projects / cvc5.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Merge pull request #92 from timothy-king/travis-cpp11
[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 /////////////////////////////////////////////////////////////////////////////
335 // NOTIFICATIONS
336 /////////////////////////////////////////////////////////////////////////////
337
338
339 void TheoryStrings::presolve() {
340 Debug("strings-presolve") << "TheoryStrings::Presolving : get fmf options " << (options::stringFMF() ? "true" : "false") << std::endl;
341
342 if(!options::stdASCII()) {
343 d_card_size = 256;
344 }
345 }
346
347
348 /////////////////////////////////////////////////////////////////////////////
349 // MODEL GENERATION
350 /////////////////////////////////////////////////////////////////////////////
351
352
353 void TheoryStrings::collectModelInfo( TheoryModel* m, bool fullModel ) {
354 Trace("strings-model") << "TheoryStrings : Collect model info, fullModel = " << fullModel << std::endl;
355 Trace("strings-model") << "TheoryStrings : assertEqualityEngine." << std::endl;
356 m->assertEqualityEngine( &d_equalityEngine );
357 // Generate model
358 std::vector< Node > nodes;
359 getEquivalenceClasses( nodes );
360 std::map< Node, Node > processed;
361 std::vector< std::vector< Node > > col;
362 std::vector< Node > lts;
363 separateByLength( nodes, col, lts );
364 //step 1 : get all values for known lengths
365 std::vector< Node > lts_values;
366 std::map< unsigned, bool > values_used;
367 for( unsigned i=0; i<col.size(); i++ ) {
368 Trace("strings-model") << "Checking length for {";
369 for( unsigned j=0; j<col[i].size(); j++ ) {
370 if( j>0 ) {
371 Trace("strings-model") << ", ";
372 }
373 Trace("strings-model") << col[i][j];
374 }
375 Trace("strings-model") << " } (length is " << lts[i] << ")" << std::endl;
376 if( lts[i].isConst() ) {
377 lts_values.push_back( lts[i] );
378 Assert(lts[i].getConst<Rational>() <= RMAXINT, "Exceeded LONG_MAX in string model");
379 unsigned lvalue = lts[i].getConst<Rational>().getNumerator().toUnsignedInt();
380 values_used[ lvalue ] = true;
381 }else{
382 //get value for lts[i];
383 if( !lts[i].isNull() ){
384 Node v = d_valuation.getModelValue(lts[i]);
385 Trace("strings-model") << "Model value for " << lts[i] << " is " << v << std::endl;
386 lts_values.push_back( v );
387 Assert(v.getConst<Rational>() <= RMAXINT, "Exceeded LONG_MAX in string model");
388 unsigned lvalue = v.getConst<Rational>().getNumerator().toUnsignedInt();
389 values_used[ lvalue ] = true;
390 }else{
391 //Trace("strings-model-warn") << "No length for eqc " << col[i][0] << std::endl;
392 //Assert( false );
393 lts_values.push_back( Node::null() );
394 }
395 }
396 }
397 ////step 2 : assign arbitrary values for unknown lengths?
398 // confirmed by calculus invariant, see paper
399 Trace("strings-model") << "Assign to equivalence classes..." << std::endl;
400 //step 3 : assign values to equivalence classes that are pure variables
401 for( unsigned i=0; i<col.size(); i++ ){
402 std::vector< Node > pure_eq;
403 Trace("strings-model") << "The equivalence classes ";
404 for( unsigned j=0; j<col[i].size(); j++ ) {
405 Trace("strings-model") << col[i][j] << " ";
406 //check if col[i][j] has only variables
407 EqcInfo* ei = getOrMakeEqcInfo( col[i][j], false );
408 Node cst = ei ? ei->d_const_term : Node::null();
409 if( cst.isNull() ){
410 Assert( d_normal_forms.find( col[i][j] )!=d_normal_forms.end() );
411 if( d_normal_forms[col[i][j]].size()==1 ){//&& d_normal_forms[col[i][j]][0]==col[i][j] ){
412 pure_eq.push_back( col[i][j] );
413 }
414 }else{
415 processed[col[i][j]] = cst;
416 }
417 }
418 Trace("strings-model") << "have length " << lts_values[i] << std::endl;
419
420 //assign a new length if necessary
421 if( !pure_eq.empty() ){
422 if( lts_values[i].isNull() ){
423 unsigned lvalue = 0;
424 while( values_used.find( lvalue )!=values_used.end() ){
425 lvalue++;
426 }
427 Trace("strings-model") << "*** Decide to make length of " << lvalue << std::endl;
428 lts_values[i] = NodeManager::currentNM()->mkConst( Rational( lvalue ) );
429 values_used[ lvalue ] = true;
430 }
431 Trace("strings-model") << "Need to assign values of length " << lts_values[i] << " to equivalence classes ";
432 for( unsigned j=0; j<pure_eq.size(); j++ ){
433 Trace("strings-model") << pure_eq[j] << " ";
434 }
435 Trace("strings-model") << std::endl;
436
437
438 //use type enumerator
439 Assert(lts_values[i].getConst<Rational>() <= RMAXINT, "Exceeded LONG_MAX in string model");
440 StringEnumeratorLength sel(lts_values[i].getConst<Rational>().getNumerator().toUnsignedInt());
441 for( unsigned j=0; j<pure_eq.size(); j++ ){
442 Assert( !sel.isFinished() );
443 Node c = *sel;
444 while( d_equalityEngine.hasTerm( c ) ){
445 ++sel;
446 Assert( !sel.isFinished() );
447 c = *sel;
448 }
449 ++sel;
450 Trace("strings-model") << "*** Assigned constant " << c << " for " << pure_eq[j] << std::endl;
451 processed[pure_eq[j]] = c;
452 m->assertEquality( pure_eq[j], c, true );
453 }
454 }
455 }
456 Trace("strings-model") << "String Model : Pure Assigned." << std::endl;
457 //step 4 : assign constants to all other equivalence classes
458 for( unsigned i=0; i<nodes.size(); i++ ){
459 if( processed.find( nodes[i] )==processed.end() ){
460 Assert( d_normal_forms.find( nodes[i] )!=d_normal_forms.end() );
461 Trace("strings-model") << "Construct model for " << nodes[i] << " based on normal form ";
462 for( unsigned j=0; j<d_normal_forms[nodes[i]].size(); j++ ) {
463 if( j>0 ) Trace("strings-model") << " ++ ";
464 Trace("strings-model") << d_normal_forms[nodes[i]][j];
465 Node r = getRepresentative( d_normal_forms[nodes[i]][j] );
466 if( !r.isConst() && processed.find( r )==processed.end() ){
467 Trace("strings-model") << "(UNPROCESSED)";
468 }
469 }
470 Trace("strings-model") << std::endl;
471 std::vector< Node > nc;
472 for( unsigned j=0; j<d_normal_forms[nodes[i]].size(); j++ ) {
473 Node r = getRepresentative( d_normal_forms[nodes[i]][j] );
474 Assert( r.isConst() || processed.find( r )!=processed.end() );
475 nc.push_back(r.isConst() ? r : processed[r]);
476 }
477 Node cc = mkConcat( nc );
478 Assert( cc.getKind()==kind::CONST_STRING );
479 Trace("strings-model") << "*** Determined constant " << cc << " for " << nodes[i] << std::endl;
480 processed[nodes[i]] = cc;
481 m->assertEquality( nodes[i], cc, true );
482 }
483 }
484 //Trace("strings-model") << "String Model : Assigned." << std::endl;
485 Trace("strings-model") << "String Model : Finished." << std::endl;
486 }
487
488 /////////////////////////////////////////////////////////////////////////////
489 // MAIN SOLVER
490 /////////////////////////////////////////////////////////////////////////////
491
492
493 void TheoryStrings::preRegisterTerm(TNode n) {
494 if( d_pregistered_terms_cache.find(n) == d_pregistered_terms_cache.end() ) {
495 d_pregistered_terms_cache.insert(n);
496 //check for logic exceptions
497 if( !options::stringExp() ){
498 if( n.getKind()==kind::STRING_STRIDOF ||
499 n.getKind() == kind::STRING_ITOS || n.getKind() == kind::STRING_U16TOS || n.getKind() == kind::STRING_U32TOS ||
500 n.getKind() == kind::STRING_STOI || n.getKind() == kind::STRING_STOU16 || n.getKind() == kind::STRING_STOU32 ||
501 n.getKind() == kind::STRING_STRREPL || n.getKind() == kind::STRING_STRCTN ){
502 std::stringstream ss;
503 ss << "Term of kind " << n.getKind() << " not supported in default mode, try --strings-exp";
504 throw LogicException(ss.str());
505 }
506 }
507 switch( n.getKind() ) {
508 case kind::EQUAL: {
509 d_equalityEngine.addTriggerEquality(n);
510 break;
511 }
512 case kind::STRING_IN_REGEXP: {
513 d_out->requirePhase(n, true);
514 d_equalityEngine.addTriggerPredicate(n);
515 d_equalityEngine.addTerm(n[0]);
516 d_equalityEngine.addTerm(n[1]);
517 break;
518 }
519 default: {
520 TypeNode tn = n.getType();
521 if( tn.isString() ) {
522 registerTerm( n, 0 );
523 // FMF
524 if( n.getKind() == kind::VARIABLE && options::stringFMF() ){
525 d_input_vars.insert(n);
526 }
527 d_equalityEngine.addTerm(n);
528 } else if (tn.isBoolean()) {
529 // Get triggered for both equal and dis-equal
530 d_equalityEngine.addTriggerPredicate(n);
531 } else {
532 // Function applications/predicates
533 d_equalityEngine.addTerm(n);
534 if( options::stringExp() ){
535 //collect extended functions here: some may not be asserted to strings (such as those with return type Int),
536 // but we need to record them so they are treated properly
537 std::map< Node, bool > visited;
538 collectExtendedFuncTerms( n, visited );
539 }
540 }
541 //concat terms do not contribute to theory combination? TODO: verify
542 if( n.hasOperator() && kindToTheoryId( n.getKind() )==THEORY_STRINGS && n.getKind()!=kind::STRING_CONCAT ){
543 d_functionsTerms.push_back( n );
544 }
545 }
546 }
547 }
548 }
549
550 Node TheoryStrings::expandDefinition(LogicRequest &logicRequest, Node node) {
551 Trace("strings-exp-def") << "TheoryStrings::expandDefinition : " << node << std::endl;
552 return node;
553 }
554
555
556 void TheoryStrings::check(Effort e) {
557 if (done() && e<EFFORT_FULL) {
558 return;
559 }
560
561 TimerStat::CodeTimer checkTimer(d_checkTime);
562
563 bool polarity;
564 TNode atom;
565
566 /*if(getLogicInfo().hasEverything()) {
567 WarningOnce() << "WARNING: strings not supported in default configuration (ALL_SUPPORTED).\n"
568 << "To suppress this warning in the future use proper logic symbol, e.g. (set-logic QF_S)." << std::endl;
569 }
570 }*/
571
572 if( !done() && !hasTerm( d_emptyString ) ) {
573 preRegisterTerm( d_emptyString );
574 }
575
576 // Trace("strings-process") << "Theory of strings, check : " << e << std::endl;
577 Trace("strings-check") << "Theory of strings, check : " << e << std::endl;
578 while ( !done() && !d_conflict ) {
579 // Get all the assertions
580 Assertion assertion = get();
581 TNode fact = assertion.assertion;
582
583 Trace("strings-assertion") << "get assertion: " << fact << endl;
584 polarity = fact.getKind() != kind::NOT;
585 atom = polarity ? fact : fact[0];
586
587 //assert pending fact
588 assertPendingFact( atom, polarity, fact );
589 }
590 doPendingFacts();
591
592 if( !d_conflict && ( ( e == EFFORT_FULL && !d_valuation.needCheck() ) || ( e==EFFORT_STANDARD && options::stringEager() ) ) ) {
593 Trace("strings-check") << "Theory of strings full effort check " << std::endl;
594
595 if(Trace.isOn("strings-eqc")) {
596 for( unsigned t=0; t<2; t++ ) {
597 eq::EqClassesIterator eqcs2_i = eq::EqClassesIterator( &d_equalityEngine );
598 Trace("strings-eqc") << (t==0 ? "STRINGS:" : "OTHER:") << std::endl;
599 while( !eqcs2_i.isFinished() ){
600 Node eqc = (*eqcs2_i);
601 bool print = (t==0 && eqc.getType().isString() ) || (t==1 && !eqc.getType().isString() );
602 if (print) {
603 eq::EqClassIterator eqc2_i = eq::EqClassIterator( eqc, &d_equalityEngine );
604 Trace("strings-eqc") << "Eqc( " << eqc << " ) : { ";
605 while( !eqc2_i.isFinished() ) {
606 if( (*eqc2_i)!=eqc && (*eqc2_i).getKind()!=kind::EQUAL ){
607 Trace("strings-eqc") << (*eqc2_i) << " ";
608 }
609 ++eqc2_i;
610 }
611 Trace("strings-eqc") << " } " << std::endl;
612 EqcInfo * ei = getOrMakeEqcInfo( eqc, false );
613 if( ei ){
614 Trace("strings-eqc-debug") << "* Length term : " << ei->d_length_term.get() << std::endl;
615 Trace("strings-eqc-debug") << "* Cardinality lemma k : " << ei->d_cardinality_lem_k.get() << std::endl;
616 Trace("strings-eqc-debug") << "* Normalization length lemma : " << ei->d_normalized_length.get() << std::endl;
617 }
618 }
619 ++eqcs2_i;
620 }
621 Trace("strings-eqc") << std::endl;
622 }
623 Trace("strings-eqc") << std::endl;
624 }
625
626 bool addedLemma = false;
627 bool addedFact;
628 do{
629 Trace("strings-process") << "----check, next round---" << std::endl;
630 checkInit();
631 Trace("strings-process") << "Done check init, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
632 if( !hasProcessed() ){
633 checkExtfEval();
634 Trace("strings-process") << "Done check extended functions eval, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
635 if( !hasProcessed() ){
636 checkFlatForms();
637 Trace("strings-process") << "Done check flat forms, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
638 if( !hasProcessed() && e==EFFORT_FULL ){
639 checkNormalForms();
640 Trace("strings-process") << "Done check normal forms, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
641 if( !hasProcessed() ){
642 if( options::stringEagerLen() ){
643 checkLengthsEqc();
644 Trace("strings-process") << "Done check lengths, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
645 }
646 if( !hasProcessed() ){
647 if( options::stringExp() && !options::stringGuessModel() ){
648 checkExtfReductions( 2 );
649 Trace("strings-process") << "Done check extended functions reduction 2, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
650 }
651 if( !hasProcessed() ){
652 checkMemberships();
653 Trace("strings-process") << "Done check memberships, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
654 if( !hasProcessed() ){
655 checkCardinality();
656 Trace("strings-process") << "Done check cardinality, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
657 }
658 }
659 }
660 }
661 }
662 }
663 }
664 //flush the facts
665 addedFact = !d_pending.empty();
666 addedLemma = !d_lemma_cache.empty();
667 doPendingFacts();
668 doPendingLemmas();
669 }while( !d_conflict && !addedLemma && addedFact );
670
671 Trace("strings-check") << "Theory of strings done full effort check " << addedLemma << " " << d_conflict << std::endl;
672 }else if( e==EFFORT_LAST_CALL ){
673 Assert( !hasProcessed() );
674 Trace("strings-check") << "Theory of strings last call effort check " << std::endl;
675 checkExtfEval( 3 );
676 checkExtfReductions( 2 );
677 doPendingFacts();
678 doPendingLemmas();
679 Trace("strings-process") << "Done check extended functions reduction 2, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
680 }
681 Trace("strings-check") << "Theory of strings, done check : " << e << std::endl;
682 Assert( d_pending.empty() );
683 Assert( d_lemma_cache.empty() );
684 }
685
686 bool TheoryStrings::needsCheckLastEffort() {
687 if( options::stringGuessModel() ){
688 return d_has_extf.get();
689 }else{
690 return false;
691 }
692 }
693
694 void TheoryStrings::checkExtfReductions( int effort ) {
695 std::vector< Node > extf;
696 d_extt->getActive( extf );
697 for( unsigned i=0; i<extf.size(); i++ ){
698 Node n = extf[i];
699 if( d_extf_info_tmp[n].d_model_active ){
700 Assert( d_extf_info_tmp.find( n )!=d_extf_info_tmp.end() );
701 if( checkExtfReduction( n, d_extf_info_tmp[n].d_pol, effort ) ){
702 d_extt->markReduced( n );
703 }
704 if( hasProcessed() ){
705 return;
706 }
707 }
708 }
709 }
710
711 bool TheoryStrings::checkExtfReduction( Node atom, int pol, int effort ) {
712 //determine the effort level to process the extf at
713 // 0 - at assertion time, 1+ - after no other reduction is applicable
714 int r_effort = -1;
715 if( atom.getKind()==kind::STRING_STRCTN ){
716 if( pol==1 ){
717 r_effort = 1;
718 }else{
719 Assert( pol==-1 );
720 if( effort==2 ){
721 Node x = atom[0];
722 Node s = atom[1];
723 std::vector< Node > lexp;
724 Node lenx = getLength( x, lexp );
725 Node lens = getLength( s, lexp );
726 if( areEqual( lenx, lens ) ){
727 Trace("strings-extf-debug") << " resolve extf : " << atom << " based on equal lengths disequality." << std::endl;
728 //we can reduce to disequality when lengths are equal
729 if( !areDisequal( x, s ) ){
730 lexp.push_back( lenx.eqNode(lens) );
731 lexp.push_back( atom.negate() );
732 Node xneqs = x.eqNode(s).negate();
733 sendInference( lexp, xneqs, "NEG-CTN-EQL", true );
734 }
735 return true;
736 }else if( !areDisequal( lenx, lens ) ){
737 //split on their lenths
738 sendSplit( lenx, lens, "NEG-CTN-SP" );
739 }else{
740 r_effort = 2;
741 }
742 }
743 }
744 }else{
745 if( options::stringLazyPreproc() ){
746 if( atom.getKind()==kind::STRING_SUBSTR ){
747 r_effort = 1;
748 }else if( atom.getKind()!=kind::STRING_IN_REGEXP ){
749 r_effort = 2;
750 }
751 }
752 }
753 if( effort==r_effort ){
754 Node c_atom = pol==-1 ? atom.negate() : atom;
755 if( d_preproc_cache.find( c_atom )==d_preproc_cache.end() ){
756 d_preproc_cache[ c_atom ] = true;
757 Trace("strings-process-debug") << "Process reduction for " << atom << ", pol = " << pol << std::endl;
758 if( atom.getKind()==kind::STRING_STRCTN && pol==1 ){
759 Node x = atom[0];
760 Node s = atom[1];
761 //positive contains reduces to a equality
762 Node sk1 = mkSkolemCached( x, s, sk_id_ctn_pre, "sc1" );
763 Node sk2 = mkSkolemCached( x, s, sk_id_ctn_post, "sc2" );
764 Node eq = Rewriter::rewrite( x.eqNode( mkConcat( sk1, s, sk2 ) ) );
765 std::vector< Node > exp_vec;
766 exp_vec.push_back( atom );
767 sendInference( d_empty_vec, exp_vec, eq, "POS-CTN", true );
768 //we've reduced this atom
769 Trace("strings-extf-debug") << " resolve extf : " << atom << " based on positive contain reduction." << std::endl;
770 return true;
771 }else{
772 // for STRING_SUBSTR, STRING_STRCTN with pol=-1,
773 // STRING_STRIDOF, STRING_ITOS, STRING_U16TOS, STRING_U32TOS, STRING_STOI, STRING_STOU16, STRING_STOU32, STRING_STRREPL
774 std::vector< Node > new_nodes;
775 Node res = d_preproc.simplify( atom, new_nodes );
776 Assert( res!=atom );
777 new_nodes.push_back( NodeManager::currentNM()->mkNode( res.getType().isBoolean() ? kind::IFF : kind::EQUAL, res, atom ) );
778 Node nnlem = new_nodes.size()==1 ? new_nodes[0] : NodeManager::currentNM()->mkNode( kind::AND, new_nodes );
779 nnlem = Rewriter::rewrite( nnlem );
780 Trace("strings-red-lemma") << "Reduction_" << effort << " lemma : " << nnlem << std::endl;
781 Trace("strings-red-lemma") << "...from " << atom << std::endl;
782 sendInference( d_empty_vec, nnlem, "Reduction", true );
783 //we've reduced this atom
784 Trace("strings-extf-debug") << " resolve extf : " << atom << " based on reduction." << std::endl;
785 return true;
786 }
787 }
788 }
789 return false;
790 }
791
792 TheoryStrings::EqcInfo::EqcInfo( context::Context* c ) : d_const_term(c), d_length_term(c), d_cardinality_lem_k(c), d_normalized_length(c) {
793
794 }
795
796 TheoryStrings::EqcInfo * TheoryStrings::getOrMakeEqcInfo( Node eqc, bool doMake ) {
797 std::map< Node, EqcInfo* >::iterator eqc_i = d_eqc_info.find( eqc );
798 if( eqc_i==d_eqc_info.end() ){
799 if( doMake ){
800 EqcInfo* ei = new EqcInfo( getSatContext() );
801 d_eqc_info[eqc] = ei;
802 return ei;
803 }else{
804 return NULL;
805 }
806 }else{
807 return (*eqc_i).second;
808 }
809 }
810
811
812 /** Conflict when merging two constants */
813 void TheoryStrings::conflict(TNode a, TNode b){
814 if( !d_conflict ){
815 Debug("strings-conflict") << "Making conflict..." << std::endl;
816 d_conflict = true;
817 Node conflictNode;
818 if (a.getKind() == kind::CONST_BOOLEAN) {
819 conflictNode = explain( a.iffNode(b) );
820 } else {
821 conflictNode = explain( a.eqNode(b) );
822 }
823 Trace("strings-conflict") << "CONFLICT: Eq engine conflict : " << conflictNode << std::endl;
824 d_out->conflict( conflictNode );
825 }
826 }
827
828 /** called when a new equivalance class is created */
829 void TheoryStrings::eqNotifyNewClass(TNode t){
830 if( t.getKind() == kind::CONST_STRING ){
831 EqcInfo * ei =getOrMakeEqcInfo( t, true );
832 ei->d_const_term = t;
833 }
834 if( t.getKind() == kind::STRING_LENGTH ){
835 Trace("strings-debug") << "New length eqc : " << t << std::endl;
836 Node r = d_equalityEngine.getRepresentative(t[0]);
837 EqcInfo * ei = getOrMakeEqcInfo( r, true );
838 ei->d_length_term = t[0];
839 //we care about the length of this string
840 registerTerm( t[0], 1 );
841 }
842 }
843
844 /** called when two equivalance classes will merge */
845 void TheoryStrings::eqNotifyPreMerge(TNode t1, TNode t2){
846 EqcInfo * e2 = getOrMakeEqcInfo(t2, false);
847 if( e2 ){
848 EqcInfo * e1 = getOrMakeEqcInfo( t1 );
849 //add information from e2 to e1
