma = !d_lemma_cache.empty();
750 doPendingFacts();
751 doPendingLemmas();
752 }while( !d_conflict && !addedLemma && addedFact );
753
754 Trace("strings-check") << "Theory of strings done full effort check " << addedLemma << " " << d_conflict << std::endl;
755 }else if( e==EFFORT_LAST_CALL ){
756 Assert( !hasProcessed() );
757 Trace("strings-check") << "Theory of strings last call effort check " << std::endl;
758 checkExtfEval( 3 );
759 checkExtfReductions( 2 );
760 doPendingFacts();
761 doPendingLemmas();
762 Trace("strings-process") << "Done check extended functions reduction 2, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
763 }
764 Trace("strings-check") << "Theory of strings, done check : " << e << std::endl;
765 Assert( d_pending.empty() );
766 Assert( d_lemma_cache.empty() );
767 }
768
769 bool TheoryStrings::needsCheckLastEffort() {
770 if( options::stringGuessModel() ){
771 return d_has_extf.get();
772 }else{
773 return false;
774 }
775 }
776
777 void TheoryStrings::checkExtfReductions( int effort ) {
778 //standardize this?
779 //std::vector< Node > nred;
780 //d_extt->doReductions( effort, nred, false );
781
782 std::vector< Node > extf;
783 d_extt->getActive( extf );
784 Trace("strings-process") << "checking " << extf.size() << " active extf" << std::endl;
785 for( unsigned i=0; i<extf.size(); i++ ){
786 Node n = extf[i];
787 Trace("strings-process") << "Check " << n << ", active in model=" << d_extf_info_tmp[n].d_model_active << std::endl;
788 Node nr;
789 int ret = getReduction( effort, n, nr );
790 Assert( nr.isNull() );
791 if( ret!=0 ){
792 d_extt->markReduced( extf[i] );
793 if( options::stringOpt1() && hasProcessed() ){
794 return;
795 }
796 }
797 }
798 }
799
800 TheoryStrings::EqcInfo::EqcInfo( context::Context* c ) : d_length_term(c), d_cardinality_lem_k(c), d_normalized_length(c) {
801
802 }
803
804 TheoryStrings::EqcInfo * TheoryStrings::getOrMakeEqcInfo( Node eqc, bool doMake ) {
805 std::map< Node, EqcInfo* >::iterator eqc_i = d_eqc_info.find( eqc );
806 if( eqc_i==d_eqc_info.end() ){
807 if( doMake ){
808 EqcInfo* ei = new EqcInfo( getSatContext() );
809 d_eqc_info[eqc] = ei;
810 return ei;
811 }else{
812 return NULL;
813 }
814 }else{
815 return (*eqc_i).second;
816 }
817 }
818
819
820 /** Conflict when merging two constants */
821 void TheoryStrings::conflict(TNode a, TNode b){
822 if( !d_conflict ){
823 Debug("strings-conflict") << "Making conflict..." << std::endl;
824 d_conflict = true;
825 Node conflictNode;
826 if (a.getKind() == kind::CONST_BOOLEAN) {
827 conflictNode = explain( a.iffNode(b) );
828 } else {
829 conflictNode = explain( a.eqNode(b) );
830 }
831 Trace("strings-conflict") << "CONFLICT: Eq engine conflict : " << conflictNode << std::endl;
832 d_out->conflict( conflictNode );
833 }
834 }
835
836 /** called when a new equivalance class is created */
837 void TheoryStrings::eqNotifyNewClass(TNode t){
838 if( t.getKind() == kind::STRING_LENGTH ){
839 Trace("strings-debug") << "New length eqc : " << t << std::endl;
840 Node r = d_equalityEngine.getRepresentative(t[0]);
841 EqcInfo * ei = getOrMakeEqcInfo( r, true );
842 ei->d_length_term = t[0];
843 //we care about the length of this string
844 registerTerm( t[0], 1 );
845 }else{
846 //d_extt->registerTerm( t );
847 }
848 }
849
850 /** called when two equivalance classes will merge */
851 void TheoryStrings::eqNotifyPreMerge(TNode t1, TNode t2){
852 EqcInfo * e2 = getOrMakeEqcInfo(t2, false);
853 if( e2 ){
854 EqcInfo * e1 = getOrMakeEqcInfo( t1 );
855 //add information from e2 to e1
856 if( !e2->d_length_term.get().isNull() ){
857 e1->d_length_term.set( e2->d_length_term );
858 }
859 if( e2->d_cardinality_lem_k.get()>e1->d_cardinality_lem_k.get() ) {
860 e1->d_cardinality_lem_k.set( e2->d_cardinality_lem_k );
861 }
862 if( !e2->d_normalized_length.get().isNull() ){
863 e1->d_normalized_length.set( e2->d_normalized_length );
864 }
865 }
866 }
867
868 /** called when two equivalance classes have merged */
869 void TheoryStrings::eqNotifyPostMerge(TNode t1, TNode t2) {
870
871 }
872
873 /** called when two equivalance classes are disequal */
874 void TheoryStrings::eqNotifyDisequal(TNode t1, TNode t2, TNode reason) {
875 if( t1.getType().isString() ){
876 //store disequalities between strings, may need to check if their lengths are equal/disequal
877 d_ee_disequalities.push_back( t1.eqNode( t2 ) );
878 }
879 }
880
881 void TheoryStrings::addCarePairs( quantifiers::TermArgTrie * t1, quantifiers::TermArgTrie * t2, unsigned arity, unsigned depth ) {
882 if( depth==arity ){
883 if( t2!=NULL ){
884 Node f1 = t1->getNodeData();
885 Node f2 = t2->getNodeData();
886 if( !d_equalityEngine.areEqual( f1, f2 ) ){
887 Trace("strings-cg-debug") << "TheoryStrings::computeCareGraph(): checking function " << f1 << " and " << f2 << std::endl;
888 vector< pair<TNode, TNode> > currentPairs;
889 for (unsigned k = 0; k < f1.getNumChildren(); ++ k) {
890 TNode x = f1[k];
891 TNode y = f2[k];
892 Assert( d_equalityEngine.hasTerm(x) );
893 Assert( d_equalityEngine.hasTerm(y) );
894 Assert( !d_equalityEngine.areDisequal( x, y, false ) );
895 if( !d_equalityEngine.areEqual( x, y ) ){
896 if( d_equalityEngine.isTriggerTerm(x, THEORY_STRINGS) && d_equalityEngine.isTriggerTerm(y, THEORY_STRINGS) ){
897 TNode x_shared = d_equalityEngine.getTriggerTermRepresentative(x, THEORY_STRINGS);
898 TNode y_shared = d_equalityEngine.getTriggerTermRepresentative(y, THEORY_STRINGS);
899 EqualityStatus eqStatus = d_valuation.getEqualityStatus(x_shared, y_shared);
900 if( eqStatus==EQUALITY_FALSE_AND_PROPAGATED || eqStatus==EQUALITY_FALSE || eqStatus==EQUALITY_FALSE_IN_MODEL ){
901 //an argument is disequal, we are done
902 return;
903 }else{
904 currentPairs.push_back(make_pair(x_shared, y_shared));
905 }
906 }
907 }
908 }
909 for (unsigned c = 0; c < currentPairs.size(); ++ c) {
910 Trace("strings-cg-pair") << "TheoryStrings::computeCareGraph(): pair : " << currentPairs[c].first << " " << currentPairs[c].second << std::endl;
911 addCarePair(currentPairs[c].first, currentPairs[c].second);
912 }
913 }
914 }
915 }else{
916 if( t2==NULL ){
917 if( depth<(arity-1) ){
918 //add care pairs internal to each child
919 for( std::map< TNode, quantifiers::TermArgTrie >::iterator it = t1->d_data.begin(); it != t1->d_data.end(); ++it ){
920 addCarePairs( &it->second, NULL, arity, depth+1 );
921 }
922 }
923 //add care pairs based on each pair of non-disequal arguments
924 for( std::map< TNode, quantifiers::TermArgTrie >::iterator it = t1->d_data.begin(); it != t1->d_data.end(); ++it ){
925 std::map< TNode, quantifiers::TermArgTrie >::iterator it2 = it;
926 ++it2;
927 for( ; it2 != t1->d_data.end(); ++it2 ){
928 if( !d_equalityEngine.areDisequal(it->first, it2->first, false) ){
929 addCarePairs( &it->second, &it2->second, arity, depth+1 );
930 }
931 }
932 }
933 }else{
934 //add care pairs based on product of indices, non-disequal arguments
935 for( std::map< TNode, quantifiers::TermArgTrie >::iterator it = t1->d_data.begin(); it != t1->d_data.end(); ++it ){
936 for( std::map< TNode, quantifiers::TermArgTrie >::iterator it2 = t2->d_data.begin(); it2 != t2->d_data.end(); ++it2 ){
937 if( !d_equalityEngine.areDisequal(it->first, it2->first, false) ){
938 addCarePairs( &it->second, &it2->second, arity, depth+1 );
939 }
940 }
941 }
942 }
943 }
944 }
945
946 void TheoryStrings::computeCareGraph(){
947 //computing the care graph here is probably still necessary, due to operators that take non-string arguments TODO: verify
948 Trace("strings-cg") << "TheoryStrings::computeCareGraph(): Build term indices..." << std::endl;
949 std::map< Node, quantifiers::TermArgTrie > index;
950 std::map< Node, unsigned > arity;
951 unsigned functionTerms = d_functionsTerms.size();
952 for (unsigned i = 0; i < functionTerms; ++ i) {
953 TNode f1 = d_functionsTerms[i];
954 Trace("strings-cg") << "...build for " << f1 << std::endl;
955 Node op = f1.getOperator();
956 std::vector< TNode > reps;
957 bool has_trigger_arg = false;
958 for( unsigned j=0; j<f1.getNumChildren(); j++ ){
959 reps.push_back( d_equalityEngine.getRepresentative( f1[j] ) );
960 if( d_equalityEngine.isTriggerTerm( f1[j], THEORY_STRINGS ) ){
961 has_trigger_arg = true;
962 }
963 }
964 if( has_trigger_arg ){
965 index[op].addTerm( f1, reps );
966 arity[op] = reps.size();
967 }
968 }
969 //for each index
970 for( std::map< Node, quantifiers::TermArgTrie >::iterator itii = index.begin(); itii != index.end(); ++itii ){
971 Trace("strings-cg") << "TheoryStrings::computeCareGraph(): Process index " << itii->first << "..." << std::endl;
972 addCarePairs( &itii->second, NULL, arity[ itii->first ], 0 );
973 }
974 }
975
976 void TheoryStrings::assertPendingFact(Node atom, bool polarity, Node exp) {
977 Trace("strings-pending") << "Assert pending fact : " << atom << " " << polarity << " from " << exp << std::endl;
978 Assert(atom.getKind() != kind::OR, "Infer error: a split.");
979 if( atom.getKind()==kind::EQUAL ){
980 Trace("strings-pending-debug") << " Register term" << std::endl;
981 for( unsigned j=0; j<2; j++ ) {
982 if( !d_equalityEngine.hasTerm( atom[j] ) && atom[j].getType().isString() ) {
983 registerTerm( atom[j], 0 );
984 }
985 }
986 Trace("strings-pending-debug") << " Now assert equality" << std::endl;
987 d_equalityEngine.assertEquality( atom, polarity, exp );
988 Trace("strings-pending-debug") << " Finished assert equality" << std::endl;
989 } else {
990 d_equalityEngine.assertPredicate( atom, polarity, exp );
991 //process extf
992 if( atom.getKind()==kind::STRING_IN_REGEXP ){
993 if( polarity && atom[1].getKind()==kind::REGEXP_RANGE ){
994 if( d_extf_infer_cache_u.find( atom )==d_extf_infer_cache_u.end() ){
995 d_extf_infer_cache_u.insert( atom );
996 //length of first argument is one
997 Node conc = d_one.eqNode( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, atom[0] ) );
998 Node lem = NodeManager::currentNM()->mkNode( kind::OR, atom.negate(), conc );
999 Trace("strings-lemma") << "Strings::Lemma RE-Range-Len : " << lem << std::endl;
1000 d_out->lemma( lem );
1001 }
1002 }
1003 }
1004 //register the atom here, since it may not create a new equivalence class
1005 //d_extt->registerTerm( atom );
1006 }
1007 Trace("strings-pending-debug") << " Now collect terms" << std::endl;
1008 //collect extended function terms in the atom
1009 d_extt->registerTermRec( atom );
1010 Trace("strings-pending-debug") << " Finished collect terms" << std::endl;
1011 }
1012
1013 void TheoryStrings::doPendingFacts() {
1014 size_t i=0;
1015 while( !d_conflict && i<d_pending.size() ) {
1016 Node fact = d_pending[i];
1017 Node exp = d_pending_exp[ fact ];
1018 if(fact.getKind() == kind::AND) {
1019 for(size_t j=0; j<fact.getNumChildren(); j++) {
1020 bool polarity = fact[j].getKind() != kind::NOT;
1021 TNode atom = polarity ? fact[j] : fact[j][0];
1022 assertPendingFact(atom, polarity, exp);
1023 }
1024 } else {
1025 bool polarity = fact.getKind() != kind::NOT;
1026 TNode atom = polarity ? fact : fact[0];
1027 assertPendingFact(atom, polarity, exp);
1028 }
1029 i++;
1030 }
1031 d_pending.clear();
1032 d_pending_exp.clear();
1033 }
1034
1035 void TheoryStrings::doPendingLemmas() {
1036 if( !d_conflict && !d_lemma_cache.empty() ){
1037 for( unsigned i=0; i<d_lemma_cache.size(); i++ ){
1038 Trace("strings-pending") << "Process pending lemma : " << d_lemma_cache[i] << std::endl;
1039 d_out->lemma( d_lemma_cache[i] );
1040 }
1041 for( std::map< Node, bool >::iterator it = d_pending_req_phase.begin(); it != d_pending_req_phase.end(); ++it ){
1042 Trace("strings-pending") << "Require phase : " << it->first << ", polarity = " << it->second << std::endl;
1043 d_out->requirePhase( it->first, it->second );
1044 }
1045 }
1046 d_lemma_cache.clear();
1047 d_pending_req_phase.clear();
1048 }
1049
1050 bool TheoryStrings::hasProcessed() {
1051 return d_conflict || !d_lemma_cache.empty() || !d_pending.empty();
1052 }
1053
1054 void TheoryStrings::addToExplanation( Node a, Node b, std::vector< Node >& exp ) {
1055 if( a!=b ){
1056 Debug("strings-explain") << "Add to explanation : " << a << " == " << b << std::endl;
1057 Assert( areEqual( a, b ) );
1058 exp.push_back( a.eqNode( b ) );
1059 }
1060 }
1061
1062 void TheoryStrings::addToExplanation( Node lit, std::vector< Node >& exp ) {
1063 if( !lit.isNull() ){
1064 exp.push_back( lit );
1065 }
1066 }
1067
1068 void TheoryStrings::checkInit() {
1069 //build term index
1070 d_eqc_to_const.clear();
1071 d_eqc_to_const_base.clear();
1072 d_eqc_to_const_exp.clear();
1073 d_eqc_to_len_term.clear();
1074 d_term_index.clear();
1075 d_strings_eqc.clear();
1076
1077 std::map< Kind, unsigned > ncongruent;
1078 std::map< Kind, unsigned > congruent;
1079 d_emptyString_r = getRepresentative( d_emptyString );
1080 eq::EqClassesIterator eqcs_i = eq::EqClassesIterator( &d_equalityEngine );
1081 while( !eqcs_i.isFinished() ){
1082 Node eqc = (*eqcs_i);
1083 TypeNode tn = eqc.getType();
1084 if( !tn.isRegExp() ){
1085 if( tn.isString() ){
1086 d_strings_eqc.push_back( eqc );
1087 }
1088 Node var;
1089 eq::EqClassIterator eqc_i = eq::EqClassIterator( eqc, &d_equalityEngine );
1090 while( !eqc_i.isFinished() ) {
1091 Node n = *eqc_i;
1092 if( n.isConst() ){
1093 d_eqc_to_const[eqc] = n;
1094 d_eqc_to_const_base[eqc] = n;
1095 d_eqc_to_const_exp[eqc] = Node::null();
1096 }else if( tn.isInteger() ){
1097 if( n.getKind()==kind::STRING_LENGTH ){
1098 Node nr = getRepresentative( n[0] );
1099 d_eqc_to_len_term[nr] = n[0];
1100 }
1101 }else if( n.getNumChildren()>0 ){
1102 Kind k = n.getKind();
1103 if( k!=kind::EQUAL ){
1104 if( d_congruent.find( n )==d_congruent.end() ){
1105 std::vector< Node > c;
1106 Node nc = d_term_index[k].add( n, 0, this, d_emptyString_r, c );
1107 if( nc!=n ){
1108 //check if we have inferred a new equality by removal of empty components
1109 if( n.getKind()==kind::STRING_CONCAT && !areEqual( nc, n ) ){
1110 std::vector< Node > exp;
1111 unsigned count[2] = { 0, 0 };
1112 while( count[0]<nc.getNumChildren() || count[1]<n.getNumChildren() ){
1113 //explain empty prefixes
1114 for( unsigned t=0; t<2; t++ ){
1115 Node nn = t==0 ? nc : n;
1116 while( count[t]<nn.getNumChildren() &&
1117 ( nn[count[t]]==d_emptyString || areEqual( nn[count[t]], d_emptyString ) ) ){
1118 if( nn[count[t]]!=d_emptyString ){
1119 exp.push_back( nn[count[t]].eqNode( d_emptyString ) );
1120 }
1121 count[t]++;
1122 }
1123 }
1124 //explain equal components
1125 if( count[0]<nc.getNumChildren() ){
1126 Assert( count[1]<n.getNumChildren() );
1127 if( nc[count[0]]!=n[count[1]] ){
1128 exp.push_back( nc[count[0]].eqNode( n[count[1]] ) );
1129 }
1130 count[0]++;
1131 count[1]++;
1132 }
1133 }
1134 //infer the equality
1135 sendInference( exp, n.eqNode( nc ), "I_Norm" );
1136 }else if( d_extt->hasFunctionKind( n.getKind() ) ){
1137 //mark as congruent : only process if neither has been reduced
1138 d_extt->markCongruent( nc, n );
1139 }
1140 //this node is congruent to another one, we can ignore it
1141 Trace("strings-process-debug") << " congruent term : " << n << std::endl;
1142 d_congruent.insert( n );
1143 congruent[k]++;
1144 }else if( k==kind::STRING_CONCAT && c.size()==1 ){
1145 Trace("strings-process-debug") << " congruent term by singular : " << n << " " << c[0] << std::endl;
1146 //singular case
1147 if( !areEqual( c[0], n ) ){
1148 std::vector< Node > exp;
1149 //explain empty components
1150 bool foundNEmpty = false;
1151 for( unsigned i=0; i<n.getNumChildren(); i++ ){
1152 if( areEqual( n[i], d_emptyString ) ){
1153 if( n[i]!=d_emptyString ){
1154 exp.push_back( n[i].eqNode( d_emptyString ) );
1155 }
1156 }else{
1157 Assert( !foundNEmpty );
1158 if( n[i]!=c[0] ){
1159 exp.push_back( n[i].eqNode( c[0] ) );
1160 }
1161 foundNEmpty = true;
1162 }
1163 }
1164 AlwaysAssert( foundNEmpty );
1165 //infer the equality
1166 sendInference( exp, n.eqNode( c[0] ), "I_Norm_S" );
1167 }
1168 d_congruent.insert( n );
1169 congruent[k]++;
1170 }else{
1171 ncongruent[k]++;
1172 }
1173 }else{
1174 congruent[k]++;
1175 }
1176 }
1177 }else{
1178 if( d_congruent.find( n )==d_congruent.end() ){
1179 if( var.isNull() ){
1180 var = n;
1181 }else{
1182 Trace("strings-process-debug") << " congruent variable : " << n << std::endl;
1183 d_congruent.insert( n );
1184 }
1185 }
1186 }
1187 ++eqc_i;
1188 }
1189 }
1190 ++eqcs_i;
1191 }
1192 if( Trace.isOn("strings-process") ){
1193 for( std::map< Kind, TermIndex >::iterator it = d_term_index.begin(); it != d_term_index.end(); ++it ){
1194 Trace("strings-process") << " Terms[" << it->first << "] = " << ncongruent[it->first] << "/" << (congruent[it->first]+ncongruent[it->first]) << std::endl;
1195 }
1196 }
1197 Trace("strings-process") << "Done check init, addedLemma = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
1198 //now, infer constants for equivalence classes
1199 if( !hasProcessed() ){
1200 //do fixed point
1201 unsigned prevSize;
1202 do{
1203 Trace("strings-process-debug") << "Check constant equivalence classes..." << std::endl;
1204 prevSize = d_eqc_to_const.size();
1205 std::vector< Node > vecc;
1206 checkConstantEquivalenceClasses( &d_term_index[kind::STRING_CONCAT], vecc );
1207 }while( !hasProcessed() && d_eqc_to_const.size()>prevSize );
1208 Trace("strings-process") << "Done check constant equivalence classes, addedLemma = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
1209 }
1210 }
1211
1212 void TheoryStrings::checkConstantEquivalenceClasses( TermIndex* ti, std::vector< Node >& vecc ) {
1213 Node n = ti->d_data;
1214 if( !n.isNull() ){
1215 //construct the constant
1216 Node c = mkConcat( vecc );
1217 if( !areEqual( n, c ) ){
1218 Trace("strings-debug") << "Constant eqc : " << c << " for " << n << std::endl;
1219 Trace("strings-debug") << " ";
1220 for( unsigned i=0; i<vecc.size(); i++ ){
1221 Trace("strings-debug") << vecc[i] << " ";
1222 }
1223 Trace("strings-debug") << std::endl;
1224 unsigned count = 0;
1225 unsigned countc = 0;
1226 std::vector< Node > exp;
1227 while( count<n.getNumChildren() ){
1228 while( count<n.getNumChildren() && areEqual( n[count], d_emptyString ) ){
1229 addToExplanation( n[count], d_emptyString, exp );
1230 count++;
1231 }
1232 if( count<n.getNumChildren() ){
1233 Trace("strings-debug") << "...explain " << n[count] << " " << vecc[countc] << std::endl;
1234 if( !areEqual( n[count], vecc[countc] ) ){
1235 Node nrr = getRepresentative( n[count] );
1236 Assert( !d_eqc_to_const_exp[nrr].isNull() );
1237 addToExplanation( n[count], d_eqc_to_const_base[nrr], exp );
1238 exp.push_back( d_eqc_to_const_exp[nrr] );
1239 }else{
1240 addToExplanation( n[count], vecc[countc], exp );
1241 }
1242 countc++;
1243 count++;
1244 }
1245 }
1246 //exp contains an explanation of n==c
1247 Assert( countc==vecc.size() );
1248 if( hasTerm( c ) ){
1249 sendInference( exp, n.eqNode( c ), "I_CONST_MERGE" );
1250 return;
1251 }else if( !hasProcessed() ){
1252 Node nr = getRepresentative( n );
1253 std::map< Node, Node >::iterator it = d_eqc_to_const.find( nr );
1254 if( it==d_eqc_to_const.end() ){
1255 Trace("strings-debug") << "Set eqc const " << n << " to " << c << std::endl;
1256 d_eqc_to_const[nr] = c;
1257 d_eqc_to_const_base[nr] = n;
1258 d_eqc_to_const_exp[nr] = mkAnd( exp );
1259 }else if( c!=it->second ){
1260 //conflict
1261 Trace("strings-debug") << "Conflict, other constant was " << it->second << ", this constant was " << c << std::endl;
1262 if( d_eqc_to_const_exp[nr].isNull() ){
1263 // n==c ^ n == c' => false
1264 addToExplanation( n, it->second, exp );
1265 }else{
1266 // n==c ^ n == d_eqc_to_const_base[nr] == c' => false
1267 exp.push_back( d_eqc_to_const_exp[nr] );
1268 addToExplanation( n, d_eqc_to_const_base[nr], exp );
1269 }
1270 sendInference( exp, d_false, "I_CONST_CONFLICT" );
1271 return;
1272 }else{
1273 Trace("strings-debug") << "Duplicate constant." << std::endl;
1274 }
1275 }
1276 }
1277 }
1278 for( std::map< TNode, TermIndex >::iterator it = ti->d_children.begin(); it != ti->d_children.end(); ++it ){
1279 std::map< Node, Node >::iterator itc = d_eqc_to_const.find( it->first );
1280 if( itc!=d_eqc_to_const.end() ){
1281 vecc.push_back( itc->second );
1282 checkConstantEquivalenceClasses( &it->second, vecc );
1283 vecc.pop_back();
1284 if( hasProcessed() ){
1285 break;
1286 }
1287 }
1288 }
1289 }
1290
1291 void TheoryStrings::checkExtfEval( int effort ) {
1292 Trace("strings-extf-list") << "Active extended functions, effort=" << effort << " : " << std::endl;
1293 d_extf_info_tmp.clear();
1294 bool has_nreduce = false;
1295 std::vector< Node > terms;
1296 std::vector< Node > sterms;
1297 std::vector< std::vector< Node > > exp;
1298 d_extt->getActive( terms );
1299 d_extt->getSubstitutedTerms( effort, terms, sterms, exp );
1300 for( unsigned i=0; i<terms.size(); i++ ){
1301 Node n = terms[i];
1302 Node sn = sterms[i];
1303 //setup information about extf
1304 d_extf_info_tmp[n].init();
1305 std::map< Node, ExtfInfoTmp >::iterator itit = d_extf_info_tmp.find( n );
1306 if( n.getType().isBoolean() ){
1307 if( areEqual( n, d_true ) ){
1308 itit->second.d_pol = 1;
1309 }else if( areEqual( n, d_false ) ){
1310 itit->second.d_pol = -1;
1311 }
1312 }
1313 Trace("strings-extf-debug") << "Check extf " << n << " == " << sn << ", pol = " << itit->second.d_pol << ", effort=" << effort << "..." << std::endl;
1314 //do the inference
1315 Node to_reduce;
1316 if( n!=sn ){
1317 itit->second.d_exp.insert( itit->second.d_exp.end(), exp[i].begin(), exp[i].end() );
1318 // inference is rewriting the substituted node
1319 Node nrc = Rewriter::rewrite( sn );
1320 //if rewrites to a constant, then do the inference and mark as reduced
1321 if( nrc.isConst() ){
1322 if( effort<3 ){
1323 d_extt->markReduced( n );
1324 Trace("strings-extf-debug") << " resolvable by evaluation..." << std::endl;
1325 std::vector< Node > exps;
1326 Trace("strings-extf-debug") << " get symbolic definition..." << std::endl;
1327 Node nrs = getSymbolicDefinition( sn, exps );
1328 if( !nrs.isNull() ){
1329 Trace("strings-extf-debug") << " rewrite " << nrs << "..." << std::endl;
1330 nrs = Rewriter::rewrite( nrs );
1331 //ensure the symbolic form is non-trivial
1332 if( nrs.isConst() ){
1333 Trace("strings-extf-debug") << " symbolic definition is trivial..." << std::endl;
1334 nrs = Node::null();
1335 }
1336 }else{
1337 Trace("strings-extf-debug") << " could not infer symbolic definition." << std::endl;
1338 }
1339 Node conc;
1340 if( !nrs.isNull() ){
1341 Trace("strings-extf-debug") << " symbolic def : " << nrs << std::endl;
1342 if( !areEqual( nrs, nrc ) ){
1343 //infer symbolic unit
1344 if( n.getType().isBoolean() ){
1345 conc = nrc==d_true ? nrs : nrs.negate();
1346 }else{
1347 conc = nrs.eqNode( nrc );
1348 }
1349 itit->second.d_exp.clear();
1350 }
1351 }else{
1352 if( !areEqual( n, nrc ) ){
1353 if( n.getType().isBoolean() ){
1354 if( areEqual( n, nrc==d_true ? d_false : d_true ) ){
1355 itit->second.d_exp.push_back( nrc==d_true ? n.negate() : n );
1356 conc = d_false;
1357 }else{
1358 conc = nrc==d_true ? n : n.negate();
1359 }
1360 }else{
1361 conc = n.eqNode( nrc );
1362 }
1363 }
1364 }
1365 if( !conc.isNull() ){
1366 Trace("strings-extf") << " resolve extf : " << sn << " -> " << nrc << std::endl;
1367 sendInference( itit->second.d_exp, conc, effort==0 ? "EXTF" : "EXTF-N", true );
1368 if( d_conflict ){
1369 Trace("strings-extf-debug") << " conflict, return." << std::endl;
1370 return;
1371 }
1372 }
1373 }else{
1374 //check if it is already equal, if so, mark as reduced. Otherwise, do nothing.