850 if( !e2->d_const_term.get().isNull() ){
851 e1->d_const_term.set( e2->d_const_term );
852 }
853 if( !e2->d_length_term.get().isNull() ){
854 e1->d_length_term.set( e2->d_length_term );
855 }
856 if( e2->d_cardinality_lem_k.get()>e1->d_cardinality_lem_k.get() ) {
857 e1->d_cardinality_lem_k.set( e2->d_cardinality_lem_k );
858 }
859 if( !e2->d_normalized_length.get().isNull() ){
860 e1->d_normalized_length.set( e2->d_normalized_length );
861 }
862 }
863 }
864
865 /** called when two equivalance classes have merged */
866 void TheoryStrings::eqNotifyPostMerge(TNode t1, TNode t2) {
867
868 }
869
870 /** called when two equivalance classes are disequal */
871 void TheoryStrings::eqNotifyDisequal(TNode t1, TNode t2, TNode reason) {
872 if( t1.getType().isString() ){
873 //store disequalities between strings, may need to check if their lengths are equal/disequal
874 d_ee_disequalities.push_back( t1.eqNode( t2 ) );
875 }
876 }
877
878 void TheoryStrings::addCarePairs( quantifiers::TermArgTrie * t1, quantifiers::TermArgTrie * t2, unsigned arity, unsigned depth ) {
879 if( depth==arity ){
880 if( t2!=NULL ){
881 Node f1 = t1->getNodeData();
882 Node f2 = t2->getNodeData();
883 if( !d_equalityEngine.areEqual( f1, f2 ) ){
884 Trace("strings-cg-debug") << "TheoryStrings::computeCareGraph(): checking function " << f1 << " and " << f2 << std::endl;
885 vector< pair<TNode, TNode> > currentPairs;
886 for (unsigned k = 0; k < f1.getNumChildren(); ++ k) {
887 TNode x = f1[k];
888 TNode y = f2[k];
889 Assert( d_equalityEngine.hasTerm(x) );
890 Assert( d_equalityEngine.hasTerm(y) );
891 Assert( !d_equalityEngine.areDisequal( x, y, false ) );
892 if( !d_equalityEngine.areEqual( x, y ) ){
893 if( d_equalityEngine.isTriggerTerm(x, THEORY_STRINGS) && d_equalityEngine.isTriggerTerm(y, THEORY_STRINGS) ){
894 TNode x_shared = d_equalityEngine.getTriggerTermRepresentative(x, THEORY_STRINGS);
895 TNode y_shared = d_equalityEngine.getTriggerTermRepresentative(y, THEORY_STRINGS);
896 EqualityStatus eqStatus = d_valuation.getEqualityStatus(x_shared, y_shared);
897 if( eqStatus==EQUALITY_FALSE_AND_PROPAGATED || eqStatus==EQUALITY_FALSE || eqStatus==EQUALITY_FALSE_IN_MODEL ){
898 //an argument is disequal, we are done
899 return;
900 }else{
901 currentPairs.push_back(make_pair(x_shared, y_shared));
902 }
903 }
904 }
905 }
906 for (unsigned c = 0; c < currentPairs.size(); ++ c) {
907 Trace("strings-cg-pair") << "TheoryStrings::computeCareGraph(): pair : " << currentPairs[c].first << " " << currentPairs[c].second << std::endl;
908 addCarePair(currentPairs[c].first, currentPairs[c].second);
909 }
910 }
911 }
912 }else{
913 if( t2==NULL ){
914 if( depth<(arity-1) ){
915 //add care pairs internal to each child
916 for( std::map< TNode, quantifiers::TermArgTrie >::iterator it = t1->d_data.begin(); it != t1->d_data.end(); ++it ){
917 addCarePairs( &it->second, NULL, arity, depth+1 );
918 }
919 }
920 //add care pairs based on each pair of non-disequal arguments
921 for( std::map< TNode, quantifiers::TermArgTrie >::iterator it = t1->d_data.begin(); it != t1->d_data.end(); ++it ){
922 std::map< TNode, quantifiers::TermArgTrie >::iterator it2 = it;
923 ++it2;
924 for( ; it2 != t1->d_data.end(); ++it2 ){
925 if( !d_equalityEngine.areDisequal(it->first, it2->first, false) ){
926 addCarePairs( &it->second, &it2->second, arity, depth+1 );
927 }
928 }
929 }
930 }else{
931 //add care pairs based on product of indices, non-disequal arguments
932 for( std::map< TNode, quantifiers::TermArgTrie >::iterator it = t1->d_data.begin(); it != t1->d_data.end(); ++it ){
933 for( std::map< TNode, quantifiers::TermArgTrie >::iterator it2 = t2->d_data.begin(); it2 != t2->d_data.end(); ++it2 ){
934 if( !d_equalityEngine.areDisequal(it->first, it2->first, false) ){
935 addCarePairs( &it->second, &it2->second, arity, depth+1 );
936 }
937 }
938 }
939 }
940 }
941 }
942
943 void TheoryStrings::computeCareGraph(){
944 //computing the care graph here is probably still necessary, due to operators that take non-string arguments TODO: verify
945 Trace("strings-cg") << "TheoryStrings::computeCareGraph(): Build term indices..." << std::endl;
946 std::map< Node, quantifiers::TermArgTrie > index;
947 std::map< Node, unsigned > arity;
948 unsigned functionTerms = d_functionsTerms.size();
949 for (unsigned i = 0; i < functionTerms; ++ i) {
950 TNode f1 = d_functionsTerms[i];
951 Trace("strings-cg") << "...build for " << f1 << std::endl;
952 Node op = f1.getOperator();
953 std::vector< TNode > reps;
954 bool has_trigger_arg = false;
955 for( unsigned j=0; j<f1.getNumChildren(); j++ ){
956 reps.push_back( d_equalityEngine.getRepresentative( f1[j] ) );
957 if( d_equalityEngine.isTriggerTerm( f1[j], THEORY_STRINGS ) ){
958 has_trigger_arg = true;
959 }
960 }
961 if( has_trigger_arg ){
962 index[op].addTerm( f1, reps );
963 arity[op] = reps.size();
964 }
965 }
966 //for each index
967 for( std::map< Node, quantifiers::TermArgTrie >::iterator itii = index.begin(); itii != index.end(); ++itii ){
968 Trace("strings-cg") << "TheoryStrings::computeCareGraph(): Process index " << itii->first << "..." << std::endl;
969 addCarePairs( &itii->second, NULL, arity[ itii->first ], 0 );
970 }
971 }
972
973 void TheoryStrings::assertPendingFact(Node atom, bool polarity, Node exp) {
974 Trace("strings-pending") << "Assert pending fact : " << atom << " " << polarity << " from " << exp << std::endl;
975 Assert(atom.getKind() != kind::OR, "Infer error: a split.");
976 if( atom.getKind()==kind::EQUAL ){
977 Trace("strings-pending-debug") << " Register term" << std::endl;
978 for( unsigned j=0; j<2; j++ ) {
979 if( !d_equalityEngine.hasTerm( atom[j] ) && atom[j].getType().isString() ) {
980 registerTerm( atom[j], 0 );
981 }
982 }
983 Trace("strings-pending-debug") << " Now assert equality" << std::endl;
984 d_equalityEngine.assertEquality( atom, polarity, exp );
985 Trace("strings-pending-debug") << " Finished assert equality" << std::endl;
986 } else {
987 d_equalityEngine.assertPredicate( atom, polarity, exp );
988 //process extf
989 if( atom.getKind()==kind::STRING_IN_REGEXP ){
990 d_extt->registerTerm( atom );
991 if( polarity && atom[1].getKind()==kind::REGEXP_RANGE ){
992 if( d_extf_infer_cache_u.find( atom )==d_extf_infer_cache_u.end() ){
993 d_extf_infer_cache_u.insert( atom );
994 //length of first argument is one
995 Node conc = d_one.eqNode( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, atom[0] ) );
996 Node lem = NodeManager::currentNM()->mkNode( kind::OR, atom.negate(), conc );
997 Trace("strings-lemma") << "Strings::Lemma RE-Range-Len : " << lem << std::endl;
998 d_out->lemma( lem );
999 }
1000 }
1001 }
1002 }
1003 Trace("strings-pending-debug") << " Now collect terms" << std::endl;
1004 //collect extended function terms in the atom
1005 std::map< Node, bool > visited;
1006 collectExtendedFuncTerms( atom, visited );
1007 Trace("strings-pending-debug") << " Finished collect terms" << std::endl;
1008 }
1009
1010 void TheoryStrings::doPendingFacts() {
1011 size_t i=0;
1012 while( !d_conflict && i<d_pending.size() ) {
1013 Node fact = d_pending[i];
1014 Node exp = d_pending_exp[ fact ];
1015 if(fact.getKind() == kind::AND) {
1016 for(size_t j=0; j<fact.getNumChildren(); j++) {
1017 bool polarity = fact[j].getKind() != kind::NOT;
1018 TNode atom = polarity ? fact[j] : fact[j][0];
1019 assertPendingFact(atom, polarity, exp);
1020 }
1021 } else {
1022 bool polarity = fact.getKind() != kind::NOT;
1023 TNode atom = polarity ? fact : fact[0];
1024 assertPendingFact(atom, polarity, exp);
1025 }
1026 i++;
1027 }
1028 d_pending.clear();
1029 d_pending_exp.clear();
1030 }
1031
1032 void TheoryStrings::doPendingLemmas() {
1033 if( !d_conflict && !d_lemma_cache.empty() ){
1034 for( unsigned i=0; i<d_lemma_cache.size(); i++ ){
1035 Trace("strings-pending") << "Process pending lemma : " << d_lemma_cache[i] << std::endl;
1036 d_out->lemma( d_lemma_cache[i] );
1037 }
1038 for( std::map< Node, bool >::iterator it = d_pending_req_phase.begin(); it != d_pending_req_phase.end(); ++it ){
1039 Trace("strings-pending") << "Require phase : " << it->first << ", polarity = " << it->second << std::endl;
1040 d_out->requirePhase( it->first, it->second );
1041 }
1042 }
1043 d_lemma_cache.clear();
1044 d_pending_req_phase.clear();
1045 }
1046
1047 bool TheoryStrings::hasProcessed() {
1048 return d_conflict || !d_lemma_cache.empty() || !d_pending.empty();
1049 }
1050
1051 void TheoryStrings::addToExplanation( Node a, Node b, std::vector< Node >& exp ) {
1052 if( a!=b ){
1053 Debug("strings-explain") << "Add to explanation : " << a << " == " << b << std::endl;
1054 Assert( areEqual( a, b ) );
1055 exp.push_back( a.eqNode( b ) );
1056 }
1057 }
1058
1059 void TheoryStrings::addToExplanation( Node lit, std::vector< Node >& exp ) {
1060 if( !lit.isNull() ){
1061 exp.push_back( lit );
1062 }
1063 }
1064
1065 void TheoryStrings::checkInit() {
1066 //build term index
1067 d_eqc_to_const.clear();
1068 d_eqc_to_const_base.clear();
1069 d_eqc_to_const_exp.clear();
1070 d_eqc_to_len_term.clear();
1071 d_term_index.clear();
1072 d_strings_eqc.clear();
1073
1074 std::map< Kind, unsigned > ncongruent;
1075 std::map< Kind, unsigned > congruent;
1076 d_emptyString_r = getRepresentative( d_emptyString );
1077 eq::EqClassesIterator eqcs_i = eq::EqClassesIterator( &d_equalityEngine );
1078 while( !eqcs_i.isFinished() ){
1079 Node eqc = (*eqcs_i);
1080 TypeNode tn = eqc.getType();
1081 if( !tn.isRegExp() ){
1082 if( tn.isString() ){
1083 d_strings_eqc.push_back( eqc );
1084 }
1085 Node var;
1086 eq::EqClassIterator eqc_i = eq::EqClassIterator( eqc, &d_equalityEngine );
1087 while( !eqc_i.isFinished() ) {
1088 Node n = *eqc_i;
1089 if( n.isConst() ){
1090 d_eqc_to_const[eqc] = n;
1091 d_eqc_to_const_base[eqc] = n;
1092 d_eqc_to_const_exp[eqc] = Node::null();
1093 }else if( tn.isInteger() ){
1094 if( n.getKind()==kind::STRING_LENGTH ){
1095 Node nr = getRepresentative( n[0] );
1096 d_eqc_to_len_term[nr] = n[0];
1097 }
1098 }else if( n.getNumChildren()>0 ){
1099 Kind k = n.getKind();
1100 if( k!=kind::EQUAL ){
1101 if( d_congruent.find( n )==d_congruent.end() ){
1102 std::vector< Node > c;
1103 Node nc = d_term_index[k].add( n, 0, this, d_emptyString_r, c );
1104 if( nc!=n ){
1105 //check if we have inferred a new equality by removal of empty components
1106 if( n.getKind()==kind::STRING_CONCAT && !areEqual( nc, n ) ){
1107 std::vector< Node > exp;
1108 unsigned count[2] = { 0, 0 };
1109 while( count[0]<nc.getNumChildren() || count[1]<n.getNumChildren() ){
1110 //explain empty prefixes
1111 for( unsigned t=0; t<2; t++ ){
1112 Node nn = t==0 ? nc : n;
1113 while( count[t]<nn.getNumChildren() &&
1114 ( nn[count[t]]==d_emptyString || areEqual( nn[count[t]], d_emptyString ) ) ){
1115 if( nn[count[t]]!=d_emptyString ){
1116 exp.push_back( nn[count[t]].eqNode( d_emptyString ) );
1117 }
1118 count[t]++;
1119 }
1120 }
1121 //explain equal components
1122 if( count[0]<nc.getNumChildren() ){
1123 Assert( count[1]<n.getNumChildren() );
1124 if( nc[count[0]]!=n[count[1]] ){
1125 exp.push_back( nc[count[0]].eqNode( n[count[1]] ) );
1126 }
1127 count[0]++;
1128 count[1]++;
1129 }
1130 }
1131 //infer the equality
1132 sendInference( exp, n.eqNode( nc ), "I_Norm" );
1133 }else{
1134 //mark as congruent : only process if neither has been reduced
1135 d_extt->markCongruent( nc, n );
1136 }
1137 //this node is congruent to another one, we can ignore it
1138 Trace("strings-process-debug") << " congruent term : " << n << std::endl;
1139 d_congruent.insert( n );
1140 congruent[k]++;
1141 }else if( k==kind::STRING_CONCAT && c.size()==1 ){
1142 Trace("strings-process-debug") << " congruent term by singular : " << n << " " << c[0] << std::endl;
1143 //singular case
1144 if( !areEqual( c[0], n ) ){
1145 std::vector< Node > exp;
1146 //explain empty components
1147 bool foundNEmpty = false;
1148 for( unsigned i=0; i<n.getNumChildren(); i++ ){
1149 if( areEqual( n[i], d_emptyString ) ){
1150 if( n[i]!=d_emptyString ){
1151 exp.push_back( n[i].eqNode( d_emptyString ) );
1152 }
1153 }else{
1154 Assert( !foundNEmpty );
1155 if( n[i]!=c[0] ){
1156 exp.push_back( n[i].eqNode( c[0] ) );
1157 }
1158 foundNEmpty = true;
1159 }
1160 }
1161 AlwaysAssert( foundNEmpty );
1162 //infer the equality
1163 sendInference( exp, n.eqNode( c[0] ), "I_Norm_S" );
1164 }
1165 d_congruent.insert( n );
1166 congruent[k]++;
1167 }else{
1168 ncongruent[k]++;
1169 }
1170 }else{
1171 congruent[k]++;
1172 }
1173 }
1174 }else{
1175 if( d_congruent.find( n )==d_congruent.end() ){
1176 if( var.isNull() ){
1177 var = n;
1178 }else{
1179 Trace("strings-process-debug") << " congruent variable : " << n << std::endl;
1180 d_congruent.insert( n );
1181 }
1182 }
1183 }
1184 ++eqc_i;
1185 }
1186 }
1187 ++eqcs_i;
1188 }
1189 if( Trace.isOn("strings-process") ){
1190 for( std::map< Kind, TermIndex >::iterator it = d_term_index.begin(); it != d_term_index.end(); ++it ){
1191 Trace("strings-process") << " Terms[" << it->first << "] = " << ncongruent[it->first] << "/" << (congruent[it->first]+ncongruent[it->first]) << std::endl;
1192 }
1193 }
1194 Trace("strings-process") << "Done check init, addedLemma = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
1195 //now, infer constants for equivalence classes
1196 if( !hasProcessed() ){
1197 //do fixed point
1198 unsigned prevSize;
1199 do{
1200 Trace("strings-process-debug") << "Check constant equivalence classes..." << std::endl;
1201 prevSize = d_eqc_to_const.size();
1202 std::vector< Node > vecc;
1203 checkConstantEquivalenceClasses( &d_term_index[kind::STRING_CONCAT], vecc );
1204 }while( !hasProcessed() && d_eqc_to_const.size()>prevSize );
1205 Trace("strings-process") << "Done check constant equivalence classes, addedLemma = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
1206 }
1207 }
1208
1209 void TheoryStrings::checkConstantEquivalenceClasses( TermIndex* ti, std::vector< Node >& vecc ) {
1210 Node n = ti->d_data;
1211 if( !n.isNull() ){
1212 //construct the constant
1213 Node c = mkConcat( vecc );
1214 if( !areEqual( n, c ) ){
1215 Trace("strings-debug") << "Constant eqc : " << c << " for " << n << std::endl;
1216 Trace("strings-debug") << " ";
1217 for( unsigned i=0; i<vecc.size(); i++ ){
1218 Trace("strings-debug") << vecc[i] << " ";
1219 }
1220 Trace("strings-debug") << std::endl;
1221 unsigned count = 0;
1222 unsigned countc = 0;
1223 std::vector< Node > exp;
1224 while( count<n.getNumChildren() ){
1225 while( count<n.getNumChildren() && areEqual( n[count], d_emptyString ) ){
1226 addToExplanation( n[count], d_emptyString, exp );
1227 count++;
1228 }
1229 if( count<n.getNumChildren() ){
1230 Trace("strings-debug") << "...explain " << n[count] << " " << vecc[countc] << std::endl;
1231 if( !areEqual( n[count], vecc[countc] ) ){
1232 Node nrr = getRepresentative( n[count] );
1233 Assert( !d_eqc_to_const_exp[nrr].isNull() );
1234 addToExplanation( n[count], d_eqc_to_const_base[nrr], exp );
1235 exp.push_back( d_eqc_to_const_exp[nrr] );
1236 }else{
1237 addToExplanation( n[count], vecc[countc], exp );
1238 }
1239 countc++;
1240 count++;
1241 }
1242 }
1243 //exp contains an explanation of n==c
1244 Assert( countc==vecc.size() );
1245 if( hasTerm( c ) ){
1246 sendInference( exp, n.eqNode( c ), "I_CONST_MERGE" );
1247 return;
1248 }else if( !hasProcessed() ){
1249 Node nr = getRepresentative( n );
1250 std::map< Node, Node >::iterator it = d_eqc_to_const.find( nr );
1251 if( it==d_eqc_to_const.end() ){
1252 Trace("strings-debug") << "Set eqc const " << n << " to " << c << std::endl;
1253 d_eqc_to_const[nr] = c;
1254 d_eqc_to_const_base[nr] = n;
1255 d_eqc_to_const_exp[nr] = mkAnd( exp );
1256 }else if( c!=it->second ){
1257 //conflict
1258 Trace("strings-debug") << "Conflict, other constant was " << it->second << ", this constant was " << c << std::endl;
1259 if( d_eqc_to_const_exp[nr].isNull() ){
1260 // n==c ^ n == c' => false
1261 addToExplanation( n, it->second, exp );
1262 }else{
1263 // n==c ^ n == d_eqc_to_const_base[nr] == c' => false
1264 exp.push_back( d_eqc_to_const_exp[nr] );
1265 addToExplanation( n, d_eqc_to_const_base[nr], exp );
1266 }
1267 sendInference( exp, d_false, "I_CONST_CONFLICT" );
1268 return;
1269 }else{
1270 Trace("strings-debug") << "Duplicate constant." << std::endl;
1271 }
1272 }
1273 }
1274 }
1275 for( std::map< TNode, TermIndex >::iterator it = ti->d_children.begin(); it != ti->d_children.end(); ++it ){
1276 std::map< Node, Node >::iterator itc = d_eqc_to_const.find( it->first );
1277 if( itc!=d_eqc_to_const.end() ){
1278 vecc.push_back( itc->second );
1279 checkConstantEquivalenceClasses( &it->second, vecc );
1280 vecc.pop_back();
1281 if( hasProcessed() ){
1282 break;
1283 }
1284 }
1285 }
1286 }
1287
1288 void TheoryStrings::checkExtfEval( int effort ) {
1289 Trace("strings-extf-list") << "Active extended functions, effort=" << effort << " : " << std::endl;
1290 d_extf_info_tmp.clear();
1291 bool has_nreduce = false;
1292 std::vector< Node > terms;
1293 std::vector< Node > sterms;
1294 std::vector< std::vector< Node > > exp;
1295 d_extt->getInferences( effort, terms, sterms, exp );
1296 for( unsigned i=0; i<terms.size(); i++ ){
1297 Node n = terms[i];
1298 Node sn = sterms[i];
1299 //setup information about extf
1300 d_extf_info_tmp[n].init();
1301 std::map< Node, ExtfInfoTmp >::iterator itit = d_extf_info_tmp.find( n );
1302 if( n.getType().isBoolean() ){
1303 if( areEqual( n, d_true ) ){
1304 itit->second.d_pol = 1;
1305 }else if( areEqual( n, d_false ) ){
1306 itit->second.d_pol = -1;
1307 }
1308 }
1309 Trace("strings-extf-debug") << "Check extf " << n << " == " << sn << ", pol = " << itit->second.d_pol << ", effort=" << effort << "..." << std::endl;
1310 //do the inference
1311 Node to_reduce;
1312 if( n!=sn ){
1313 itit->second.d_exp.insert( itit->second.d_exp.end(), exp[i].begin(), exp[i].end() );
1314 // inference is rewriting the substituted node
1315 Node nrc = Rewriter::rewrite( sn );
1316 //if rewrites to a constant, then do the inference and mark as reduced
1317 if( nrc.isConst() ){
1318 if( effort<3 ){
1319 d_extt->markReduced( n );
1320 Trace("strings-extf-debug") << " resolvable by evaluation..." << std::endl;
1321 std::vector< Node > exps;
1322 Trace("strings-extf-debug") << " get symbolic definition..." << std::endl;
1323 Node nrs = getSymbolicDefinition( sn, exps );
1324 if( !nrs.isNull() ){
1325 Trace("strings-extf-debug") << " rewrite " << nrs << "..." << std::endl;
1326 nrs = Rewriter::rewrite( nrs );
1327 //ensure the symbolic form is non-trivial
1328 if( nrs.isConst() ){
1329 Trace("strings-extf-debug") << " symbolic definition is trivial..." << std::endl;
1330 nrs = Node::null();
1331 }
1332 }else{
1333 Trace("strings-extf-debug") << " could not infer symbolic definition." << std::endl;
1334 }
1335 Node conc;
1336 if( !nrs.isNull() ){
1337 Trace("strings-extf-debug") << " symbolic def : " << nrs << std::endl;
1338 if( !areEqual( nrs, nrc ) ){
1339 //infer symbolic unit
1340 if( n.getType().isBoolean() ){
1341 conc = nrc==d_true ? nrs : nrs.negate();
1342 }else{
1343 conc = nrs.eqNode( nrc );
1344 }
1345 itit->second.d_exp.clear();
1346 }
1347 }else{
1348 if( !areEqual( n, nrc ) ){
1349 if( n.getType().isBoolean() ){
1350 if( areEqual( n, nrc==d_true ? d_false : d_true ) ){
1351 itit->second.d_exp.push_back( nrc==d_true ? n.negate() : n );
1352 conc = d_false;
1353 }else{
1354 conc = nrc==d_true ? n : n.negate();
1355 }
1356 }else{
1357 conc = n.eqNode( nrc );
1358 }
1359 }
1360 }
1361 if( !conc.isNull() ){
1362 Trace("strings-extf") << " resolve extf : " << sn << " -> " << nrc << std::endl;
1363 sendInference( itit->second.d_exp, conc, effort==0 ? "EXTF" : "EXTF-N", true );
1364 if( d_conflict ){
1365 Trace("strings-extf-debug") << " conflict, return." << std::endl;
1366 return;
1367 }
1368 }
1369 }else{
1370 //check if it is already equal, if so, mark as reduced. Otherwise, do nothing.
1371 if( areEqual( n, nrc ) ){
1372 Trace("strings-extf") << " resolved extf, since satisfied by model: " << n << std::endl;
1373 itit->second.d_model_active = false;
1374 }
1375 }
1376 //if it reduces to a conjunction, infer each and reduce
1377 }else if( ( nrc.getKind()==kind::OR && itit->second.d_pol==-1 ) || ( nrc.getKind()==kind::AND && itit->second.d_pol==1 ) ){
1378 Assert( effort<3 );
1379 d_extt->markReduced( n );
1380 itit->second.d_exp.push_back( itit->second.d_pol==-1 ? n.negate() : n );
1381 Trace("strings-extf-debug") << " decomposable..." << std::endl;
1382 Trace("strings-extf") << " resolve extf : " << sn << " -> " << nrc << ", pol = " << itit->second.d_pol << std::endl;
1383 for( unsigned i=0; i<nrc.getNumChildren(); i++ ){
1384 sendInference( itit->second.d_exp, itit->second.d_pol==-1 ? nrc[i].negate() : nrc[i], effort==0 ? "EXTF_d" : "EXTF_d-N" );
1385 }
1386 }else{
1387 to_reduce = nrc;
1388 }
1389 }else{
1390 to_reduce = sterms[i];
1391 }
1392 //if not reduced
1393 if( !to_reduce.isNull() ){
1394 Assert( effort<3 );
1395 if( effort==1 ){
1396 Trace("strings-extf") << " cannot rewrite extf : " << to_reduce << std::endl;
1397 }
1398 checkExtfInference( n, to_reduce, itit->second, effort );
1399 if( Trace.isOn("strings-extf-list") ){
1400 Trace("strings-extf-list") << " * " << to_reduce;
1401 if( itit->second.d_pol!=0 ){
1402 Trace("strings-extf-list") << ", pol = " << itit->second.d_pol;
1403 }
1404 if( n!=to_reduce ){
1405 Trace("strings-extf-list") << ", from " << n;
1406 }
1407 Trace("strings-extf-list") << std::endl;
1408 }
1409 if( d_extt->isActive( n ) && itit->second.d_model_active ){
1410 has_nreduce = true;
1411 }
1412 }
1413 }
1414 d_has_extf = has_nreduce;
1415 }
1416
1417 void TheoryStrings::checkExtfInference( Node n, Node nr, ExtfInfoTmp& in, int effort ){
1418 //make additional inferences that do not contribute to the reduction of n, but may help show a refutation
1419 if( in.d_pol!=0 ){
1420 //add original to explanation
1421 in.d_exp.push_back( in.d_pol==1 ? n : n.negate() );
1422
1423 //d_extf_infer_cache stores whether we have made the inferences associated with a node n,
1424 // this may need to be generalized if multiple inferences apply
1425
1426 if( nr.getKind()==kind::STRING_STRCTN ){
1427 if( ( in.d_pol==1 && nr[1].getKind()==kind::STRING_CONCAT ) || ( in.d_pol==-1 && nr[0].getKind()==kind::STRING_CONCAT ) ){
1428 if( d_extf_infer_cache.find( nr )==d_extf_infer_cache.end() ){
1429 d_extf_infer_cache.insert( nr );
1430
1431 //one argument does (not) contain each of the components of the other argument
1432 int index = in.d_pol==1 ? 1 : 0;
1433 std::vector< Node > children;
1434 children.push_back( nr[0] );
1435 children.push_back( nr[1] );
1436 //Node exp_n = mkAnd( exp );
1437 for( unsigned i=0; i<nr[index].getNumChildren(); i++ ){
1438 children[index] = nr[index][i];
1439 Node conc = NodeManager::currentNM()->mkNode( kind::STRING_STRCTN, children );
1440 //can mark as reduced, since model for n => model for conc
1441 d_extt->markReduced( conc );
1442 sendInference( in.d_exp, in.d_pol==1 ? conc : conc.negate(), "CTN_Decompose" );
1443 }
1444
1445 }
1446 }else{
1447 //store this (reduced) assertion
1448 //Assert( effort==0 || nr[0]==getRepresentative( nr[0] ) );
1449 bool pol = in.d_pol==1;
1450 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() ){
1451 Trace("strings-extf-debug") << " store contains info : " << nr[0] << " " << pol << " " << nr[1] << std::endl;
1452 d_extf_info_tmp[nr[0]].d_ctn[pol].push_back( nr[1] );
1453 d_extf_info_tmp[nr[0]].d_ctn_from[pol].push_back( n );
1454 //transitive closure for contains
1455 bool opol = !pol;
1456 for( unsigned i=0; i<d_extf_info_tmp[nr[0]].d_ctn[opol].size(); i++ ){
1457 Node onr = d_extf_info_tmp[nr[0]].d_ctn[opol][i];
1458 Node conc = NodeManager::currentNM()->mkNode( kind::STRING_STRCTN, pol ? nr[1] : onr, pol ? onr : nr[1] );
1459 conc = Rewriter::rewrite( conc );
1460 bool do_infer = false;
1461 if( conc.getKind()==kind::EQUAL ){
1462 do_infer = !areDisequal( conc[0], conc[1] );
1463 }else{
1464 do_infer = !areEqual( conc, d_false );
1465 }
1466 if( do_infer ){
1467 conc = conc.negate();
1468 std::vector< Node > exp_c;
1469 exp_c.insert( exp_c.end(), in.d_exp.begin(), in.d_exp.end() );
1470 Node ofrom = d_extf_info_tmp[nr[0]].d_ctn_from[opol][i];
1471 Assert( d_extf_info_tmp.find( ofrom )!=d_extf_info_tmp.end() );
1472 exp_c.insert( exp_c.end(), d_extf_info_tmp[ofrom].d_exp.begin(), d_extf_info_tmp[ofrom].d_exp.end() );
1473 sendInference( exp_c, conc, "CTN_Trans" );
1474 }
1475 }
1476 }else{
1477 Trace("strings-extf-debug") << " redundant." << std::endl;
1478 d_extt->markReduced( n );
1479 }
1480 }
1481 }
1482 }
1483 }
1484
1485 void TheoryStrings::collectVars( Node n, std::vector< Node >& vars, std::map< Node, bool >& visited ) {
1486 if( !n.isConst() ){
1487 if( visited.find( n )==visited.end() ){
1488 visited[n] = true;
1489 if( n.getNumChildren()>0 ){
1490 for( unsigned i=0; i<n.getNumChildren(); i++ ){
1491 collectVars( n[i], vars, visited );
1492 }
1493 }else{
1494 //Node nr = getRepresentative( n );
1495 //vars[nr].push_back( n );
1496 vars.push_back( n );
1497 }
1498 }
1499 }
1500 }
1501
1502 Node TheoryStrings::getSymbolicDefinition( Node n, std::vector< Node >& exp ) {
1503 if( n.getNumChildren()==0 ){
1504 NodeNodeMap::const_iterator it = d_proxy_var.find( n );
1505 if( it==d_proxy_var.end() ){
1506 return Node::null();
1507 }else{
1508 Node eq = n.eqNode( (*it).second );
1509 eq = Rewriter::rewrite( eq );
1510 if( std::find( exp.begin(), exp.end(), eq )==exp.end() ){
1511 exp.push_back( eq );
1512 }
1513 return (*it).second;
1514 }
1515 }else{
1516 std::vector< Node > children;
1517 if (n.getMetaKind() == kind::metakind::PARAMETERIZED) {
1518 children.push_back( n.getOperator() );
1519 }
1520 for( unsigned i=0; i<n.getNumChildren(); i++ ){
1521 if( n.getKind()==kind::STRING_IN_REGEXP && i==1 ){
1522 children.push_back( n[i] );
1523 }else{
1524 Node ns = getSymbolicDefinition( n[i], exp );
1525 if( ns.isNull() ){
1526 return Node::null();
1527 }else{
1528 children.push_back( ns );
1529 }
1530 }
1531 }
1532 return NodeManager::currentNM()->mkNode( n.getKind(), children );
1533 }
1534 }
1535
1536 Node TheoryStrings::getConstantEqc( Node eqc ) {
1537 std::map< Node, Node >::iterator it = d_eqc_to_const.find( eqc );
1538 if( it!=d_eqc_to_const.end() ){
1539 return it->second;
1540 }else{
1541 return Node::null();
1542 }
1543 }
1544
1545 void TheoryStrings::debugPrintFlatForms( const char * tc ){
1546 for( unsigned k=0; k<d_strings_eqc.size(); k++ ){
1547 Node eqc = d_strings_eqc[k];
1548 if( d_eqc[eqc].size()>1 ){
1549 Trace( tc ) << "EQC [" << eqc << "]" << std::endl;
1550 }else{
1551 Trace( tc ) << "eqc [" << eqc << "]";
1552 }
1553 std::map< Node, Node >::iterator itc = d_eqc_to_const.find( eqc );
1554 if( itc!=d_eqc_to_const.end() ){
1555 Trace( tc ) << " C: " << itc->second;
1556 if( d_eqc[eqc].size()>1 ){
1557 Trace( tc ) << std::endl;
1558 }
1559 }
1560 if( d_eqc[eqc].size()>1 ){
1561 for( unsigned i=0; i<d_eqc[eqc].size(); i++ ){
1562 Node n = d_eqc[eqc][i];
1563 Trace( tc ) << " ";
1564 for( unsigned j=0; j<d_flat_form[n].size(); j++ ){
1565 Node fc = d_flat_form[n][j];
1566 itc = d_eqc_to_const.find( fc );
1567 Trace( tc ) << " ";
1568 if( itc!=d_eqc_to_const.end() ){
1569 Trace( tc ) << itc->second;
1570 }else{
1571 Trace( tc ) << fc;
1572 }
1573 }
1574 if( n!=eqc ){
1575 Trace( tc ) << ", from " << n;
1576 }
1577 Trace( tc ) << std::endl;
1578 }
1579 }else{
1580 Trace( tc ) << std::endl;
1581 }
1582 }
1583 Trace( tc ) << std::endl;
1584 }
1585
1586 void TheoryStrings::debugPrintNormalForms( const char * tc ) {
1587 }
1588
1589 struct sortConstLength {
1590 std::map< Node, unsigned > d_const_length;
1591 bool operator() (Node i, Node j) {
1592 std::map< Node, unsigned >::iterator it_i = d_const_length.find( i );
1593 std::map< Node, unsigned >::iterator it_j = d_const_length.find( j );
1594 if( it_i==d_const_length.end() ){
1595 if( it_j==d_const_length.end() ){
1596 return i<j;
1597 }else{
1598 return false;
1599 }
1600 }else{
1601 if( it_j==d_const_length.end() ){
1602 return true;
1603 }else{
1604 return it_i->second<it_j->second;
1605 }
1606 }
1607 }
1608 };
1609
1610
1611 void TheoryStrings::checkFlatForms() {
1612 //first check for cycles, while building ordering of equivalence classes
1613 d_eqc.clear();
1614 d_flat_form.clear();
1615 d_flat_form_index.clear();
1616 Trace("strings-process") << "Check equivalence classes cycles...." << std::endl;
1617 //rebuild strings eqc based on acyclic ordering
1618 std::vector< Node > eqc;
1619 eqc.insert( eqc.end(), d_strings_eqc.begin(), d_strings_eqc.end() );
1620 d_strings_eqc.clear();
1621 if( options::stringBinaryCsp() ){
1622 //sort: process smallest constants first (necessary if doing binary splits)
1623 sortConstLength scl;
1624 for( unsigned i=0; i<eqc.size(); i++ ){
1625 std::map< Node, Node >::iterator itc = d_eqc_to_const.find( eqc[i] );
1626 if( itc!=d_eqc_to_const.end() ){
1627 scl.d_const_length[eqc[i]] = itc->second.getConst<String>().size();
1628 }
1629 }
1630 std::sort( eqc.begin(), eqc.end(), scl );
1631 }
1632 for( unsigned i=0; i<eqc.size(); i++ ){
1633 std::vector< Node > curr;
1634 std::vector< Node > exp;
1635 checkCycles( eqc[i], curr, exp );
1636 if( hasProcessed() ){
1637 return;
1638 }
1639 }
1640 Trace("strings-process-debug") << "Done check cycles, lemmas = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << std::endl;
1641 if( !hasProcessed() ){
1642 //debug print flat forms
1643 if( Trace.isOn("strings-ff") ){
1644 Trace("strings-ff") << "Flat forms : " << std::endl;
1645 debugPrintFlatForms( "strings-ff" );
1646 }
1647
1648 //inferences without recursively expanding flat forms
1649
1650 //(1) approximate equality by containment, infer conflicts
1651 for( unsigned k=0; k<d_strings_eqc.size(); k++ ){
1652 Node eqc = d_strings_eqc[k];
1653 Node c = getConstantEqc( eqc );
1654 if( !c.isNull() ){
1655 //if equivalence class is constant, all component constants in flat forms must be contained in it, in order
1656 std::map< Node, std::vector< Node > >::iterator it = d_eqc.find( eqc );
1657 if( it!=d_eqc.end() ){
1658 for( unsigned i=0; i<it->second.size(); i++ ){
1659 Node n = it->second[i];
1660 int firstc, lastc;
1661 if( !TheoryStringsRewriter::canConstantContainList( c, d_flat_form[n], firstc, lastc ) ){
1662 Trace("strings-ff-debug") << "Flat form for " << n << " cannot be contained in constant " << c << std::endl;
1663 Trace("strings-ff-debug") << " indices = " << firstc << "/" << lastc << std::endl;
1664 //conflict, explanation is n = base ^ base = c ^ relevant porition of ( n = f[n] )
1665 std::vector< Node > exp;
1666 Assert( d_eqc_to_const_base.find( eqc )!=d_eqc_to_const_base.end() );
1667 addToExplanation( n, d_eqc_to_const_base[eqc], exp );
1668 Assert( d_eqc_to_const_exp.find( eqc )!=d_eqc_to_const_exp.end() );
1669 if( !d_eqc_to_const_exp[eqc].isNull() ){
1670 exp.push_back( d_eqc_to_const_exp[eqc] );
1671 }
1672 for( int e=firstc; e<=lastc; e++ ){
1673 if( d_flat_form[n][e].isConst() ){
1674 Assert( e>=0 && e<(int)d_flat_form_index[n].size() );
1675 Assert( d_flat_form_index[n][e]>=0 && d_flat_form_index[n][e]<(int)n.getNumChildren() );
1676 addToExplanation( d_flat_form[n][e], n[d_flat_form_index[n][e]], exp );
1677 }
1678 }
1679 Node conc = d_false;
1680 sendInference( exp, conc, "F_NCTN" );
1681 return;
1682 }
1683 }
1684 }
1685 }
1686 }
1687
1688 //(2) scan lists, unification to infer conflicts and equalities
1689 for( unsigned k=0; k<d_strings_eqc.size(); k++ ){
1690 Node eqc = d_strings_eqc[k];
1691 std::map< Node, std::vector< Node > >::iterator it = d_eqc.find( eqc );
1692 if( it!=d_eqc.end() && it->second.size()>1 ){
1693 //iterate over start index
1694 for( unsigned start=0; start<it->second.size()-1; start++ ){
1695 for( unsigned r=0; r<2; r++ ){
1696 unsigned count = 0;
1697 std::vector< Node > inelig;
1698 for( unsigned i=0; i<=start; i++ ){
1699 inelig.push_back( it->second[start] );
1700 }
1701 Node a = it->second[start];
1702 Node b;
1703 do{
1704 std::vector< Node > exp;
1705 //std::vector< Node > exp_n;
1706 Node conc;
1707 int inf_type = -1;
1708 if( count==d_flat_form[a].size() ){
1709 for( unsigned i=start+1; i<it->second.size(); i++ ){
1710 b = it->second[i];
1711 if( std::find( inelig.begin(), inelig.end(), b )==inelig.end() ){
1712 if( count<d_flat_form[b].size() ){
1713 //endpoint
1714 std::vector< Node > conc_c;
1715 for( unsigned j=count; j<d_flat_form[b].size(); j++ ){
1716 conc_c.push_back( b[d_flat_form_index[b][j]].eqNode( d_emptyString ) );
1717 }
1718 Assert( !conc_c.empty() );
1719 conc = mkAnd( conc_c );
1720 inf_type = 2;
1721 Assert( count>0 );
1722 //swap, will enforce is empty past current
1723 a = it->second[i]; b = it->second[start];
1724 count--;
1725 break;
1726 }
1727 inelig.push_back( it->second[i] );
1728 }
1729 }
1730 }else{
1731 Node curr = d_flat_form[a][count];
1732 Node curr_c = getConstantEqc( curr );
1733 Node ac = a[d_flat_form_index[a][count]];
1734 std::vector< Node > lexp;
1735 Node lcurr = getLength( ac, lexp );
1736 for( unsigned i=1; i<it->second.size(); i++ ){
1737 b = it->second[i];
1738 if( std::find( inelig.begin(), inelig.end(), b )==inelig.end() ){
1739 if( count==d_flat_form[b].size() ){
1740 inelig.push_back( b );
1741 //endpoint
1742 std::vector< Node > conc_c;
1743 for( unsigned j=count; j<d_flat_form[a].size(); j++ ){
1744 conc_c.push_back( a[d_flat_form_index[a][j]].eqNode( d_emptyString ) );
1745 }
1746 Assert( !conc_c.empty() );
1747 conc = mkAnd( conc_c );
1748 inf_type = 2;
1749 Assert( count>0 );
1750 count--;
1751 break;
1752 }else{
1753 Node cc = d_flat_form[b][count];
1754 if( cc!=curr ){
1755 Node bc = b[d_flat_form_index[b][count]];
1756 inelig.push_back( b );
1757 Assert( !areEqual( curr, cc ) );
1758 Node cc_c = getConstantEqc( cc );
1759 if( !curr_c.isNull() && !cc_c.isNull() ){
1760 //check for constant conflict
1761 int index;
1762 Node s = TheoryStringsRewriter::splitConstant( cc_c, curr_c, index, r==1 );
1763 if( s.isNull() ){
1764 addToExplanation( ac, d_eqc_to_const_base[curr], exp );
1765 addToExplanation( d_eqc_to_const_exp[curr], exp );
1766 addToExplanation( bc, d_eqc_to_const_base[cc], exp );
1767 addToExplanation( d_eqc_to_const_exp[cc], exp );
1768 conc = d_false;
1769 inf_type = 0;
1770 break;
1771 }
1772 }else if( (d_flat_form[a].size()-1)==count && (d_flat_form[b].size()-1)==count ){
1773 conc = ac.eqNode( bc );
1774 inf_type = 3;
1775 break;
1776 }else{
1777 //if lengths are the same, apply LengthEq
1778 std::vector< Node > lexp2;
1779 Node lcc = getLength( bc, lexp2 );
1780 if( areEqual( lcurr, lcc ) ){
1781 Trace("strings-ff-debug") << "Infer " << ac << " == " << bc << " since " << lcurr << " == " << lcc << std::endl;
1782 //exp_n.push_back( getLength( curr, true ).eqNode( getLength( cc, true ) ) );
1783 Trace("strings-ff-debug") << "Explanation for " << lcurr << " is ";
1784 for( unsigned j=0; j<lexp.size(); j++ ) { Trace("strings-ff-debug") << lexp[j] << std::endl; }
1785 Trace("strings-ff-debug") << "Explanation for " << lcc << " is ";
1786 for( unsigned j=0; j<lexp2.size(); j++ ) { Trace("strings-ff-debug") << lexp2[j] << std::endl; }
1787 exp.insert( exp.end(), lexp.begin(), lexp.end() );
1788 exp.insert( exp.end(), lexp2.begin(), lexp2.end() );
1789 addToExplanation( lcurr, lcc, exp );
1790 conc = ac.eqNode( bc );
1791 inf_type = 1;
1792 break;
1793 }
1794 }
1795 }
1796 }
1797 }
1798 }
1799 }
1800 if( !conc.isNull() ){
1801 Trace("strings-ff-debug") << "Found inference : " << conc << " based on equality " << a << " == " << b << " " << r << " " << inf_type << std::endl;
1802 addToExplanation( a, b, exp );
1803 //explain why prefixes up to now were the same
1804 for( unsigned j=0; j<count; j++ ){
1805 Trace("strings-ff-debug") << "Add at " << d_flat_form_index[a][j] << " " << d_flat_form_index[b][j] << std::endl;
1806 addToExplanation( a[d_flat_form_index[a][j]], b[d_flat_form_index[b][j]], exp );
1807 }
1808 //explain why other components up to now are empty
1809 for( unsigned t=0; t<2; t++ ){
1810 Node c = t==0 ? a : b;
1811 int jj;
1812 if( inf_type==3 || ( t==1 && inf_type==2 ) ){
1813 //explain all the empty components for F_EndpointEq, all for the short end for F_EndpointEmp
1814 jj = r==0 ? c.getNumChildren() : -1;
1815 }else{
1816 jj = t==0 ? d_flat_form_index[a][count] : d_flat_form_index[b][count];
1817 }
1818 if( r==0 ){
1819 for( int j=0; j<jj; j++ ){
1820 if( areEqual( c[j], d_emptyString ) ){
1821 addToExplanation( c[j], d_emptyString, exp );
1822 }
1823 }
1824 }else{
1825 for( int j=(c.getNumChildren()-1); j>jj; --j ){
1826 if( areEqual( c[j], d_emptyString ) ){
1827 addToExplanation( c[j], d_emptyString, exp );
1828 }
1829 }
1830 }
1831 }
1832 //notice that F_EndpointEmp is not typically applied, since strict prefix equality ( a.b = a ) where a,b non-empty
1833 // is conflicting by arithmetic len(a.b)=len(a)+len(b)!=len(a) when len(b)!=0.