1375 if( areEqual( n, nrc ) ){
1376 Trace("strings-extf") << " resolved extf, since satisfied by model: " << n << std::endl;
1377 itit->second.d_model_active = false;
1378 }
1379 }
1380 //if it reduces to a conjunction, infer each and reduce
1381 }else if( ( nrc.getKind()==kind::OR && itit->second.d_pol==-1 ) || ( nrc.getKind()==kind::AND && itit->second.d_pol==1 ) ){
1382 Assert( effort<3 );
1383 d_extt->markReduced( n );
1384 itit->second.d_exp.push_back( itit->second.d_pol==-1 ? n.negate() : n );
1385 Trace("strings-extf-debug") << " decomposable..." << std::endl;
1386 Trace("strings-extf") << " resolve extf : " << sn << " -> " << nrc << ", pol = " << itit->second.d_pol << std::endl;
1387 for( unsigned i=0; i<nrc.getNumChildren(); i++ ){
1388 sendInference( itit->second.d_exp, itit->second.d_pol==-1 ? nrc[i].negate() : nrc[i], effort==0 ? "EXTF_d" : "EXTF_d-N" );
1389 }
1390 }else{
1391 to_reduce = nrc;
1392 }
1393 }else{
1394 to_reduce = sterms[i];
1395 }
1396 //if not reduced
1397 if( !to_reduce.isNull() ){
1398 Assert( effort<3 );
1399 if( effort==1 ){
1400 Trace("strings-extf") << " cannot rewrite extf : " << to_reduce << std::endl;
1401 }
1402 checkExtfInference( n, to_reduce, itit->second, effort );
1403 if( Trace.isOn("strings-extf-list") ){
1404 Trace("strings-extf-list") << " * " << to_reduce;
1405 if( itit->second.d_pol!=0 ){
1406 Trace("strings-extf-list") << ", pol = " << itit->second.d_pol;
1407 }
1408 if( n!=to_reduce ){
1409 Trace("strings-extf-list") << ", from " << n;
1410 }
1411 Trace("strings-extf-list") << std::endl;
1412 }
1413 if( d_extt->isActive( n ) && itit->second.d_model_active ){
1414 has_nreduce = true;
1415 }
1416 }
1417 }
1418 d_has_extf = has_nreduce;
1419 }
1420
1421 void TheoryStrings::checkExtfInference( Node n, Node nr, ExtfInfoTmp& in, int effort ){
1422 //make additional inferences that do not contribute to the reduction of n, but may help show a refutation
1423 if( in.d_pol!=0 ){
1424 //add original to explanation
1425 in.d_exp.push_back( in.d_pol==1 ? n : n.negate() );
1426
1427 //d_extf_infer_cache stores whether we have made the inferences associated with a node n,
1428 // this may need to be generalized if multiple inferences apply
1429
1430 if( nr.getKind()==kind::STRING_STRCTN ){
1431 if( ( in.d_pol==1 && nr[1].getKind()==kind::STRING_CONCAT ) || ( in.d_pol==-1 && nr[0].getKind()==kind::STRING_CONCAT ) ){
1432 if( d_extf_infer_cache.find( nr )==d_extf_infer_cache.end() ){
1433 d_extf_infer_cache.insert( nr );
1434
1435 //one argument does (not) contain each of the components of the other argument
1436 int index = in.d_pol==1 ? 1 : 0;
1437 std::vector< Node > children;
1438 children.push_back( nr[0] );
1439 children.push_back( nr[1] );
1440 //Node exp_n = mkAnd( exp );
1441 for( unsigned i=0; i<nr[index].getNumChildren(); i++ ){
1442 children[index] = nr[index][i];
1443 Node conc = NodeManager::currentNM()->mkNode( kind::STRING_STRCTN, children );
1444 //can mark as reduced, since model for n => model for conc
1445 d_extt->markReduced( conc );
1446 sendInference( in.d_exp, in.d_pol==1 ? conc : conc.negate(), "CTN_Decompose" );
1447 }
1448
1449 }
1450 }else{
1451 //store this (reduced) assertion
1452 //Assert( effort==0 || nr[0]==getRepresentative( nr[0] ) );
1453 bool pol = in.d_pol==1;
1454 if( std::find( d_extf_info_tmp[nr[0]].d_ctn[pol].begin(), d_extf_info_tmp[nr[0]].d_ctn[pol].end(), nr[1] )==d_extf_info_tmp[nr[0]].d_ctn[pol].end() ){
1455 Trace("strings-extf-debug") << " store contains info : " << nr[0] << " " << pol << " " << nr[1] << std::endl;
1456 d_extf_info_tmp[nr[0]].d_ctn[pol].push_back( nr[1] );
1457 d_extf_info_tmp[nr[0]].d_ctn_from[pol].push_back( n );
1458 //transitive closure for contains
1459 bool opol = !pol;
1460 for( unsigned i=0; i<d_extf_info_tmp[nr[0]].d_ctn[opol].size(); i++ ){
1461 Node onr = d_extf_info_tmp[nr[0]].d_ctn[opol][i];
1462 Node conc = NodeManager::currentNM()->mkNode( kind::STRING_STRCTN, pol ? nr[1] : onr, pol ? onr : nr[1] );
1463 conc = Rewriter::rewrite( conc );
1464 bool do_infer = false;
1465 if( conc.getKind()==kind::EQUAL ){
1466 do_infer = !areDisequal( conc[0], conc[1] );
1467 }else{
1468 do_infer = !areEqual( conc, d_false );
1469 }
1470 if( do_infer ){
1471 conc = conc.negate();
1472 std::vector< Node > exp_c;
1473 exp_c.insert( exp_c.end(), in.d_exp.begin(), in.d_exp.end() );
1474 Node ofrom = d_extf_info_tmp[nr[0]].d_ctn_from[opol][i];
1475 Assert( d_extf_info_tmp.find( ofrom )!=d_extf_info_tmp.end() );
1476 exp_c.insert( exp_c.end(), d_extf_info_tmp[ofrom].d_exp.begin(), d_extf_info_tmp[ofrom].d_exp.end() );
1477 sendInference( exp_c, conc, "CTN_Trans" );
1478 }
1479 }
1480 }else{
1481 Trace("strings-extf-debug") << " redundant." << std::endl;
1482 d_extt->markReduced( n );
1483 }
1484 }
1485 }
1486 }
1487 }
1488
1489 void TheoryStrings::collectVars( Node n, std::vector< Node >& vars, std::map< Node, bool >& visited ) {
1490 if( !n.isConst() ){
1491 if( visited.find( n )==visited.end() ){
1492 visited[n] = true;
1493 if( n.getNumChildren()>0 ){
1494 for( unsigned i=0; i<n.getNumChildren(); i++ ){
1495 collectVars( n[i], vars, visited );
1496 }
1497 }else{
1498 //Node nr = getRepresentative( n );
1499 //vars[nr].push_back( n );
1500 vars.push_back( n );
1501 }
1502 }
1503 }
1504 }
1505
1506 Node TheoryStrings::getSymbolicDefinition( Node n, std::vector< Node >& exp ) {
1507 if( n.getNumChildren()==0 ){
1508 NodeNodeMap::const_iterator it = d_proxy_var.find( n );
1509 if( it==d_proxy_var.end() ){
1510 return Node::null();
1511 }else{
1512 Node eq = n.eqNode( (*it).second );
1513 eq = Rewriter::rewrite( eq );
1514 if( std::find( exp.begin(), exp.end(), eq )==exp.end() ){
1515 exp.push_back( eq );
1516 }
1517 return (*it).second;
1518 }
1519 }else{
1520 std::vector< Node > children;
1521 if (n.getMetaKind() == kind::metakind::PARAMETERIZED) {
1522 children.push_back( n.getOperator() );
1523 }
1524 for( unsigned i=0; i<n.getNumChildren(); i++ ){
1525 if( n.getKind()==kind::STRING_IN_REGEXP && i==1 ){
1526 children.push_back( n[i] );
1527 }else{
1528 Node ns = getSymbolicDefinition( n[i], exp );
1529 if( ns.isNull() ){
1530 return Node::null();
1531 }else{
1532 children.push_back( ns );
1533 }
1534 }
1535 }
1536 return NodeManager::currentNM()->mkNode( n.getKind(), children );
1537 }
1538 }
1539
1540 Node TheoryStrings::getConstantEqc( Node eqc ) {
1541 std::map< Node, Node >::iterator it = d_eqc_to_const.find( eqc );
1542 if( it!=d_eqc_to_const.end() ){
1543 return it->second;
1544 }else{
1545 return Node::null();
1546 }
1547 }
1548
1549 void TheoryStrings::debugPrintFlatForms( const char * tc ){
1550 for( unsigned k=0; k<d_strings_eqc.size(); k++ ){
1551 Node eqc = d_strings_eqc[k];
1552 if( d_eqc[eqc].size()>1 ){
1553 Trace( tc ) << "EQC [" << eqc << "]" << std::endl;
1554 }else{
1555 Trace( tc ) << "eqc [" << eqc << "]";
1556 }
1557 std::map< Node, Node >::iterator itc = d_eqc_to_const.find( eqc );
1558 if( itc!=d_eqc_to_const.end() ){
1559 Trace( tc ) << " C: " << itc->second;
1560 if( d_eqc[eqc].size()>1 ){
1561 Trace( tc ) << std::endl;
1562 }
1563 }
1564 if( d_eqc[eqc].size()>1 ){
1565 for( unsigned i=0; i<d_eqc[eqc].size(); i++ ){
1566 Node n = d_eqc[eqc][i];
1567 Trace( tc ) << " ";
1568 for( unsigned j=0; j<d_flat_form[n].size(); j++ ){
1569 Node fc = d_flat_form[n][j];
1570 itc = d_eqc_to_const.find( fc );
1571 Trace( tc ) << " ";
1572 if( itc!=d_eqc_to_const.end() ){
1573 Trace( tc ) << itc->second;
1574 }else{
1575 Trace( tc ) << fc;
1576 }
1577 }
1578 if( n!=eqc ){
1579 Trace( tc ) << ", from " << n;
1580 }
1581 Trace( tc ) << std::endl;
1582 }
1583 }else{
1584 Trace( tc ) << std::endl;
1585 }
1586 }
1587 Trace( tc ) << std::endl;
1588 }
1589
1590 void TheoryStrings::debugPrintNormalForms( const char * tc ) {
1591 }
1592
1593 struct sortConstLength {
1594 std::map< Node, unsigned > d_const_length;
1595 bool operator() (Node i, Node j) {
1596 std::map< Node, unsigned >::iterator it_i = d_const_length.find( i );
1597 std::map< Node, unsigned >::iterator it_j = d_const_length.find( j );
1598 if( it_i==d_const_length.end() ){
1599 if( it_j==d_const_length.end() ){
1600 return i<j;
1601 }else{
1602 return false;
1603 }
1604 }else{
1605 if( it_j==d_const_length.end() ){
1606 return true;
1607 }else{
1608 return it_i->second<it_j->second;
1609 }
1610 }
1611 }
1612 };
1613
1614
1615 void TheoryStrings::checkFlatForms() {
1616 //first check for cycles, while building ordering of equivalence classes
1617 d_eqc.clear();
1618 d_flat_form.clear();
1619 d_flat_form_index.clear();
1620 Trace("strings-process") << "Check equivalence classes cycles...." << std::endl;
1621 //rebuild strings eqc based on acyclic ordering
1622 std::vector< Node > eqc;
1623 eqc.insert( eqc.end(), d_strings_eqc.begin(), d_strings_eqc.end() );
1624 d_strings_eqc.clear();
1625 if( options::stringBinaryCsp() ){
1626 //sort: process smallest constants first (necessary if doing binary splits)
1627 sortConstLength scl;
1628 for( unsigned i=0; i<eqc.size(); i++ ){
1629 std::map< Node, Node >::iterator itc = d_eqc_to_const.find( eqc[i] );
1630 if( itc!=d_eqc_to_const.end() ){
1631 scl.d_const_length[eqc[i]] = itc->second.getConst<String>().size();
1632 }
1633 }
1634 std::sort( eqc.begin(), eqc.end(), scl );
1635 }
1636 for( unsigned i=0; i<eqc.size(); i++ ){
1637 std::vector< Node > curr;
1638 std::vector< Node > exp;
1639 checkCycles( eqc[i], curr, exp );
1640 if( hasProcessed() ){
1641 return;
1642 }
1643 }
1644 Trace("strings-process-debug") << "Done check cycles, lemmas = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << std::endl;
1645 if( !hasProcessed() ){
1646 //debug print flat forms
1647 if( Trace.isOn("strings-ff") ){
1648 Trace("strings-ff") << "Flat forms : " << std::endl;
1649 debugPrintFlatForms( "strings-ff" );
1650 }
1651
1652 //inferences without recursively expanding flat forms
1653
1654 //(1) approximate equality by containment, infer conflicts
1655 for( unsigned k=0; k<d_strings_eqc.size(); k++ ){
1656 Node eqc = d_strings_eqc[k];
1657 Node c = getConstantEqc( eqc );
1658 if( !c.isNull() ){
1659 //if equivalence class is constant, all component constants in flat forms must be contained in it, in order
1660 std::map< Node, std::vector< Node > >::iterator it = d_eqc.find( eqc );
1661 if( it!=d_eqc.end() ){
1662 for( unsigned i=0; i<it->second.size(); i++ ){
1663 Node n = it->second[i];
1664 int firstc, lastc;
1665 if( !TheoryStringsRewriter::canConstantContainList( c, d_flat_form[n], firstc, lastc ) ){
1666 Trace("strings-ff-debug") << "Flat form for " << n << " cannot be contained in constant " << c << std::endl;
1667 Trace("strings-ff-debug") << " indices = " << firstc << "/" << lastc << std::endl;
1668 //conflict, explanation is n = base ^ base = c ^ relevant porition of ( n = f[n] )
1669 std::vector< Node > exp;
1670 Assert( d_eqc_to_const_base.find( eqc )!=d_eqc_to_const_base.end() );
1671 addToExplanation( n, d_eqc_to_const_base[eqc], exp );
1672 Assert( d_eqc_to_const_exp.find( eqc )!=d_eqc_to_const_exp.end() );
1673 if( !d_eqc_to_const_exp[eqc].isNull() ){
1674 exp.push_back( d_eqc_to_const_exp[eqc] );
1675 }
1676 for( int e=firstc; e<=lastc; e++ ){
1677 if( d_flat_form[n][e].isConst() ){
1678 Assert( e>=0 && e<(int)d_flat_form_index[n].size() );
1679 Assert( d_flat_form_index[n][e]>=0 && d_flat_form_index[n][e]<(int)n.getNumChildren() );
1680 addToExplanation( d_flat_form[n][e], n[d_flat_form_index[n][e]], exp );
1681 }
1682 }
1683 Node conc = d_false;
1684 sendInference( exp, conc, "F_NCTN" );
1685 return;
1686 }
1687 }
1688 }
1689 }
1690 }
1691
1692 //(2) scan lists, unification to infer conflicts and equalities
1693 for( unsigned k=0; k<d_strings_eqc.size(); k++ ){
1694 Node eqc = d_strings_eqc[k];
1695 std::map< Node, std::vector< Node > >::iterator it = d_eqc.find( eqc );
1696 if( it!=d_eqc.end() && it->second.size()>1 ){
1697 //iterate over start index
1698 for( unsigned start=0; start<it->second.size()-1; start++ ){
1699 for( unsigned r=0; r<2; r++ ){
1700 unsigned count = 0;
1701 std::vector< Node > inelig;
1702 for( unsigned i=0; i<=start; i++ ){
1703 inelig.push_back( it->second[start] );
1704 }
1705 Node a = it->second[start];
1706 Node b;
1707 do{
1708 std::vector< Node > exp;
1709 //std::vector< Node > exp_n;
1710 Node conc;
1711 int inf_type = -1;
1712 if( count==d_flat_form[a].size() ){
1713 for( unsigned i=start+1; i<it->second.size(); i++ ){
1714 b = it->second[i];
1715 if( std::find( inelig.begin(), inelig.end(), b )==inelig.end() ){
1716 if( count<d_flat_form[b].size() ){
1717 //endpoint
1718 std::vector< Node > conc_c;
1719 for( unsigned j=count; j<d_flat_form[b].size(); j++ ){
1720 conc_c.push_back( b[d_flat_form_index[b][j]].eqNode( d_emptyString ) );
1721 }
1722 Assert( !conc_c.empty() );
1723 conc = mkAnd( conc_c );
1724 inf_type = 2;
1725 Assert( count>0 );
1726 //swap, will enforce is empty past current
1727 a = it->second[i]; b = it->second[start];
1728 count--;
1729 break;
1730 }
1731 inelig.push_back( it->second[i] );
1732 }
1733 }
1734 }else{
1735 Node curr = d_flat_form[a][count];
1736 Node curr_c = getConstantEqc( curr );
1737 Node ac = a[d_flat_form_index[a][count]];
1738 std::vector< Node > lexp;
1739 Node lcurr = getLength( ac, lexp );
1740 for( unsigned i=1; i<it->second.size(); i++ ){
1741 b = it->second[i];
1742 if( std::find( inelig.begin(), inelig.end(), b )==inelig.end() ){
1743 if( count==d_flat_form[b].size() ){
1744 inelig.push_back( b );
1745 //endpoint
1746 std::vector< Node > conc_c;
1747 for( unsigned j=count; j<d_flat_form[a].size(); j++ ){
1748 conc_c.push_back( a[d_flat_form_index[a][j]].eqNode( d_emptyString ) );
1749 }
1750 Assert( !conc_c.empty() );
1751 conc = mkAnd( conc_c );
1752 inf_type = 2;
1753 Assert( count>0 );
1754 count--;
1755 break;
1756 }else{
1757 Node cc = d_flat_form[b][count];
1758 if( cc!=curr ){
1759 Node bc = b[d_flat_form_index[b][count]];
1760 inelig.push_back( b );
1761 Assert( !areEqual( curr, cc ) );
1762 Node cc_c = getConstantEqc( cc );
1763 if( !curr_c.isNull() && !cc_c.isNull() ){
1764 //check for constant conflict
1765 int index;
1766 Node s = TheoryStringsRewriter::splitConstant( cc_c, curr_c, index, r==1 );
1767 if( s.isNull() ){
1768 addToExplanation( ac, d_eqc_to_const_base[curr], exp );
1769 addToExplanation( d_eqc_to_const_exp[curr], exp );
1770 addToExplanation( bc, d_eqc_to_const_base[cc], exp );
1771 addToExplanation( d_eqc_to_const_exp[cc], exp );
1772 conc = d_false;
1773 inf_type = 0;
1774 break;
1775 }
1776 }else if( (d_flat_form[a].size()-1)==count && (d_flat_form[b].size()-1)==count ){
1777 conc = ac.eqNode( bc );
1778 inf_type = 3;
1779 break;
1780 }else{
1781 //if lengths are the same, apply LengthEq
1782 std::vector< Node > lexp2;
1783 Node lcc = getLength( bc, lexp2 );
1784 if( areEqual( lcurr, lcc ) ){
1785 Trace("strings-ff-debug") << "Infer " << ac << " == " << bc << " since " << lcurr << " == " << lcc << std::endl;
1786 //exp_n.push_back( getLength( curr, true ).eqNode( getLength( cc, true ) ) );
1787 Trace("strings-ff-debug") << "Explanation for " << lcurr << " is ";
1788 for( unsigned j=0; j<lexp.size(); j++ ) { Trace("strings-ff-debug") << lexp[j] << std::endl; }
1789 Trace("strings-ff-debug") << "Explanation for " << lcc << " is ";
1790 for( unsigned j=0; j<lexp2.size(); j++ ) { Trace("strings-ff-debug") << lexp2[j] << std::endl; }
1791 exp.insert( exp.end(), lexp.begin(), lexp.end() );
1792 exp.insert( exp.end(), lexp2.begin(), lexp2.end() );
1793 addToExplanation( lcurr, lcc, exp );
1794 conc = ac.eqNode( bc );
1795 inf_type = 1;
1796 break;
1797 }
1798 }
1799 }
1800 }
1801 }
1802 }
1803 }
1804 if( !conc.isNull() ){
1805 Trace("strings-ff-debug") << "Found inference : " << conc << " based on equality " << a << " == " << b << " " << r << " " << inf_type << std::endl;
1806 addToExplanation( a, b, exp );
1807 //explain why prefixes up to now were the same
1808 for( unsigned j=0; j<count; j++ ){
1809 Trace("strings-ff-debug") << "Add at " << d_flat_form_index[a][j] << " " << d_flat_form_index[b][j] << std::endl;
1810 addToExplanation( a[d_flat_form_index[a][j]], b[d_flat_form_index[b][j]], exp );
1811 }
1812 //explain why other components up to now are empty
1813 for( unsigned t=0; t<2; t++ ){
1814 Node c = t==0 ? a : b;
1815 int jj;
1816 if( inf_type==3 || ( t==1 && inf_type==2 ) ){
1817 //explain all the empty components for F_EndpointEq, all for the short end for F_EndpointEmp
1818 jj = r==0 ? c.getNumChildren() : -1;
1819 }else{
1820 jj = t==0 ? d_flat_form_index[a][count] : d_flat_form_index[b][count];
1821 }
1822 if( r==0 ){
1823 for( int j=0; j<jj; j++ ){
1824 if( areEqual( c[j], d_emptyString ) ){
1825 addToExplanation( c[j], d_emptyString, exp );
1826 }
1827 }
1828 }else{
1829 for( int j=(c.getNumChildren()-1); j>jj; --j ){
1830 if( areEqual( c[j], d_emptyString ) ){
1831 addToExplanation( c[j], d_emptyString, exp );
1832 }
1833 }
1834 }
1835 }
1836 //notice that F_EndpointEmp is not typically applied, since strict prefix equality ( a.b = a ) where a,b non-empty
1837 // is conflicting by arithmetic len(a.b)=len(a)+len(b)!=len(a) when len(b)!=0.