1834 sendInference( exp, conc, inf_type==0 ? "F_Const" : ( inf_type==1 ? "F_Unify" : ( inf_type==2 ? "F_EndpointEmp" : "F_EndpointEq" ) ) );
1835 if( d_conflict ){
1836 return;
1837 }else{
1838 break;
1839 }
1840 }
1841 count++;
1842 }while( inelig.size()<it->second.size() );
1843
1844 for( unsigned i=0; i<it->second.size(); i++ ){
1845 std::reverse( d_flat_form[it->second[i]].begin(), d_flat_form[it->second[i]].end() );
1846 std::reverse( d_flat_form_index[it->second[i]].begin(), d_flat_form_index[it->second[i]].end() );
1847 }
1848 }
1849 }
1850 }
1851 }
1852 if( !hasProcessed() ){
1853 // simple extended func reduction
1854 Trace("strings-process") << "Check extended function reduction effort=1..." << std::endl;
1855 checkExtfReductions( 1 );
1856 Trace("strings-process") << "Done check extended function reduction" << std::endl;
1857 }
1858 }
1859 }
1860
1861 Node TheoryStrings::checkCycles( Node eqc, std::vector< Node >& curr, std::vector< Node >& exp ){
1862 if( std::find( curr.begin(), curr.end(), eqc )!=curr.end() ){
1863 // a loop
1864 return eqc;
1865 }else if( std::find( d_strings_eqc.begin(), d_strings_eqc.end(), eqc )==d_strings_eqc.end() ){
1866 curr.push_back( eqc );
1867 //look at all terms in this equivalence class
1868 eq::EqClassIterator eqc_i = eq::EqClassIterator( eqc, &d_equalityEngine );
1869 while( !eqc_i.isFinished() ) {
1870 Node n = (*eqc_i);
1871 if( d_congruent.find( n )==d_congruent.end() ){
1872 if( n.getKind() == kind::STRING_CONCAT ){
1873 Trace("strings-cycle") << eqc << " check term : " << n << " in " << eqc << std::endl;
1874 if( eqc!=d_emptyString_r ){
1875 d_eqc[eqc].push_back( n );
1876 }
1877 for( unsigned i=0; i<n.getNumChildren(); i++ ){
1878 Node nr = getRepresentative( n[i] );
1879 if( eqc==d_emptyString_r ){
1880 //for empty eqc, ensure all components are empty
1881 if( nr!=d_emptyString_r ){
1882 std::vector< Node > exp;
1883 exp.push_back( n.eqNode( d_emptyString ) );
1884 sendInference( exp, n[i].eqNode( d_emptyString ), "I_CYCLE_E" );
1885 return Node::null();
1886 }
1887 }else{
1888 if( nr!=d_emptyString_r ){
1889 d_flat_form[n].push_back( nr );
1890 d_flat_form_index[n].push_back( i );
1891 }
1892 //for non-empty eqc, recurse and see if we find a loop
1893 Node ncy = checkCycles( nr, curr, exp );
1894 if( !ncy.isNull() ){
1895 Trace("strings-cycle") << eqc << " cycle: " << ncy << " at " << n << "[" << i << "] : " << n[i] << std::endl;
1896 addToExplanation( n, eqc, exp );
1897 addToExplanation( nr, n[i], exp );
1898 if( ncy==eqc ){
1899 //can infer all other components must be empty
1900 for( unsigned j=0; j<n.getNumChildren(); j++ ){
1901 //take first non-empty
1902 if( j!=i && !areEqual( n[j], d_emptyString ) ){
1903 sendInference( exp, n[j].eqNode( d_emptyString ), "I_CYCLE" );
1904 return Node::null();
1905 }
1906 }
1907 Trace("strings-error") << "Looping term should be congruent : " << n << " " << eqc << " " << ncy << std::endl;
1908 //should find a non-empty component, otherwise would have been singular congruent (I_Norm_S)
1909 Assert( false );
1910 }else{
1911 return ncy;
1912 }
1913 }else{
1914 if( hasProcessed() ){
1915 return Node::null();
1916 }
1917 }
1918 }
1919 }
1920 }
1921 }
1922 ++eqc_i;
1923 }
1924 curr.pop_back();
1925 //now we can add it to the list of equivalence classes
1926 d_strings_eqc.push_back( eqc );
1927 }else{
1928 //already processed
1929 }
1930 return Node::null();
1931 }
1932
1933
1934 void TheoryStrings::checkNormalForms(){
1935 if( !options::stringEagerLen() ){
1936 for( unsigned i=0; i<d_strings_eqc.size(); i++ ) {
1937 Node eqc = d_strings_eqc[i];
1938 eq::EqClassIterator eqc_i = eq::EqClassIterator( eqc, &d_equalityEngine );
1939 while( !eqc_i.isFinished() ) {
1940 Node n = (*eqc_i);
1941 if( d_congruent.find( n )==d_congruent.end() ){
1942 registerTerm( n, 2 );
1943 }
1944 ++eqc_i;
1945 }
1946 }
1947 }
1948 if( !hasProcessed() ){
1949 Trace("strings-process") << "Normalize equivalence classes...." << std::endl;
1950 //calculate normal forms for each equivalence class, possibly adding splitting lemmas
1951 d_normal_forms.clear();
1952 d_normal_forms_exp.clear();
1953 std::map< Node, Node > nf_to_eqc;
1954 std::map< Node, Node > eqc_to_nf;
1955 std::map< Node, Node > eqc_to_exp;
1956 for( unsigned i=0; i<d_strings_eqc.size(); i++ ) {
1957 Node eqc = d_strings_eqc[i];
1958 Trace("strings-process-debug") << "- Verify normal forms are the same for " << eqc << std::endl;
1959 normalizeEquivalenceClass( eqc );
1960 Trace("strings-debug") << "Finished normalizing eqc..." << std::endl;
1961 if( hasProcessed() ){
1962 return;
1963 }else{
1964 Node nf_term = mkConcat( d_normal_forms[eqc] );
1965 std::map< Node, Node >::iterator itn = nf_to_eqc.find( nf_term );
1966 if( itn!=nf_to_eqc.end() ){
1967 //two equivalence classes have same normal form, merge
1968 std::vector< Node > nf_exp;
1969 nf_exp.push_back( mkAnd( d_normal_forms_exp[eqc] ) );
1970 nf_exp.push_back( eqc_to_exp[itn->second] );
1971 Node eq = d_normal_forms_base[eqc].eqNode( d_normal_forms_base[itn->second] );
1972 sendInference( nf_exp, eq, "Normal_Form" );
1973 } else {
1974 nf_to_eqc[nf_term] = eqc;
1975 eqc_to_nf[eqc] = nf_term;
1976 eqc_to_exp[eqc] = mkAnd( d_normal_forms_exp[eqc] );
1977 }
1978 }
1979 Trace("strings-process-debug") << "Done verifying normal forms are the same for " << eqc << std::endl;
1980 }
1981 if( !hasProcessed() ){
1982 if(Trace.isOn("strings-nf")) {
1983 Trace("strings-nf") << "**** Normal forms are : " << std::endl;
1984 for( std::map< Node, Node >::iterator it = eqc_to_exp.begin(); it != eqc_to_exp.end(); ++it ){
1985 Trace("strings-nf") << " N[" << it->first << "] (base " << d_normal_forms_base[it->first] << ") = " << eqc_to_nf[it->first] << std::endl;
1986 Trace("strings-nf") << " exp: " << it->second << std::endl;
1987 }
1988 Trace("strings-nf") << std::endl;
1989 }
1990 checkExtfEval( 1 );
1991 Trace("strings-process-debug") << "Done check extended functions re-eval, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
1992 if( !hasProcessed() ){
1993 if( !options::stringEagerLen() ){
1994 checkLengthsEqc();
1995 if( hasProcessed() ){
1996 return;
1997 }
1998 }
1999 //process disequalities between equivalence classes
2000 checkDeqNF();
2001 Trace("strings-process-debug") << "Done check disequalities, addedFact = " << !d_pending.empty() << " " << !d_lemma_cache.empty() << ", d_conflict = " << d_conflict << std::endl;
2002 }
2003 }
2004 Trace("strings-solve") << "Finished check normal forms, #lemmas = " << d_lemma_cache.size() << ", conflict = " << d_conflict << std::endl;
2005 }
2006 }
2007
2008 //compute d_normal_forms_(base,exp,exp_depend)[eqc]
2009 void TheoryStrings::normalizeEquivalenceClass( Node eqc ) {
2010 Trace("strings-process-debug") << "Process equivalence class " << eqc << std::endl;
2011 if( areEqual( eqc, d_emptyString ) ) {
2012 #ifdef CVC4_ASSERTIONS
2013 for( unsigned j=0; j<d_eqc[eqc].size(); j++ ){
2014 Node n = d_eqc[eqc][j];
2015 for( unsigned i=0; i<n.getNumChildren(); i++ ){
2016 Assert( areEqual( n[i], d_emptyString ) );
2017 }
2018 }
2019 #endif
2020 //do nothing
2021 Trace("strings-process-debug") << "Return process equivalence class " << eqc << " : empty." << std::endl;
2022 d_normal_forms_base[eqc] = d_emptyString;
2023 d_normal_forms[eqc].clear();
2024 d_normal_forms_exp[eqc].clear();
2025 } else {
2026 Assert( d_normal_forms.find(eqc)==d_normal_forms.end() );
2027 //phi => t = s1 * ... * sn
2028 // normal form for each non-variable term in this eqc (s1...sn)
2029 std::vector< std::vector< Node > > normal_forms;
2030 // explanation for each normal form (phi)
2031 std::vector< std::vector< Node > > normal_forms_exp;
2032 // dependency information
2033 std::vector< std::map< Node, std::map< bool, int > > > normal_forms_exp_depend;
2034 // record terms for each normal form (t)
2035 std::vector< Node > normal_form_src;
2036 // get normal forms
2037 getNormalForms(eqc, normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend);
2038 if( hasProcessed() ){
2039 return;
2040 }
2041 // process the normal forms
2042 processNEqc( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend );
2043 if( hasProcessed() ){
2044 return;
2045 }
2046 //debugPrintNormalForms( "strings-solve", eqc, normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend );
2047
2048 //construct the normal form
2049 Assert( !normal_forms.empty() );
2050
2051 int nf_index = 0;
2052 std::vector< Node >::iterator itn = std::find( normal_form_src.begin(), normal_form_src.end(), eqc );
2053 if( itn!=normal_form_src.end() ){
2054 nf_index = itn - normal_form_src.begin();
2055 Trace("strings-solve-debug2") << "take normal form " << nf_index << std::endl;
2056 Assert( normal_form_src[nf_index]==eqc );
2057 }else{
2058 //just take the first normal form
2059 Trace("strings-solve-debug2") << "take the first normal form" << std::endl;
2060 }
2061 d_normal_forms[eqc].insert( d_normal_forms[eqc].end(), normal_forms[nf_index].begin(), normal_forms[nf_index].end() );
2062 d_normal_forms_exp[eqc].insert( d_normal_forms_exp[eqc].end(), normal_forms_exp[nf_index].begin(), normal_forms_exp[nf_index].end() );
2063 Trace("strings-solve-debug2") << "take normal form ... done" << std::endl;
2064 d_normal_forms_base[eqc] = normal_form_src[nf_index];
2065 //track dependencies
2066 for( unsigned i=0; i<normal_forms_exp[nf_index].size(); i++ ){
2067 Node exp = normal_forms_exp[nf_index][i];
2068 for( unsigned r=0; r<2; r++ ){
2069 d_normal_forms_exp_depend[eqc][exp][r==0] = normal_forms_exp_depend[nf_index][exp][r==0];
2070 }
2071 }
2072 Trace("strings-process-debug") << "Return process equivalence class " << eqc << " : returned, size = " << d_normal_forms[eqc].size() << std::endl;
2073 }
2074 }
2075
2076 void TheoryStrings::getNormalForms( Node &eqc, std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2077 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend ) {
2078 //constant for equivalence class
2079 Node eqc_non_c = eqc;
2080 Trace("strings-process-debug") << "Get normal forms " << eqc << std::endl;
2081 eq::EqClassIterator eqc_i = eq::EqClassIterator( eqc, &d_equalityEngine );
2082 while( !eqc_i.isFinished() ){
2083 Node n = (*eqc_i);
2084 if( d_congruent.find( n )==d_congruent.end() ){
2085 if( n.getKind() == kind::CONST_STRING || n.getKind() == kind::STRING_CONCAT ){
2086 Trace("strings-process-debug") << "Get Normal Form : Process term " << n << " in eqc " << eqc << std::endl;
2087 std::vector< Node > nf_n;
2088 std::vector< Node > nf_exp_n;
2089 std::map< Node, std::map< bool, int > > nf_exp_depend_n;
2090 if( n.getKind()==kind::CONST_STRING ){
2091 if( n!=d_emptyString ) {
2092 nf_n.push_back( n );
2093 }
2094 }else if( n.getKind()==kind::STRING_CONCAT ){
2095 for( unsigned i=0; i<n.getNumChildren(); i++ ) {
2096 Node nr = d_equalityEngine.getRepresentative( n[i] );
2097 Trace("strings-process-debug") << "Normalizing subterm " << n[i] << " = " << nr << std::endl;
2098 Assert( d_normal_forms.find( nr )!=d_normal_forms.end() );
2099 unsigned orig_size = nf_n.size();
2100 unsigned add_size = d_normal_forms[nr].size();
2101 //if not the empty string, add to current normal form
2102 if( !d_normal_forms[nr].empty() ){
2103 for( unsigned r=0; r<d_normal_forms[nr].size(); r++ ) {
2104 if( Trace.isOn("strings-error") ) {
2105 if( d_normal_forms[nr][r].getKind()==kind::STRING_CONCAT ){
2106 Trace("strings-error") << "Strings::Error: From eqc = " << eqc << ", " << n << " index " << i << ", bad normal form : ";
2107 for( unsigned rr=0; rr<d_normal_forms[nr].size(); rr++ ) {
2108 Trace("strings-error") << d_normal_forms[nr][rr] << " ";
2109 }
2110 Trace("strings-error") << std::endl;
2111 }
2112 }
2113 Assert( d_normal_forms[nr][r].getKind()!=kind::STRING_CONCAT );
2114 }
2115 nf_n.insert( nf_n.end(), d_normal_forms[nr].begin(), d_normal_forms[nr].end() );
2116 }
2117
2118 for( unsigned j=0; j<d_normal_forms_exp[nr].size(); j++ ){
2119 Node exp = d_normal_forms_exp[nr][j];
2120 nf_exp_n.push_back( exp );
2121 //track depends
2122 for( unsigned k=0; k<2; k++ ){
2123 int prev_dep = d_normal_forms_exp_depend[nr][exp][k==1];
2124 if( k==0 ){
2125 nf_exp_depend_n[exp][false] = orig_size + prev_dep;
2126 }else if( k==1 ){
2127 //store forward index (converted back to reverse index below)
2128 nf_exp_depend_n[exp][true] = orig_size + ( add_size - prev_dep );
2129 }
2130 }
2131 }
2132 if( d_normal_forms_base[nr]!=n[i] ){
2133 Assert( d_normal_forms_base.find( nr )!=d_normal_forms_base.end() );
2134 Node eq = n[i].eqNode( d_normal_forms_base[nr] );
2135 nf_exp_n.push_back( eq );
2136 //track depends
2137 nf_exp_depend_n[eq][false] = orig_size;
2138 nf_exp_depend_n[eq][true] = orig_size + add_size;
2139 }
2140 }
2141 //convert forward indices to reverse indices
2142 int total_size = nf_n.size();
2143 for( std::map< Node, std::map< bool, int > >::iterator it = nf_exp_depend_n.begin(); it != nf_exp_depend_n.end(); ++it ){
2144 it->second[true] = total_size - it->second[true];
2145 Assert( it->second[true]>=0 );
2146 }
2147 }
2148 //if not equal to self
2149 if( nf_n.size()>1 || ( nf_n.size()==1 && nf_n[0].getKind()==kind::CONST_STRING ) ){
2150 if( nf_n.size()>1 ) {
2151 for( unsigned i=0; i<nf_n.size(); i++ ){
2152 if( Trace.isOn("strings-error") ){
2153 Trace("strings-error") << "Cycle for normal form ";
2154 printConcat(nf_n,"strings-error");
2155 Trace("strings-error") << "..." << nf_n[i] << std::endl;
2156 }
2157 Assert( !areEqual( nf_n[i], n ) );
2158 }
2159 }
2160 normal_forms.push_back(nf_n);
2161 normal_form_src.push_back(n);
2162 normal_forms_exp.push_back(nf_exp_n);
2163 normal_forms_exp_depend.push_back(nf_exp_depend_n);
2164 }else{
2165 //this was redundant: combination of self + empty string(s)
2166 Node nn = nf_n.size()==0 ? d_emptyString : nf_n[0];
2167 Assert( areEqual( nn, eqc ) );
2168 }
2169 }else{
2170 eqc_non_c = n;
2171 }
2172 }
2173 ++eqc_i;
2174 }
2175
2176 if( normal_forms.empty() ) {
2177 Trace("strings-solve-debug2") << "construct the normal form" << std::endl;
2178 //do not choose a concat here use "eqc_non_c" (in this case they have non-trivial explanation why they normalize to self)
2179 std::vector< Node > eqc_non_c_nf;
2180 getConcatVec( eqc_non_c, eqc_non_c_nf );
2181 normal_forms.push_back( eqc_non_c_nf );
2182 normal_form_src.push_back( eqc_non_c );
2183 normal_forms_exp.push_back( std::vector< Node >() );
2184 normal_forms_exp_depend.push_back( std::map< Node, std::map< bool, int > >() );
2185 }else{
2186 if(Trace.isOn("strings-solve")) {
2187 Trace("strings-solve") << "--- Normal forms for equivlance class " << eqc << " : " << std::endl;
2188 for( unsigned i=0; i<normal_forms.size(); i++ ) {
2189 Trace("strings-solve") << "#" << i << " (from " << normal_form_src[i] << ") : ";
2190 for( unsigned j=0; j<normal_forms[i].size(); j++ ) {
2191 if(j>0) {
2192 Trace("strings-solve") << ", ";
2193 }
2194 Trace("strings-solve") << normal_forms[i][j];
2195 }
2196 Trace("strings-solve") << std::endl;
2197 Trace("strings-solve") << " Explanation is : ";
2198 if(normal_forms_exp[i].size() == 0) {
2199 Trace("strings-solve") << "NONE";
2200 } else {
2201 for( unsigned j=0; j<normal_forms_exp[i].size(); j++ ) {
2202 if(j>0) {
2203 Trace("strings-solve") << " AND ";
2204 }
2205 Trace("strings-solve") << normal_forms_exp[i][j];
2206 }
2207 Trace("strings-solve") << std::endl;
2208 Trace("strings-solve") << "WITH DEPENDENCIES : " << std::endl;
2209 for( unsigned j=0; j<normal_forms_exp[i].size(); j++ ) {
2210 Trace("strings-solve") << " " << normal_forms_exp[i][j] << " -> ";
2211 Trace("strings-solve") << normal_forms_exp_depend[i][normal_forms_exp[i][j]][false] << ",";
2212 Trace("strings-solve") << normal_forms_exp_depend[i][normal_forms_exp[i][j]][true] << std::endl;
2213 }
2214 }
2215 Trace("strings-solve") << std::endl;
2216
2217 }
2218 } else {
2219 Trace("strings-solve") << "--- Single normal form for equivalence class " << eqc << std::endl;
2220 }
2221
2222 //if equivalence class is constant, approximate as containment, infer conflicts
2223 Node c = getConstantEqc( eqc );
2224 if( !c.isNull() ){
2225 Trace("strings-solve") << "Eqc is constant " << c << std::endl;
2226 for( unsigned i=0; i<normal_forms.size(); i++ ) {
2227 int firstc, lastc;
2228 if( !TheoryStringsRewriter::canConstantContainList( c, normal_forms[i], firstc, lastc ) ){
2229 Node n = normal_form_src[i];
2230 //conflict
2231 Trace("strings-solve") << "Normal form for " << n << " cannot be contained in constant " << c << std::endl;
2232 //conflict, explanation is n = base ^ base = c ^ relevant porition of ( n = N[n] )
2233 std::vector< Node > exp;
2234 Assert( d_eqc_to_const_base.find( eqc )!=d_eqc_to_const_base.end() );
2235 addToExplanation( n, d_eqc_to_const_base[eqc], exp );
2236 Assert( d_eqc_to_const_exp.find( eqc )!=d_eqc_to_const_exp.end() );
2237 if( !d_eqc_to_const_exp[eqc].isNull() ){
2238 exp.push_back( d_eqc_to_const_exp[eqc] );
2239 }
2240 //TODO: this can be minimized based on firstc/lastc, normal_forms_exp_depend
2241 exp.insert( exp.end(), normal_forms_exp[i].begin(), normal_forms_exp[i].end() );
2242 Node conc = d_false;
2243 sendInference( exp, conc, "N_NCTN" );
2244 }
2245 }
2246 }
2247 }
2248 }
2249
2250 void TheoryStrings::getExplanationVectorForPrefix( std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend,
2251 unsigned i, int index, bool isRev, std::vector< Node >& curr_exp ) {
2252 if( index==-1 || !options::stringMinPrefixExplain() ){
2253 curr_exp.insert(curr_exp.end(), normal_forms_exp[i].begin(), normal_forms_exp[i].end() );
2254 }else{
2255 for( unsigned k=0; k<normal_forms_exp[i].size(); k++ ){
2256 Node exp = normal_forms_exp[i][k];
2257 int dep = normal_forms_exp_depend[i][exp][isRev];
2258 if( dep<=index ){
2259 curr_exp.push_back( exp );
2260 Trace("strings-explain-prefix-debug") << " include : " << exp << std::endl;
2261 }else{
2262 Trace("strings-explain-prefix-debug") << " exclude : " << exp << std::endl;
2263 }
2264 }
2265 }
2266 }
2267
2268 void TheoryStrings::getExplanationVectorForPrefixEq( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2269 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend,
2270 unsigned i, unsigned j, int index_i, int index_j, bool isRev, std::vector< Node >& curr_exp ) {
2271 Trace("strings-explain-prefix") << "Get explanation for prefix " << index_i << ", " << index_j << " of normal forms " << i << " and " << j << ", reverse = " << isRev << std::endl;
2272 for( unsigned r=0; r<2; r++ ){
2273 getExplanationVectorForPrefix( normal_forms_exp, normal_forms_exp_depend, r==0 ? i : j, r==0 ? index_i : index_j, isRev, curr_exp );
2274 }
2275 Trace("strings-explain-prefix") << "Included " << curr_exp.size() << " / " << ( normal_forms_exp[i].size() + normal_forms_exp[j].size() ) << std::endl;
2276 addToExplanation( normal_form_src[i], normal_form_src[j], curr_exp );
2277 }
2278
2279
2280 void TheoryStrings::processNEqc( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2281 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend ){
2282 //the possible inferences
2283 std::vector< InferInfo > pinfer;
2284 // loop over all pairs
2285 for(unsigned i=0; i<normal_forms.size()-1; i++) {
2286 //unify each normalform[j] with normal_forms[i]
2287 for(unsigned j=i+1; j<normal_forms.size(); j++ ) {
2288 //ensure that normal_forms[i] and normal_forms[j] are the same modulo equality, add to pinfer if not
2289 Trace("strings-solve") << "Strings: Process normal form #" << i << " against #" << j << "..." << std::endl;
2290 if( isNormalFormPair( normal_form_src[i], normal_form_src[j] ) ) {
2291 Trace("strings-solve") << "Strings: Already cached." << std::endl;
2292 }else{
2293 //process the reverse direction first (check for easy conflicts and inferences)
2294 unsigned rindex = 0;
2295 processReverseNEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, rindex, 0, pinfer );
2296 if( hasProcessed() ){
2297 return;
2298 }else if( !pinfer.empty() && pinfer.back().d_id==1 ){
2299 break;
2300 }
2301 //AJR: for less aggressive endpoint inference
2302 //rindex = 0;
2303
2304 unsigned index = 0;
2305 processSimpleNEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, false, rindex, pinfer );
2306 if( hasProcessed() ){
2307 return;
2308 }else if( !pinfer.empty() && pinfer.back().d_id==1 ){
2309 break;
2310 }
2311 }
2312 }
2313 }
2314 if( !pinfer.empty() ){
2315 //now, determine which of the possible inferences we want to add
2316 int use_index = -1;
2317 Trace("strings-solve") << "Possible inferences (" << pinfer.size() << ") : " << std::endl;
2318 unsigned min_id = 9;
2319 unsigned max_index = 0;
2320 for( unsigned i=0; i<pinfer.size(); i++ ){
2321 Trace("strings-solve") << "From " << pinfer[i].d_i << " / " << pinfer[i].d_j << " (rev=" << pinfer[i].d_rev << ") : ";
2322 Trace("strings-solve") << pinfer[i].d_conc << " by " << pinfer[i].getId() << std::endl;
2323 if( use_index==-1 || pinfer[i].d_id<min_id || ( pinfer[i].d_id==min_id && pinfer[i].d_index>max_index ) ){
2324 min_id = pinfer[i].d_id;
2325 max_index = pinfer[i].d_index;
2326 use_index = i;
2327 }
2328 }
2329 //send the inference
2330 sendInference( pinfer[use_index].d_ant, pinfer[use_index].d_antn, pinfer[use_index].d_conc, pinfer[use_index].getId(), pinfer[use_index].sendAsLemma() );
2331 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 ){
2332 for( unsigned i=0; i<it->second.size(); i++ ){
2333 if( it->first==0 ){
2334 sendLengthLemma( it->second[i] );
2335 }else if( it->first==1 ){
2336 registerNonEmptySkolem( it->second[i] );
2337 }
2338 }
2339 }
2340 }
2341 }
2342
2343 bool TheoryStrings::InferInfo::sendAsLemma() {
2344 return true;
2345 }
2346
2347 void TheoryStrings::processReverseNEq( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2348 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend,
2349 unsigned i, unsigned j, unsigned& index, unsigned rproc, std::vector< InferInfo >& pinfer ) {
2350 //reverse normal form of i, j
2351 std::reverse( normal_forms[i].begin(), normal_forms[i].end() );
2352 std::reverse( normal_forms[j].begin(), normal_forms[j].end() );
2353
2354 processSimpleNEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, true, rproc, pinfer );
2355
2356 //reverse normal form of i, j
2357 std::reverse( normal_forms[i].begin(), normal_forms[i].end() );
2358 std::reverse( normal_forms[j].begin(), normal_forms[j].end() );
2359 }
2360
2361 //rproc is the # is the size of suffix that is identical
2362 void TheoryStrings::processSimpleNEq( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2363 std::vector< std::vector< Node > > &normal_forms_exp, std::vector< std::map< Node, std::map< bool, int > > >& normal_forms_exp_depend,
2364 unsigned i, unsigned j, unsigned& index, bool isRev, unsigned rproc, std::vector< InferInfo >& pinfer ) {
2365 Assert( rproc<=normal_forms[i].size() && rproc<=normal_forms[j].size() );
2366 bool success;
2367 do {
2368 success = false;
2369 //if we are at the end
2370 if( index==(normal_forms[i].size()-rproc) || index==(normal_forms[j].size()-rproc) ){
2371 if( index==(normal_forms[i].size()-rproc) && index==(normal_forms[j].size()-rproc) ){
2372 //we're done
2373 }else{
2374 //the remainder must be empty
2375 unsigned k = index==(normal_forms[i].size()-rproc) ? j : i;
2376 unsigned index_k = index;
2377 //Node eq_exp = mkAnd( curr_exp );
2378 std::vector< Node > curr_exp;
2379 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, -1, -1, isRev, curr_exp );
2380 while( !d_conflict && index_k<(normal_forms[k].size()-rproc) ){
2381 //can infer that this string must be empty
2382 Node eq = normal_forms[k][index_k].eqNode( d_emptyString );
2383 //Trace("strings-lemma") << "Strings: Infer " << eq << " from " << eq_exp << std::endl;
2384 Assert( !areEqual( d_emptyString, normal_forms[k][index_k] ) );
2385 sendInference( curr_exp, eq, "N_EndpointEmp" );
2386 index_k++;
2387 }
2388 }
2389 }else{
2390 Trace("strings-solve-debug") << "Process " << normal_forms[i][index] << " ... " << normal_forms[j][index] << std::endl;
2391 if( normal_forms[i][index]==normal_forms[j][index] ){
2392 Trace("strings-solve-debug") << "Simple Case 1 : strings are equal" << std::endl;
2393 index++;
2394 success = true;
2395 }else{
2396 Assert( !areEqual(normal_forms[i][index], normal_forms[j][index]) );
2397 std::vector< Node > temp_exp;
2398 Node length_term_i = getLength( normal_forms[i][index], temp_exp );
2399 Node length_term_j = getLength( normal_forms[j][index], temp_exp );
2400 //check length(normal_forms[i][index]) == length(normal_forms[j][index])
2401 if( areEqual( length_term_i, length_term_j ) ){
2402 Trace("strings-solve-debug") << "Simple Case 2 : string lengths are equal" << std::endl;
2403 Node eq = normal_forms[i][index].eqNode( normal_forms[j][index] );
2404 //eq = Rewriter::rewrite( eq );
2405 Node length_eq = length_term_i.eqNode( length_term_j );
2406 //temp_exp.insert(temp_exp.end(), curr_exp.begin(), curr_exp.end() );
2407 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, index, isRev, temp_exp );
2408 temp_exp.push_back(length_eq);
2409 sendInference( temp_exp, eq, "N_Unify" );
2410 return;
2411 }else if( ( normal_forms[i][index].getKind()!=kind::CONST_STRING && index==normal_forms[i].size()-rproc-1 ) ||
2412 ( normal_forms[j][index].getKind()!=kind::CONST_STRING && index==normal_forms[j].size()-rproc-1 ) ){
2413 Trace("strings-solve-debug") << "Simple Case 3 : at endpoint" << std::endl;
2414 std::vector< Node > antec;
2415 //antec.insert(antec.end(), curr_exp.begin(), curr_exp.end() );
2416 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, -1, -1, isRev, antec );
2417 std::vector< Node > eqn;
2418 for( unsigned r=0; r<2; r++ ) {
2419 int index_k = index;
2420 int k = r==0 ? i : j;
2421 std::vector< Node > eqnc;
2422 for( unsigned index_l=index_k; index_l<(normal_forms[k].size()-rproc); index_l++ ) {
2423 if(isRev) {
2424 eqnc.insert(eqnc.begin(), normal_forms[k][index_l] );
2425 } else {
2426 eqnc.push_back( normal_forms[k][index_l] );
2427 }
2428 }
2429 eqn.push_back( mkConcat( eqnc ) );
2430 }
2431 if( !areEqual( eqn[0], eqn[1] ) ){
2432 sendInference( antec, eqn[0].eqNode( eqn[1] ), "N_EndpointEq", true );
2433 return;
2434 }else{
2435 Assert( normal_forms[i].size()==normal_forms[j].size() );
2436 index = normal_forms[i].size()-rproc;
2437 }
2438 }else if( normal_forms[i][index].isConst() && normal_forms[j][index].isConst() ){
2439 Node const_str = normal_forms[i][index];
2440 Node other_str = normal_forms[j][index];
2441 Trace("strings-solve-debug") << "Simple Case 3 : Const Split : " << const_str << " vs " << other_str << " at index " << index << ", isRev = " << isRev << std::endl;
2442 unsigned len_short = const_str.getConst<String>().size() <= other_str.getConst<String>().size() ? const_str.getConst<String>().size() : other_str.getConst<String>().size();
2443 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);
2444 if( isSameFix ) {
2445 //same prefix/suffix
2446 //k is the index of the string that is shorter
2447 int k = const_str.getConst<String>().size()<other_str.getConst<String>().size() ? i : j;
2448 int l = const_str.getConst<String>().size()<other_str.getConst<String>().size() ? j : i;
2449 //update the nf exp dependencies
2450 //notice this is not critical for soundness: not doing the below incrementing will only lead to overapproximating when antecedants are required in explanations
2451 for( std::map< Node, std::map< bool, int > >::iterator itnd = normal_forms_exp_depend[l].begin(); itnd != normal_forms_exp_depend[l].end(); ++itnd ){
2452 for( std::map< bool, int >::iterator itnd2 = itnd->second.begin(); itnd2 != itnd->second.end(); ++itnd2 ){
2453 //see if this can be incremented: it can if it is not relevant to the current index
2454 Assert( itnd2->second>=0 && itnd2->second<=(int)normal_forms[l].size() );
2455 bool increment = (itnd2->first==isRev) ? itnd2->second>(int)index : ( (int)normal_forms[l].size()-1-itnd2->second )<(int)index;
2456 if( increment ){
2457 normal_forms_exp_depend[l][itnd->first][itnd2->first] = itnd2->second + 1;
2458 }
2459 }
2460 }
2461 if( isRev ){
2462 int new_len = normal_forms[l][index].getConst<String>().size() - len_short;
2463 Node remainderStr = NodeManager::currentNM()->mkConst( normal_forms[l][index].getConst<String>().substr(0, new_len) );
2464 Trace("strings-solve-debug-test") << "Break normal form of " << normal_forms[l][index] << " into " << normal_forms[k][index] << ", " << remainderStr << std::endl;
2465 normal_forms[l].insert( normal_forms[l].begin()+index + 1, remainderStr );
2466 }else{
2467 Node remainderStr = NodeManager::currentNM()->mkConst(normal_forms[l][index].getConst<String>().substr(len_short));
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 }
2471 normal_forms[l][index] = normal_forms[k][index];
2472 index++;
2473 success = true;
2474 }else{
2475 //conflict
2476 std::vector< Node > antec;
2477 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, index, isRev, antec );
2478 sendInference( antec, d_false, "N_Const", true );
2479 return;
2480 }
2481 }else{
2482 //construct the candidate inference "info"
2483 InferInfo info;
2484 info.d_index = index;
2485 //for debugging
2486 info.d_i = i;
2487 info.d_j = j;
2488 info.d_rev = isRev;
2489 bool info_valid = false;
2490 Assert( index<normal_forms[i].size()-rproc && index<normal_forms[j].size()-rproc );
2491 std::vector< Node > lexp;
2492 Node length_term_i = getLength( normal_forms[i][index], lexp );
2493 Node length_term_j = getLength( normal_forms[j][index], lexp );
2494 //split on equality between string lengths (note that splitting on equality between strings is worse since it is harder to process)
2495 if( !areDisequal( length_term_i, length_term_j ) && !areEqual( length_term_i, length_term_j ) &&
2496 normal_forms[i][index].getKind()!=kind::CONST_STRING && normal_forms[j][index].getKind()!=kind::CONST_STRING ){ //AJR: remove the latter 2 conditions?
2497 Trace("strings-solve-debug") << "Non-simple Case 1 : string lengths neither equal nor disequal" << std::endl;
2498 //try to make the lengths equal via splitting on demand
2499 Node length_eq = NodeManager::currentNM()->mkNode( kind::EQUAL, length_term_i, length_term_j );
2500 length_eq = Rewriter::rewrite( length_eq );
2501 //set info
2502 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, length_eq, length_eq.negate() );
2503 info.d_pending_phase[ length_eq ] = true;
2504 info.d_id = 3;
2505 info_valid = true;
2506 }else{
2507 Trace("strings-solve-debug") << "Non-simple Case 2 : must compare strings" << std::endl;
2508 int loop_in_i = -1;
2509 int loop_in_j = -1;
2510 if( detectLoop( normal_forms, i, j, index, loop_in_i, loop_in_j, rproc ) ){
2511 if( !isRev ){ //FIXME
2512 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, -1, -1, isRev, info.d_ant );
2513 //set info
2514 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 ) ){
2515 info_valid = true;
2516 }
2517 }
2518 }else{
2519 //AJR: length entailment here?