1838 sendInference( exp, conc, inf_type==0 ? "F_Const" : ( inf_type==1 ? "F_Unify" : ( inf_type==2 ? "F_EndpointEmp" : "F_EndpointEq" ) ) );
1839 if( d_conflict ){
1840 return;
1841 }else{
1842 break;
1843 }
1844 }
1845 count++;
1846 }while( inelig.size()<it->second.size() );
1847
1848 for( unsigned i=0; i<it->second.size(); i++ ){
1849 std::reverse( d_flat_form[it->second[i]].begin(), d_flat_form[it->second[i]].end() );
1850 std::reverse( d_flat_form_index[it->second[i]].begin(), d_flat_form_index[it->second[i]].end() );
1851 }
1852 }
1853 }
1854 }
1855 }
1856 if( !hasProcessed() ){
1857 // simple extended func reduction
1858 Trace("strings-process") << "Check extended function reduction effort=1..." << std::endl;
1859 checkExtfReductions( 1 );
1860 Trace("strings-process") << "Done check extended function reduction" << std::endl;
1861 }
1862 }
1863 }
1864
1865 Node TheoryStrings::checkCycles( Node eqc, std::vector< Node >& curr, std::vector< Node >& exp ){
1866 if( std::find( curr.begin(), curr.end(), eqc )!=curr.end() ){
1867 // a loop
1868 return eqc;
1869 }else if( std::find( d_strings_eqc.begin(), d_strings_eqc.end(), eqc )==d_strings_eqc.end() ){
1870 curr.push_back( eqc );
1871 //look at all terms in this equivalence class
1872 eq::EqClassIterator eqc_i = eq::EqClassIterator( eqc, &d_equalityEngine );
1873 while( !eqc_i.isFinished() ) {
1874 Node n = (*eqc_i);
1875 if( d_congruent.find( n )==d_congruent.end() ){
1876 if( n.getKind() == kind::STRING_CONCAT ){
1877 Trace("strings-cycle") << eqc << " check term : " << n << " in " << eqc << std::endl;
1878 if( eqc!=d_emptyString_r ){
1879 d_eqc[eqc].push_back( n );
1880 }
1881 for( unsigned i=0; i<n.getNumChildren(); i++ ){
1882 Node nr = getRepresentative( n[i] );
1883 if( eqc==d_emptyString_r ){
1884 //for empty eqc, ensure all components are empty
1885 if( nr!=d_emptyString_r ){
1886 std::vector< Node > exp;
1887 exp.push_back( n.eqNode( d_emptyString ) );
1888 sendInference( exp, n[i].eqNode( d_emptyString ), "I_CYCLE_E" );
1889 return Node::null();
1890 }
1891 }else{
1892 if( nr!=d_emptyString_r ){
1893 d_flat_form[n].push_back( nr );
1894 d_flat_form_index[n].push_back( i );
1895 }
1896 //for non-empty eqc, recurse and see if we find a loop
1897 Node ncy = checkCycles( nr, curr, exp );
1898 if( !ncy.isNull() ){
1899 Trace("strings-cycle") << eqc << " cycle: " << ncy << " at " << n << "[" << i << "] : " << n[i] << std::endl;
1900 addToExplanation( n, eqc, exp );
1901 addToExplanation( nr, n[i], exp );
1902 if( ncy==eqc ){
1903 //can infer all other components must be empty
1904 for( unsigned j=0; j<n.getNumChildren(); j++ ){
1905 //take first non-empty
1906 if( j!=i && !areEqual( n[j], d_emptyString ) ){
1907 sendInference( exp, n[j].eqNode( d_emptyString ), "I_CYCLE" );
1908 return Node::null();
1909 }
1910 }
1911 Trace("strings-error") << "Looping term should be congruent : " << n << " " << eqc << " " << ncy << std::endl;
1912 //should find a non-empty component, otherwise would have been singular congruent (I_Norm_S)
1913 Assert( false );
1914 }else{
1915 return ncy;
1916 }
1917 }else{
1918 if( hasProcessed() ){
1919 return Node::null();
1920 }
1921 }
1922 }
1923 }
1924 }
1925 }
1926 ++eqc_i;
1927 }
1928 curr.pop_back();
1929 //now we can add it to the list of equivalence classes
1930 d_strings_eqc.push_back( eqc );
1931 }else{
1932 //already processed
1933 }
1934 return Node::null();
1935 }
1936
1937
1938 void TheoryStrings::checkNormalForms(){
1939 if( !options::stringEagerLen() ){
1940 for( unsigned i=0; i<d_strings_eqc.size(); i++ ) {
1941 Node eqc = d_strings_eqc[i];
1942 eq::EqClassIterator eqc_i = eq::EqClassIterator( eqc, &d_equalityEngine );
1943 while( !eqc_i.isFinished() ) {
1944 Node n = (*eqc_i);
1945 if( d_congruent.find( n )==d_congruent.end() ){
1946 registerTerm( n, 2 );
1947 }
1948 ++eqc_i;
1949 }
1950 }
1951 }
1952 if( !hasProcessed() ){
1953 Trace("strings-process") << "Normalize equivalence classes...." << std::endl;
1954 //calculate normal forms for each equivalence class, possibly adding splitting lemmas
1955 d_normal_forms.clear();
1956 d_normal_forms_exp.clear();
1957 std::map< Node, Node > nf_to_eqc;
1958 std::map< Node, Node > eqc_to_nf;
1959 std::map< Node, Node > eqc_to_exp;
1960 for( unsigned i=0; i<d_strings_eqc.size(); i++ ) {
1961 Node eqc = d_strings_eqc[i];
1962 Trace("strings-process-debug") << "- Verify normal forms are the same for " << eqc << std::endl;
1963 normalizeEquivalenceClass( eqc );
1964 Trace("strings-debug") << "Finished normalizing eqc..." << std::endl;
1965 if( hasProcessed() ){
1966 return;
1967 }else{
1968 Node nf_term = mkConcat( d_normal_forms[eqc] );
1969 std::map< Node, Node >::iterator itn = nf_to_eqc.find( nf_term );
1970 if( itn!=nf_to_eqc.end() ){
1971 //two equivalence classes have same normal form, merge
1972 std::vector< Node > nf_exp;
1973 nf_exp.push_back( mkAnd( d_normal_forms_exp[eqc] ) );
1974 nf_exp.push_back( eqc_to_exp[itn->second] );
1975 Node eq = d_normal_forms_base[eqc].eqNode( d_normal_forms_base[itn->second] );
1976 sendInference( nf_exp, eq, "Normal_Form" );
1977 } else {
1978 nf_to_eqc[nf_term] = eqc;
1979 eqc_to_nf[eqc] = nf_term;
1980 eqc_to_exp[eqc] = mkAnd( d_normal_forms_exp[eqc] );
1981 }
1982 }
1983 Trace("strings-process-debug") << "Done verifying normal forms are the same for " << eqc << std::endl;
1984 }
1985 if( !hasProcessed() ){
1986 if(Trace.isOn("strings-nf")) {
1987 Trace("strings-nf") << "**** Normal forms are : " << std::endl;
1988 for( std::map< Node, Node >::iterator it = eqc_to_exp.begin(); it != eqc_to_exp.end(); ++it ){
1989 Trace("strings-nf") << " N[" << it->first << "] (base " << d_normal_forms_base[it->first] << ") = " << eqc_to_nf[it->first] << std::endl;
1990 Trace("strings-nf") << " exp: " << it->second << std::endl;
1991 }
1992 Trace("strings-nf") << std::endl;
1993 }
1994 checkExtfEval( 1 );
1995 Trace("strings-process-debug") << "Done check extended functions re-eval, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
1996 if( !hasProcessed() ){
1997 if( !options::stringEagerLen() ){
1998 checkLengthsEqc();
1999 if( hasProcessed() ){
2000 return;
2001 }
2002 }
2003 //process disequalities between equivalence classes
2004 checkDeqNF();
2005 Trace("strings-process-debug") << "Done check disequalities, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
2006 }
2007 }
2008 Trace("strings-solve") << "Finished check normal forms, #lemmas = " << d_lemma_cache.size() << ", conflict = " << d_conflict << std::endl;
2009 }
2010 }
2011
2012 //compute d_normal_forms_(base,exp,exp_depend)[eqc]
2013 void TheoryStrings::normalizeEquivalenceClass( Node eqc ) {
2014 Trace("strings-process-debug") << "Process equivalence class " << eqc << std::endl;
2015 if( areEqual( eqc, d_emptyString ) ) {
2016 #ifdef CVC4_ASSERTIONS
2017 for( unsigned j=0; j<d_eqc[eqc].size(); j++ ){
2018 Node n = d_eqc[eqc][j];
2019 for( unsigned i=0; i<n.getNumChildren(); i++ ){
2020 Assert( areEqual( n[i], d_emptyString ) );
2021 }
2022 }
2023 #endif
2024 //do nothing
2025 Trace("strings-process-debug") << "Return process equivalence class " << eqc << " : empty." << std::endl;
2026 d_normal_forms_base[eqc] = d_emptyString;
2027 d_normal_forms[eqc].clear();
2028 d_normal_forms_exp[eqc].clear();
2029 } else {
2030 Assert( d_normal_forms.find(eqc)==d_normal_forms.end() );
2031 //phi => t = s1 * ... * sn
2032 // normal form for each non-variable term in this eqc (s1...sn)
2033 std::vector< std::vector< Node > > normal_forms;
2034 // explanation for each normal form (phi)
2035 std::vector< std::vector< Node > > normal_forms_exp;
2036 // dependency information
2037 std::vector< std::map< Node, std::map< bool, int > > > normal_forms_exp_depend;
2038 // record terms for each normal form (t)
2039 std::vector< Node > normal_form_src;
2040 // get normal forms
2041 getNormalForms(eqc, normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend);
2042 if( hasProcessed() ){
2043 return;
2044 }
2045 // process the normal forms
2046 processNEqc( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend );
2047 if( hasProcessed() ){
2048 return;
2049 }
2050 //debugPrintNormalForms( "strings-solve", eqc, normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend );
2051
2052 //construct the normal form
2053 Assert( !normal_forms.empty() );
2054
2055 int nf_index = 0;
2056 std::vector< Node >::iterator itn = std::find( normal_form_src.begin(), normal_form_src.end(), eqc );
2057 if( itn!=normal_form_src.end() ){
2058 nf_index = itn - normal_form_src.begin();
2059 Trace("strings-solve-debug2") << "take normal form " << nf_index << std::endl;
2060 Assert( normal_form_src[nf_index]==eqc );
2061 }else{
2062 //just take the first normal form
2063 Trace("strings-solve-debug2") << "take the first normal form" << std::endl;
2064 }
2065 d_normal_forms[eqc].insert( d_normal_forms[eqc].end(), normal_forms[nf_index].begin(), normal_forms[nf_index].end() );
2066 d_normal_forms_exp[eqc].insert( d_normal_forms_exp[eqc].end(), normal_forms_exp[nf_index].begin(), normal_forms_exp[nf_index].end() );
2067 Trace("strings-solve-debug2") << "take normal form ... done" << std::endl;
2068 d_normal_forms_base[eqc] = normal_form_src[nf_index];
2069 //track dependencies
2070 for( unsigned i=0; i<normal_forms_exp[nf_index].size(); i++ ){
2071 Node exp = normal_forms_exp[nf_index][i];
2072 for( unsigned r=0; r<2; r++ ){
2073 d_normal_forms_exp_depend[eqc][exp][r==0] = normal_forms_exp_depend[nf_index][exp][r==0];
2074 }
2075 }
2076 Trace("strings-process-debug") << "Return process equivalence class " << eqc << " : returned, size = " << d_normal_forms[eqc].size() << std::endl;
2077 }
2078 }
2079
2080 void TheoryStrings::getNormalForms( Node &eqc, std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2081 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend ) {
2082 //constant for equivalence class
2083 Node eqc_non_c = eqc;
2084 Trace("strings-process-debug") << "Get normal forms " << eqc << std::endl;
2085 eq::EqClassIterator eqc_i = eq::EqClassIterator( eqc, &d_equalityEngine );
2086 while( !eqc_i.isFinished() ){
2087 Node n = (*eqc_i);
2088 if( d_congruent.find( n )==d_congruent.end() ){
2089 if( n.getKind() == kind::CONST_STRING || n.getKind() == kind::STRING_CONCAT ){
2090 Trace("strings-process-debug") << "Get Normal Form : Process term " << n << " in eqc " << eqc << std::endl;
2091 std::vector< Node > nf_n;
2092 std::vector< Node > nf_exp_n;
2093 std::map< Node, std::map< bool, int > > nf_exp_depend_n;
2094 if( n.getKind()==kind::CONST_STRING ){
2095 if( n!=d_emptyString ) {
2096 nf_n.push_back( n );
2097 }
2098 }else if( n.getKind()==kind::STRING_CONCAT ){
2099 for( unsigned i=0; i<n.getNumChildren(); i++ ) {
2100 Node nr = d_equalityEngine.getRepresentative( n[i] );
2101 Trace("strings-process-debug") << "Normalizing subterm " << n[i] << " = " << nr << std::endl;
2102 Assert( d_normal_forms.find( nr )!=d_normal_forms.end() );
2103 unsigned orig_size = nf_n.size();
2104 unsigned add_size = d_normal_forms[nr].size();
2105 //if not the empty string, add to current normal form
2106 if( !d_normal_forms[nr].empty() ){
2107 for( unsigned r=0; r<d_normal_forms[nr].size(); r++ ) {
2108 if( Trace.isOn("strings-error") ) {
2109 if( d_normal_forms[nr][r].getKind()==kind::STRING_CONCAT ){
2110 Trace("strings-error") << "Strings::Error: From eqc = " << eqc << ", " << n << " index " << i << ", bad normal form : ";
2111 for( unsigned rr=0; rr<d_normal_forms[nr].size(); rr++ ) {
2112 Trace("strings-error") << d_normal_forms[nr][rr] << " ";
2113 }
2114 Trace("strings-error") << std::endl;
2115 }
2116 }
2117 Assert( d_normal_forms[nr][r].getKind()!=kind::STRING_CONCAT );
2118 }
2119 nf_n.insert( nf_n.end(), d_normal_forms[nr].begin(), d_normal_forms[nr].end() );
2120 }
2121
2122 for( unsigned j=0; j<d_normal_forms_exp[nr].size(); j++ ){
2123 Node exp = d_normal_forms_exp[nr][j];
2124 nf_exp_n.push_back( exp );
2125 //track depends
2126 for( unsigned k=0; k<2; k++ ){
2127 int prev_dep = d_normal_forms_exp_depend[nr][exp][k==1];
2128 if( k==0 ){
2129 nf_exp_depend_n[exp][false] = orig_size + prev_dep;
2130 }else if( k==1 ){
2131 //store forward index (converted back to reverse index below)
2132 nf_exp_depend_n[exp][true] = orig_size + ( add_size - prev_dep );
2133 }
2134 }
2135 }
2136 if( d_normal_forms_base[nr]!=n[i] ){
2137 Assert( d_normal_forms_base.find( nr )!=d_normal_forms_base.end() );
2138 Node eq = n[i].eqNode( d_normal_forms_base[nr] );
2139 nf_exp_n.push_back( eq );
2140 //track depends
2141 nf_exp_depend_n[eq][false] = orig_size;
2142 nf_exp_depend_n[eq][true] = orig_size + add_size;
2143 }
2144 }
2145 //convert forward indices to reverse indices
2146 int total_size = nf_n.size();
2147 for( std::map< Node, std::map< bool, int > >::iterator it = nf_exp_depend_n.begin(); it != nf_exp_depend_n.end(); ++it ){
2148 it->second[true] = total_size - it->second[true];
2149 Assert( it->second[true]>=0 );
2150 }
2151 }
2152 //if not equal to self
2153 if( nf_n.size()>1 || ( nf_n.size()==1 && nf_n[0].getKind()==kind::CONST_STRING ) ){
2154 if( nf_n.size()>1 ) {
2155 for( unsigned i=0; i<nf_n.size(); i++ ){
2156 if( Trace.isOn("strings-error") ){
2157 Trace("strings-error") << "Cycle for normal form ";
2158 printConcat(nf_n,"strings-error");
2159 Trace("strings-error") << "..." << nf_n[i] << std::endl;
2160 }
2161 Assert( !areEqual( nf_n[i], n ) );
2162 }
2163 }
2164 normal_forms.push_back(nf_n);
2165 normal_form_src.push_back(n);
2166 normal_forms_exp.push_back(nf_exp_n);
2167 normal_forms_exp_depend.push_back(nf_exp_depend_n);
2168 }else{
2169 //this was redundant: combination of self + empty string(s)
2170 Node nn = nf_n.size()==0 ? d_emptyString : nf_n[0];
2171 Assert( areEqual( nn, eqc ) );
2172 }
2173 }else{
2174 eqc_non_c = n;
2175 }
2176 }
2177 ++eqc_i;
2178 }
2179
2180 if( normal_forms.empty() ) {
2181 Trace("strings-solve-debug2") << "construct the normal form" << std::endl;
2182 //do not choose a concat here use "eqc_non_c" (in this case they have non-trivial explanation why they normalize to self)
2183 std::vector< Node > eqc_non_c_nf;
2184 getConcatVec( eqc_non_c, eqc_non_c_nf );
2185 normal_forms.push_back( eqc_non_c_nf );
2186 normal_form_src.push_back( eqc_non_c );
2187 normal_forms_exp.push_back( std::vector< Node >() );
2188 normal_forms_exp_depend.push_back( std::map< Node, std::map< bool, int > >() );
2189 }else{
2190 if(Trace.isOn("strings-solve")) {
2191 Trace("strings-solve") << "--- Normal forms for equivlance class " << eqc << " : " << std::endl;
2192 for( unsigned i=0; i<normal_forms.size(); i++ ) {
2193 Trace("strings-solve") << "#" << i << " (from " << normal_form_src[i] << ") : ";
2194 for( unsigned j=0; j<normal_forms[i].size(); j++ ) {
2195 if(j>0) {
2196 Trace("strings-solve") << ", ";
2197 }
2198 Trace("strings-solve") << normal_forms[i][j];
2199 }
2200 Trace("strings-solve") << std::endl;
2201 Trace("strings-solve") << " Explanation is : ";
2202 if(normal_forms_exp[i].size() == 0) {
2203 Trace("strings-solve") << "NONE";
2204 } else {
2205 for( unsigned j=0; j<normal_forms_exp[i].size(); j++ ) {
2206 if(j>0) {
2207 Trace("strings-solve") << " AND ";
2208 }
2209 Trace("strings-solve") << normal_forms_exp[i][j];
2210 }
2211 Trace("strings-solve") << std::endl;
2212 Trace("strings-solve") << "WITH DEPENDENCIES : " << std::endl;
2213 for( unsigned j=0; j<normal_forms_exp[i].size(); j++ ) {
2214 Trace("strings-solve") << " " << normal_forms_exp[i][j] << " -> ";
2215 Trace("strings-solve") << normal_forms_exp_depend[i][normal_forms_exp[i][j]][false] << ",";
2216 Trace("strings-solve") << normal_forms_exp_depend[i][normal_forms_exp[i][j]][true] << std::endl;
2217 }
2218 }
2219 Trace("strings-solve") << std::endl;
2220
2221 }
2222 } else {
2223 Trace("strings-solve") << "--- Single normal form for equivalence class " << eqc << std::endl;
2224 }
2225
2226 //if equivalence class is constant, approximate as containment, infer conflicts
2227 Node c = getConstantEqc( eqc );
2228 if( !c.isNull() ){
2229 Trace("strings-solve") << "Eqc is constant " << c << std::endl;
2230 for( unsigned i=0; i<normal_forms.size(); i++ ) {
2231 int firstc, lastc;
2232 if( !TheoryStringsRewriter::canConstantContainList( c, normal_forms[i], firstc, lastc ) ){
2233 Node n = normal_form_src[i];
2234 //conflict
2235 Trace("strings-solve") << "Normal form for " << n << " cannot be contained in constant " << c << std::endl;
2236 //conflict, explanation is n = base ^ base = c ^ relevant porition of ( n = N[n] )
2237 std::vector< Node > exp;
2238 Assert( d_eqc_to_const_base.find( eqc )!=d_eqc_to_const_base.end() );
2239 addToExplanation( n, d_eqc_to_const_base[eqc], exp );
2240 Assert( d_eqc_to_const_exp.find( eqc )!=d_eqc_to_const_exp.end() );
2241 if( !d_eqc_to_const_exp[eqc].isNull() ){
2242 exp.push_back( d_eqc_to_const_exp[eqc] );
2243 }
2244 //TODO: this can be minimized based on firstc/lastc, normal_forms_exp_depend
2245 exp.insert( exp.end(), normal_forms_exp[i].begin(), normal_forms_exp[i].end() );
2246 Node conc = d_false;
2247 sendInference( exp, conc, "N_NCTN" );
2248 }
2249 }
2250 }
2251 }
2252 }
2253
2254 void TheoryStrings::getExplanationVectorForPrefix( std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend,
2255 unsigned i, int index, bool isRev, std::vector< Node >& curr_exp ) {
2256 if( index==-1 || !options::stringMinPrefixExplain() ){
2257 curr_exp.insert(curr_exp.end(), normal_forms_exp[i].begin(), normal_forms_exp[i].end() );
2258 }else{
2259 for( unsigned k=0; k<normal_forms_exp[i].size(); k++ ){
2260 Node exp = normal_forms_exp[i][k];
2261 int dep = normal_forms_exp_depend[i][exp][isRev];
2262 if( dep<=index ){
2263 curr_exp.push_back( exp );
2264 Trace("strings-explain-prefix-debug") << " include : " << exp << std::endl;
2265 }else{
2266 Trace("strings-explain-prefix-debug") << " exclude : " << exp << std::endl;
2267 }
2268 }
2269 }
2270 }
2271
2272 void TheoryStrings::getExplanationVectorForPrefixEq( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2273 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend,
2274 unsigned i, unsigned j, int index_i, int index_j, bool isRev, std::vector< Node >& curr_exp ) {
2275 Trace("strings-explain-prefix") << "Get explanation for prefix " << index_i << ", " << index_j << " of normal forms " << i << " and " << j << ", reverse = " << isRev << std::endl;
2276 for( unsigned r=0; r<2; r++ ){
2277 getExplanationVectorForPrefix( normal_forms_exp, normal_forms_exp_depend, r==0 ? i : j, r==0 ? index_i : index_j, isRev, curr_exp );
2278 }
2279 Trace("strings-explain-prefix") << "Included " << curr_exp.size() << " / " << ( normal_forms_exp[i].size() + normal_forms_exp[j].size() ) << std::endl;
2280 addToExplanation( normal_form_src[i], normal_form_src[j], curr_exp );
2281 }
2282
2283
2284 void TheoryStrings::processNEqc( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2285 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend ){
2286 //the possible inferences
2287 std::vector< InferInfo > pinfer;
2288 // loop over all pairs
2289 for(unsigned i=0; i<normal_forms.size()-1; i++) {
2290 //unify each normalform[j] with normal_forms[i]
2291 for(unsigned j=i+1; j<normal_forms.size(); j++ ) {
2292 //ensure that normal_forms[i] and normal_forms[j] are the same modulo equality, add to pinfer if not
2293 Trace("strings-solve") << "Strings: Process normal form #" << i << " against #" << j << "..." << std::endl;
2294 if( isNormalFormPair( normal_form_src[i], normal_form_src[j] ) ) {
2295 Trace("strings-solve") << "Strings: Already cached." << std::endl;
2296 }else{