2520 if( normal_forms[i][index].getKind() == kind::CONST_STRING || normal_forms[j][index].getKind() == kind::CONST_STRING ){
2521 unsigned const_k = normal_forms[i][index].getKind() == kind::CONST_STRING ? i : j;
2522 unsigned nconst_k = normal_forms[i][index].getKind() == kind::CONST_STRING ? j : i;
2523 Node other_str = normal_forms[nconst_k][index];
2524 Assert( other_str.getKind()!=kind::CONST_STRING, "Other string is not constant." );
2525 Assert( other_str.getKind()!=kind::STRING_CONCAT, "Other string is not CONCAT." );
2526 if( !d_equalityEngine.areDisequal( other_str, d_emptyString, true ) ){
2527 Node eq = other_str.eqNode( d_emptyString );
2528 //set info
2529 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, eq, eq.negate() );
2530 info.d_id = 4;
2531 info_valid = true;
2532 }else{
2533 if( !isRev ){ //FIXME
2534 Node xnz = other_str.eqNode( d_emptyString ).negate();
2535 unsigned index_nc_k = index+1;
2536 //Node next_const_str = TheoryStringsRewriter::collectConstantStringAt( normal_forms[nconst_k], index_nc_k, false );
2537 unsigned start_index_nc_k = index+1;
2538 Node next_const_str = TheoryStringsRewriter::getNextConstantAt( normal_forms[nconst_k], start_index_nc_k, index_nc_k, false );
2539 if( !next_const_str.isNull() ) {
2540 unsigned index_c_k = index;
2541 Node const_str = TheoryStringsRewriter::collectConstantStringAt( normal_forms[const_k], index_c_k, false );
2542 Assert( !const_str.isNull() );
2543 CVC4::String stra = const_str.getConst<String>();
2544 CVC4::String strb = next_const_str.getConst<String>();
2545 //since non-empty, we start with charecter #1
2546 size_t p;
2547 if( isRev ){
2548 CVC4::String stra1 = stra.prefix( stra.size()-1 );
2549 p = stra.size() - stra1.roverlap(strb);
2550 Trace("strings-csp-debug") << "Compute roverlap : " << const_str << " " << next_const_str << std::endl;
2551 size_t p2 = stra1.rfind(strb);
2552 p = p2==std::string::npos ? p : ( p>p2+1? p2+1 : p );
2553 Trace("strings-csp-debug") << "overlap : " << stra1 << " " << strb << " returned " << p << " " << p2 << " " << (p2==std::string::npos) << std::endl;
2554 }else{
2555 CVC4::String stra1 = stra.substr( 1 );
2556 p = stra.size() - stra1.overlap(strb);
2557 Trace("strings-csp-debug") << "Compute overlap : " << const_str << " " << next_const_str << std::endl;
2558 size_t p2 = stra1.find(strb);
2559 p = p2==std::string::npos ? p : ( p>p2+1? p2+1 : p );
2560 Trace("strings-csp-debug") << "overlap : " << stra1 << " " << strb << " returned " << p << " " << p2 << " " << (p2==std::string::npos) << std::endl;
2561 }
2562 if( p>1 ){
2563 if( start_index_nc_k==index+1 ){
2564 info.d_ant.push_back( xnz );
2565 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend,
2566 const_k, nconst_k, index_c_k, index_nc_k, isRev, info.d_ant );
2567 Node prea = p==stra.size() ? const_str : NodeManager::currentNM()->mkConst( isRev ? stra.suffix( p ) : stra.prefix( p ) );
2568 Node sk = mkSkolemCached( other_str, prea, isRev ? sk_id_c_spt_rev : sk_id_c_spt, "c_spt", -1 );
2569 Trace("strings-csp") << "Const Split: " << prea << " is removed from " << stra << " due to " << strb << ", p=" << p << std::endl;
2570 //set info
2571 info.d_conc = other_str.eqNode( isRev ? mkConcat( sk, prea ) : mkConcat(prea, sk) );
2572 info.d_new_skolem[0].push_back( sk );
2573 info.d_id = 1;
2574 info_valid = true;
2575 }
2576 /* FIXME for isRev, speculative
2577 else if( options::stringLenPropCsp() ){
2578 //propagate length constraint
2579 std::vector< Node > cc;
2580 for( unsigned i=index; i<start_index_nc_k; i++ ){
2581 cc.push_back( normal_forms[nconst_k][i] );
2582 }
2583 Node lt = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, mkConcat( cc ) );
2584 conc = NodeManager::currentNM()->mkNode( kind::GEQ, lt, NodeManager::currentNM()->mkConst( Rational(p) ) );
2585 sendInference( ant, conc, "S-Split(CSP-P)-lprop", true );
2586 }
2587 */
2588 }
2589 }
2590 if( !info_valid ){
2591 info.d_ant.push_back( xnz );
2592 Node const_str = normal_forms[const_k][index];
2593 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, index, isRev, info.d_ant );
2594 CVC4::String stra = const_str.getConst<String>();
2595 if( options::stringBinaryCsp() && stra.size()>3 ){
2596 //split string in half
2597 Node c_firstHalf = NodeManager::currentNM()->mkConst( isRev ? stra.substr( stra.size()/2 ) : stra.substr(0, stra.size()/2 ) );
2598 Node sk = mkSkolemCached( other_str, c_firstHalf , isRev ? sk_id_vc_bin_spt_rev : sk_id_vc_bin_spt, "cb_spt", -1 );
2599 Trace("strings-csp") << "Const Split: " << c_firstHalf << " is removed from " << const_str << " (binary) " << std::endl;
2600 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, other_str.eqNode( isRev ? mkConcat( sk, c_firstHalf ) : mkConcat( c_firstHalf, sk ) ),
2601 NodeManager::currentNM()->mkNode( kind::AND,
2602 sk.eqNode( d_emptyString ).negate(),
2603 c_firstHalf.eqNode( isRev ? mkConcat( sk, other_str ) : mkConcat( other_str, sk ) ) ) );
2604 info.d_new_skolem[0].push_back( sk );
2605 info.d_id = 5;
2606 info_valid = true;
2607 }else{
2608 // normal v/c split
2609 Node firstChar = stra.size() == 1 ? const_str : NodeManager::currentNM()->mkConst( isRev ? stra.suffix( 1 ) : stra.prefix( 1 ) );
2610 Node sk = mkSkolemCached( other_str, firstChar, isRev ? sk_id_vc_spt_rev : sk_id_vc_spt, "c_spt", -1 );
2611 Trace("strings-csp") << "Const Split: " << firstChar << " is removed from " << const_str << " (serial) " << std::endl;
2612 info.d_conc = other_str.eqNode( isRev ? mkConcat( sk, firstChar ) : mkConcat(firstChar, sk) );
2613 info.d_new_skolem[0].push_back( sk );
2614 info.d_id = 6;
2615 info_valid = true;
2616 }
2617 }
2618 }
2619 }
2620 }else{
2621 int lentTestSuccess = -1;
2622 Node lentTestExp;
2623 if( options::stringCheckEntailLen() ){
2624 //check entailment
2625 for( unsigned e=0; e<2; e++ ){
2626 Node t = e==0 ? normal_forms[i][index] : normal_forms[j][index];
2627 //do not infer constants are larger than variables
2628 if( t.getKind()!=kind::CONST_STRING ){
2629 Node lt1 = e==0 ? length_term_i : length_term_j;
2630 Node lt2 = e==0 ? length_term_j : length_term_i;
2631 Node ent_lit = Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::GT, lt1, lt2 ) );
2632 std::pair<bool, Node> et = d_valuation.entailmentCheck( THEORY_OF_TYPE_BASED, ent_lit );
2633 if( et.first ){
2634 Trace("strings-entail") << "Strings entailment : " << ent_lit << " is entailed in the current context." << std::endl;
2635 Trace("strings-entail") << " explanation was : " << et.second << std::endl;
2636 lentTestSuccess = e;
2637 lentTestExp = et.second;
2638 break;
2639 }
2640 }
2641 }
2642 }
2643
2644 getExplanationVectorForPrefixEq( normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend, i, j, index, index, isRev, info.d_ant );
2645 //x!=e /\ y!=e
2646 for(unsigned xory=0; xory<2; xory++) {
2647 Node x = xory==0 ? normal_forms[i][index] : normal_forms[j][index];
2648 Node xgtz = x.eqNode( d_emptyString ).negate();
2649 if( d_equalityEngine.areDisequal( x, d_emptyString, true ) ) {
2650 info.d_ant.push_back( xgtz );
2651 } else {
2652 info.d_antn.push_back( xgtz );
2653 }
2654 }
2655 Node sk = mkSkolemCached( normal_forms[i][index], normal_forms[j][index], isRev ? sk_id_v_spt_rev : sk_id_v_spt, "v_spt", -1 );
2656 //must add length requirement
2657 info.d_new_skolem[1].push_back( sk );
2658 Node eq1 = normal_forms[i][index].eqNode( isRev ? mkConcat(sk, normal_forms[j][index]) : mkConcat(normal_forms[j][index], sk) );
2659 Node eq2 = normal_forms[j][index].eqNode( isRev ? mkConcat(sk, normal_forms[i][index]) : mkConcat(normal_forms[i][index], sk) );
2660
2661 if( lentTestSuccess!=-1 ){
2662 info.d_antn.push_back( lentTestExp );
2663 info.d_conc = lentTestSuccess==0 ? eq1 : eq2;
2664 info.d_id = 2;
2665 info_valid = true;
2666 }else{
2667 Node ldeq = NodeManager::currentNM()->mkNode( kind::EQUAL, length_term_i, length_term_j ).negate();
2668 if( d_equalityEngine.areDisequal( length_term_i, length_term_j, true ) ){
2669 info.d_ant.push_back( ldeq );
2670 }else{
2671 info.d_antn.push_back(ldeq);
2672 }
2673 //set info
2674 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, eq1, eq2 );
2675 info.d_id = 7;
2676 info_valid = true;
2677 }
2678 }
2679 }
2680 }
2681 if( info_valid ){
2682 pinfer.push_back( info );
2683 Assert( !success );
2684 }
2685 }
2686 }
2687 }
2688 }while( success );
2689 }
2690
2691 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 ){
2692 int has_loop[2] = { -1, -1 };
2693 if( options::stringLB() != 2 ) {
2694 for( unsigned r=0; r<2; r++ ) {
2695 int n_index = (r==0 ? i : j);
2696 int other_n_index = (r==0 ? j : i);
2697 if( normal_forms[other_n_index][index].getKind() != kind::CONST_STRING ) {
2698 for( unsigned lp = index+1; lp<normal_forms[n_index].size()-rproc; lp++ ){
2699 if( normal_forms[n_index][lp]==normal_forms[other_n_index][index] ){
2700 has_loop[r] = lp;
2701 break;
2702 }
2703 }
2704 }
2705 }
2706 }
2707 if( has_loop[0]!=-1 || has_loop[1]!=-1 ) {
2708 loop_in_i = has_loop[0];
2709 loop_in_j = has_loop[1];
2710 return true;
2711 } else {
2712 Trace("strings-solve-debug") << "No loops detected." << std::endl;
2713 return false;
2714 }
2715 }
2716
2717 //xs(zy)=t(yz)xr
2718 bool TheoryStrings::processLoop( std::vector< std::vector< Node > > &normal_forms, std::vector< Node > &normal_form_src,
2719 int i, int j, int loop_n_index, int other_n_index, int loop_index, int index, InferInfo& info ){
2720 if( options::stringAbortLoop() ){
2721 Message() << "Looping word equation encountered." << std::endl;
2722 exit( 1 );
2723 }else{
2724 Node conc;
2725 Trace("strings-loop") << "Detected possible loop for " << normal_forms[loop_n_index][loop_index] << std::endl;
2726 Trace("strings-loop") << " ... (X)= " << normal_forms[other_n_index][index] << std::endl;
2727
2728 Trace("strings-loop") << " ... T(Y.Z)= ";
2729 std::vector< Node > vec_t;
2730 for(int lp=index; lp<loop_index; ++lp) {
2731 if(lp != index) Trace("strings-loop") << " ++ ";
2732 Trace("strings-loop") << normal_forms[loop_n_index][lp];
2733 vec_t.push_back( normal_forms[loop_n_index][lp] );
2734 }
2735 Node t_yz = mkConcat( vec_t );
2736 Trace("strings-loop") << " (" << t_yz << ")" << std::endl;
2737 Trace("strings-loop") << " ... S(Z.Y)= ";
2738 std::vector< Node > vec_s;
2739 for(int lp=index+1; lp<(int)normal_forms[other_n_index].size(); ++lp) {
2740 if(lp != index+1) Trace("strings-loop") << " ++ ";
2741 Trace("strings-loop") << normal_forms[other_n_index][lp];
2742 vec_s.push_back( normal_forms[other_n_index][lp] );
2743 }
2744 Node s_zy = mkConcat( vec_s );
2745 Trace("strings-loop") << " (" << s_zy << ")" << std::endl;
2746 Trace("strings-loop") << " ... R= ";
2747 std::vector< Node > vec_r;
2748 for(int lp=loop_index+1; lp<(int)normal_forms[loop_n_index].size(); ++lp) {
2749 if(lp != loop_index+1) Trace("strings-loop") << " ++ ";
2750 Trace("strings-loop") << normal_forms[loop_n_index][lp];
2751 vec_r.push_back( normal_forms[loop_n_index][lp] );
2752 }
2753 Node r = mkConcat( vec_r );
2754 Trace("strings-loop") << " (" << r << ")" << std::endl;
2755
2756 //Trace("strings-loop") << "Lemma Cache: " << normal_form_src[i] << " vs " << normal_form_src[j] << std::endl;
2757 //TODO: can be more general
2758 if( s_zy.isConst() && r.isConst() && r!=d_emptyString) {
2759 int c;
2760 bool flag = true;
2761 if(s_zy.getConst<String>().tailcmp( r.getConst<String>(), c ) ) {
2762 if( c>=0) {
2763 s_zy = NodeManager::currentNM()->mkConst( s_zy.getConst<String>().substr(0, c) );
2764 r = d_emptyString;
2765 vec_r.clear();
2766 Trace("strings-loop") << "Strings::Loop: Refactor S(Z.Y)= " << s_zy << ", c=" << c << std::endl;
2767 flag = false;
2768 }
2769 }
2770 if( flag ){
2771 Trace("strings-loop") << "Strings::Loop: tails are different." << std::endl;
2772 sendInference( info.d_ant, conc, "Loop Conflict", true );
2773 return false;
2774 }
2775 }
2776
2777 //require that x is non-empty
2778 Node split_eq;
2779 if( !areDisequal( normal_forms[loop_n_index][loop_index], d_emptyString ) ){
2780 //try to make normal_forms[loop_n_index][loop_index] equal to empty to avoid loop
2781 split_eq = normal_forms[loop_n_index][loop_index].eqNode( d_emptyString );
2782 }else if( !areDisequal( t_yz, d_emptyString ) && t_yz.getKind()!=kind::CONST_STRING ) {
2783 //try to make normal_forms[loop_n_index][loop_index] equal to empty to avoid loop
2784 split_eq = t_yz.eqNode( d_emptyString );
2785 }
2786 if( !split_eq.isNull() ){
2787 info.d_conc = NodeManager::currentNM()->mkNode( kind::OR, split_eq, split_eq.negate() );
2788 info.d_id = 4;
2789 return true;
2790 }else{
2791 //need to break
2792 info.d_ant.push_back( normal_forms[loop_n_index][loop_index].eqNode( d_emptyString ).negate() );
2793 if( t_yz.getKind()!=kind::CONST_STRING ) {
2794 info.d_ant.push_back( t_yz.eqNode( d_emptyString ).negate() );
2795 }
2796 Node ant = mkExplain( info.d_ant );
2797 if( d_loop_antec.find( ant ) == d_loop_antec.end() ){
2798 d_loop_antec.insert( ant );
2799 info.d_ant.clear();
2800 info.d_antn.push_back( ant );
2801
2802 Node str_in_re;
2803 if( s_zy == t_yz &&
2804 r == d_emptyString &&
2805 s_zy.isConst() &&
2806 s_zy.getConst<String>().isRepeated()
2807 ) {
2808 Node rep_c = NodeManager::currentNM()->mkConst( s_zy.getConst<String>().substr(0, 1) );
2809 Trace("strings-loop") << "Special case (X)=" << normal_forms[other_n_index][index] << " " << std::endl;
2810 Trace("strings-loop") << "... (C)=" << rep_c << " " << std::endl;
2811 //special case
2812 str_in_re = NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP, normal_forms[other_n_index][index],
2813 NodeManager::currentNM()->mkNode( kind::REGEXP_STAR,
2814 NodeManager::currentNM()->mkNode( kind::STRING_TO_REGEXP, rep_c ) ) );
2815 conc = str_in_re;
2816 } else if(t_yz.isConst()) {
2817 Trace("strings-loop") << "Strings::Loop: Const Normal Breaking." << std::endl;
2818 CVC4::String s = t_yz.getConst< CVC4::String >();
2819 unsigned size = s.size();
2820 std::vector< Node > vconc;
2821 for(unsigned len=1; len<=size; len++) {
2822 Node y = NodeManager::currentNM()->mkConst(s.substr(0, len));
2823 Node z = NodeManager::currentNM()->mkConst(s.substr(len, size - len));
2824 Node restr = s_zy;
2825 Node cc;
2826 if(r != d_emptyString) {
2827 std::vector< Node > v2(vec_r);
2828 v2.insert(v2.begin(), y);
2829 v2.insert(v2.begin(), z);
2830 restr = mkConcat( z, y );
2831 cc = Rewriter::rewrite(s_zy.eqNode( mkConcat( v2 ) ));
2832 } else {
2833 cc = Rewriter::rewrite(s_zy.eqNode( mkConcat( z, y) ));
2834 }
2835 if(cc == d_false) {
2836 continue;
2837 }
2838 Node conc2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, normal_forms[other_n_index][index],
2839 NodeManager::currentNM()->mkNode(kind::REGEXP_CONCAT,
2840 NodeManager::currentNM()->mkNode(kind::STRING_TO_REGEXP, y),
2841 NodeManager::currentNM()->mkNode(kind::REGEXP_STAR,
2842 NodeManager::currentNM()->mkNode(kind::STRING_TO_REGEXP, restr))));
2843 cc = cc==d_true ? conc2 : NodeManager::currentNM()->mkNode( kind::AND, cc, conc2 );
2844 d_regexp_ant[conc2] = ant;
2845 vconc.push_back(cc);
2846 }
2847 conc = vconc.size()==0 ? Node::null() : vconc.size()==1 ? vconc[0] : NodeManager::currentNM()->mkNode(kind::OR, vconc);
2848 } else {
2849 Trace("strings-loop") << "Strings::Loop: Normal Loop Breaking." << std::endl;
2850 //right
2851 Node sk_w= mkSkolemS( "w_loop" );
2852 Node sk_y= mkSkolemS( "y_loop", 1 );
2853 Node sk_z= mkSkolemS( "z_loop" );
2854 //t1 * ... * tn = y * z
2855 Node conc1 = t_yz.eqNode( mkConcat( sk_y, sk_z ) );
2856 // s1 * ... * sk = z * y * r
2857 vec_r.insert(vec_r.begin(), sk_y);
2858 vec_r.insert(vec_r.begin(), sk_z);
2859 Node conc2 = s_zy.eqNode( mkConcat( vec_r ) );
2860 Node conc3 = normal_forms[other_n_index][index].eqNode( mkConcat( sk_y, sk_w ) );
2861 Node restr = r == d_emptyString ? s_zy : mkConcat( sk_z, sk_y );
2862 str_in_re = NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP, sk_w,
2863 NodeManager::currentNM()->mkNode( kind::REGEXP_STAR,
2864 NodeManager::currentNM()->mkNode( kind::STRING_TO_REGEXP, restr ) ) );
2865
2866 std::vector< Node > vec_conc;
2867 vec_conc.push_back(conc1); vec_conc.push_back(conc2); vec_conc.push_back(conc3);
2868 vec_conc.push_back(str_in_re);
2869 //vec_conc.push_back(sk_y.eqNode(d_emptyString).negate());//by mkskolems
2870 conc = NodeManager::currentNM()->mkNode( kind::AND, vec_conc );
2871 } // normal case
2872
2873 //set its antecedant to ant, to say when it is relevant
2874 if(!str_in_re.isNull()) {
2875 d_regexp_ant[str_in_re] = ant;
2876 }
2877 //we will be done
2878 addNormalFormPair( normal_form_src[i], normal_form_src[j] );
2879 if( options::stringProcessLoop() ){
2880 info.d_conc = conc;
2881 info.d_id = 8;
2882 return true;
2883 }else{
2884 d_out->setIncomplete();
2885 }
2886 }else{
2887 Trace("strings-loop") << "Strings::Loop: loop lemma for " << ant << " has already added." << std::endl;
2888 addNormalFormPair( normal_form_src[i], normal_form_src[j] );
2889 }
2890 }
2891 }
2892 return false;
2893 }
2894
2895 //return true for lemma, false if we succeed
2896 void TheoryStrings::processDeq( Node ni, Node nj ) {
2897 //Assert( areDisequal( ni, nj ) );
2898 if( d_normal_forms[ni].size()>1 || d_normal_forms[nj].size()>1 ){
2899 std::vector< Node > nfi;
2900 nfi.insert( nfi.end(), d_normal_forms[ni].begin(), d_normal_forms[ni].end() );
2901 std::vector< Node > nfj;
2902 nfj.insert( nfj.end(), d_normal_forms[nj].begin(), d_normal_forms[nj].end() );
2903
2904 int revRet = processReverseDeq( nfi, nfj, ni, nj );
2905 if( revRet!=0 ){
2906 return;
2907 }
2908
2909 nfi.clear();
2910 nfi.insert( nfi.end(), d_normal_forms[ni].begin(), d_normal_forms[ni].end() );
2911 nfj.clear();
2912 nfj.insert( nfj.end(), d_normal_forms[nj].begin(), d_normal_forms[nj].end() );
2913
2914 unsigned index = 0;
2915 while( index<nfi.size() || index<nfj.size() ){
2916 int ret = processSimpleDeq( nfi, nfj, ni, nj, index, false );
2917 if( ret!=0 ) {
2918 return;
2919 }else{
2920 Assert( index<nfi.size() && index<nfj.size() );
2921 Node i = nfi[index];
2922 Node j = nfj[index];
2923 Trace("strings-solve-debug") << "...Processing(DEQ) " << i << " " << j << std::endl;
2924 if( !areEqual( i, j ) ){
2925 Assert( i.getKind()!=kind::CONST_STRING || j.getKind()!=kind::CONST_STRING );
2926 std::vector< Node > lexp;
2927 Node li = getLength( i, lexp );
2928 Node lj = getLength( j, lexp );
2929 if( areDisequal( li, lj ) ){
2930 if( i.getKind()==kind::CONST_STRING || j.getKind()==kind::CONST_STRING ){
2931 //check if empty
2932 Node const_k = i.getKind() == kind::CONST_STRING ? i : j;
2933 Node nconst_k = i.getKind() == kind::CONST_STRING ? j : i;
2934 Node lnck = i.getKind() == kind::CONST_STRING ? lj : li;
2935 if( !d_equalityEngine.areDisequal( nconst_k, d_emptyString, true ) ){
2936 Node eq = nconst_k.eqNode( d_emptyString );
2937 Node conc = NodeManager::currentNM()->mkNode( kind::OR, eq, eq.negate() );
2938 sendInference( d_empty_vec, conc, "D-DISL-Emp-Split" );
2939 return;
2940 }else{
2941 //split on first character
2942 CVC4::String str = const_k.getConst<String>();
2943 Node firstChar = str.size() == 1 ? const_k : NodeManager::currentNM()->mkConst( str.prefix( 1 ) );
2944 if( areEqual( lnck, d_one ) ){
2945 if( areDisequal( firstChar, nconst_k ) ){
2946 return;
2947 }else if( !areEqual( firstChar, nconst_k ) ){
2948 //splitting on demand : try to make them disequal
2949 Node eq = firstChar.eqNode( nconst_k );
2950 sendSplit( firstChar, nconst_k, "S-Split(DEQL-Const)" );
2951 eq = Rewriter::rewrite( eq );
2952 d_pending_req_phase[ eq ] = false;
2953 return;
2954 }
2955 }else{
2956 Node sk = mkSkolemCached( nconst_k, firstChar, sk_id_dc_spt, "dc_spt", 2 );
2957 Node skr = mkSkolemCached( nconst_k, firstChar, sk_id_dc_spt_rem, "dc_spt_rem" );
2958 Node eq1 = nconst_k.eqNode( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, sk, skr ) );
2959 eq1 = Rewriter::rewrite( eq1 );
2960 Node eq2 = nconst_k.eqNode( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, firstChar, skr ) );
2961 std::vector< Node > antec;
2962 antec.insert( antec.end(), d_normal_forms_exp[ni].begin(), d_normal_forms_exp[ni].end() );
2963 antec.insert( antec.end(), d_normal_forms_exp[nj].begin(), d_normal_forms_exp[nj].end() );
2964 antec.push_back( nconst_k.eqNode( d_emptyString ).negate() );
2965 sendInference( antec, NodeManager::currentNM()->mkNode( kind::OR,
2966 NodeManager::currentNM()->mkNode( kind::AND, eq1, sk.eqNode( firstChar ).negate() ), eq2 ), "D-DISL-CSplit" );
2967 d_pending_req_phase[ eq1 ] = true;
2968 return;
2969 }
2970 }
2971 }else{
2972 Trace("strings-solve") << "Non-Simple Case 1 : add lemma " << std::endl;
2973 //must add lemma
2974 std::vector< Node > antec;
2975 std::vector< Node > antec_new_lits;
2976 antec.insert( antec.end(), d_normal_forms_exp[ni].begin(), d_normal_forms_exp[ni].end() );
2977 antec.insert( antec.end(), d_normal_forms_exp[nj].begin(), d_normal_forms_exp[nj].end() );
2978 //check disequal
2979 if( areDisequal( ni, nj ) ){
2980 antec.push_back( ni.eqNode( nj ).negate() );
2981 }else{
2982 antec_new_lits.push_back( ni.eqNode( nj ).negate() );
2983 }
2984 antec_new_lits.push_back( li.eqNode( lj ).negate() );
2985 std::vector< Node > conc;
2986 Node sk1 = mkSkolemCached( i, j, sk_id_deq_x, "x_dsplit" );
2987 Node sk2 = mkSkolemCached( i, j, sk_id_deq_y, "y_dsplit" );
2988 Node sk3 = mkSkolemCached( i, j, sk_id_deq_z, "z_dsplit", 1 );
2989 //Node nemp = sk3.eqNode(d_emptyString).negate();
2990 //conc.push_back(nemp);
2991 Node lsk1 = mkLength( sk1 );
2992 conc.push_back( lsk1.eqNode( li ) );
2993 Node lsk2 = mkLength( sk2 );
2994 conc.push_back( lsk2.eqNode( lj ) );
2995 conc.push_back( NodeManager::currentNM()->mkNode( kind::OR, j.eqNode( mkConcat( sk1, sk3 ) ), i.eqNode( mkConcat( sk2, sk3 ) ) ) );
2996 sendInference( antec, antec_new_lits, NodeManager::currentNM()->mkNode( kind::AND, conc ), "D-DISL-Split" );
2997 ++(d_statistics.d_deq_splits);
2998 return;
2999 }
3000 }else if( areEqual( li, lj ) ){
3001 Assert( !areDisequal( i, j ) );
3002 //splitting on demand : try to make them disequal
3003 Node eq = i.eqNode( j );
3004 sendSplit( i, j, "S-Split(DEQL)" );
3005 eq = Rewriter::rewrite( eq );
3006 d_pending_req_phase[ eq ] = false;
3007 return;
3008 }else{
3009 //splitting on demand : try to make lengths equal
3010 Node eq = li.eqNode( lj );
3011 sendSplit( li, lj, "D-Split" );
3012 eq = Rewriter::rewrite( eq );
3013 d_pending_req_phase[ eq ] = true;
3014 return;
3015 }
3016 }
3017 index++;
3018 }
3019 }
3020 Assert( false );
3021 }
3022 }
3023
3024 int TheoryStrings::processReverseDeq( std::vector< Node >& nfi, std::vector< Node >& nfj, Node ni, Node nj ) {
3025 //reverse normal form of i, j
3026 std::reverse( nfi.begin(), nfi.end() );
3027 std::reverse( nfj.begin(), nfj.end() );
3028
3029 unsigned index = 0;
3030 int ret = processSimpleDeq( nfi, nfj, ni, nj, index, true );
3031
3032 //reverse normal form of i, j
3033 std::reverse( nfi.begin(), nfi.end() );
3034 std::reverse( nfj.begin(), nfj.end() );
3035
3036 return ret;
3037 }
3038
3039 int TheoryStrings::processSimpleDeq( std::vector< Node >& nfi, std::vector< Node >& nfj, Node ni, Node nj, unsigned& index, bool isRev ){
3040 //see if one side is constant, if so, we can approximate as containment
3041 for( unsigned i=0; i<2; i++ ){
3042 Node c = getConstantEqc( i==0 ? ni : nj );
3043 if( !c.isNull() ){
3044 int findex, lindex;
3045 if( !TheoryStringsRewriter::canConstantContainList( c, i==0 ? nfj : nfi, findex, lindex ) ){
3046 return 1;
3047 }
3048 }
3049 }
3050 while( index<nfi.size() || index<nfj.size() ) {
3051 if( index>=nfi.size() || index>=nfj.size() ){
3052 Trace("strings-solve-debug") << "Disequality normalize empty" << std::endl;
3053 std::vector< Node > ant;
3054 //we have a conflict : because the lengths are equal, the remainder needs to be empty, which will lead to a conflict
3055 Node lni = getLengthExp( ni, ant, d_normal_forms_base[ni] );
3056 Node lnj = getLengthExp( nj, ant, d_normal_forms_base[nj] );
3057 ant.push_back( lni.eqNode( lnj ) );
3058 ant.insert( ant.end(), d_normal_forms_exp[ni].begin(), d_normal_forms_exp[ni].end() );
3059 ant.insert( ant.end(), d_normal_forms_exp[nj].begin(), d_normal_forms_exp[nj].end() );
3060 std::vector< Node > cc;
3061 std::vector< Node >& nfk = index>=nfi.size() ? nfj : nfi;
3062 for( unsigned index_k=index; index_k<nfk.size(); index_k++ ){
3063 cc.push_back( nfk[index_k].eqNode( d_emptyString ) );
3064 }
3065 Node conc = cc.size()==1 ? cc[0] : NodeManager::currentNM()->mkNode( kind::AND, cc );
3066 conc = Rewriter::rewrite( conc );
3067 sendInference( ant, conc, "Disequality Normalize Empty", true);
3068 return -1;
3069 }else{
3070 Node i = nfi[index];
3071 Node j = nfj[index];
3072 Trace("strings-solve-debug") << "...Processing(QED) " << i << " " << j << std::endl;
3073 if( !areEqual( i, j ) ) {
3074 if( i.getKind()==kind::CONST_STRING && j.getKind()==kind::CONST_STRING ) {
3075 unsigned int len_short = i.getConst<String>().size() < j.getConst<String>().size() ? i.getConst<String>().size() : j.getConst<String>().size();
3076 bool isSameFix = isRev ? i.getConst<String>().rstrncmp(j.getConst<String>(), len_short): i.getConst<String>().strncmp(j.getConst<String>(), len_short);
3077 if( isSameFix ) {
3078 //same prefix/suffix
3079 //k is the index of the string that is shorter
3080 Node nk = i.getConst<String>().size() < j.getConst<String>().size() ? i : j;
3081 Node nl = i.getConst<String>().size() < j.getConst<String>().size() ? j : i;
3082 Node remainderStr;
3083 if( isRev ){
3084 int new_len = nl.getConst<String>().size() - len_short;
3085 remainderStr = NodeManager::currentNM()->mkConst( nl.getConst<String>().substr(0, new_len) );
3086 Trace("strings-solve-debug-test") << "Rev. Break normal form of " << nl << " into " << nk << ", " << remainderStr << std::endl;
3087 } else {
3088 remainderStr = NodeManager::currentNM()->mkConst( nl.getConst<String>().substr( len_short ) );
3089 Trace("strings-solve-debug-test") << "Break normal form of " << nl << " into " << nk << ", " << remainderStr << std::endl;
3090 }
3091 if( i.getConst<String>().size() < j.getConst<String>().size() ) {
3092 nfj.insert( nfj.begin() + index + 1, remainderStr );
3093 nfj[index] = nfi[index];
3094 } else {
3095 nfi.insert( nfi.begin() + index + 1, remainderStr );
3096 nfi[index] = nfj[index];
3097 }
3098 }else{
3099 return 1;
3100 }
3101 }else{
3102 std::vector< Node > lexp;
3103 Node li = getLength( i, lexp );
3104 Node lj = getLength( j, lexp );
3105 if( areEqual( li, lj ) && areDisequal( i, j ) ){
3106 Trace("strings-solve") << "Simple Case 2 : found equal length disequal sub strings " << i << " " << j << std::endl;
3107 //we are done: D-Remove
3108 return 1;
3109 }else{
3110 return 0;
3111 }
3112 }
3113 }
3114 index++;
3115 }
3116 }
3117 return 0;
3118 }
3119
3120 void TheoryStrings::addNormalFormPair( Node n1, Node n2 ){
3121 if( !isNormalFormPair( n1, n2 ) ){
3122 int index = 0;
3123 NodeIntMap::const_iterator it = d_nf_pairs.find( n1 );
3124 if( it!=d_nf_pairs.end() ){
3125 index = (*it).second;
3126 }
3127 d_nf_pairs[n1] = index + 1;
3128 if( index<(int)d_nf_pairs_data[n1].size() ){
3129 d_nf_pairs_data[n1][index] = n2;
3130 }else{
3131 d_nf_pairs_data[n1].push_back( n2 );
3132 }
3133 Assert( isNormalFormPair( n1, n2 ) );
3134 } else {
3135 Trace("strings-nf-debug") << "Already a normal form pair " << n1 << " " << n2 << std::endl;
3136 }
3137 }
3138
3139 bool TheoryStrings::isNormalFormPair( Node n1, Node n2 ) {
3140 //TODO: modulo equality?
3141 return isNormalFormPair2( n1, n2 ) || isNormalFormPair2( n2, n1 );
3142 }
3143
3144 bool TheoryStrings::isNormalFormPair2( Node n1, Node n2 ) {
3145 //Trace("strings-debug") << "is normal form pair. " << n1 << " " << n2 << std::endl;
3146 NodeIntMap::const_iterator it = d_nf_pairs.find( n1 );
3147 if( it!=d_nf_pairs.end() ){
3148 Assert( d_nf_pairs_data.find( n1 )!=d_nf_pairs_data.end() );
3149 for( int i=0; i<(*it).second; i++ ){
3150 Assert( i<(int)d_nf_pairs_data[n1].size() );
3151 if( d_nf_pairs_data[n1][i]==n2 ){
3152 return true;
3153 }
3154 }
3155 }
3156 return false;
3157 }
3158
3159 void TheoryStrings::registerTerm( Node n, int effort ) {
3160 // 0 : upon preregistration or internal assertion
3161 // 1 : upon occurrence in length term
3162 // 2 : before normal form computation
3163 // 3 : called on normal form terms
3164 bool do_register = false;
3165 if( options::stringEagerLen() ){
3166 do_register = effort==0;
3167 }else{
3168 do_register = effort>0 || n.getKind()!=kind::STRING_CONCAT;
3169 }
3170 if( do_register ){
3171 if(d_registered_terms_cache.find(n) == d_registered_terms_cache.end()) {
3172 d_registered_terms_cache.insert(n);
3173 Debug("strings-register") << "TheoryStrings::registerTerm() " << n << ", effort = " << effort << std::endl;
3174 if(n.getType().isString()) {
3175 //register length information:
3176 // for variables, split on empty vs positive length
3177 // for concat/const/replace, introduce proxy var and state length relation
3178 Node lsum;
3179 bool processed = false;
3180 if( n.getKind()!=kind::STRING_CONCAT && n.getKind()!=kind::CONST_STRING ) {
3181 if( d_length_lemma_terms_cache.find( n )==d_length_lemma_terms_cache.end() ){
3182 Node lsumb = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, n );
3183 lsum = Rewriter::rewrite( lsumb );
3184 // can register length term if it does not rewrite
3185 if( lsum==lsumb ){
3186 sendLengthLemma( n );
3187 processed = true;
3188 }
3189 }else{
3190 processed = true;
3191 }
3192 }
3193 if( !processed ){
3194 Node sk = mkSkolemS( "lsym", -1 );
3195 StringsProxyVarAttribute spva;
3196 sk.setAttribute(spva,true);
3197 Node eq = Rewriter::rewrite( sk.eqNode(n) );
3198 Trace("strings-lemma") << "Strings::Lemma LENGTH Term : " << eq << std::endl;
3199 d_proxy_var[n] = sk;
3200 Trace("strings-assert") << "(assert " << eq << ")" << std::endl;
3201 d_out->lemma(eq);
3202 Node skl = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, sk );
3203 if( n.getKind()==kind::STRING_CONCAT ){
3204 std::vector<Node> node_vec;
3205 for( unsigned i=0; i<n.getNumChildren(); i++ ) {
3206 if( n[i].getAttribute(StringsProxyVarAttribute()) ){
3207 Assert( d_proxy_var_to_length.find( n[i] )!=d_proxy_var_to_length.end() );
3208 node_vec.push_back( d_proxy_var_to_length[n[i]] );
3209 }else{
3210 Node lni = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, n[i] );
3211 node_vec.push_back(lni);
3212 }
3213 }
3214 lsum = NodeManager::currentNM()->mkNode( kind::PLUS, node_vec );
3215 lsum = Rewriter::rewrite( lsum );
3216 }else if( n.getKind()==kind::CONST_STRING ){
3217 lsum = NodeManager::currentNM()->mkConst( ::CVC4::Rational( n.getConst<String>().size() ) );
3218 }
3219 Assert( !lsum.isNull() );
3220 d_proxy_var_to_length[sk] = lsum;
3221 Node ceq = Rewriter::rewrite( skl.eqNode( lsum ) );
3222 Trace("strings-lemma") << "Strings::Lemma LENGTH : " << ceq << std::endl;
3223 Trace("strings-lemma-debug") << " prerewrite : " << skl.eqNode( lsum ) << std::endl;
3224 Trace("strings-assert") << "(assert " << ceq << ")" << std::endl;
3225 d_out->lemma(ceq);
3226
3227 }
3228 } else {
3229 AlwaysAssert(false, "String Terms only in registerTerm.");
3230 }
3231 }
3232 }
3233 }
3234
3235 void TheoryStrings::sendInference( std::vector< Node >& exp, std::vector< Node >& exp_n, Node eq, const char * c, bool asLemma ) {
3236 eq = eq.isNull() ? d_false : Rewriter::rewrite( eq );
3237 if( eq!=d_true ){
3238 if( Trace.isOn("strings-infer-debug") ){
3239 Trace("strings-infer-debug") << "By " << c << ", infer : " << eq << " from: " << std::endl;
3240 for( unsigned i=0; i<exp.size(); i++ ){
3241 Trace("strings-infer-debug") << " " << exp[i] << std::endl;
3242 }
3243 for( unsigned i=0; i<exp_n.size(); i++ ){
3244 Trace("strings-infer-debug") << " N:" << exp_n[i] << std::endl;
3245 }
3246 //Trace("strings-infer-debug") << "as lemma : " << asLemma << std::endl;
3247 }
3248 //check if we should send a lemma or an inference
3249 if( asLemma || eq==d_false || eq.getKind()==kind::OR || !exp_n.empty() || options::stringInferAsLemmas() ){
3250 Node eq_exp;
3251 if( options::stringRExplainLemmas() ){
3252 eq_exp = mkExplain( exp, exp_n );
3253 }else{
3254 if( exp.empty() ){
3255 eq_exp = mkAnd( exp_n );
3256 }else if( exp_n.empty() ){
3257 eq_exp = mkAnd( exp );
3258 }else{
3259 std::vector< Node > ev;
3260 ev.insert( ev.end(), exp.begin(), exp.end() );
3261 ev.insert( ev.end(), exp_n.begin(), exp_n.end() );
3262 eq_exp = NodeManager::currentNM()->mkNode( kind::AND, ev );
3263 }
3264 }
3265 sendLemma( eq_exp, eq, c );
3266 }else{
3267 sendInfer( mkAnd( exp ), eq, c );
3268 }
3269 }
3270 }
3271
3272 void TheoryStrings::sendInference( std::vector< Node >& exp, Node eq, const char * c, bool asLemma ) {
3273 std::vector< Node > exp_n;
3274 sendInference( exp, exp_n, eq, c, asLemma );
3275 }
3276
3277 void TheoryStrings::sendLemma( Node ant, Node conc, const char * c ) {
3278 if( conc.isNull() || conc == d_false ) {
3279 d_out->conflict(ant);
3280 Trace("strings-conflict") << "Strings::Conflict : " << c << " : " << ant << std::endl;
3281 Trace("strings-lemma") << "Strings::Conflict : " << c << " : " << ant << std::endl;
3282 Trace("strings-assert") << "(assert (not " << ant << ")) ; conflict " << c << std::endl;
3283 d_conflict = true;
3284 } else {
3285 Node lem;
3286 if( ant == d_true ) {
3287 lem = conc;
3288 }else{
3289 lem = NodeManager::currentNM()->mkNode( kind::IMPLIES, ant, conc );
3290 }
3291 Trace("strings-lemma") << "Strings::Lemma " << c << " : " << lem << std::endl;
3292 Trace("strings-assert") << "(assert " << lem << ") ; lemma " << c << std::endl;
3293 d_lemma_cache.push_back( lem );
3294 }
3295 }
3296
3297 void TheoryStrings::sendInfer( Node eq_exp, Node eq, const char * c ) {
3298 if( options::stringInferSym() ){
3299 std::vector< Node > vars;
3300 std::vector< Node > subs;
3301 std::vector< Node > unproc;
3302 inferSubstitutionProxyVars( eq_exp, vars, subs, unproc );
3303 if( unproc.empty() ){
3304 Trace("strings-lemma-debug") << "Strings::Infer " << eq << " from " << eq_exp << " by " << c << std::endl;
3305 Node eqs = eq.substitute( vars.begin(), vars.end(), subs.begin(), subs.end() );
3306 Trace("strings-lemma-debug") << "Strings::Infer Alternate : " << eqs << std::endl;
3307 for( unsigned i=0; i<vars.size(); i++ ){
3308 Trace("strings-lemma-debug") << " " << vars[i] << " -> " << subs[i] << std::endl;
3309 }
3310 sendLemma( d_true, eqs, c );
3311 return;
3312 }else{
3313 for( unsigned i=0; i<unproc.size(); i++ ){
3314 Trace("strings-lemma-debug") << " non-trivial exp : " << unproc[i] << std::endl;
3315 }
3316 }
3317 }
3318 Trace("strings-lemma") << "Strings::Infer " << eq << " from " << eq_exp << " by " << c << std::endl;
3319 Trace("strings-assert") << "(assert (=> " << eq_exp << " " << eq << ")) ; infer " << c << std::endl;
3320 d_pending.push_back( eq );
3321 d_pending_exp[eq] = eq_exp;
3322 d_infer.push_back( eq );
3323 d_infer_exp.push_back( eq_exp );
3324
3325 }
3326
3327 void TheoryStrings::sendSplit( Node a, Node b, const char * c, bool preq ) {
3328 Node eq = a.eqNode( b );
3329 eq = Rewriter::rewrite( eq );
3330 Node neq = NodeManager::currentNM()->mkNode( kind::NOT, eq );
3331 Node lemma_or = NodeManager::currentNM()->mkNode( kind::OR, eq, neq );
3332 Trace("strings-lemma") << "Strings::Lemma " << c << " SPLIT : " << lemma_or << std::endl;
3333 d_lemma_cache.push_back(lemma_or);
3334 d_pending_req_phase[eq] = preq;
3335 ++(d_statistics.d_splits);
3336 }
3337
3338
3339 void TheoryStrings::sendLengthLemma( Node n ){
3340 Node n_len = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, n);
3341 if( options::stringSplitEmp() || !options::stringLenGeqZ() ){
3342 Node n_len_eq_z = n_len.eqNode( d_zero );
3343 Node n_len_eq_z_2 = n.eqNode( d_emptyString );
3344 n_len_eq_z = Rewriter::rewrite( n_len_eq_z );
3345 n_len_eq_z_2 = Rewriter::rewrite( n_len_eq_z_2 );
3346 Node n_len_geq_zero = NodeManager::currentNM()->mkNode( kind::OR, NodeManager::currentNM()->mkNode( kind::AND, n_len_eq_z, n_len_eq_z_2 ),
3347 NodeManager::currentNM()->mkNode( kind::GT, n_len, d_zero) );
3348 Trace("strings-lemma") << "Strings::Lemma LENGTH >= 0 : " << n_len_geq_zero << std::endl;
3349 d_out->lemma(n_len_geq_zero);
3350 d_out->requirePhase( n_len_eq_z, true );
3351 d_out->requirePhase( n_len_eq_z_2, true );
3352 }
3353 //AJR: probably a good idea
3354 if( options::stringLenGeqZ() ){
3355 Node n_len_geq = NodeManager::currentNM()->mkNode( kind::GEQ, n_len, d_zero);
3356 n_len_geq = Rewriter::rewrite( n_len_geq );
3357 d_out->lemma( n_len_geq );
3358 }
3359 }
3360