2297 //process the reverse direction first (check for easy conflicts and inferences)
2298 unsigned rindex = 0;
2299 processReverseNEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, rindex, 0, pinfer );
2300 if( hasProcessed() ){
2301 return;
2302 }else if( !pinfer.empty() && pinfer.back().d_id==1 ){
2303 break;
2304 }
2305 //AJR: for less aggressive endpoint inference
2306 //rindex = 0;
2307
2308 unsigned index = 0;
2309 processSimpleNEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, false, rindex, pinfer );
2310 if( hasProcessed() ){
2311 return;
2312 }else if( !pinfer.empty() && pinfer.back().d_id==1 ){
2313 break;
2314 }
2315 }
2316 }
2317 }
2318 if( !pinfer.empty() ){
2319 //now, determine which of the possible inferences we want to add
2320 int use_index = -1;
2321 Trace("strings-solve") << "Possible inferences (" << pinfer.size() << ") : " << std::endl;
2322 unsigned min_id = 9;
2323 unsigned max_index = 0;
2324 for( unsigned i=0; i<pinfer.size(); i++ ){
2325 Trace("strings-solve") << "From " << pinfer[i].d_i << " / " << pinfer[i].d_j << " (rev=" << pinfer[i].d_rev << ") : ";
2326 Trace("strings-solve") << pinfer[i].d_conc << " by " << pinfer[i].getId() << std::endl;
2327 if( use_index==-1 || pinfer[i].d_id<min_id || ( pinfer[i].d_id==min_id && pinfer[i].d_index>max_index ) ){
2328 min_id = pinfer[i].d_id;
2329 max_index = pinfer[i].d_index;
2330 use_index = i;
2331 }
2332 }
2333 //send the inference
2334 sendInference( pinfer[use_index].d_ant, pinfer[use_index].d_antn, pinfer[use_index].d_conc, pinfer[use_index].getId(), pinfer[use_index].sendAsLemma() );
2335 for( std::map< int, std::vector< Node > >::iterator it = pinfer[use_index].d_new_skolem.begin(); it != pinfer[use_index].d_new_skolem.end(); ++it ){
2336 for( unsigned i=0; i<it->second.size(); i++ ){
2337 if( it->first==0 ){
2338 sendLengthLemma( it->second[i] );
2339 }else if( it->first==1 ){
2340 registerNonEmptySkolem( it->second[i] );
2341 }
2342 }
2343 }
2344 }
2345 }
2346
2347 bool TheoryStrings::InferInfo::sendAsLemma() {
2348 return true;
2349 }
2350
2351 void TheoryStrings::processReverseNEq( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2352 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend,
2353 unsigned i, unsigned j, unsigned& index, unsigned rproc, std::vector< InferInfo >& pinfer ) {
2354 //reverse normal form of i, j
2355 std::reverse( normal_forms[i].begin(), normal_forms[i].end() );
2356 std::reverse( normal_forms[j].begin(), normal_forms[j].end() );
2357
2358 processSimpleNEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, true, rproc, pinfer );
2359
2360 //reverse normal form of i, j
2361 std::reverse( normal_forms[i].begin(), normal_forms[i].end() );
2362 std::reverse( normal_forms[j].begin(), normal_forms[j].end() );
2363 }
2364
2365 //rproc is the # is the size of suffix that is identical
2366 void TheoryStrings::processSimpleNEq( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2367 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend,
2368 unsigned i, unsigned j, unsigned& index, bool isRev, unsigned rproc, std::vector< InferInfo >& pinfer ) {
2369 Assert( rproc<=normal_forms[i].size() && rproc<=normal_forms[j].size() );
2370 bool success;
2371 do {
2372 success = false;
2373 //if we are at the end
2374 if( index==(normal_forms[i].size()-rproc) || index==(normal_forms[j].size()-rproc) ){
2375 if( index==(normal_forms[i].size()-rproc) && index==(normal_forms[j].size()-rproc) ){
2376 //we're done
2377 }else{
2378 //the remainder must be empty
2379 unsigned k = index==(normal_forms[i].size()-rproc) ? j : i;
2380 unsigned index_k = index;
2381 //Node eq_exp = mkAnd( curr_exp );
2382 std::vector< Node > curr_exp;
2383 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, -1, -1, isRev, curr_exp );
2384 while( !d_conflict && index_k<(normal_forms[k].size()-rproc) ){
2385 //can infer that this string must be empty
2386 Node eq = normal_forms[k][index_k].eqNode( d_emptyString );
2387 //Trace("strings-lemma") << "Strings: Infer " << eq << " from " << eq_exp << std::endl;
2388 Assert( !areEqual( d_emptyString, normal_forms[k][index_k] ) );
2389 sendInference( curr_exp, eq, "N_EndpointEmp" );
2390 index_k++;
2391 }
2392 }
2393 }else{
2394 Trace("strings-solve-debug") << "Process " << normal_forms[i][index] << " ... " << normal_forms[j][index] << std::endl;
2395 if( normal_forms[i][index]==normal_forms[j][index] ){
2396 Trace("strings-solve-debug") << "Simple Case 1 : strings are equal" << std::endl;
2397 index++;
2398 success = true;
2399 }else{
2400 Assert( !areEqual(normal_forms[i][index], normal_forms[j][index]) );
2401 std::vector< Node > temp_exp;
2402 Node length_term_i = getLength( normal_forms[i][index], temp_exp );
2403 Node length_term_j = getLength( normal_forms[j][index], temp_exp );
2404 //check length(normal_forms[i][index]) == length(normal_forms[j][index])
2405 if( areEqual( length_term_i, length_term_j ) ){
2406 Trace("strings-solve-debug") << "Simple Case 2 : string lengths are equal" << std::endl;
2407 Node eq = normal_forms[i][index].eqNode( normal_forms[j][index] );
2408 //eq = Rewriter::rewrite( eq );
2409 Node length_eq = length_term_i.eqNode( length_term_j );
2410 //temp_exp.insert(temp_exp.end(), curr_exp.begin(), curr_exp.end() );
2411 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, index, isRev, temp_exp );
2412 temp_exp.push_back(length_eq);
2413 sendInference( temp_exp, eq, "N_Unify" );
2414 return;
2415 }else if( ( normal_forms[i][index].getKind()!=kind::CONST_STRING && index==normal_forms[i].size()-rproc-1 ) ||
2416 ( normal_forms[j][index].getKind()!=kind::CONST_STRING && index==normal_forms[j].size()-rproc-1 ) ){
2417 Trace("strings-solve-debug") << "Simple Case 3 : at endpoint" << std::endl;
2418 std::vector< Node > antec;
2419 //antec.insert(antec.end(), curr_exp.begin(), curr_exp.end() );
2420 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, -1, -1, isRev, antec );
2421 std::vector< Node > eqn;
2422 for( unsigned r=0; r<2; r++ ) {
2423 int index_k = index;
2424 int k = r==0 ? i : j;
2425 std::vector< Node > eqnc;
2426 for( unsigned index_l=index_k; index_l<(normal_forms[k].size()-rproc); index_l++ ) {
2427 if(isRev) {
2428 eqnc.insert(eqnc.begin(), normal_forms[k][index_l] );
2429 } else {
2430 eqnc.push_back( normal_forms[k][index_l] );
2431 }
2432 }
2433 eqn.push_back( mkConcat( eqnc ) );
2434 }
2435 if( !areEqual( eqn[0], eqn[1] ) ){
2436 sendInference( antec, eqn[0].eqNode( eqn[1] ), "N_EndpointEq", true );
2437 return;
2438 }else{
2439 Assert( normal_forms[i].size()==normal_forms[j].size() );
2440 index = normal_forms[i].size()-rproc;
2441 }
2442 }else if( normal_forms[i][index].isConst() && normal_forms[j][index].isConst() ){
2443 Node const_str = normal_forms[i][index];
2444 Node other_str = normal_forms[j][index];
2445 Trace("strings-solve-debug") << "Simple Case 3 : Const Split : " << const_str << " vs " << other_str << " at index " << index << ", isRev = " << isRev << std::endl;
2446 unsigned len_short = const_str.getConst<String>().size() <= other_str.getConst<String>().size() ? const_str.getConst<String>().size() : other_str.getConst<String>().size();
2447 bool isSameFix = isRev ? const_str.getConst<String>().rstrncmp(other_str.getConst<String>(), len_short): const_str.getConst<String>().strncmp(other_str.getConst<String>(), len_short);
2448 if( isSameFix ) {
2449 //same prefix/suffix
2450 //k is the index of the string that is shorter
2451 int k = const_str.getConst<String>().size()<other_str.getConst<String>().size() ? i : j;
2452 int l = const_str.getConst<String>().size()<other_str.getConst<String>().size() ? j : i;
2453 //update the nf exp dependencies
2454 //notice this is not critical for soundness: not doing the below incrementing will only lead to overapproximating when antecedants are required in explanations
2455 for( std::map< Node, std::map< bool, int > >::iterator itnd = normal_forms_exp_depend[l].begin(); itnd != normal_forms_exp_depend[l].end(); ++itnd ){
2456 for( std::map< bool, int >::iterator itnd2 = itnd->second.begin(); itnd2 != itnd->second.end(); ++itnd2 ){
2457 //see if this can be incremented: it can if it is not relevant to the current index
2458 Assert( itnd2->second>=0 && itnd2->second<=(int)normal_forms[l].size() );
2459 bool increment = (itnd2->first==isRev) ? itnd2->second>(int)index : ( (int)normal_forms[l].size()-1-itnd2->second )<(int)index;
2460 if( increment ){
2461 normal_forms_exp_depend[l][itnd->first][itnd2->first] = itnd2->second + 1;
2462 }
2463 }
2464 }
2465 if( isRev ){
2466 int new_len = normal_forms[l][index].getConst<String>().size() - len_short;
2467 Node remainderStr = NodeManager::currentNM()->mkConst( normal_forms[l][index].getConst<String>().substr(0, new_len) );
2468 Trace("strings-solve-debug-test") << "Break normal form of " << normal_forms[l][index] << " into " << normal_forms[k][index] << ", " << remainderStr << std::endl;
2469 normal_forms[l].insert( normal_forms[l].begin()+index + 1, remainderStr );
2470 }else{
2471 Node remainderStr = NodeManager::currentNM()->mkConst(normal_forms[l][index].getConst<String>().substr(len_short));
2472 Trace("strings-solve-debug-test") << "Break normal form of " << normal_forms[l][index] << " into " << normal_forms[k][index] << ", " << remainderStr << std::endl;
2473 normal_forms[l].insert( normal_forms[l].begin()+index + 1, remainderStr );
2474 }
2475 normal_forms[l][index] = normal_forms[k][index];
2476 index++;
2477 success = true;
2478 }else{
2479 //conflict
2480 std::vector< Node > antec;
2481 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, index, isRev, antec );
2482 sendInference( antec, d_false, "N_Const", true );
2483 return;
2484 }
2485 }else{
2486 //construct the candidate inference "info"
2487 InferInfo info;
2488 info.d_index = index;
2489 //for debugging
2490 info.d_i = i;
2491 info.d_j = j;
2492 info.d_rev = isRev;
2493 bool info_valid = false;
2494 Assert( index<normal_forms[i].size()-rproc && index<normal_forms[j].size()-rproc );
2495 std::vector< Node > lexp;
2496 Node length_term_i = getLength( normal_forms[i][index], lexp );
2497 Node length_term_j = getLength( normal_forms[j][index], lexp );
2498 //split on equality between string lengths (note that splitting on equality between strings is worse since it is harder to process)
2499 if( !areDisequal( length_term_i, length_term_j ) && !areEqual( length_term_i, length_term_j ) &&
2500 normal_forms[i][index].getKind()!=kind::CONST_STRING && normal_forms[j][index].getKind()!=kind::CONST_STRING ){ //AJR: remove the latter 2 conditions?
2501 Trace("strings-solve-debug") << "Non-simple Case 1 : string lengths neither equal nor disequal" << std::endl;
2502 //try to make the lengths equal via splitting on demand
2503 Node length_eq = NodeManager::currentNM()->mkNode( kind::EQUAL, length_term_i, length_term_j );
2504 length_eq = Rewriter::rewrite( length_eq );
2505 //set info
2506 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, length_eq, length_eq.negate() );
2507 info.d_pending_phase[ length_eq ] = true;
2508 info.d_id = 3;
2509 info_valid = true;
2510 }else{
2511 Trace("strings-solve-debug") << "Non-simple Case 2 : must compare strings" << std::endl;
2512 int loop_in_i = -1;
2513 int loop_in_j = -1;
2514 if( detectLoop( normal_forms, i, j, index, loop_in_i, loop_in_j, rproc ) ){
2515 if( !isRev ){ //FIXME
2516 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, -1, -1, isRev, info.d_ant );
2517 //set info
2518 if( processLoop( normal_forms, normal_form_src, i, j, loop_in_i!=-1 ? i : j, loop_in_i!=-1 ? j : i, loop_in_i!=-1 ? loop_in_i : loop_in_j, index, info ) ){
2519 info_valid = true;
2520 }
2521 }
2522 }else{
2523 //AJR: length entailment here?
2524 if( normal_forms[i][index].getKind() == kind::CONST_STRING || normal_forms[j][index].getKind() == kind::CONST_STRING ){
2525 unsigned const_k = normal_forms[i][index].getKind() == kind::CONST_STRING ? i : j;
2526 unsigned nconst_k = normal_forms[i][index].getKind() == kind::CONST_STRING ? j : i;
2527 Node other_str = normal_forms[nconst_k][index];
2528 Assert( other_str.getKind()!=kind::CONST_STRING, "Other string is not constant." );
2529 Assert( other_str.getKind()!=kind::STRING_CONCAT, "Other string is not CONCAT." );
2530 if( !d_equalityEngine.areDisequal( other_str, d_emptyString, true ) ){
2531 Node eq = other_str.eqNode( d_emptyString );
2532 //set info
2533 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, eq, eq.negate() );
2534 info.d_id = 4;
2535 info_valid = true;
2536 }else{
2537 if( !isRev ){ //FIXME
2538 Node xnz = other_str.eqNode( d_emptyString ).negate();
2539 unsigned index_nc_k = index+1;
2540 //Node next_const_str = TheoryStringsRewriter::collectConstantStringAt( normal_forms[nconst_k], index_nc_k, false );
2541 unsigned start_index_nc_k = index+1;
2542 Node next_const_str = TheoryStringsRewriter::getNextConstantAt( normal_forms[nconst_k], start_index_nc_k, index_nc_k, false );
2543 if( !next_const_str.isNull() ) {
2544 unsigned index_c_k = index;
2545 Node const_str = TheoryStringsRewriter::collectConstantStringAt( normal_forms[const_k], index_c_k, false );
2546 Assert( !const_str.isNull() );
2547 CVC4::String stra = const_str.getConst<String>();
2548 CVC4::String strb = next_const_str.getConst<String>();
2549 //since non-empty, we start with charecter #1
2550 size_t p;
2551 if( isRev ){
2552 CVC4::String stra1 = stra.prefix( stra.size()-1 );
2553 p = stra.size() - stra1.roverlap(strb);
2554 Trace("strings-csp-debug") << "Compute roverlap : " << const_str << " " << next_const_str << std::endl;
2555 size_t p2 = stra1.rfind(strb);
2556 p = p2==std::string::npos ? p : ( p>p2+1? p2+1 : p );
2557 Trace("strings-csp-debug") << "overlap : " << stra1 << " " << strb << " returned " << p << " " << p2 << " " << (p2==std::string::npos) << std::endl;
2558 }else{
2559 CVC4::String stra1 = stra.substr( 1 );
2560 p = stra.size() - stra1.overlap(strb);
2561 Trace("strings-csp-debug") << "Compute overlap : " << const_str << " " << next_const_str << std::endl;
2562 size_t p2 = stra1.find(strb);
2563 p = p2==std::string::npos ? p : ( p>p2+1? p2+1 : p );
2564 Trace("strings-csp-debug") << "overlap : " << stra1 << " " << strb << " returned " << p << " " << p2 << " " << (p2==std::string::npos) << std::endl;
2565 }
2566 if( p>1 ){
2567 if( start_index_nc_k==index+1 ){
2568 info.d_ant.push_back( xnz );
2569 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend,
2570 const_k, nconst_k, index_c_k, index_nc_k, isRev, info.d_ant );
2571 Node prea = p==stra.size() ? const_str : NodeManager::currentNM()->mkConst( isRev ? stra.suffix( p ) : stra.prefix( p ) );
2572 Node sk = mkSkolemCached( other_str, prea, isRev ? sk_id_c_spt_rev : sk_id_c_spt, "c_spt", -1 );
2573 Trace("strings-csp") << "Const Split: " << prea << " is removed from " << stra << " due to " << strb << ", p=" << p << std::endl;
2574 //set info
2575 info.d_conc = other_str.eqNode( isRev ? mkConcat( sk, prea ) : mkConcat(prea, sk) );
2576 info.d_new_skolem[0].push_back( sk );
2577 info.d_id = 1;
2578 info_valid = true;
2579 }
2580 /* FIXME for isRev, speculative
2581 else if( options::stringLenPropCsp() ){
2582 //propagate length constraint
2583 std::vector< Node > cc;
2584 for( unsigned i=index; i<start_index_nc_k; i++ ){
2585 cc.push_back( normal_forms[nconst_k][i] );
2586 }
2587 Node lt = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, mkConcat( cc ) );
2588 conc = NodeManager::currentNM()->mkNode( kind::GEQ, lt, NodeManager::currentNM()->mkConst( Rational(p) ) );
2589 sendInference( ant, conc, "S-Split(CSP-P)-lprop", true );
2590 }
2591 */
2592 }
2593 }
2594 if( !info_valid ){
2595 info.d_ant.push_back( xnz );
2596 Node const_str = normal_forms[const_k][index];
2597 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, index, isRev, info.d_ant );
2598 CVC4::String stra = const_str.getConst<String>();
2599 if( options::stringBinaryCsp() && stra.size()>3 ){
2600 //split string in half
2601 Node c_firstHalf = NodeManager::currentNM()->mkConst( isRev ? stra.substr( stra.size()/2 ) : stra.substr(0, stra.size()/2 ) );
2602 Node sk = mkSkolemCached( other_str, c_firstHalf , isRev ? sk_id_vc_bin_spt_rev : sk_id_vc_bin_spt, "cb_spt", -1 );
2603 Trace("strings-csp") << "Const Split: " << c_firstHalf << " is removed from " << const_str << " (binary) " << std::endl;
2604 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, other_str.eqNode( isRev ? mkConcat( sk, c_firstHalf ) : mkConcat( c_firstHalf, sk ) ),
2605 NodeManager::currentNM()->mkNode( kind::AND,
2606 sk.eqNode( d_emptyString ).negate(),
2607 c_firstHalf.eqNode( isRev ? mkConcat( sk, other_str ) : mkConcat( other_str, sk ) ) ) );
2608 info.d_new_skolem[0].push_back( sk );
2609 info.d_id = 5;
2610 info_valid = true;
2611 }else{
2612 // normal v/c split
2613 Node firstChar = stra.size() == 1 ? const_str : NodeManager::currentNM()->mkConst( isRev ? stra.suffix( 1 ) : stra.prefix( 1 ) );
2614 Node sk = mkSkolemCached( other_str, firstChar, isRev ? sk_id_vc_spt_rev : sk_id_vc_spt, "c_spt", -1 );
2615 Trace("strings-csp") << "Const Split: " << firstChar << " is removed from " << const_str << " (serial) " << std::endl;
2616 info.d_conc = other_str.eqNode( isRev ? mkConcat( sk, firstChar ) : mkConcat(firstChar, sk) );
2617 info.d_new_skolem[0].push_back( sk );
2618 info.d_id = 6;
2619 info_valid = true;
2620 }
2621 }
2622 }
2623 }
2624 }else{
2625 int lentTestSuccess = -1;
2626 Node lentTestExp;
2627 if( options::stringCheckEntailLen() ){
2628 //check entailment
2629 for( unsigned e=0; e<2; e++ ){
2630 Node t = e==0 ? normal_forms[i][index] : normal_forms[j][index];
2631 //do not infer constants are larger than variables
2632 if( t.getKind()!=kind::CONST_STRING ){
2633 Node lt1 = e==0 ? length_term_i : length_term_j;
2634 Node lt2 = e==0 ? length_term_j : length_term_i;
2635 Node ent_lit = Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::GT, lt1, lt2 ) );
2636 std::pair<bool, Node> et = d_valuation.entailmentCheck( THEORY_OF_TYPE_BASED, ent_lit );
2637 if( et.first ){
2638 Trace("strings-entail") << "Strings entailment : " << ent_lit << " is entailed in the current context." << std::endl;
2639 Trace("strings-entail") << " explanation was : " << et.second << std::endl;
2640 lentTestSuccess = e;
2641 lentTestExp = et.second;
2642 break;
2643 }
2644 }
2645 }
2646 }
2647
2648 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, index, isRev, info.d_ant );
2649 //x!=e /\ y!=e
2650 for(unsigned xory=0; xory<2; xory++) {
2651 Node x = xory==0 ? normal_forms[i][index] : normal_forms[j][index];
2652 Node xgtz = x.eqNode( d_emptyString ).negate();
2653 if( d_equalityEngine.areDisequal( x, d_emptyString, true ) ) {
2654 info.d_ant.push_back( xgtz );
2655 } else {
2656 info.d_antn.push_back( xgtz );
2657 }
2658 }
2659 Node sk = mkSkolemCached( normal_forms[i][index], normal_forms[j][index], isRev ? sk_id_v_spt_rev : sk_id_v_spt, "v_spt", -1 );
2660 //must add length requirement
2661 info.d_new_skolem[1].push_back( sk );
2662 Node eq1 = normal_forms[i][index].eqNode( isRev ? mkConcat(sk, normal_forms[j][index]) : mkConcat(normal_forms[j][index], sk) );
2663 Node eq2 = normal_forms[j][index].eqNode( isRev ? mkConcat(sk, normal_forms[i][index]) : mkConcat(normal_forms[i][index], sk) );
2664
2665 if( lentTestSuccess!=-1 ){
2666 info.d_antn.push_back( lentTestExp );
2667 info.d_conc = lentTestSuccess==0 ? eq1 : eq2;
2668 info.d_id = 2;
2669 info_valid = true;
2670 }else{
2671 Node ldeq = NodeManager::currentNM()->mkNode( kind::EQUAL, length_term_i, length_term_j ).negate();
2672 if( d_equalityEngine.areDisequal( length_term_i, length_term_j, true ) ){
2673 info.d_ant.push_back( ldeq );
2674 }else{
2675 info.d_antn.push_back(ldeq);
2676 }
2677 //set info
2678 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, eq1, eq2 );
2679 info.d_id = 7;
2680 info_valid = true;
2681 }
2682 }
2683 }
2684 }
2685 if( info_valid ){
2686 pinfer.push_back( info );
2687 Assert( !success );
2688 }
2689 }
2690 }
2691 }
2692 }while( success );
2693 }
2694
2695 bool TheoryStrings::detectLoop( std::vector< std::vector< Node > > &normal_forms, int i, int j, int index, int &loop_in_i, int &loop_in_j, unsigned rproc ){
2696 int has_loop[2] = { -1, -1 };
2697 if( options::stringLB() != 2 ) {
2698 for( unsigned r=0; r<2; r++ ) {
2699 int n_index = (r==0 ? i : j);
2700 int other_n_index = (r==0 ? j : i);
2701 if( normal_forms[other_n_index][index].getKind() != kind::CONST_STRING ) {
2702 for( unsigned lp = index+1; lp<normal_forms[n_index].size()-rproc; lp++ ){
2703 if( normal_forms[n_index][lp]==normal_forms[other_n_index][index] ){
2704 has_loop[r] = lp;
2705 break;
2706 }
2707 }
2708 }
2709 }
2710 }
2711 if( has_loop[0]!=-1 || has_loop[1]!=-1 ) {
2712 loop_in_i = has_loop[0];
2713 loop_in_j = has_loop[1];
2714 return true;
2715 } else {
2716 Trace("strings-solve-debug") << "No loops detected." << std::endl;
2717 return false;
2718 }
2719 }
2720
2721 //xs(zy)=t(yz)xr
2722 bool TheoryStrings::processLoop( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2723 int i, int j, int loop_n_index, int other_n_index, int loop_index, int index, InferInfo& info ){
2724 if( options::stringAbortLoop() ){
2725 Message() << "Looping word equation encountered." << std::endl;
2726 exit( 1 );
2727 }else{
2728 Node conc;
2729 Trace("strings-loop") << "Detected possible loop for " << normal_forms[loop_n_index][loop_index] << std::endl;
2730 Trace("strings-loop") << " ... (X)= " << normal_forms[other_n_index][index] << std::endl;
2731
2732 Trace("strings-loop") << " ... T(Y.Z)= ";
2733 std::vector< Node > vec_t;
2734 for(int lp=index; lp<loop_index; ++lp) {
2735 if(lp != index) Trace("strings-loop") << " ++ ";
2736 Trace("strings-loop") << normal_forms[loop_n_index][lp];
2737 vec_t.push_back( normal_forms[loop_n_index][lp] );
2738 }
2739 Node t_yz = mkConcat( vec_t );
2740 Trace("strings-loop") << " (" << t_yz << ")" << std::endl;
2741 Trace("strings-loop") << " ... S(Z.Y)= ";
2742 std::vector< Node > vec_s;
2743 for(int lp=index+1; lp<(int)normal_forms[other_n_index].size(); ++lp) {
2744 if(lp != index+1) Trace("strings-loop") << " ++ ";
2745 Trace("strings-loop") << normal_forms[other_n_index][lp];
2746 vec_s.push_back( normal_forms[other_n_index][lp] );
2747 }
2748 Node s_zy = mkConcat( vec_s );
2749 Trace("strings-loop") << " (" << s_zy << ")" << std::endl;
2750 Trace("strings-loop") << " ... R= ";
2751 std::vector< Node > vec_r;
2752 for(int lp=loop_index+1; lp<(int)normal_forms[loop_n_index].size(); ++lp) {
2753 if(lp != loop_index+1) Trace("strings-loop") << " ++ ";
2754 Trace("strings-loop") << normal_forms[loop_n_index][lp];
2755 vec_r.push_back( normal_forms[loop_n_index][lp] );
2756 }
2757 Node r = mkConcat( vec_r );
2758 Trace("strings-loop") << " (" << r << ")" << std::endl;
2759
2760 //Trace("strings-loop") << "Lemma Cache: " << normal_form_src[i] << " vs " << normal_form_src[j] << std::endl;
2761 //TODO: can be more general
2762 if( s_zy.isConst() && r.isConst() && r!=d_emptyString) {
2763 int c;
2764 bool flag = true;
2765 if(s_zy.getConst<String>().tailcmp( r.getConst<String>(), c ) ) {
2766 if( c>=0) {
2767 s_zy = NodeManager::currentNM()->mkConst( s_zy.getConst<String>().substr(0, c) );
2768 r = d_emptyString;
2769 vec_r.clear();
2770 Trace("strings-loop") << "Strings::Loop: Refactor S(Z.Y)= " << s_zy << ", c=" << c << std::endl;
2771 flag = false;
2772 }
2773 }
2774 if( flag ){
2775 Trace("strings-loop") << "Strings::Loop: tails are different." << std::endl;
2776 sendInference( info.d_ant, conc, "Loop Conflict", true );
2777 return false;
2778 }
2779 }
2780
2781 //require that x is non-empty
2782 Node split_eq;
2783 if( !areDisequal( normal_forms[loop_n_index][loop_index], d_emptyString ) ){
2784 //try to make normal_forms[loop_n_index][loop_index] equal to empty to avoid loop
2785 split_eq = normal_forms[loop_n_index][loop_index].eqNode( d_emptyString );
2786 }else if( !areDisequal( t_yz, d_emptyString ) && t_yz.getKind()!=kind::CONST_STRING ) {
2787 //try to make normal_forms[loop_n_index][loop_index] equal to empty to avoid loop
2788 split_eq = t_yz.eqNode( d_emptyString );
2789 }
2790 if( !split_eq.isNull() ){
2791 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, split_eq, split_eq.negate() );
2792 info.d_id = 4;
2793 return true;
2794 }else{
2795 //need to break
2796 info.d_ant.push_back( normal_forms[loop_n_index][loop_index].eqNode( d_emptyString ).negate() );
2797 if( t_yz.getKind()!=kind::CONST_STRING ) {
2798 info.d_ant.push_back( t_yz.eqNode( d_emptyString ).negate() );
2799 }
2800 Node ant = mkExplain( info.d_ant );
2801 if( d_loop_antec.find( ant ) == d_loop_antec.end() ){
2802 d_loop_antec.insert( ant );
2803 info.d_ant.clear();
2804 info.d_antn.push_back( ant );
2805
2806 Node str_in_re;
2807 if( s_zy == t_yz &&
2808 r == d_emptyString &&
2809 s_zy.isConst() &&
2810 s_zy.getConst<String>().isRepeated()
2811 ) {
2812 Node rep_c = NodeManager::currentNM()->mkConst( s_zy.getConst<String>().substr(0, 1) );
2813 Trace("strings-loop") << "Special case (X)=" << normal_forms[other_n_index][index] << " " << std::endl;
2814 Trace("strings-loop") << "... (C)=" << rep_c << " " << std::endl;
2815 //special case
2816 str_in_re = NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP, normal_forms[other_n_index][index],
2817 NodeManager::currentNM()->mkNode( kind::REGEXP_STAR,
2818 NodeManager::currentNM()->mkNode( kind::STRING_TO_REGEXP, rep_c ) ) );
2819 conc = str_in_re;
2820 } else if(t_yz.isConst()) {
2821 Trace("strings-loop") << "Strings::Loop: Const Normal Breaking." << std::endl;
2822 CVC4::String s = t_yz.getConst< CVC4::String >();
2823 unsigned size = s.size();
2824 std::vector< Node > vconc;
2825 for(unsigned len=1; len<=size; len++) {
2826 Node y = NodeManager::currentNM()->mkConst(s.substr(0, len));
2827 Node z = NodeManager::currentNM()->mkConst(s.substr(len, size - len));
2828 Node restr = s_zy;
2829 Node cc;
2830 if(r != d_emptyString) {
2831 std::vector< Node > v2(vec_r);
2832 v2.insert(v2.begin(), y);
2833 v2.insert(v2.begin(), z);
2834 restr = mkConcat( z, y );
2835 cc = Rewriter::rewrite(s_zy.eqNode( mkConcat( v2 ) ));
2836 } else {
2837 cc = Rewriter::rewrite(s_zy.eqNode( mkConcat( z, y) ));
2838 }
2839 if(cc == d_false) {
2840 continue;
2841 }
2842 Node conc2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, normal_forms[other_n_index][index],
2843 NodeManager::currentNM()->mkNode(kind::REGEXP_CONCAT,
2844 NodeManager::currentNM()->mkNode(kind::STRING_TO_REGEXP, y),
2845 NodeManager::currentNM()->mkNode(kind::REGEXP_STAR,
2846 NodeManager::currentNM()->mkNode(kind::STRING_TO_REGEXP, restr))));
2847 cc = cc==d_true ? conc2 : NodeManager::currentNM()->mkNode( kind::AND, cc, conc2 );
2848 d_regexp_ant[conc2] = ant;
2849 vconc.push_back(cc);
2850 }
2851 conc = vconc.size()==0 ? Node::null() : vconc.size()==1 ? vconc[0] : NodeManager::currentNM()->mkNode(kind::OR, vconc);
2852 } else {
2853 Trace("strings-loop") << "Strings::Loop: Normal Loop Breaking." << std::endl;
2854 //right
2855 Node sk_w= mkSkolemS( "w_loop" );
2856 Node sk_y= mkSkolemS( "y_loop", 1 );
2857 Node sk_z= mkSkolemS( "z_loop" );
2858 //t1 * ... * tn = y * z
2859 Node conc1 = t_yz.eqNode( mkConcat( sk_y, sk_z ) );
2860 // s1 * ... * sk = z * y * r
2861 vec_r.insert(vec_r.begin(), sk_y);
2862 vec_r.insert(vec_r.begin(), sk_z);
2863 Node conc2 = s_zy.eqNode( mkConcat( vec_r ) );
2864 Node conc3 = normal_forms[other_n_index][index].eqNode( mkConcat( sk_y, sk_w ) );
2865 Node restr = r == d_emptyString ? s_zy : mkConcat( sk_z, sk_y );
2866 str_in_re = NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP, sk_w,
2867 NodeManager::currentNM()->mkNode( kind::REGEXP_STAR,
2868 NodeManager::currentNM()->mkNode( kind::STRING_TO_REGEXP, restr ) ) );
2869
2870 std::vector< Node > vec_conc;
2871 vec_conc.push_back(conc1); vec_conc.push_back(conc2); vec_conc.push_back(conc3);
2872 vec_conc.push_back(str_in_re);
2873 //vec_conc.push_back(sk_y.eqNode(d_emptyString).negate());//by mkskolems
2874 conc = NodeManager::currentNM()->mkNode( kind::AND, vec_conc );
2875 } // normal case
2876
2877 //set its antecedant to ant, to say when it is relevant
2878 if(!str_in_re.isNull()) {
2879 d_regexp_ant[str_in_re] = ant;
2880 }
2881 //we will be done
2882 addNormalFormPair( normal_form_src[i], normal_form_src[j] );
2883 if( options::stringProcessLoop() ){
2884 info.d_conc = conc;
2885 info.d_id = 8;
2886 return true;
2887 }else{
2888 d_out->setIncomplete();
2889 }
2890 }else{
2891 Trace("strings-loop") << "Strings::Loop: loop lemma for " << ant << " has already added." << std::endl;
2892 addNormalFormPair( normal_form_src[i], normal_form_src[j] );
2893 }
2894 }
2895 }
2896 return false;
2897 }
2898
2899 //return true for lemma, false if we succeed
2900 void TheoryStrings::processDeq( Node ni, Node nj ) {
2901 //Assert( areDisequal( ni, nj ) );
2902 if( d_normal_forms[ni].size()>1 || d_normal_forms[nj].size()>1 ){
2903 std::vector< Node > nfi;
2904 nfi.insert( nfi.end(), d_normal_forms[ni].begin(), d_normal_forms[ni].end() );
2905 std::vector< Node > nfj;
2906 nfj.insert( nfj.end(), d_normal_forms[nj].begin(), d_normal_forms[nj].end() );
2907
2908 int revRet = processReverseDeq( nfi, nfj, ni, nj );
2909 if( revRet!=0 ){
2910 return;
2911 }
2912
2913 nfi.clear();
2914 nfi.insert( nfi.end(), d_normal_forms[ni].begin(), d_normal_forms[ni].end() );
2915 nfj.clear();
2916 nfj.insert( nfj.end(), d_normal_forms[nj].begin(), d_normal_forms[nj].end() );
2917
2918 unsigned index = 0;
2919 while( index<nfi.size() || index<nfj.size() ){
2920 int ret = processSimpleDeq( nfi, nfj, ni, nj, index, false );
2921 if( ret!=0 ) {
2922 return;
2923 }else{
2924 Assert( index<nfi.size() && index<nfj.size() );
2925 Node i = nfi[index];
2926 Node j = nfj[index];
2927 Trace("strings-solve-debug") << "...Processing(DEQ) " << i << " " << j << std::endl;
2928 if( !areEqual( i, j ) ){
2929 Assert( i.getKind()!=kind::CONST_STRING || j.getKind()!=kind::CONST_STRING );
2930 std::vector< Node > lexp;
2931 Node li = getLength( i, lexp );
2932 Node lj = getLength( j, lexp );
2933 if( areDisequal( li, lj ) ){
2934 if( i.getKind()==kind::CONST_STRING || j.getKind()==kind::CONST_STRING ){
2935 //check if empty
2936 Node const_k = i.getKind() == kind::CONST_STRING ? i : j;
2937 Node nconst_k = i.getKind() == kind::CONST_STRING ? j : i;
2938 Node lnck = i.getKind() == kind::CONST_STRING ? lj : li;
2939 if( !d_equalityEngine.areDisequal( nconst_k, d_emptyString, true ) ){
2940 Node eq = nconst_k.eqNode( d_emptyString );
2941 Node conc = NodeManager::currentNM()->mkNode( kind::OR, eq, eq.negate() );
2942 sendInference( d_empty_vec, conc, "D-DISL-Emp-Split" );
2943 return;
2944 }else{
2945 //split on first character
2946 CVC4::String str = const_k.getConst<String>();
2947 Node firstChar = str.size() == 1 ? const_k : NodeManager::currentNM()->mkConst( str.prefix( 1 ) );
2948 if( areEqual( lnck, d_one ) ){
2949 if( areDisequal( firstChar, nconst_k ) ){
2950 return;
2951 }else if( !areEqual( firstChar, nconst_k ) ){
2952 //splitting on demand : try to make them disequal
2953 Node eq = firstChar.eqNode( nconst_k );
2954 sendSplit( firstChar, nconst_k, "S-Split(DEQL-Const)" );
2955 eq = Rewriter::rewrite( eq );
2956 d_pending_req_phase[ eq ] = false;
2957 return;
2958 }
2959 }else{
2960 Node sk = mkSkolemCached( nconst_k, firstChar, sk_id_dc_spt, "dc_spt", 2 );
2961 Node skr = mkSkolemCached( nconst_k, firstChar, sk_id_dc_spt_rem, "dc_spt_rem" );
2962 Node eq1 = nconst_k.eqNode( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, sk, skr ) );
2963 eq1 = Rewriter::rewrite( eq1 );
2964 Node eq2 = nconst_k.eqNode( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, firstChar, skr ) );
2965 std::vector< Node > antec;
2966 antec.insert( antec.end(), d_normal_forms_exp[ni].begin(), d_normal_forms_exp[ni].end() );
2967 antec.insert( antec.end(), d_normal_forms_exp[nj].begin(), d_normal_forms_exp[nj].end() );
2968 antec.push_back( nconst_k.eqNode( d_emptyString ).negate() );
2969 sendInference( antec, NodeManager::currentNM()->mkNode( kind::OR,
2970 NodeManager::currentNM()->mkNode( kind::AND, eq1, sk.eqNode( firstChar ).negate() ), eq2 ), "D-DISL-CSplit" );
2971 d_pending_req_phase[ eq1 ] = true;
2972 return;
2973 }
2974 }
2975 }else{
2976 Trace("strings-solve") << "Non-Simple Case 1 : add lemma " << std::endl;
2977 //must add lemma
2978 std::vector< Node > antec;
2979 std::vector< Node > antec_new_lits;
2980 antec.insert( antec.end(), d_normal_forms_exp[ni].begin(), d_normal_forms_exp[ni].end() );
2981 antec.insert( antec.end(), d_normal_forms_exp[nj].begin(), d_normal_forms_exp[nj].end() );
2982 //check disequal
2983 if( areDisequal( ni, nj ) ){
2984 antec.push_back( ni.eqNode( nj ).negate() );
2985 }else{
2986 antec_new_lits.push_back( ni.eqNode( nj ).negate() );
2987 }
2988 antec_new_lits.push_back( li.eqNode( lj ).negate() );
2989 std::vector< Node > conc;
2990 Node sk1 = mkSkolemCached( i, j, sk_id_deq_x, "x_dsplit" );
2991 Node sk2 = mkSkolemCached( i, j, sk_id_deq_y, "y_dsplit" );
2992 Node sk3 = mkSkolemCached( i, j, sk_id_deq_z, "z_dsplit", 1 );
2993 //Node nemp = sk3.eqNode(d_emptyString).negate();
2994 //conc.push_back(nemp);
2995 Node lsk1 = mkLength( sk1 );
2996 conc.push_back( lsk1.eqNode( li ) );
2997 Node lsk2 = mkLength( sk2 );
2998 conc.push_back( lsk2.eqNode( lj ) );
2999 conc.push_back( NodeManager::currentNM()->mkNode( kind::OR, j.eqNode( mkConcat( sk1, sk3 ) ), i.eqNode( mkConcat( sk2, sk3 ) ) ) );
3000 sendInference( antec, antec_new_lits, NodeManager::currentNM()->mkNode( kind::AND, conc ), "D-DISL-Split" );
3001 ++(d_statistics.d_deq_splits);
3002 return;
3003 }
3004 }else if( areEqual( li, lj ) ){
3005 Assert( !areDisequal( i, j ) );
3006 //splitting on demand : try to make them disequal
3007 Node eq = i.eqNode( j );
3008 sendSplit( i, j, "S-Split(DEQL)" );
3009 eq = Rewriter::rewrite( eq );
3010 d_pending_req_phase[ eq ] = false;
3011 return;
3012 }else{
3013 //splitting on demand : try to make lengths equal
3014 Node eq = li.eqNode( lj );
3015 sendSplit( li, lj, "D-Split" );
3016 eq = Rewriter::rewrite( eq );
3017 d_pending_req_phase[ eq ] = true;
3018 return;
3019 }
3020 }
3021 index++;
3022 }
3023 }
3024 Assert( false );
3025 }
3026 }
3027
3028 int TheoryStrings::processReverseDeq( std::vector< Node >& nfi, std::vector< Node >& nfj, Node ni, Node nj ) {
3029 //reverse normal form of i, j
3030 std::reverse( nfi.begin(), nfi.end() );
3031 std::reverse( nfj.begin(), nfj.end() );
3032
3033 unsigned index = 0;
3034 int ret = processSimpleDeq( nfi, nfj, ni, nj, index, true );
3035
3036 //reverse normal form of i, j
3037 std::reverse( nfi.begin(), nfi.end() );
3038 std::reverse( nfj.begin(), nfj.end() );
3039
3040 return ret;
3041 }
3042
3043 int TheoryStrings::processSimpleDeq( std::vector< Node >& nfi, std::vector< Node >& nfj, Node ni, Node nj, unsigned& index, bool isRev ){
3044 //see if one side is constant, if so, we can approximate as containment
3045 for( unsigned i=0; i<2; i++ ){
3046 Node c = getConstantEqc( i==0 ? ni : nj );
3047 if( !c.isNull() ){
3048 int findex, lindex;
3049 if( !TheoryStringsRewriter::canConstantContainList( c, i==0 ? nfj : nfi, findex, lindex ) ){
3050 return 1;
3051 }
3052 }
3053 }
3054 while( index<nfi.size() || index<nfj.size() ) {
3055 if( index>=nfi.size() || index>=nfj.size() ){
3056 Trace("strings-solve-debug") << "Disequality normalize empty" << std::endl;
3057 std::vector< Node > ant;
3058 //we have a conflict : because the lengths are equal, the remainder needs to be empty, which will lead to a conflict
3059 Node lni = getLengthExp( ni, ant, d_normal_forms_base[ni] );
3060 Node lnj = getLengthExp( nj, ant, d_normal_forms_base[nj] );
3061 ant.push_back( lni.eqNode( lnj ) );
3062 ant.insert( ant.end(), d_normal_forms_exp[ni].begin(), d_normal_forms_exp[ni].end() );
3063 ant.insert( ant.end(), d_normal_forms_exp[nj].begin(), d_normal_forms_exp[nj].end() );
3064 std::vector< Node > cc;
3065 std::vector< Node >& nfk = index>=nfi.size() ? nfj : nfi;
3066 for( unsigned index_k=index; index_k<nfk.size(); index_k++ ){
3067 cc.push_back( nfk[index_k].eqNode( d_emptyString ) );
3068 }
3069 Node conc = cc.size()==1 ? cc[0] : NodeManager::currentNM()->mkNode( kind::AND, cc );
3070 conc = Rewriter::rewrite( conc );
3071 sendInference( ant, conc, "Disequality Normalize Empty", true);
3072 return -1;
3073 }else{
3074 Node i = nfi[index];
3075 Node j = nfj[index];
3076 Trace("strings-solve-debug") << "...Processing(QED) " << i << " " << j << std::endl;
3077 if( !areEqual( i, j ) ) {
3078 if( i.getKind()==kind::CONST_STRING && j.getKind()==kind::CONST_STRING ) {
3079 unsigned int len_short = i.getConst<String>().size() < j.getConst<String>().size() ? i.getConst<String>().size() : j.getConst<String>().size();
3080 bool isSameFix = isRev ? i.getConst<String>().rstrncmp(j.getConst<String>(), len_short): i.getConst<String>().strncmp(j.getConst<String>(), len_short);
3081 if( isSameFix ) {
3082 //same prefix/suffix
3083 //k is the index of the string that is shorter
3084 Node nk = i.getConst<String>().size() < j.getConst<String>().size() ? i : j;
3085 Node nl = i.getConst<String>().size() < j.getConst<String>().size() ? j : i;
3086 Node remainderStr;
3087 if( isRev ){
3088 int new_len = nl.getConst<String>().size() - len_short;
3089 remainderStr = NodeManager::currentNM()->mkConst( nl.getConst<String>().substr(0, new_len) );
3090 Trace("strings-solve-debug-test") << "Rev. Break normal form of " << nl << " into " << nk << ", " << remainderStr << std::endl;
3091 } else {
3092 remainderStr = NodeManager::currentNM()->mkConst( nl.getConst<String>().substr( len_short ) );
3093 Trace("strings-solve-debug-test") << "Break normal form of " << nl << " into " << nk << ", " << remainderStr << std::endl;
3094 }
3095 if( i.getConst<String>().size() < j.getConst<String>().size() ) {
3096 nfj.insert( nfj.begin() + index + 1, remainderStr );
3097 nfj[index] = nfi[index];
3098 } else {
3099 nfi.insert( nfi.begin() + index + 1, remainderStr );
3100 nfi[index] = nfj[index];
3101 }
3102 }else{
3103 return 1;
3104 }
3105 }else{
3106 std::vector< Node > lexp;
3107 Node li = getLength( i, lexp );
3108 Node lj = getLength( j, lexp );
3109 if( areEqual( li, lj ) && areDisequal( i, j ) ){
3110 Trace("strings-solve") << "Simple Case 2 : found equal length disequal sub strings " << i << " " << j << std::endl;
3111 //we are done: D-Remove
3112 return 1;
3113 }else{
3114 return 0;
3115 }
3116 }
3117 }
3118 index++;
3119 }
3120 }
3121 return 0;
3122 }
3123
3124 void TheoryStrings::addNormalFormPair( Node n1, Node n2 ){
3125 if( !isNormalFormPair( n1, n2 ) ){
3126 int index = 0;
3127 NodeIntMap::const_iterator it = d_nf_pairs.find( n1 );
3128 if( it!=d_nf_pairs.end() ){
3129 index = (*it).second;
3130 }
3131 d_nf_pairs[n1] = index + 1;
3132 if( index<(int)d_nf_pairs_data[n1].size() ){
3133 d_nf_pairs_data[n1][index] = n2;
3134 }else{
3135 d_nf_pairs_data[n1].push_back( n2 );
3136 }
3137 Assert( isNormalFormPair( n1, n2 ) );
3138 } else {
3139 Trace("strings-nf-debug") << "Already a normal form pair " << n1 << " " << n2 << std::endl;
3140 }
3141 }
3142
3143 bool TheoryStrings::isNormalFormPair( Node n1, Node n2 ) {
3144 //TODO: modulo equality?
3145 return isNormalFormPair2( n1, n2 ) || isNormalFormPair2( n2, n1 );
3146 }
3147
3148 bool TheoryStrings::isNormalFormPair2( Node n1, Node n2 ) {
3149 //Trace("strings-debug") << "is normal form pair. " << n1 << " " << n2 << std::endl;
3150 NodeIntMap::const_iterator it = d_nf_pairs.find( n1 );
3151 if( it!=d_nf_pairs.end() ){
3152 Assert( d_nf_pairs_data.find( n1 )!=d_nf_pairs_data.end() );
3153 for( int i=0; i<(*it).second; i++ ){
3154 Assert( i<(int)d_nf_pairs_data[n1].size() );
3155 if( d_nf_pairs_data[n1][i]==n2 ){
3156 return true;
3157 }
3158 }
3159 }
3160 return false;
3161 }
3162
3163 void TheoryStrings::registerTerm( Node n, int effort ) {
3164 // 0 : upon preregistration or internal assertion
3165 // 1 : upon occurrence in length term
3166 // 2 : before normal form computation
3167 // 3 : called on normal form terms
3168 bool do_register = false;
3169 if( options::stringEagerLen() ){
3170 do_register = effort==0;
3171 }else{
3172 do_register = effort>0 || n.getKind()!=kind::STRING_CONCAT;
3173 }
3174 if( do_register ){
3175 if(d_registered_terms_cache.find(n) == d_registered_terms_cache.end()) {
3176 d_registered_terms_cache.insert(n);
3177 Debug("strings-register") << "TheoryStrings::registerTerm() " << n << ", effort = " << effort << std::endl;
3178 if(n.getType().isString()) {
3179 //register length information:
3180 // for variables, split on empty vs positive length
3181 // for concat/const/replace, introduce proxy var and state length relation
3182 Node lsum;
3183 bool processed = false;
3184 if( n.getKind()!=kind::STRING_CONCAT && n.getKind()!=kind::CONST_STRING ) {
3185 if( d_length_lemma_terms_cache.find( n )==d_length_lemma_terms_cache.end() ){
3186 Node lsumb = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, n );
3187 lsum = Rewriter::rewrite( lsumb );
3188 // can register length term if it does not rewrite
3189 if( lsum==lsumb ){
3190 sendLengthLemma( n );
3191 processed = true;
3192 }
3193 }else{
3194 processed = true;
3195 }
3196 }
3197 if( !processed ){
3198 Node sk = mkSkolemS( "lsym", -1 );
3199 StringsProxyVarAttribute spva;
3200 sk.setAttribute(spva,true);
3201 Node eq = Rewriter::rewrite( sk.eqNode(n) );
3202 Trace("strings-lemma") << "Strings::Lemma LENGTH Term : " << eq << std::endl;
3203 d_proxy_var[n] = sk;
3204 Trace("strings-assert") << "(assert " << eq << ")" << std::endl;
3205 d_out->lemma(eq);
3206 Node skl = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, sk );
3207 if( n.getKind()==kind::STRING_CONCAT ){
3208 std::vector<Node> node_vec;
3209 for( unsigned i=0; i<n.getNumChildren(); i++ ) {
3210 if( n[i].getAttribute(StringsProxyVarAttribute()) ){
3211 Assert( d_proxy_var_to_length.find( n[i] )!=d_proxy_var_to_length.end() );
3212 node_vec.push_back( d_proxy_var_to_length[n[i]] );
3213 }else{
3214 Node lni = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, n[i] );
3215 node_vec.push_back(lni);
3216 }
3217 }
3218 lsum = NodeManager::currentNM()->mkNode( kind::PLUS, node_vec );
3219 lsum = Rewriter::rewrite( lsum );
3220 }else if( n.getKind()==kind::CONST_STRING ){
3221 lsum = NodeManager::currentNM()->mkConst( ::CVC4::Rational( n.getConst<String>().size() ) );
3222 }
3223 Assert( !lsum.isNull() );
3224 d_proxy_var_to_length[sk] = lsum;
3225 Node ceq = Rewriter::rewrite( skl.eqNode( lsum ) );
3226 Trace("strings-lemma") << "Strings::Lemma LENGTH : " << ceq << std::endl;
3227 Trace("strings-lemma-debug") << " prerewrite : " << skl.eqNode( lsum ) << std::endl;
3228 Trace("strings-assert") << "(assert " << ceq << ")" << std::endl;
3229 d_out->lemma(ceq);
3230
3231 }
3232 } else {
3233 AlwaysAssert(false, "String Terms only in registerTerm.");
3234 }
3235 }
3236 }
3237 }
3238
3239 void TheoryStrings::sendInference( std::vector< Node >& exp, std::vector< Node >& exp_n, Node eq, const char * c, bool asLemma ) {
3240 eq = eq.isNull() ? d_false : Rewriter::rewrite( eq );
3241 if( eq!=d_true ){
3242 if( Trace.isOn("strings-infer-debug") ){
3243 Trace("strings-infer-debug") << "By " << c << ", infer : " << eq << " from: " << std::endl;
3244 for( unsigned i=0; i<exp.size(); i++ ){
3245 Trace("strings-infer-debug") << " " << exp[i] << std::endl;
3246 }
3247 for( unsigned i=0; i<exp_n.size(); i++ ){
3248 Trace("strings-infer-debug") << " N:" << exp_n[i] << std::endl;
3249 }
3250 //Trace("strings-infer-debug") << "as lemma : " << asLemma << std::endl;
3251 }
3252 //check if we should send a lemma or an inference
3253 if( asLemma || eq==d_false || eq.getKind()==kind::OR || !exp_n.empty() || options::stringInferAsLemmas() ){
3254 Node eq_exp;
3255 if( options::stringRExplainLemmas() ){
3256 eq_exp = mkExplain( exp, exp_n );
3257 }else{
3258 if( exp.empty() ){
3259 eq_exp = mkAnd( exp_n );
3260 }else if( exp_n.empty() ){
3261 eq_exp = mkAnd( exp );
3262 }else{
3263 std::vector< Node > ev;
3264 ev.insert( ev.end(), exp.begin(), exp.end() );
3265 ev.insert( ev.end(), exp_n.begin(), exp_n.end() );
3266 eq_exp = NodeManager::currentNM()->mkNode( kind::AND, ev );
3267 }
3268 }
3269 sendLemma( eq_exp, eq, c );
3270 }else{
3271 sendInfer( mkAnd( exp ), eq, c );
3272 }
3273 }
3274 }
3275
3276 void TheoryStrings::sendInference( std::vector< Node >& exp, Node eq, const char * c, bool asLemma ) {
3277 std::vector< Node > exp_n;
3278 sendInference( exp, exp_n, eq, c, asLemma );
3279 }
3280
3281 void TheoryStrings::sendLemma( Node ant, Node conc, const char * c ) {
3282 if( conc.isNull() || conc == d_false ) {
3283 d_out->conflict(ant);
3284 Trace("strings-conflict") << "Strings::Conflict : " << c << " : " << ant << std::endl;
3285 Trace("strings-lemma") << "Strings::Conflict : " << c << " : " << ant << std::endl;
3286 Trace("strings-assert") << "(assert (not " << ant << ")) ; conflict " << c << std::endl;
3287 d_conflict = true;
3288 } else {
3289 Node lem;
3290 if( ant == d_true ) {
3291 lem = conc;
3292 }else{
3293 lem = NodeManager::currentNM()->mkNode( kind::IMPLIES, ant, conc );
3294 }
3295 Trace("strings-lemma") << "Strings::Lemma " << c << " : " << lem << std::endl;
3296 Trace("strings-assert") << "(assert " << lem << ") ; lemma " << c << std::endl;
3297 d_lemma_cache.push_back( lem );
3298 }
3299 }
3300
3301 void TheoryStrings::sendInfer( Node eq_exp, Node eq, const char * c ) {
3302 if( options::stringInferSym() ){
3303 std::vector< Node > vars;
3304 std::vector< Node > subs;
3305 std::vector< Node > unproc;
3306 inferSubstitutionProxyVars( eq_exp, vars, subs, unproc );
3307 if( unproc.empty() ){
3308 Trace("strings-lemma-debug") << "Strings::Infer " << eq << " from " << eq_exp << " by " << c << std::endl;
3309 Node eqs = eq.substitute( vars.begin(), vars.end(), subs.begin(), subs.end() );
3310 Trace("strings-lemma-debug") << "Strings::Infer Alternate : " << eqs << std::endl;
3311 for( unsigned i=0; i<vars.size(); i++ ){
3312 Trace("strings-lemma-debug") << " " << vars[i] << " -> " << subs[i] << std::endl;
3313 }
3314 sendLemma( d_true, eqs, c );
3315 return;
3316 }else{
3317 for( unsigned i=0; i<unproc.size(); i++ ){
3318 Trace("strings-lemma-debug") << " non-trivial exp : " << unproc[i] << std::endl;
3319 }
3320 }
3321 }
3322 Trace("strings-lemma") << "Strings::Infer " << eq << " from " << eq_exp << " by " << c << std::endl;
3323 Trace("strings-assert") << "(assert (=> " << eq_exp << " " << eq << ")) ; infer " << c << std::endl;
3324 d_pending.push_back( eq );
3325 d_pending_exp[eq] = eq_exp;
3326 d_infer.push_back( eq );
3327 d_infer_exp.push_back( eq_exp );
3328
3329 }
3330
3331 void TheoryStrings::sendSplit( Node a, Node b, const char * c, bool preq ) {
3332 Node eq = a.eqNode( b );
3333 eq = Rewriter::rewrite( eq );
3334 Node neq = NodeManager::currentNM()->mkNode( kind::NOT, eq );
3335 Node lemma_or = NodeManager::currentNM()->mkNode( kind::OR, eq, neq );
3336 Trace("strings-lemma") << "Strings::Lemma " << c << " SPLIT : " << lemma_or << std::endl;
3337 d_lemma_cache.push_back(lemma_or);
3338 d_pending_req_phase[eq] = preq;
3339 ++(d_statistics.d_splits);
3340 }
3341
3342
3343 void TheoryStrings::sendLengthLemma( Node n ){
3344 Node n_len = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, n);
3345 if( options::stringSplitEmp() || !options::stringLenGeqZ() ){
3346 Node n_len_eq_z = n_len.eqNode( d_zero );
3347 Node n_len_eq_z_2 = n.eqNode( d_emptyString );
3348 n_len_eq_z = Rewriter::rewrite( n_len_eq_z );
3349 n_len_eq_z_2 = Rewriter::rewrite( n_len_eq_z_2 );
3350 Node n_len_geq_zero = NodeManager::currentNM()->mkNode( kind::OR, NodeManager::currentNM()->mkNode( kind::AND, n_len_eq_z, n_len_eq_z_2 ),
3351 NodeManager::currentNM()->mkNode( kind::GT, n_len, d_zero) );
3352 Trace("strings-lemma") << "Strings::Lemma LENGTH >= 0 : " << n_len_geq_zero << std::endl;
3353 d_out->lemma(n_len_geq_zero);
3354 d_out->requirePhase( n_len_eq_z, true );
3355 d_out->requirePhase( n_len_eq_z_2, true );
3356 }
3357 //AJR: probably a good idea
3358 if( options::stringLenGeqZ() ){
3359 Node n_len_geq = NodeManager::currentNM()->mkNode( kind::GEQ, n_len, d_zero);
3360 n_len_geq = Rewriter::rewrite( n_len_geq );
3361 d_out->lemma( n_len_geq );
3362 }
3363 }
3364
3365 void TheoryStrings::inferSubstitutionProxyVars( Node n, std::vector< Node >& vars, std::vector< Node >& subs, std::vector< Node >& unproc ) {
3366 if( n.getKind()==kind::AND ){
3367 for( unsigned i=0; i<n.getNumChildren(); i++ ){
3368 inferSubstitutionProxyVars( n[i], vars, subs, unproc );
3369 }
3370 return;
3371 }else if( n.getKind()==kind::EQUAL ){
3372 Node ns = n.substitute( vars.begin(), vars.end(), subs.begin(), subs.end() );
3373 ns = Rewriter::rewrite( ns );
3374 if( ns.getKind()==kind::EQUAL ){
3375 Node s;
3376 Node v;
3377 for( unsigned i=0; i<2; i++ ){
3378 Node ss;
3379 if( ns[i].getAttribute(StringsProxyVarAttribute()) ){
3380 ss = ns[i];
3381 }else if( ns[i].isConst() ){
3382 NodeNodeMap::const_iterator it = d_proxy_var.find( ns[i] );
3383 if( it!=d_proxy_var.end() ){
3384 ss = (*it).second;
3385 }
3386 }
3387 if( !ss.isNull() ){
3388 v = ns[1-i];
3389 if( v.getNumChildren()==0 ){
3390 if( s.isNull() ){
3391 s = ss;
3392 }else{
3393 //both sides involved in proxy var
3394 if( ss==s ){
3395 return;
3396 }else{
3397 s = Node::null();
3398 }
3399 }
3400 }
3401 }
3402 }
3403 if( !s.isNull() ){
3404 subs.push_back( s );
3405 vars.push_back( v );
3406 return;
3407 }
3408 }else{
3409 n = ns;
3410 }
3411 }
3412 if( n!=d_true ){
3413 unproc.push_back( n );
3414 }
3415 }
3416
3417
3418 Node TheoryStrings::mkConcat( Node n1, Node n2 ) {
3419 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, n1, n2 ) );
3420 }
3421
3422 Node TheoryStrings::mkConcat( Node n1, Node n2, Node n3 ) {
3423 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, n1, n2, n3 ) );
3424 }
3425
3426 Node TheoryStrings::mkConcat( const std::vector< Node >& c ) {
3427 return Rewriter::rewrite( c.size()>1 ? NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, c ) : ( c.size()==1 ? c[0] : d_emptyString ) );
3428 }
3429
3430 Node TheoryStrings::mkLength( Node t ) {
3431 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, t ) );
3432 }
3433
3434 Node TheoryStrings::mkSkolemCached( Node a, Node b, int id, const char * c, int isLenSplit ){
3435 //return mkSkolemS( c, isLenSplit );
3436 std::map< int, Node >::iterator it = d_skolem_cache[a][b].find( id );
3437 if( it==d_skolem_cache[a][b].end() ){
3438 Node sk = mkSkolemS( c, isLenSplit );
3439 d_skolem_cache[a][b][id] = sk;
3440 return sk;
3441 }else{
3442 return it->second;
3443 }
3444 }
3445
3446 //isLenSplit: -1-ignore, 0-no restriction, 1-greater than one, 2-one
3447 Node TheoryStrings::mkSkolemS( const char *c, int isLenSplit ) {
3448 Node n = NodeManager::currentNM()->mkSkolem( c, NodeManager::currentNM()->stringType(), "string sko" );
3449 d_length_lemma_terms_cache.insert( n );
3450 ++(d_statistics.d_new_skolems);
3451 if( isLenSplit==0 ){
3452 sendLengthLemma( n );
3453 } else if( isLenSplit == 1 ){
3454 registerNonEmptySkolem( n );
3455 }else if( isLenSplit==2 ){
3456 Node len_one = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, n ).eqNode( d_one );
3457 Trace("strings-lemma") << "Strings::Lemma SK-ONE : " << len_one << std::endl;
3458 Trace("strings-assert") << "(assert " << len_one << ")" << std::endl;
3459 d_out->lemma( len_one );
3460 }
3461 return n;
3462 }
3463
3464 void TheoryStrings::registerNonEmptySkolem( Node n ) {
3465 if( d_skolem_ne_reg_cache.find( n )==d_skolem_ne_reg_cache.end() ){
3466 d_skolem_ne_reg_cache.insert( n );
3467 d_equalityEngine.assertEquality(n.eqNode(d_emptyString), false, d_true);
3468 Node len_n_gt_z = NodeManager::currentNM()->mkNode(kind::GT,
3469 NodeManager::currentNM()->mkNode(kind::STRING_LENGTH, n), d_zero);
3470 Trace("strings-lemma") << "Strings::Lemma SK-NON-ZERO : " << len_n_gt_z << std::endl;
3471 Trace("strings-assert") << "(assert " << len_n_gt_z << ")" << std::endl;
3472 d_out->lemma(len_n_gt_z);
3473 }
3474 }
3475
3476 Node TheoryStrings::mkExplain( std::vector< Node >& a ) {
3477 std::vector< Node > an;
3478 return mkExplain( a, an );
3479 }
3480
3481 Node TheoryStrings::mkExplain( std::vector< Node >& a, std::vector< Node >& an ) {
3482 std::vector< TNode > antec_exp;
3483 for( unsigned i=0; i<a.size(); i++ ) {
3484 if( std::find( a.begin(), a.begin() + i, a[i] )==a.begin() + i ) {
3485 bool exp = true;
3486 Debug("strings-explain") << "Ask for explanation of " << a[i] << std::endl;
3487 //assert
3488 if(a[i].getKind() == kind::EQUAL) {
3489 //assert( hasTerm(a[i][0]) );
3490 //assert( hasTerm(a[i][1]) );
3491 Assert( areEqual(a[i][0], a[i][1]) );
3492 if( a[i][0]==a[i][1] ){
3493 exp = false;
3494 }
3495 } else if( a[i].getKind()==kind::NOT && a[i][0].getKind()==kind::EQUAL ) {
3496 Assert( hasTerm(a[i][0][0]) );
3497 Assert( hasTerm(a[i][0][1]) );
3498 AlwaysAssert( d_equalityEngine.areDisequal(a[i][0][0], a[i][0][1], true) );
3499 }else if( a[i].getKind() == kind::AND ){
3500 for( unsigned j=0; j<a[i].getNumChildren(); j++ ){
3501 a.push_back( a[i][j] );
3502 }
3503 exp = false;
3504 }
3505 if( exp ) {
3506 unsigned ps = antec_exp.size();
3507 explain(a[i], antec_exp);
3508 Debug("strings-explain") << "Done, explanation was : " << std::endl;
3509 for( unsigned j=ps; j<antec_exp.size(); j++ ) {
3510 Debug("strings-explain") << " " << antec_exp[j] << std::endl;
3511 }
3512 Debug("strings-explain") << std::endl;
3513 }
3514 }
3515 }
3516 for( unsigned i=0; i<an.size(); i++ ) {
3517 if( std::find( an.begin(), an.begin() + i, an[i] )==an.begin() + i ){
3518 Debug("strings-explain") << "Add to explanation (new literal) " << an[i] << std::endl;
3519 antec_exp.push_back(an[i]);
3520 }
3521 }
3522 Node ant;
3523 if( antec_exp.empty() ) {
3524 ant = d_true;
3525 } else if( antec_exp.size()==1 ) {
3526 ant = antec_exp[0];
3527 } else {
3528 ant = NodeManager::currentNM()->mkNode( kind::AND, antec_exp );
3529 }
3530 ant = Rewriter::rewrite( ant );
3531 return ant;
3532 }
3533
3534 Node TheoryStrings::mkAnd( std::vector< Node >& a ) {
3535 std::vector< Node > au;
3536 for( unsigned i=0; i<a.size(); i++ ){
3537 if( std::find( au.begin(), au.end(), a[i] )==au.end() ){
3538 au.push_back( a[i] );
3539 }
3540 }
3541 if( au.empty() ) {
3542 return d_true;
3543 } else if( au.size() == 1 ) {
3544 return au[0];
3545 } else {
3546 return NodeManager::currentNM()->mkNode( kind::AND, au );
3547 }
3548 }
3549
3550 void TheoryStrings::getConcatVec( Node n, std::vector< Node >& c ) {
3551 if( n.getKind()==kind::STRING_CONCAT ) {
3552 for( unsigned i=0; i<n.getNumChildren(); i++ ) {
3553 if( !areEqual( n[i], d_emptyString ) ) {
3554 c.push_back( n[i] );
3555 }
3556 }
3557 }else{
3558 c.push_back( n );
3559 }
3560 }
3561
3562 void TheoryStrings::checkDeqNF() {
3563 std::vector< std::vector< Node > > cols;
3564 std::vector< Node > lts;
3565 std::map< Node, std::map< Node, bool > > processed;
3566
3567 //for each pair of disequal strings, must determine whether their lengths are equal or disequal
3568 for( NodeList::const_iterator id = d_ee_disequalities.begin(); id != d_ee_disequalities.end(); ++id ) {
3569 Node eq = *id;
3570 Node n[2];
3571 for( unsigned i=0; i<2; i++ ){
3572 n[i] = d_equalityEngine.getRepresentative( eq[i] );
3573 }
3574 if( processed[n[0]].find( n[1] )==processed[n[0]].end() ){
3575 processed[n[0]][n[1]] = true;
3576 Node lt[2];
3577 for( unsigned i=0; i<2; i++ ){
3578 EqcInfo* ei = getOrMakeEqcInfo( n[i], false );
3579 lt[i] = ei ? ei->d_length_term : Node::null();
3580 if( lt[i].isNull() ){
3581 lt[i] = eq[i];
3582 }
3583 lt[i] = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, lt[i] );
3584 }
3585 if( !areEqual( lt[0], lt[1] ) && !areDisequal( lt[0], lt[1] ) ){
3586 sendSplit( lt[0], lt[1], "DEQ-LENGTH-SP" );
3587 }
3588 }
3589 }
3590
3591 if( !hasProcessed() ){
3592 separateByLength( d_strings_eqc, cols, lts );
3593 for( unsigned i=0; i<cols.size(); i++ ){
3594 if( cols[i].size()>1 && d_lemma_cache.empty() ){
3595 Trace("strings-solve") << "- Verify disequalities are processed for " << cols[i][0] << ", normal form : ";
3596 printConcat( d_normal_forms[cols[i][0]], "strings-solve" );
3597 Trace("strings-solve") << "... #eql = " << cols[i].size() << std::endl;
3598 //must ensure that normal forms are disequal
3599 for( unsigned j=0; j<cols[i].size(); j++ ){
3600 for( unsigned k=(j+1); k<cols[i].size(); k++ ){
3601 //for strings that are disequal, but have the same length
3602 if( areDisequal( cols[i][j], cols[i][k] ) ){
3603 Assert( !d_conflict );
3604 Trace("strings-solve") << "- Compare " << cols[i][j] << " ";
3605 printConcat( d_normal_forms[cols[i][j]], "strings-solve" );
3606 Trace("strings-solve") << " against " << cols[i][k] << " ";
3607 printConcat( d_normal_forms[cols[i][k]], "strings-solve" );
3608 Trace("strings-solve") << "..." << std::endl;
3609 processDeq( cols[i][j], cols[i][k] );
3610 if( hasProcessed() ){
3611 return;
3612 }
3613 }
3614 }
3615 }
3616 }
3617 }
3618 }
3619 }
3620
3621 void TheoryStrings::checkLengthsEqc() {
3622 if( options::stringLenNorm() ){
3623 for( unsigned i=0; i<d_strings_eqc.size(); i++ ){
3624 //if( d_normal_forms[nodes[i]].size()>1 ) {
3625 Trace("strings-process-debug") << "Process length constraints for " << d_strings_eqc[i] << std::endl;
3626 //check if there is a length term for this equivalence class
3627 EqcInfo* ei = getOrMakeEqcInfo( d_strings_eqc[i], false );
3628 Node lt = ei ? ei->d_length_term : Node::null();
3629 if( !lt.isNull() ) {
3630 Node llt = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, lt );
3631 //now, check if length normalization has occurred
3632 if( ei->d_normalized_length.get().isNull() ) {
3633 Node nf = mkConcat( d_normal_forms[d_strings_eqc[i]] );
3634 if( Trace.isOn("strings-process-debug") ){
3635 Trace("strings-process-debug") << " normal form is " << nf << " from base " << d_normal_forms_base[d_strings_eqc[i]] << std::endl;
3636 Trace("strings-process-debug") << " normal form exp is: " << std::endl;
3637 for( unsigned j=0; j<d_normal_forms_exp[d_strings_eqc[i]].size(); j++ ){
3638 Trace("strings-process-debug") << " " << d_normal_forms_exp[d_strings_eqc[i]][j] << std::endl;
3639 }
3640 }
3641
3642 //if not, add the lemma
3643 std::vector< Node > ant;
3644 ant.insert( ant.end(), d_normal_forms_exp[d_strings_eqc[i]].begin(), d_normal_forms_exp[d_strings_eqc[i]].end() );
3645 ant.push_back( d_normal_forms_base[d_strings_eqc[i]].eqNode( lt ) );
3646 Node lc = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, nf );
3647 Node lcr = Rewriter::rewrite( lc );
3648 Trace("strings-process-debug") << "Rewrote length " << lc << " to " << lcr << std::endl;
3649 Node eq = llt.eqNode( lcr );
3650 if( llt!=lcr ){
3651 ei->d_normalized_length.set( eq );
3652 sendInference( ant, eq, "LEN-NORM", true );
3653 }
3654 }
3655 }else{
3656 Trace("strings-process-debug") << "No length term for eqc " << d_strings_eqc[i] << " " << d_eqc_to_len_term[d_strings_eqc[i]] << std::endl;
3657 if( !options::stringEagerLen() ){
3658 Node c = mkConcat( d_normal_forms[d_strings_eqc[i]] );
3659 registerTerm( c, 3 );
3660 /*
3661 if( !c.isConst() ){
3662 NodeNodeMap::const_iterator it = d_proxy_var.find( c );
3663 if( it!=d_proxy_var.end() ){
3664 Node pv = (*it).second;
3665 Assert( d_proxy_var_to_length.find( pv )!=d_proxy_var_to_length.end() );
3666 Node pvl = d_proxy_var_to_length[pv];
3667 Node ceq = Rewriter::rewrite( mkLength( pv ).eqNode( pvl ) );
3668 sendInference( d_empty_vec, ceq, "LEN-NORM-I", true );
3669 }
3670 }
3671 */
3672 }
3673 }
3674 //} else {
3675 // Trace("strings-process-debug") << "Do not process length constraints for " << nodes[i] << " " << d_normal_forms[nodes[i]].size() << std::endl;
3676 //}
3677 }
3678 }
3679 }
3680
3681 void TheoryStrings::checkCardinality() {
3682 //int cardinality = options::stringCharCardinality();
3683 //Trace("strings-solve-debug2") << "get cardinality: " << cardinality << endl;
3684
3685 //AJR: this will create a partition of eqc, where each collection has length that are pairwise propagated to be equal.
3686 // we do not require disequalities between the lengths of each collection, since we split on disequalities between lengths of string terms that are disequal (DEQ-LENGTH-SP).
3687 // TODO: revisit this?
3688 std::vector< std::vector< Node > > cols;
3689 std::vector< Node > lts;
3690 separateByLength( d_strings_eqc, cols, lts );
3691
3692 for( unsigned i = 0; i<cols.size(); ++i ) {
3693 Node lr = lts[i];
3694 Trace("strings-card") << "Number of strings with length equal to " << lr << " is " << cols[i].size() << std::endl;
3695 if( cols[i].size() > 1 ) {
3696 // size > c^k
3697 unsigned card_need = 1;
3698 double curr = (double)cols[i].size();
3699 while( curr>d_card_size ){
3700 curr = curr/(double)d_card_size;
3701 card_need++;
3702 }
3703 Trace("strings-card") << "Need length " << card_need << " for this number of strings (where alphabet size is " << d_card_size << ")." << std::endl;
3704 Node cmp = NodeManager::currentNM()->mkNode( kind::GEQ, lr, NodeManager::currentNM()->mkConst( Rational( card_need ) ) );
3705 cmp = Rewriter::rewrite( cmp );
3706 if( cmp!=d_true ){
3707 unsigned int int_k = (unsigned int)card_need;
3708 for( std::vector< Node >::iterator itr1 = cols[i].begin();
3709 itr1 != cols[i].end(); ++itr1) {
3710 for( std::vector< Node >::iterator itr2 = itr1 + 1;
3711 itr2 != cols[i].end(); ++itr2) {
3712 if(!areDisequal( *itr1, *itr2 )) {
3713 // add split lemma
3714 sendSplit( *itr1, *itr2, "CARD-SP" );
3715 return;
3716 }
3717 }
3718 }
3719 EqcInfo* ei = getOrMakeEqcInfo( lr, true );
3720 Trace("strings-card") << "Previous cardinality used for " << lr << " is " << ((int)ei->d_cardinality_lem_k.get()-1) << std::endl;
3721 if( int_k+1 > ei->d_cardinality_lem_k.get() ){
3722 Node k_node = NodeManager::currentNM()->mkConst( ::CVC4::Rational( int_k ) );
3723 //add cardinality lemma
3724 Node dist = NodeManager::currentNM()->mkNode( kind::DISTINCT, cols[i] );
3725 std::vector< Node > vec_node;
3726 vec_node.push_back( dist );
3727 for( std::vector< Node >::iterator itr1 = cols[i].begin();
3728 itr1 != cols[i].end(); ++itr1) {
3729 Node len = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, *itr1 );
3730 if( len!=lr ) {
3731 Node len_eq_lr = len.eqNode(lr);
3732 vec_node.push_back( len_eq_lr );
3733 }
3734 }
3735 Node len = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, cols[i][0] );
3736 Node cons = NodeManager::currentNM()->mkNode( kind::GEQ, len, k_node );
3737 cons = Rewriter::rewrite( cons );
3738 ei->d_cardinality_lem_k.set( int_k+1 );
3739 if( cons!=d_true ){
3740 sendInference( d_empty_vec, vec_node, cons, "CARDINALITY", true );
3741 return;
3742 }
3743 }
3744 }
3745 }
3746 }
3747 }
3748
3749 void TheoryStrings::getEquivalenceClasses( std::vector< Node >& eqcs ) {
3750 eq::EqClassesIterator eqcs_i = eq::EqClassesIterator( &d_equalityEngine );
3751 while( !eqcs_i.isFinished() ) {
3752 Node eqc = (*eqcs_i);
3753 //if eqc.getType is string
3754 if (eqc.getType().isString()) {
3755 eqcs.push_back( eqc );
3756 }
3757 ++eqcs_i;
3758 }
3759 }
3760
3761 void TheoryStrings::separateByLength(std::vector< Node >& n,
3762 std::vector< std::vector< Node > >& cols,
3763 std::vector< Node >& lts ) {
3764 unsigned leqc_counter = 0;
3765 std::map< Node, unsigned > eqc_to_leqc;
3766 std::map< unsigned, Node > leqc_to_eqc;
3767 std::map< unsigned, std::vector< Node > > eqc_to_strings;
3768 for( unsigned i=0; i<n.size(); i++ ) {
3769 Node eqc = n[i];
3770 Assert( d_equalityEngine.getRepresentative(eqc)==eqc );
3771 EqcInfo* ei = getOrMakeEqcInfo( eqc, false );
3772 Node lt = ei ? ei->d_length_term : Node::null();
3773 if( !lt.isNull() ){
3774 lt = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, lt );
3775 Node r = d_equalityEngine.getRepresentative( lt );
3776 if( eqc_to_leqc.find( r )==eqc_to_leqc.end() ){
3777 eqc_to_leqc[r] = leqc_counter;
3778 leqc_to_eqc[leqc_counter] = r;
3779 leqc_counter++;
3780 }
3781 eqc_to_strings[ eqc_to_leqc[r] ].push_back( eqc );
3782 }else{
3783 eqc_to_strings[leqc_counter].push_back( eqc );
3784 leqc_counter++;
3785 }
3786 }
3787 for( std::map< unsigned, std::vector< Node > >::iterator it = eqc_to_strings.begin(); it != eqc_to_strings.end(); ++it ){
3788 cols.push_back( std::vector< Node >() );
3789 cols.back().insert( cols.back().end(), it->second.begin(), it->second.end() );
3790 lts.push_back( leqc_to_eqc[it->first] );
3791 }
3792 }
3793
3794 void TheoryStrings::printConcat( std::vector< Node >& n, const char * c ) {
3795 for( unsigned i=0; i<n.size(); i++ ){
3796 if( i>0 ) Trace(c) << " ++ ";
3797 Trace(c) << n[i];
3798 }
3799 }
3800
3801
3802
3803 //// Finite Model Finding
3804
3805 Node TheoryStrings::getNextDecisionRequest( unsigned& priority ) {
3806 if( options::stringFMF() && !d_conflict ){
3807 Node in_var_lsum = d_input_var_lsum.get();
3808 //Trace("strings-fmf-debug") << "Strings::FMF: Assertion Level = " << d_valuation.getAssertionLevel() << std::endl;
3809 //initialize the term we will minimize
3810 if( in_var_lsum.isNull() && !d_input_vars.empty() ){
3811 Trace("strings-fmf-debug") << "Input variables: ";
3812 std::vector< Node > ll;
3813 for(NodeSet::key_iterator itr = d_input_vars.key_begin();
3814 itr != d_input_vars.key_end(); ++itr) {
3815 Trace("strings-fmf-debug") << " " << (*itr) ;
3816 ll.push_back( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, *itr ) );
3817 }
3818 Trace("strings-fmf-debug") << std::endl;
3819 in_var_lsum = ll.size()==1 ? ll[0] : NodeManager::currentNM()->mkNode( kind::PLUS, ll );
3820 in_var_lsum = Rewriter::rewrite( in_var_lsum );
3821 d_input_var_lsum.set( in_var_lsum );
3822 }
3823 if( !in_var_lsum.isNull() ){
3824 //Trace("strings-fmf") << "Get next decision request." << std::endl;
3825 //check if we need to decide on something
3826 int decideCard = d_curr_cardinality.get();
3827 if( d_cardinality_lits.find( decideCard )!=d_cardinality_lits.end() ){
3828 bool value;
3829 Node cnode = d_cardinality_lits[ d_curr_cardinality.get() ];
3830 if( d_valuation.hasSatValue( cnode, value ) ) {
3831 if( !value ){
3832 d_curr_cardinality.set( d_curr_cardinality.get() + 1 );
3833 decideCard = d_curr_cardinality.get();
3834 Trace("strings-fmf-debug") << "Has false SAT value, increment and decide." << std::endl;
3835 }else{
3836 decideCard = -1;
3837 Trace("strings-fmf-debug") << "Has true SAT value, do not decide." << std::endl;
3838 }
3839 }else{
3840 Trace("strings-fmf-debug") << "No SAT value, decide." << std::endl;
3841 }
3842 }
3843 if( decideCard!=-1 ){
3844 if( d_cardinality_lits.find( decideCard )==d_cardinality_lits.end() ){
3845 Node lit = NodeManager::currentNM()->mkNode( kind::LEQ, in_var_lsum, NodeManager::currentNM()->mkConst( Rational( decideCard ) ) );
3846 lit = Rewriter::rewrite( lit );
3847 d_cardinality_lits[decideCard] = lit;
3848 Node lem = NodeManager::currentNM()->mkNode( kind::OR, lit, lit.negate() );
3849 Trace("strings-fmf") << "Strings::FMF: Add decision lemma " << lem << ", decideCard = " << decideCard << std::endl;
3850 d_out->lemma( lem );
3851 d_out->requirePhase( lit, true );
3852 }
3853 Node lit = d_cardinality_lits[ decideCard ];
3854 Trace("strings-fmf") << "Strings::FMF: Decide positive on " << lit << std::endl;
3855 priority = 1;
3856 return lit;
3857 }
3858 }
3859 }
3860 return Node::null();
3861 }
3862
3863 Node TheoryStrings::ppRewrite(TNode atom) {
3864 Trace("strings-ppr") << "TheoryStrings::ppRewrite " << atom << std::endl;
3865 if( !options::stringLazyPreproc() ){
3866 //eager preprocess here
3867 std::vector< Node > new_nodes;
3868 Node ret = d_preproc.processAssertion( atom, new_nodes );
3869 if( ret!=atom ){
3870 Trace("strings-ppr") << " rewrote " << atom << " -> " << ret << ", with " << new_nodes.size() << " lemmas." << std::endl;
3871 for( unsigned i=0; i<new_nodes.size(); i++ ){
3872 Trace("strings-ppr") << " lemma : " << new_nodes[i] << std::endl;
3873 d_out->lemma( new_nodes[i] );
3874 }
3875 return ret;
3876 }else{
3877 Assert( new_nodes.empty() );
3878 }
3879 }
3880 return atom;
3881 }
3882
3883 // Stats
3884 TheoryStrings::Statistics::Statistics():
3885 d_splits("TheoryStrings::NumOfSplitOnDemands", 0),
3886 d_eq_splits("TheoryStrings::NumOfEqSplits", 0),
3887 d_deq_splits("TheoryStrings::NumOfDiseqSplits", 0),
3888 d_loop_lemmas("TheoryStrings::NumOfLoops", 0),
3889 d_new_skolems("TheoryStrings::NumOfNewSkolems", 0)
3890 {
3891 smtStatisticsRegistry()->registerStat(&d_splits);
3892 smtStatisticsRegistry()->registerStat(&d_eq_splits);
3893 smtStatisticsRegistry()->registerStat(&d_deq_splits);
3894 smtStatisticsRegistry()->registerStat(&d_loop_lemmas);
3895 smtStatisticsRegistry()->registerStat(&d_new_skolems);
3896 }
3897
3898 TheoryStrings::Statistics::~Statistics(){
3899 smtStatisticsRegistry()->unregisterStat(&d_splits);
3900 smtStatisticsRegistry()->unregisterStat(&d_eq_splits);
3901 smtStatisticsRegistry()->unregisterStat(&d_deq_splits);
3902 smtStatisticsRegistry()->unregisterStat(&d_loop_lemmas);
3903 smtStatisticsRegistry()->unregisterStat(&d_new_skolems);
3904 }
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925 //// Regular Expressions
3926
3927
3928 unsigned TheoryStrings::getNumMemberships( Node n, bool isPos ) {
3929 if( isPos ){
3930 NodeIntMap::const_iterator it = d_pos_memberships.find( n );
3931 if( it!=d_pos_memberships.end() ){
3932 return (*it).second;
3933 }
3934 }else{
3935 NodeIntMap::const_iterator it = d_neg_memberships.find( n );
3936 if( it!=d_neg_memberships.end() ){
3937 return (*it).second;
3938 }
3939 }
3940 return 0;
3941 }
3942
3943 Node TheoryStrings::getMembership( Node n, bool isPos, unsigned i ) {
3944 return isPos ? d_pos_memberships_data[n][i] : d_neg_memberships_data[n][i];
3945 }
3946
3947 Node TheoryStrings::mkRegExpAntec(Node atom, Node ant) {
3948 if(d_regexp_ant.find(atom) == d_regexp_ant.end()) {
3949 return Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, ant, atom) );
3950 } else {
3951 Node n = d_regexp_ant[atom];
3952 return Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, ant, n) );
3953 }
3954 }
3955
3956 Node TheoryStrings::normalizeRegexp(Node r) {
3957 Node nf_r = r;
3958 if(d_nf_regexps.find(r) != d_nf_regexps.end()) {
3959 nf_r = d_nf_regexps[r];
3960 } else {
3961 std::vector< Node > nf_exp;
3962 if(!d_regexp_opr.checkConstRegExp(r)) {
3963 switch( r.getKind() ) {
3964 case kind::REGEXP_EMPTY:
3965 case kind::REGEXP_SIGMA: {
3966 break;
3967 }
3968 case kind::STRING_TO_REGEXP: {
3969 if(r[0].isConst()) {
3970 break;
3971 } else {
3972 if(d_normal_forms.find( r[0] ) != d_normal_forms.end()) {
3973 nf_r = mkConcat( d_normal_forms[r[0]] );
3974 Debug("regexp-nf") << "Term: " << r[0] << " has a normal form " << nf_r << std::endl;
3975 nf_exp.insert(nf_exp.end(), d_normal_forms_exp[r[0]].begin(), d_normal_forms_exp[r[0]].end());
3976 nf_r = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::STRING_TO_REGEXP, nf_r) );
3977 }
3978 }
3979 }
3980 case kind::REGEXP_CONCAT:
3981 case kind::REGEXP_UNION:
3982 case kind::REGEXP_INTER: {
3983 bool flag = false;
3984 std::vector< Node > vec_nodes;
3985 for(unsigned i=0; i<r.getNumChildren(); ++i) {
3986 Node rtmp = normalizeRegexp(r[i]);
3987 vec_nodes.push_back(rtmp);
3988 if(rtmp != r[i]) {
3989 flag = true;
3990 }
3991 }
3992 if(flag) {
3993 Node rtmp = vec_nodes.size()==1 ? vec_nodes[0] : NodeManager::currentNM()->mkNode(r.getKind(), vec_nodes);
3994 nf_r = Rewriter::rewrite( rtmp );
3995 }
3996 }
3997 case kind::REGEXP_STAR: {
3998 Node rtmp = normalizeRegexp(r[0]);
3999 if(rtmp != r[0]) {
4000 rtmp = NodeManager::currentNM()->mkNode(kind::REGEXP_STAR, rtmp);
4001 nf_r = Rewriter::rewrite( rtmp );
4002 }
4003 }
4004 default: {
4005 Unreachable();
4006 }
4007 }
4008 }
4009 d_nf_regexps[r] = nf_r;
4010 d_nf_regexps_exp[r] = nf_exp;
4011 }
4012 return nf_r;
4013 }
4014
4015 bool TheoryStrings::normalizePosMemberships(std::map< Node, std::vector< Node > > &memb_with_exps) {
4016 std::map< Node, std::vector< Node > > unprocessed_x_exps;
4017 std::map< Node, std::vector< Node > > unprocessed_memberships;
4018 std::map< Node, std::vector< Node > > unprocessed_memberships_bases;
4019 bool addLemma = false;
4020
4021 Trace("regexp-check") << "Normalizing Positive Memberships ... " << std::endl;
4022
4023 for( NodeIntMap::const_iterator itr_xr = d_pos_memberships.begin(); itr_xr != d_pos_memberships.end(); ++itr_xr ){
4024 Node x = (*itr_xr).first;
4025 Node nf_x = x;
4026 std::vector< Node > nf_x_exp;
4027 if(d_normal_forms.find( x ) != d_normal_forms.end()) {
4028 //nf_x = mkConcat( d_normal_forms[x] );
4029 nf_x_exp.insert(nf_x_exp.end(), d_normal_forms_exp[x].begin(), d_normal_forms_exp[x].end());
4030 //Debug("regexp-nf") << "Term: " << x << " has a normal form " << ret << std::endl;
4031 } else {
4032 Assert(false);
4033 }
4034 Trace("regexp-nf") << "Checking Memberships for N(" << x << ") = " << nf_x << " :" << std::endl;
4035
4036 std::vector< Node > vec_x;
4037 std::vector< Node > vec_r;
4038 unsigned n_pmem = (*itr_xr).second;
4039 Assert( getNumMemberships( x, true )==n_pmem );
4040 for( unsigned k=0; k<n_pmem; k++ ){
4041 Node r = getMembership( x, true, k );
4042 Node nf_r = normalizeRegexp( r ); //AJR: fixed (was normalizing mem #0 always)
4043 Node memb = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, nf_x, nf_r);
4044 if(d_processed_memberships.find(memb) == d_processed_memberships.end()) {
4045 if(d_regexp_opr.checkConstRegExp(nf_r)) {
4046 vec_x.push_back(x);
4047 vec_r.push_back(r);
4048 } else {
4049 Trace("regexp-nf") << "Handling Symbolic Regexp for N(" << r << ") = " << nf_r << std::endl;
4050 //TODO: handle symbolic ones
4051 addLemma = true;
4052 }
4053 d_processed_memberships.insert(memb);
4054 }
4055 }
4056 if(!vec_x.empty()) {
4057 if(unprocessed_x_exps.find(nf_x) == unprocessed_x_exps.end()) {
4058 unprocessed_x_exps[nf_x] = nf_x_exp;
4059 unprocessed_memberships[nf_x] = vec_r;
4060 unprocessed_memberships_bases[nf_x] = vec_x;
4061 } else {
4062 unprocessed_x_exps[nf_x].insert(unprocessed_x_exps[nf_x].end(), nf_x_exp.begin(), nf_x_exp.end());
4063 unprocessed_memberships[nf_x].insert(unprocessed_memberships[nf_x].end(), vec_r.begin(), vec_r.end());
4064 unprocessed_memberships_bases[nf_x].insert(unprocessed_memberships_bases[nf_x].end(), vec_x.begin(), vec_x.end());
4065 }
4066 }
4067 }
4068 //Intersection
4069 for(std::map< Node, std::vector< Node > >::const_iterator itr = unprocessed_memberships.begin();
4070 itr != unprocessed_memberships.end(); ++itr) {
4071 Node nf_x = itr->first;
4072 std::vector< Node > exp( unprocessed_x_exps[nf_x] );
4073 Node r = itr->second[0];
4074 //get nf_r
4075 Node inter_r = d_nf_regexps[r];
4076 exp.insert(exp.end(), d_nf_regexps_exp[r].begin(), d_nf_regexps_exp[r].end());
4077 Node x = unprocessed_memberships_bases[itr->first][0];
4078 Node memb = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x, r);
4079 exp.push_back(memb);
4080 for(std::size_t i=1; i < itr->second.size(); i++) {
4081 //exps
4082 Node r2 = itr->second[i];
4083 Node inter_r2 = d_nf_regexps[r2];
4084 exp.insert(exp.end(), d_nf_regexps_exp[r2].begin(), d_nf_regexps_exp[r2].end());
4085 Node x2 = unprocessed_memberships_bases[itr->first][i];
4086 memb = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x2, r2);
4087 exp.push_back(memb);
4088 //intersection
4089 bool spflag = false;
4090 inter_r = d_regexp_opr.intersect(inter_r, inter_r2, spflag);
4091 if(inter_r == d_emptyRegexp) {
4092 //conflict
4093 Node conc;
4094 sendInference( d_empty_vec, exp, conc, "INTERSECT CONFLICT", true );
4095 addLemma = true;
4096 break;
4097 }
4098 }
4099 //infer
4100 if(!d_conflict) {
4101 memb = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, nf_x, inter_r) );
4102 memb_with_exps[memb] = exp;
4103 } else {
4104 break;
4105 }
4106 }
4107
4108 return addLemma;
4109 }
4110
4111 bool TheoryStrings::applyRConsume( CVC4::String &s, Node &r) {
4112 Trace("regexp-derivative") << "TheoryStrings::derivative: s=" << s << ", r= " << r << std::endl;
4113 Assert( d_regexp_opr.checkConstRegExp(r) );
4114
4115 if( !s.isEmptyString() ) {
4116 Node dc = r;
4117
4118 for(unsigned i=0; i<s.size(); ++i) {
4119 CVC4::String c = s.substr(i, 1);
4120 Node dc2;
4121 int rt = d_regexp_opr.derivativeS(dc, c, dc2);
4122 dc = dc2;
4123 if(rt == 0) {
4124 Unreachable();
4125 } else if(rt == 2) {
4126 return false;
4127 }
4128 }
4129 r = dc;
4130 }
4131
4132 return true;
4133 }
4134
4135 Node TheoryStrings::applyRSplit(Node s1, Node s2, Node r) {
4136 Assert(d_regexp_opr.checkConstRegExp(r));
4137
4138 std::vector< std::pair< Node, Node > > vec_can;
4139 d_regexp_opr.splitRegExp(r, vec_can);
4140 //TODO: lazy cache or eager?
4141 std::vector< Node > vec_or;
4142
4143 for(unsigned int i=0; i<vec_can.size(); i++) {
4144 Node m1 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s1, vec_can[i].first);
4145 Node m2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s2, vec_can[i].second);
4146 Node c = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, m1, m2) );
4147 vec_or.push_back( c );
4148 }
4149 Node conc = vec_or.size()==0? Node::null() : vec_or.size()==1 ? vec_or[0] : Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::OR, vec_or) );
4150 return conc;
4151 }
4152
4153 bool TheoryStrings::applyRLen(std::map< Node, std::vector< Node > > &XinR_with_exps) {
4154 if(XinR_with_exps.size() > 0) {
4155 //TODO: get vector, var, store.
4156 return true;
4157 } else {
4158 return false;
4159 }
4160 }
4161
4162 bool TheoryStrings::checkMembershipsWithoutLength(
4163 std::map< Node, std::vector< Node > > &memb_with_exps,
4164 std::map< Node, std::vector< Node > > &XinR_with_exps) {
4165 for(std::map< Node, std::vector< Node > >::iterator itr = memb_with_exps.begin(); itr != memb_with_exps.end(); ++itr) {
4166 Node memb = itr->first;
4167 Node s = memb[0];
4168 Node r = memb[1];
4169 if(s.isConst()) {
4170 memb = Rewriter::rewrite( memb );
4171 if(memb == d_false) {
4172 Node conc;
4173 sendInference(d_empty_vec, itr->second, conc, "MEMBERSHIP CONFLICT", true);
4174 //addLemma = true;
4175 return true;
4176 } else {
4177 Assert(memb == d_true);
4178 }
4179 } else if(s.getKind() == kind::VARIABLE) {
4180 //add to XinR
4181 XinR_with_exps[itr->first] = itr->second;
4182 } else {
4183 Assert(s.getKind() == kind::STRING_CONCAT);
4184 Node conc;
4185 for( unsigned i=0; i<s.getNumChildren(); i++ ) {
4186 if(s[i].isConst()) {
4187 CVC4::String str( s[0].getConst< String >() );
4188 //R-Consume, see Tianyi's thesis
4189 if(!applyRConsume(str, r)) {
4190 sendInference(d_empty_vec, itr->second, conc, "R-Consume CONFLICT", true);
4191 //addLemma = true;
4192 return true;
4193 }
4194 } else {
4195 //R-Split, see Tianyi's thesis
4196 if(i == s.getNumChildren() - 1) {
4197 //add to XinR
4198 Node memb2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s[i], r);
4199 XinR_with_exps[itr->first] = itr->second;
4200 } else {
4201 Node s1 = s[i];
4202 std::vector< Node > vec_s2;
4203 for( unsigned j=i+1; j<s.getNumChildren(); j++ ) {
4204 vec_s2.push_back(s[j]);
4205 }
4206 Node s2 = mkConcat(vec_s2);
4207 conc = applyRSplit(s1, s2, r);
4208 if(conc == d_true) {
4209 break;
4210 } else if(conc.isNull() || conc == d_false) {
4211 conc = Node::null();
4212 sendInference(d_empty_vec, itr->second, conc, "R-Split Conflict", true);
4213 //addLemma = true;
4214 return true;
4215 } else {
4216 sendInference(d_empty_vec, itr->second, conc, "R-Split", true);
4217 //addLemma = true;
4218 return true;
4219 }
4220 }
4221 }
4222 }
4223 }
4224 }
4225 return false;
4226 }
4227
4228 bool TheoryStrings::checkMemberships2() {
4229 bool addedLemma = false;
4230 d_nf_regexps.clear();
4231 d_nf_regexps_exp.clear();
4232 std::map< Node, std::vector< Node > > memb_with_exps;
4233 std::map< Node, std::vector< Node > > XinR_with_exps;
4234
4235 addedLemma = normalizePosMemberships( memb_with_exps );
4236 if(!d_conflict) {
4237 // main procedure
4238 addedLemma |= checkMembershipsWithoutLength( memb_with_exps, XinR_with_exps );
4239 //TODO: check addlemma
4240 if (!addedLemma && !d_conflict) {
4241 for(std::map< Node, std::vector< Node > >::const_iterator itr = XinR_with_exps.begin();
4242 itr != XinR_with_exps.end(); ++itr) {
4243 std::vector<Node> vec_or;
4244 d_regexp_opr.disjunctRegExp( itr->first, vec_or );
4245 Node tmp = NodeManager::currentNM()->mkNode(kind::REGEXP_UNION, vec_or);
4246 Trace("regexp-process") << "Got r: " << itr->first << " to " << tmp << std::endl;
4247 /*
4248 if(r.getKind() == kind::REGEXP_STAR) {
4249 //TODO: apply R-Len
4250 addedLemma = applyRLen(XinR_with_exps);
4251 } else {
4252 //TODO: split
4253 }
4254 */
4255 }
4256 Assert(false); //TODO:tmp
4257 }
4258 }
4259
4260 return addedLemma;
4261 }
4262
4263 void TheoryStrings::checkMemberships() {
4264 //add the memberships
4265 std::vector< Node > mems;
4266 d_extt->getActive( mems, kind::STRING_IN_REGEXP );
4267 for( unsigned i=0; i<mems.size(); i++ ){
4268 Node n = mems[i];
4269 Assert( d_extf_info_tmp.find( n )!=d_extf_info_tmp.end() );
4270 Assert( d_extf_info_tmp[n].d_pol==1 || d_extf_info_tmp[n].d_pol==-1 );
4271 bool pol = d_extf_info_tmp[n].d_pol==1;
4272 Trace("strings-process-debug") << " add membership : " << n << ", pol = " << pol << std::endl;
4273 addMembership( pol ? n : n.negate() );
4274 }
4275
4276 bool addedLemma = false;
4277 bool changed = false;
4278 std::vector< Node > processed;
4279 std::vector< Node > cprocessed;
4280
4281 Trace("regexp-debug") << "Checking Memberships ... " << std::endl;
4282 //if(options::stringEIT()) {
4283 //TODO: Opt for normal forms
4284 for( NodeIntMap::const_iterator itr_xr = d_pos_memberships.begin(); itr_xr != d_pos_memberships.end(); ++itr_xr ){
4285 bool spflag = false;
4286 Node x = (*itr_xr).first;
4287 Trace("regexp-debug") << "Checking Memberships for " << x << std::endl;
4288 if(d_inter_index.find(x) == d_inter_index.end()) {
4289 d_inter_index[x] = 0;
4290 }
4291 int cur_inter_idx = d_inter_index[x];
4292 unsigned n_pmem = (*itr_xr).second;
4293 Assert( getNumMemberships( x, true )==n_pmem );
4294 if( cur_inter_idx != (int)n_pmem ) {
4295 if( n_pmem == 1) {
4296 d_inter_cache[x] = getMembership( x, true, 0 );
4297 d_inter_index[x] = 1;
4298 Trace("regexp-debug") << "... only one choice " << std::endl;
4299 } else if(n_pmem > 1) {
4300 Node r;
4301 if(d_inter_cache.find(x) != d_inter_cache.end()) {
4302 r = d_inter_cache[x];
4303 }
4304 if(r.isNull()) {
4305 r = getMembership( x, true, 0 );
4306 cur_inter_idx = 1;
4307 }
4308
4309 unsigned k_start = cur_inter_idx;
4310 Trace("regexp-debug") << "... staring from : " << cur_inter_idx << ", we have " << n_pmem << std::endl;
4311 for(unsigned k = k_start; k<n_pmem; k++) {
4312 Node r2 = getMembership( x, true, k );
4313 r = d_regexp_opr.intersect(r, r2, spflag);
4314 if(spflag) {
4315 break;
4316 } else if(r == d_emptyRegexp) {
4317 std::vector< Node > vec_nodes;
4318 for( unsigned kk=0; kk<=k; kk++ ){
4319 Node rr = getMembership( x, true, kk );
4320 Node n = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x, rr);
4321 vec_nodes.push_back( n );
4322 }
4323 Node conc;
4324 sendInference(vec_nodes, conc, "INTERSECT CONFLICT", true);
4325 addedLemma = true;
4326 break;
4327 }
4328 if(d_conflict) {
4329 break;
4330 }
4331 }
4332 //updates
4333 if(!d_conflict && !spflag) {
4334 d_inter_cache[x] = r;
4335 d_inter_index[x] = (int)n_pmem;
4336 }
4337 }
4338 }
4339 }
4340 //}
4341
4342 Trace("regexp-debug") << "... No Intersect Conflict in Memberships, addedLemma: " << addedLemma << std::endl;
4343 if(!addedLemma) {
4344 for( unsigned i=0; i<d_regexp_memberships.size(); i++ ) {
4345 //check regular expression membership
4346 Node assertion = d_regexp_memberships[i];
4347 if( d_regexp_ucached.find(assertion) == d_regexp_ucached.end()
4348 && d_regexp_ccached.find(assertion) == d_regexp_ccached.end() ) {
4349 Trace("strings-regexp") << "We have regular expression assertion : " << assertion << std::endl;
4350 Node atom = assertion.getKind()==kind::NOT ? assertion[0] : assertion;
4351 bool polarity = assertion.getKind()!=kind::NOT;
4352 bool flag = true;
4353 Node x = atom[0];
4354 Node r = atom[1];
4355 std::vector< Node > rnfexp;
4356
4357 //if(options::stringOpt1()) {
4358 if(true){
4359 if(!x.isConst()) {
4360 x = getNormalString( x, rnfexp);
4361 changed = true;
4362 }
4363 if(!d_regexp_opr.checkConstRegExp(r)) {
4364 r = getNormalSymRegExp(r, rnfexp);
4365 changed = true;
4366 }
4367 Trace("strings-regexp-nf") << "Term " << atom << " is normalized to " << x << " IN " << r << std::endl;
4368 if(changed) {
4369 Node tmp = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x, r) );
4370 if(!polarity) {
4371 tmp = tmp.negate();
4372 }
4373 if(tmp == d_true) {
4374 d_regexp_ccached.insert(assertion);
4375 continue;
4376 } else if(tmp == d_false) {
4377 Node antec = mkRegExpAntec(assertion, mkExplain(rnfexp));
4378 Node conc = Node::null();
4379 sendLemma(antec, conc, "REGEXP NF Conflict");
4380 addedLemma = true;
4381 break;
4382 }
4383 }
4384 }
4385
4386 if( polarity ) {
4387 flag = checkPDerivative(x, r, atom, addedLemma, processed, cprocessed, rnfexp);
4388 if(options::stringOpt2() && flag) {
4389 if(d_regexp_opr.checkConstRegExp(r) && x.getKind()==kind::STRING_CONCAT) {
4390 std::vector< std::pair< Node, Node > > vec_can;
4391 d_regexp_opr.splitRegExp(r, vec_can);
4392 //TODO: lazy cache or eager?
4393 std::vector< Node > vec_or;
4394 std::vector< Node > vec_s2;
4395 for(unsigned int s2i=1; s2i<x.getNumChildren(); s2i++) {
4396 vec_s2.push_back(x[s2i]);
4397 }
4398 Node s1 = x[0];
4399 Node s2 = mkConcat(vec_s2);
4400 for(unsigned int i=0; i<vec_can.size(); i++) {
4401 Node m1 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s1, vec_can[i].first);
4402 Node m2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s2, vec_can[i].second);
4403 Node c = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, m1, m2) );
4404 vec_or.push_back( c );
4405 }
4406 Node conc = vec_or.size()==1 ? vec_or[0] : Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::OR, vec_or) );
4407 //Trace("regexp-split") << "R " << r << " to " << conc << std::endl;
4408 Node antec = mkRegExpAntec(atom, mkExplain(rnfexp));
4409 if(conc == d_true) {
4410 if(changed) {
4411 cprocessed.push_back( assertion );
4412 } else {
4413 processed.push_back( assertion );
4414 }
4415 } else {
4416 sendLemma(antec, conc, "RegExp-CST-SP");
4417 }
4418 addedLemma = true;
4419 flag = false;
4420 }
4421 }
4422 } else {
4423 if(! options::stringExp()) {
4424 throw LogicException("Strings Incomplete (due to Negative Membership) by default, try --strings-exp option.");
4425 }
4426 }
4427 if(flag) {
4428 //check if the term is atomic
4429 Node xr = getRepresentative( x );
4430 //Trace("strings-regexp") << xr << " is rep of " << x << std::endl;
4431 //Assert( d_normal_forms.find( xr )!=d_normal_forms.end() );
4432 //TODO
4433 if( true || r.getKind()!=kind::REGEXP_STAR || ( d_normal_forms[xr].size()==1 && x.getKind()!=kind::STRING_CONCAT ) ){
4434 Trace("strings-regexp") << "Unroll/simplify membership of atomic term " << xr << std::endl;
4435 //if so, do simple unrolling
4436 std::vector< Node > nvec;
4437
4438 /*if(xr.isConst()) {
4439 Node tmp = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, xr, r) );
4440 if(tmp==d_true || tmp==d_false) {
4441 if(!polarity) {
4442 tmp = tmp==d_true? d_false : d_true;
4443 }
4444 nvec.push_back( tmp );
4445 }
4446 }*/
4447
4448 if(nvec.empty()) {
4449 d_regexp_opr.simplify(atom, nvec, polarity);
4450 }
4451 Node antec = assertion;
4452 if(d_regexp_ant.find(assertion) != d_regexp_ant.end()) {
4453 antec = d_regexp_ant[assertion];
4454 for(std::vector< Node >::const_iterator itr=nvec.begin(); itr<nvec.end(); itr++) {
4455 if(itr->getKind() == kind::STRING_IN_REGEXP) {
4456 if(d_regexp_ant.find( *itr ) == d_regexp_ant.end()) {
4457 d_regexp_ant[ *itr ] = antec;
4458 }
4459 }
4460 }
4461 }
4462 antec = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, antec, mkExplain(rnfexp)) );
4463 Node conc = nvec.size()==1 ? nvec[0] : NodeManager::currentNM()->mkNode(kind::AND, nvec);
4464 conc = Rewriter::rewrite(conc);
4465 sendLemma( antec, conc, "REGEXP_Unfold" );
4466 addedLemma = true;
4467 if(changed) {
4468 cprocessed.push_back( assertion );
4469 } else {
4470 processed.push_back( assertion );
4471 }
4472 //d_regexp_ucached[assertion] = true;
4473 }else{
4474 Trace("strings-regexp") << "Unroll/simplify membership of non-atomic term " << xr << " = ";
4475 for( unsigned j=0; j<d_normal_forms[xr].size(); j++ ){
4476 Trace("strings-regexp") << d_normal_forms[xr][j] << " ";
4477 }
4478 Trace("strings-regexp") << ", polarity = " << polarity << std::endl;
4479 //otherwise, distribute unrolling over parts
4480 Node p1;
4481 Node p2;
4482 if( d_normal_forms[xr].size()>1 ){
4483 p1 = d_normal_forms[xr][0];
4484 std::vector< Node > cc;
4485 cc.insert( cc.begin(), d_normal_forms[xr].begin() + 1, d_normal_forms[xr].end() );
4486 p2 = mkConcat( cc );
4487 }
4488
4489 Trace("strings-regexp-debug") << "Construct antecedant..." << std::endl;
4490 std::vector< Node > antec;
4491 std::vector< Node > antecn;
4492 antec.insert( antec.begin(), d_normal_forms_exp[xr].begin(), d_normal_forms_exp[xr].end() );
4493 if( x!=xr ){
4494 antec.push_back( x.eqNode( xr ) );
4495 }
4496 antecn.push_back( assertion );
4497 Node ant = mkExplain( antec, antecn );
4498 Trace("strings-regexp-debug") << "Construct conclusion..." << std::endl;
4499 Node conc;
4500 if( polarity ){
4501 if( d_normal_forms[xr].size()==0 ){
4502 conc = d_true;
4503 }else if( d_normal_forms[xr].size()==1 ){
4504 Trace("strings-regexp-debug") << "Case 1\n";
4505 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, d_normal_forms[xr][0], r);
4506 }else{
4507 Trace("strings-regexp-debug") << "Case 2\n";
4508 std::vector< Node > conc_c;
4509 Node s11 = mkSkolemS( "s11" );
4510 Node s12 = mkSkolemS( "s12" );
4511 Node s21 = mkSkolemS( "s21" );
4512 Node s22 = mkSkolemS( "s22" );
4513 conc = p1.eqNode( mkConcat(s11, s12) );
4514 conc_c.push_back(conc);
4515 conc = p2.eqNode( mkConcat(s21, s22) );
4516 conc_c.push_back(conc);
4517 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s11, r);
4518 conc_c.push_back(conc);
4519 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, mkConcat(s12, s21), r[0]);
4520 conc_c.push_back(conc);
4521 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s22, r);
4522 conc_c.push_back(conc);
4523 conc = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, conc_c));
4524 Node eqz = Rewriter::rewrite(x.eqNode(d_emptyString));
4525 conc = NodeManager::currentNM()->mkNode(kind::OR, eqz, conc);
4526 d_pending_req_phase[eqz] = true;
4527 }
4528 }else{
4529 if( d_normal_forms[xr].size()==0 ){
4530 conc = d_false;
4531 }else if( d_normal_forms[xr].size()==1 ){
4532 Trace("strings-regexp-debug") << "Case 3\n";
4533 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, d_normal_forms[xr][0], r).negate();
4534 }else{
4535 Trace("strings-regexp-debug") << "Case 4\n";
4536 Node len1 = NodeManager::currentNM()->mkNode(kind::STRING_LENGTH, p1);
4537 Node len2 = NodeManager::currentNM()->mkNode(kind::STRING_LENGTH, p2);
4538 Node bi = NodeManager::currentNM()->mkBoundVar(NodeManager::currentNM()->integerType());
4539 Node bj = NodeManager::currentNM()->mkBoundVar(NodeManager::currentNM()->integerType());
4540 Node b1v = NodeManager::currentNM()->mkNode(kind::BOUND_VAR_LIST, bi, bj);
4541 Node g1 = NodeManager::currentNM()->mkNode(kind::AND,
4542 NodeManager::currentNM()->mkNode(kind::GEQ, bi, d_zero),
4543 NodeManager::currentNM()->mkNode(kind::GEQ, len1, bi),
4544 NodeManager::currentNM()->mkNode(kind::GEQ, bj, d_zero),
4545 NodeManager::currentNM()->mkNode(kind::GEQ, len2, bj));
4546 Node s11 = NodeManager::currentNM()->mkNode(kind::STRING_SUBSTR, p1, d_zero, bi);
4547 Node s12 = NodeManager::currentNM()->mkNode(kind::STRING_SUBSTR, p1, bi, NodeManager::currentNM()->mkNode(kind::MINUS, len1, bi));
4548 Node s21 = NodeManager::currentNM()->mkNode(kind::STRING_SUBSTR, p2, d_zero, bj);
4549 Node s22 = NodeManager::currentNM()->mkNode(kind::STRING_SUBSTR, p2, bj, NodeManager::currentNM()->mkNode(kind::MINUS, len2, bj));
4550 Node cc1 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s11, r).negate();
4551 Node cc2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, mkConcat(s12, s21), r[0]).negate();
4552 Node cc3 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s22, r).negate();
4553 conc = NodeManager::currentNM()->mkNode(kind::OR, cc1, cc2, cc3);
4554 conc = NodeManager::currentNM()->mkNode(kind::IMPLIES, g1, conc);
4555 conc = NodeManager::currentNM()->mkNode(kind::FORALL, b1v, conc);
4556 conc = NodeManager::currentNM()->mkNode(kind::AND, x.eqNode(d_emptyString).negate(), conc);
4557 }
4558 }
4559 if( conc!=d_true ){
4560 ant = mkRegExpAntec(assertion, ant);
4561 sendLemma(ant, conc, "REGEXP CSTAR");
4562 addedLemma = true;
4563 if( conc==d_false ){
4564 d_regexp_ccached.insert( assertion );
4565 }else{
4566 cprocessed.push_back( assertion );
4567 }
4568 }else{
4569 d_regexp_ccached.insert(assertion);
4570 }
4571 }
4572 }
4573 }
4574 if(d_conflict) {
4575 break;
4576 }
4577 }
4578 }
4579 if( addedLemma ) {
4580 if( !d_conflict ){
4581 for( unsigned i=0; i<processed.size(); i++ ) {
4582 d_regexp_ucached.insert(processed[i]);
4583 }
4584 for( unsigned i=0; i<cprocessed.size(); i++ ) {
4585 d_regexp_ccached.insert(cprocessed[i]);
4586 }
4587 }
4588 }
4589 }
4590
4591 bool TheoryStrings::checkPDerivative(Node x, Node r, Node atom, bool &addedLemma,
4592 std::vector< Node > &processed, std::vector< Node > &cprocessed, std::vector< Node > &nf_exp) {
4593
4594 Node antnf = mkExplain(nf_exp);
4595
4596 if(areEqual(x, d_emptyString)) {
4597 Node exp;
4598 switch(d_regexp_opr.delta(r, exp)) {
4599 case 0: {
4600 Node antec = mkRegExpAntec(atom, x.eqNode(d_emptyString));
4601 antec = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, antec, antnf));
4602 sendLemma(antec, exp, "RegExp Delta");
4603 addedLemma = true;
4604 d_regexp_ccached.insert(atom);
4605 return false;
4606 }
4607 case 1: {
4608 d_regexp_ccached.insert(atom);
4609 break;
4610 }
4611 case 2: {
4612 Node antec = mkRegExpAntec(atom, x.eqNode(d_emptyString));
4613 antec = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, antec, antnf));
4614 Node conc = Node::null();
4615 sendLemma(antec, conc, "RegExp Delta CONFLICT");
4616 addedLemma = true;
4617 d_regexp_ccached.insert(atom);
4618 return false;
4619 }
4620 default:
4621 //Impossible
4622 break;
4623 }
4624 } else {
4625 /*Node xr = getRepresentative( x );
4626 if(x != xr) {
4627 Node n = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, xr, r);
4628 Node nn = Rewriter::rewrite( n );
4629 if(nn == d_true) {
4630 d_regexp_ccached.insert(atom);
4631 return false;
4632 } else if(nn == d_false) {
4633 Node antec = mkRegExpAntec(atom, x.eqNode(xr));
4634 Node conc = Node::null();
4635 sendLemma(antec, conc, "RegExp Delta CONFLICT");
4636 addedLemma = true;
4637 d_regexp_ccached.insert(atom);
4638 return false;
4639 }
4640 }*/
4641 Node sREant = mkRegExpAntec(atom, d_true);
4642 sREant = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, sREant, antnf));
4643 if(deriveRegExp( x, r, sREant )) {
4644 addedLemma = true;
4645 processed.push_back( atom );
4646 return false;
4647 }
4648 }
4649 return true;
4650 }
4651
4652 CVC4::String TheoryStrings::getHeadConst( Node x ) {
4653 if( x.isConst() ) {
4654 return x.getConst< String >();
4655 } else if( x.getKind() == kind::STRING_CONCAT ) {
4656 if( x[0].isConst() ) {
4657 return x[0].getConst< String >();
4658 } else {
4659 return d_emptyString.getConst< String >();
4660 }
4661 } else {
4662 return d_emptyString.getConst< String >();
4663 }
4664 }
4665
4666 bool TheoryStrings::deriveRegExp( Node x, Node r, Node ant ) {
4667 // TODO cstr in vre
4668 Assert(x != d_emptyString);
4669 Trace("regexp-derive") << "TheoryStrings::deriveRegExp: x=" << x << ", r= " << r << std::endl;
4670 //if(x.isConst()) {
4671 // Node n = NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP, x, r );
4672 // Node r = Rewriter::rewrite( n );
4673 // if(n != r) {
4674 // sendLemma(ant, r, "REGEXP REWRITE");
4675 // return true;
4676 // }
4677 //}
4678 CVC4::String s = getHeadConst( x );
4679 if( !s.isEmptyString() && d_regexp_opr.checkConstRegExp( r ) ) {
4680 Node conc = Node::null();
4681 Node dc = r;
4682 bool flag = true;
4683 for(unsigned i=0; i<s.size(); ++i) {
4684 CVC4::String c = s.substr(i, 1);
4685 Node dc2;
4686 int rt = d_regexp_opr.derivativeS(dc, c, dc2);
4687 dc = dc2;
4688 if(rt == 0) {
4689 //TODO
4690 } else if(rt == 2) {
4691 // CONFLICT
4692 flag = false;
4693 break;
4694 }
4695 }
4696 // send lemma
4697 if(flag) {
4698 if(x.isConst()) {
4699 Assert(false, "Impossible: TheoryStrings::deriveRegExp: const string in const regular expression.");
4700 return false;
4701 } else {
4702 Assert( x.getKind() == kind::STRING_CONCAT );
4703 std::vector< Node > vec_nodes;
4704 for(unsigned int i=1; i<x.getNumChildren(); ++i ) {
4705 vec_nodes.push_back( x[i] );
4706 }
4707 Node left = mkConcat( vec_nodes );
4708 left = Rewriter::rewrite( left );
4709 conc = NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP, left, dc );
4710
4711 /*std::vector< Node > sdc;
4712 d_regexp_opr.simplify(conc, sdc, true);
4713 if(sdc.size() == 1) {
4714 conc = sdc[0];
4715 } else {
4716 conc = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, conc));
4717 }*/
4718 }
4719 }
4720 sendLemma(ant, conc, "RegExp-Derive");
4721 return true;
4722 } else {
4723 return false;
4724 }
4725 }
4726
4727 void TheoryStrings::addMembership(Node assertion) {
4728 bool polarity = assertion.getKind() != kind::NOT;
4729 TNode atom = polarity ? assertion : assertion[0];
4730 Node x = atom[0];
4731 Node r = atom[1];
4732 if(polarity) {
4733 int index = 0;
4734 NodeIntMap::const_iterator it = d_pos_memberships.find( x );
4735 if( it!=d_nf_pairs.end() ){
4736 index = (*it).second;
4737 for( int k=0; k<index; k++ ){
4738 if( k<(int)d_pos_memberships_data[x].size() ){
4739 if( d_pos_memberships_data[x][k]==r ){
4740 return;
4741 }
4742 }else{
4743 break;
4744 }
4745 }
4746 }
4747 d_pos_memberships[x] = index + 1;
4748 if( index<(int)d_pos_memberships_data[x].size() ){
4749 d_pos_memberships_data[x][index] = r;
4750 }else{
4751 d_pos_memberships_data[x].push_back( r );
4752 }
4753 } else if(!options::stringIgnNegMembership()) {
4754 /*if(options::stringEIT() && d_regexp_opr.checkConstRegExp(r)) {
4755 int rt;
4756 Node r2 = d_regexp_opr.complement(r, rt);
4757 Node a = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x, r2);
4758 }*/
4759 int index = 0;
4760 NodeIntMap::const_iterator it = d_neg_memberships.find( x );
4761 if( it!=d_nf_pairs.end() ){
4762 index = (*it).second;
4763 for( int k=0; k<index; k++ ){
4764 if( k<(int)d_neg_memberships_data[x].size() ){
4765 if( d_neg_memberships_data[x][k]==r ){
4766 return;
4767 }
4768 }else{
4769 break;
4770 }
4771 }
4772 }
4773 d_neg_memberships[x] = index + 1;
4774 if( index<(int)d_neg_memberships_data[x].size() ){
4775 d_neg_memberships_data[x][index] = r;
4776 }else{
4777 d_neg_memberships_data[x].push_back( r );
4778 }
4779 }
4780 // old
4781 if(polarity || !options::stringIgnNegMembership()) {
4782 d_regexp_memberships.push_back( assertion );
4783 }
4784 }
4785
4786 Node TheoryStrings::getNormalString( Node x, std::vector< Node >& nf_exp ){
4787 if( !x.isConst() ){
4788 Node xr = getRepresentative( x );
4789 if( d_normal_forms.find( xr ) != d_normal_forms.end() ){
4790 Node ret = mkConcat( d_normal_forms[xr] );
4791 nf_exp.insert( nf_exp.end(), d_normal_forms_exp[xr].begin(), d_normal_forms_exp[xr].end() );
4792 addToExplanation( x, d_normal_forms_base[xr], nf_exp );
4793 Trace("strings-debug") << "Term: " << x << " has a normal form " << ret << std::endl;
4794 return ret;
4795 } else {
4796 if(x.getKind() == kind::STRING_CONCAT) {
4797 std::vector< Node > vec_nodes;
4798 for(unsigned i=0; i<x.getNumChildren(); i++) {
4799 Node nc = getNormalString( x[i], nf_exp );
4800 vec_nodes.push_back( nc );
4801 }
4802 return mkConcat( vec_nodes );
4803 }
4804 }
4805 }
4806 return x;
4807 }
4808
4809 Node TheoryStrings::getNormalSymRegExp(Node r, std::vector<Node> &nf_exp) {
4810 Node ret = r;
4811 switch( r.getKind() ) {
4812 case kind::REGEXP_EMPTY:
4813 case kind::REGEXP_SIGMA:
4814 break;
4815 case kind::STRING_TO_REGEXP: {
4816 if(!r[0].isConst()) {
4817 Node tmp = getNormalString( r[0], nf_exp );
4818 if(tmp != r[0]) {
4819 ret = NodeManager::currentNM()->mkNode(kind::STRING_TO_REGEXP, tmp);
4820 }
4821 }
4822 break;
4823 }
4824 case kind::REGEXP_CONCAT: {
4825 std::vector< Node > vec_nodes;
4826 for(unsigned i=0; i<r.getNumChildren(); ++i) {
4827 vec_nodes.push_back( getNormalSymRegExp(r[i], nf_exp) );
4828 }
4829 ret = mkConcat(vec_nodes);
4830 break;
4831 }
4832 case kind::REGEXP_UNION: {
4833 std::vector< Node > vec_nodes;
4834 for(unsigned i=0; i<r.getNumChildren(); ++i) {
4835 vec_nodes.push_back( getNormalSymRegExp(r[i], nf_exp) );
4836 }
4837 ret = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::REGEXP_UNION, vec_nodes) );
4838 break;
4839 }
4840 case kind::REGEXP_INTER: {
4841 std::vector< Node > vec_nodes;
4842 for(unsigned i=0; i<r.getNumChildren(); ++i) {
4843 vec_nodes.push_back( getNormalSymRegExp(r[i], nf_exp) );
4844 }
4845 ret = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::REGEXP_INTER, vec_nodes) );
4846 break;
4847 }
4848 case kind::REGEXP_STAR: {
4849 ret = getNormalSymRegExp( r[0], nf_exp );
4850 ret = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::REGEXP_STAR, ret) );
4851 break;
4852 }
4853 //case kind::REGEXP_PLUS:
4854 //case kind::REGEXP_OPT:
4855 //case kind::REGEXP_RANGE:
4856 default: {
4857 Trace("strings-error") << "Unsupported term: " << r << " in normalization SymRegExp." << std::endl;
4858 Assert( false );
4859 //return Node::null();
4860 }
4861 }
4862 return ret;
4863 }
4864
4865 }/* CVC4::theory::strings namespace */
4866 }/* CVC4::theory namespace */
4867 }/* CVC4 namespace */