3361 void TheoryStrings::inferSubstitutionProxyVars( Node n, std::vector< Node >& vars, std::vector< Node >& subs, std::vector< Node >& unproc ) {
3362 if( n.getKind()==kind::AND ){
3363 for( unsigned i=0; i<n.getNumChildren(); i++ ){
3364 inferSubstitutionProxyVars( n[i], vars, subs, unproc );
3365 }
3366 return;
3367 }else if( n.getKind()==kind::EQUAL ){
3368 Node ns = n.substitute( vars.begin(), vars.end(), subs.begin(), subs.end() );
3369 ns = Rewriter::rewrite( ns );
3370 if( ns.getKind()==kind::EQUAL ){
3371 Node s;
3372 Node v;
3373 for( unsigned i=0; i<2; i++ ){
3374 Node ss;
3375 if( ns[i].getAttribute(StringsProxyVarAttribute()) ){
3376 ss = ns[i];
3377 }else if( ns[i].isConst() ){
3378 NodeNodeMap::const_iterator it = d_proxy_var.find( ns[i] );
3379 if( it!=d_proxy_var.end() ){
3380 ss = (*it).second;
3381 }
3382 }
3383 if( !ss.isNull() ){
3384 v = ns[1-i];
3385 if( v.getNumChildren()==0 ){
3386 if( s.isNull() ){
3387 s = ss;
3388 }else{
3389 //both sides involved in proxy var
3390 if( ss==s ){
3391 return;
3392 }else{
3393 s = Node::null();
3394 }
3395 }
3396 }
3397 }
3398 }
3399 if( !s.isNull() ){
3400 subs.push_back( s );
3401 vars.push_back( v );
3402 return;
3403 }
3404 }else{
3405 n = ns;
3406 }
3407 }
3408 if( n!=d_true ){
3409 unproc.push_back( n );
3410 }
3411 }
3412
3413
3414 Node TheoryStrings::mkConcat( Node n1, Node n2 ) {
3415 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, n1, n2 ) );
3416 }
3417
3418 Node TheoryStrings::mkConcat( Node n1, Node n2, Node n3 ) {
3419 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, n1, n2, n3 ) );
3420 }
3421
3422 Node TheoryStrings::mkConcat( const std::vector< Node >& c ) {
3423 return Rewriter::rewrite( c.size()>1 ? NodeManager::currentNM()->mkNode( kind::STRING_CONCAT, c ) : ( c.size()==1 ? c[0] : d_emptyString ) );
3424 }
3425
3426 Node TheoryStrings::mkLength( Node t ) {
3427 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, t ) );
3428 }
3429
3430 Node TheoryStrings::mkSkolemCached( Node a, Node b, int id, const char * c, int isLenSplit ){
3431 //return mkSkolemS( c, isLenSplit );
3432 std::map< int, Node >::iterator it = d_skolem_cache[a][b].find( id );
3433 if( it==d_skolem_cache[a][b].end() ){
3434 Node sk = mkSkolemS( c, isLenSplit );
3435 d_skolem_cache[a][b][id] = sk;
3436 return sk;
3437 }else{
3438 return it->second;
3439 }
3440 }
3441
3442 //isLenSplit: -1-ignore, 0-no restriction, 1-greater than one, 2-one
3443 Node TheoryStrings::mkSkolemS( const char *c, int isLenSplit ) {
3444 Node n = NodeManager::currentNM()->mkSkolem( c, NodeManager::currentNM()->stringType(), "string sko" );
3445 d_length_lemma_terms_cache.insert( n );
3446 ++(d_statistics.d_new_skolems);
3447 if( isLenSplit==0 ){
3448 sendLengthLemma( n );
3449 } else if( isLenSplit == 1 ){
3450 registerNonEmptySkolem( n );
3451 }else if( isLenSplit==2 ){
3452 Node len_one = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, n ).eqNode( d_one );
3453 Trace("strings-lemma") << "Strings::Lemma SK-ONE : " << len_one << std::endl;
3454 Trace("strings-assert") << "(assert " << len_one << ")" << std::endl;
3455 d_out->lemma( len_one );
3456 }
3457 return n;
3458 }
3459
3460 void TheoryStrings::registerNonEmptySkolem( Node n ) {
3461 if( d_skolem_ne_reg_cache.find( n )==d_skolem_ne_reg_cache.end() ){
3462 d_skolem_ne_reg_cache.insert( n );
3463 d_equalityEngine.assertEquality(n.eqNode(d_emptyString), false, d_true);
3464 Node len_n_gt_z = NodeManager::currentNM()->mkNode(kind::GT,
3465 NodeManager::currentNM()->mkNode(kind::STRING_LENGTH, n), d_zero);
3466 Trace("strings-lemma") << "Strings::Lemma SK-NON-ZERO : " << len_n_gt_z << std::endl;
3467 Trace("strings-assert") << "(assert " << len_n_gt_z << ")" << std::endl;
3468 d_out->lemma(len_n_gt_z);
3469 }
3470 }
3471
3472 Node TheoryStrings::mkExplain( std::vector< Node >& a ) {
3473 std::vector< Node > an;
3474 return mkExplain( a, an );
3475 }
3476
3477 Node TheoryStrings::mkExplain( std::vector< Node >& a, std::vector< Node >& an ) {
3478 std::vector< TNode > antec_exp;
3479 for( unsigned i=0; i<a.size(); i++ ) {
3480 if( std::find( a.begin(), a.begin() + i, a[i] )==a.begin() + i ) {
3481 bool exp = true;
3482 Debug("strings-explain") << "Ask for explanation of " << a[i] << std::endl;
3483 //assert
3484 if(a[i].getKind() == kind::EQUAL) {
3485 //assert( hasTerm(a[i][0]) );
3486 //assert( hasTerm(a[i][1]) );
3487 Assert( areEqual(a[i][0], a[i][1]) );
3488 if( a[i][0]==a[i][1] ){
3489 exp = false;
3490 }
3491 } else if( a[i].getKind()==kind::NOT && a[i][0].getKind()==kind::EQUAL ) {
3492 Assert( hasTerm(a[i][0][0]) );
3493 Assert( hasTerm(a[i][0][1]) );
3494 AlwaysAssert( d_equalityEngine.areDisequal(a[i][0][0], a[i][0][1], true) );
3495 }else if( a[i].getKind() == kind::AND ){
3496 for( unsigned j=0; j<a[i].getNumChildren(); j++ ){
3497 a.push_back( a[i][j] );
3498 }
3499 exp = false;
3500 }
3501 if( exp ) {
3502 unsigned ps = antec_exp.size();
3503 explain(a[i], antec_exp);
3504 Debug("strings-explain") << "Done, explanation was : " << std::endl;
3505 for( unsigned j=ps; j<antec_exp.size(); j++ ) {
3506 Debug("strings-explain") << " " << antec_exp[j] << std::endl;
3507 }
3508 Debug("strings-explain") << std::endl;
3509 }
3510 }
3511 }
3512 for( unsigned i=0; i<an.size(); i++ ) {
3513 if( std::find( an.begin(), an.begin() + i, an[i] )==an.begin() + i ){
3514 Debug("strings-explain") << "Add to explanation (new literal) " << an[i] << std::endl;
3515 antec_exp.push_back(an[i]);
3516 }
3517 }
3518 Node ant;
3519 if( antec_exp.empty() ) {
3520 ant = d_true;
3521 } else if( antec_exp.size()==1 ) {
3522 ant = antec_exp[0];
3523 } else {
3524 ant = NodeManager::currentNM()->mkNode( kind::AND, antec_exp );
3525 }
3526 ant = Rewriter::rewrite( ant );
3527 return ant;
3528 }
3529
3530 Node TheoryStrings::mkAnd( std::vector< Node >& a ) {
3531 std::vector< Node > au;
3532 for( unsigned i=0; i<a.size(); i++ ){
3533 if( std::find( au.begin(), au.end(), a[i] )==au.end() ){
3534 au.push_back( a[i] );
3535 }
3536 }
3537 if( au.empty() ) {
3538 return d_true;
3539 } else if( au.size() == 1 ) {
3540 return au[0];
3541 } else {
3542 return NodeManager::currentNM()->mkNode( kind::AND, au );
3543 }
3544 }
3545
3546 void TheoryStrings::getConcatVec( Node n, std::vector< Node >& c ) {
3547 if( n.getKind()==kind::STRING_CONCAT ) {
3548 for( unsigned i=0; i<n.getNumChildren(); i++ ) {
3549 if( !areEqual( n[i], d_emptyString ) ) {
3550 c.push_back( n[i] );
3551 }
3552 }
3553 }else{
3554 c.push_back( n );
3555 }
3556 }
3557
3558 void TheoryStrings::checkDeqNF() {
3559 std::vector< std::vector< Node > > cols;
3560 std::vector< Node > lts;
3561 std::map< Node, std::map< Node, bool > > processed;
3562
3563 //for each pair of disequal strings, must determine whether their lengths are equal or disequal
3564 for( NodeList::const_iterator id = d_ee_disequalities.begin(); id != d_ee_disequalities.end(); ++id ) {
3565 Node eq = *id;
3566 Node n[2];
3567 for( unsigned i=0; i<2; i++ ){
3568 n[i] = d_equalityEngine.getRepresentative( eq[i] );
3569 }
3570 if( processed[n[0]].find( n[1] )==processed[n[0]].end() ){
3571 processed[n[0]][n[1]] = true;
3572 Node lt[2];
3573 for( unsigned i=0; i<2; i++ ){
3574 EqcInfo* ei = getOrMakeEqcInfo( n[i], false );
3575 lt[i] = ei ? ei->d_length_term : Node::null();
3576 if( lt[i].isNull() ){
3577 lt[i] = eq[i];
3578 }
3579 lt[i] = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, lt[i] );
3580 }
3581 if( !areEqual( lt[0], lt[1] ) && !areDisequal( lt[0], lt[1] ) ){
3582 sendSplit( lt[0], lt[1], "DEQ-LENGTH-SP" );
3583 }
3584 }
3585 }
3586
3587 if( !hasProcessed() ){
3588 separateByLength( d_strings_eqc, cols, lts );
3589 for( unsigned i=0; i<cols.size(); i++ ){
3590 if( cols[i].size()>1 && d_lemma_cache.empty() ){
3591 Trace("strings-solve") << "- Verify disequalities are processed for " << cols[i][0] << ", normal form : ";
3592 printConcat( d_normal_forms[cols[i][0]], "strings-solve" );
3593 Trace("strings-solve") << "... #eql = " << cols[i].size() << std::endl;
3594 //must ensure that normal forms are disequal
3595 for( unsigned j=0; j<cols[i].size(); j++ ){
3596 for( unsigned k=(j+1); k<cols[i].size(); k++ ){
3597 //for strings that are disequal, but have the same length
3598 if( areDisequal( cols[i][j], cols[i][k] ) ){
3599 Assert( !d_conflict );
3600 Trace("strings-solve") << "- Compare " << cols[i][j] << " ";
3601 printConcat( d_normal_forms[cols[i][j]], "strings-solve" );
3602 Trace("strings-solve") << " against " << cols[i][k] << " ";
3603 printConcat( d_normal_forms[cols[i][k]], "strings-solve" );
3604 Trace("strings-solve") << "..." << std::endl;
3605 processDeq( cols[i][j], cols[i][k] );
3606 if( hasProcessed() ){
3607 return;
3608 }
3609 }
3610 }
3611 }
3612 }
3613 }
3614 }
3615 }
3616
3617 void TheoryStrings::checkLengthsEqc() {
3618 if( options::stringLenNorm() ){
3619 for( unsigned i=0; i<d_strings_eqc.size(); i++ ){
3620 //if( d_normal_forms[nodes[i]].size()>1 ) {
3621 Trace("strings-process-debug") << "Process length constraints for " << d_strings_eqc[i] << std::endl;
3622 //check if there is a length term for this equivalence class
3623 EqcInfo* ei = getOrMakeEqcInfo( d_strings_eqc[i], false );
3624 Node lt = ei ? ei->d_length_term : Node::null();
3625 if( !lt.isNull() ) {
3626 Node llt = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, lt );
3627 //now, check if length normalization has occurred
3628 if( ei->d_normalized_length.get().isNull() ) {
3629 Node nf = mkConcat( d_normal_forms[d_strings_eqc[i]] );
3630 if( Trace.isOn("strings-process-debug") ){
3631 Trace("strings-process-debug") << " normal form is " << nf << " from base " << d_normal_forms_base[d_strings_eqc[i]] << std::endl;
3632 Trace("strings-process-debug") << " normal form exp is: " << std::endl;
3633 for( unsigned j=0; j<d_normal_forms_exp[d_strings_eqc[i]].size(); j++ ){
3634 Trace("strings-process-debug") << " " << d_normal_forms_exp[d_strings_eqc[i]][j] << std::endl;
3635 }
3636 }
3637
3638 //if not, add the lemma
3639 std::vector< Node > ant;
3640 ant.insert( ant.end(), d_normal_forms_exp[d_strings_eqc[i]].begin(), d_normal_forms_exp[d_strings_eqc[i]].end() );
3641 ant.push_back( d_normal_forms_base[d_strings_eqc[i]].eqNode( lt ) );
3642 Node lc = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, nf );
3643 Node lcr = Rewriter::rewrite( lc );
3644 Trace("strings-process-debug") << "Rewrote length " << lc << " to " << lcr << std::endl;
3645 Node eq = llt.eqNode( lcr );
3646 if( llt!=lcr ){
3647 ei->d_normalized_length.set( eq );
3648 sendInference( ant, eq, "LEN-NORM", true );
3649 }
3650 }
3651 }else{
3652 Trace("strings-process-debug") << "No length term for eqc " << d_strings_eqc[i] << " " << d_eqc_to_len_term[d_strings_eqc[i]] << std::endl;
3653 if( !options::stringEagerLen() ){
3654 Node c = mkConcat( d_normal_forms[d_strings_eqc[i]] );
3655 registerTerm( c, 3 );
3656 /*
3657 if( !c.isConst() ){
3658 NodeNodeMap::const_iterator it = d_proxy_var.find( c );
3659 if( it!=d_proxy_var.end() ){
3660 Node pv = (*it).second;
3661 Assert( d_proxy_var_to_length.find( pv )!=d_proxy_var_to_length.end() );
3662 Node pvl = d_proxy_var_to_length[pv];
3663 Node ceq = Rewriter::rewrite( mkLength( pv ).eqNode( pvl ) );
3664 sendInference( d_empty_vec, ceq, "LEN-NORM-I", true );
3665 }
3666 }
3667 */
3668 }
3669 }
3670 //} else {
3671 // Trace("strings-process-debug") << "Do not process length constraints for " << nodes[i] << " " << d_normal_forms[nodes[i]].size() << std::endl;
3672 //}
3673 }
3674 }
3675 }
3676
3677 void TheoryStrings::checkCardinality() {
3678 //int cardinality = options::stringCharCardinality();
3679 //Trace("strings-solve-debug2") << "get cardinality: " << cardinality << endl;
3680
3681 //AJR: this will create a partition of eqc, where each collection has length that are pairwise propagated to be equal.
3682 // 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).
3683 // TODO: revisit this?
3684 std::vector< std::vector< Node > > cols;
3685 std::vector< Node > lts;
3686 separateByLength( d_strings_eqc, cols, lts );
3687
3688 for( unsigned i = 0; i<cols.size(); ++i ) {
3689 Node lr = lts[i];
3690 Trace("strings-card") << "Number of strings with length equal to " << lr << " is " << cols[i].size() << std::endl;
3691 if( cols[i].size() > 1 ) {
3692 // size > c^k
3693 unsigned card_need = 1;
3694 double curr = (double)cols[i].size();
3695 while( curr>d_card_size ){
3696 curr = curr/(double)d_card_size;
3697 card_need++;
3698 }
3699 Trace("strings-card") << "Need length " << card_need << " for this number of strings (where alphabet size is " << d_card_size << ")." << std::endl;
3700 Node cmp = NodeManager::currentNM()->mkNode( kind::GEQ, lr, NodeManager::currentNM()->mkConst( Rational( card_need ) ) );
3701 cmp = Rewriter::rewrite( cmp );
3702 if( cmp!=d_true ){
3703 unsigned int int_k = (unsigned int)card_need;
3704 for( std::vector< Node >::iterator itr1 = cols[i].begin();
3705 itr1 != cols[i].end(); ++itr1) {
3706 for( std::vector< Node >::iterator itr2 = itr1 + 1;
3707 itr2 != cols[i].end(); ++itr2) {
3708 if(!areDisequal( *itr1, *itr2 )) {
3709 // add split lemma
3710 sendSplit( *itr1, *itr2, "CARD-SP" );
3711 return;
3712 }
3713 }
3714 }
3715 EqcInfo* ei = getOrMakeEqcInfo( lr, true );
3716 Trace("strings-card") << "Previous cardinality used for " << lr << " is " << ((int)ei->d_cardinality_lem_k.get()-1) << std::endl;
3717 if( int_k+1 > ei->d_cardinality_lem_k.get() ){
3718 Node k_node = NodeManager::currentNM()->mkConst( ::CVC4::Rational( int_k ) );
3719 //add cardinality lemma
3720 Node dist = NodeManager::currentNM()->mkNode( kind::DISTINCT, cols[i] );
3721 std::vector< Node > vec_node;
3722 vec_node.push_back( dist );
3723 for( std::vector< Node >::iterator itr1 = cols[i].begin();
3724 itr1 != cols[i].end(); ++itr1) {
3725 Node len = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, *itr1 );
3726 if( len!=lr ) {
3727 Node len_eq_lr = len.eqNode(lr);
3728 vec_node.push_back( len_eq_lr );
3729 }
3730 }
3731 Node len = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, cols[i][0] );
3732 Node cons = NodeManager::currentNM()->mkNode( kind::GEQ, len, k_node );
3733 cons = Rewriter::rewrite( cons );
3734 ei->d_cardinality_lem_k.set( int_k+1 );
3735 if( cons!=d_true ){
3736 sendInference( d_empty_vec, vec_node, cons, "CARDINALITY", true );
3737 return;
3738 }
3739 }
3740 }
3741 }
3742 }
3743 }
3744
3745 void TheoryStrings::getEquivalenceClasses( std::vector< Node >& eqcs ) {
3746 eq::EqClassesIterator eqcs_i = eq::EqClassesIterator( &d_equalityEngine );
3747 while( !eqcs_i.isFinished() ) {
3748 Node eqc = (*eqcs_i);
3749 //if eqc.getType is string
3750 if (eqc.getType().isString()) {
3751 eqcs.push_back( eqc );
3752 }
3753 ++eqcs_i;
3754 }
3755 }
3756
3757 void TheoryStrings::separateByLength(std::vector< Node >& n,
3758 std::vector< std::vector< Node > >& cols,
3759 std::vector< Node >& lts ) {
3760 unsigned leqc_counter = 0;
3761 std::map< Node, unsigned > eqc_to_leqc;
3762 std::map< unsigned, Node > leqc_to_eqc;
3763 std::map< unsigned, std::vector< Node > > eqc_to_strings;
3764 for( unsigned i=0; i<n.size(); i++ ) {
3765 Node eqc = n[i];
3766 Assert( d_equalityEngine.getRepresentative(eqc)==eqc );
3767 EqcInfo* ei = getOrMakeEqcInfo( eqc, false );
3768 Node lt = ei ? ei->d_length_term : Node::null();
3769 if( !lt.isNull() ){
3770 lt = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, lt );
3771 Node r = d_equalityEngine.getRepresentative( lt );
3772 if( eqc_to_leqc.find( r )==eqc_to_leqc.end() ){
3773 eqc_to_leqc[r] = leqc_counter;
3774 leqc_to_eqc[leqc_counter] = r;
3775 leqc_counter++;
3776 }
3777 eqc_to_strings[ eqc_to_leqc[r] ].push_back( eqc );
3778 }else{
3779 eqc_to_strings[leqc_counter].push_back( eqc );
3780 leqc_counter++;
3781 }
3782 }
3783 for( std::map< unsigned, std::vector< Node > >::iterator it = eqc_to_strings.begin(); it != eqc_to_strings.end(); ++it ){
3784 cols.push_back( std::vector< Node >() );
3785 cols.back().insert( cols.back().end(), it->second.begin(), it->second.end() );
3786 lts.push_back( leqc_to_eqc[it->first] );
3787 }
3788 }
3789
3790 void TheoryStrings::printConcat( std::vector< Node >& n, const char * c ) {
3791 for( unsigned i=0; i<n.size(); i++ ){
3792 if( i>0 ) Trace(c) << " ++ ";
3793 Trace(c) << n[i];
3794 }
3795 }
3796
3797
3798
3799 //// Finite Model Finding
3800
3801 Node TheoryStrings::getNextDecisionRequest() {
3802 if( options::stringFMF() && !d_conflict ){
3803 Node in_var_lsum = d_input_var_lsum.get();
3804 //Trace("strings-fmf-debug") << "Strings::FMF: Assertion Level = " << d_valuation.getAssertionLevel() << std::endl;
3805 //initialize the term we will minimize
3806 if( in_var_lsum.isNull() && !d_input_vars.empty() ){
3807 Trace("strings-fmf-debug") << "Input variables: ";
3808 std::vector< Node > ll;
3809 for(NodeSet::key_iterator itr = d_input_vars.key_begin();
3810 itr != d_input_vars.key_end(); ++itr) {
3811 Trace("strings-fmf-debug") << " " << (*itr) ;
3812 ll.push_back( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, *itr ) );
3813 }
3814 Trace("strings-fmf-debug") << std::endl;
3815 in_var_lsum = ll.size()==1 ? ll[0] : NodeManager::currentNM()->mkNode( kind::PLUS, ll );
3816 in_var_lsum = Rewriter::rewrite( in_var_lsum );
3817 d_input_var_lsum.set( in_var_lsum );
3818 }
3819 if( !in_var_lsum.isNull() ){
3820 //Trace("strings-fmf") << "Get next decision request." << std::endl;
3821 //check if we need to decide on something
3822 int decideCard = d_curr_cardinality.get();
3823 if( d_cardinality_lits.find( decideCard )!=d_cardinality_lits.end() ){
3824 bool value;
3825 Node cnode = d_cardinality_lits[ d_curr_cardinality.get() ];
3826 if( d_valuation.hasSatValue( cnode, value ) ) {
3827 if( !value ){
3828 d_curr_cardinality.set( d_curr_cardinality.get() + 1 );
3829 decideCard = d_curr_cardinality.get();
3830 Trace("strings-fmf-debug") << "Has false SAT value, increment and decide." << std::endl;
3831 }else{
3832 decideCard = -1;
3833 Trace("strings-fmf-debug") << "Has true SAT value, do not decide." << std::endl;
3834 }
3835 }else{
3836 Trace("strings-fmf-debug") << "No SAT value, decide." << std::endl;
3837 }
3838 }
3839 if( decideCard!=-1 ){
3840 if( d_cardinality_lits.find( decideCard )==d_cardinality_lits.end() ){
3841 Node lit = NodeManager::currentNM()->mkNode( kind::LEQ, in_var_lsum, NodeManager::currentNM()->mkConst( Rational( decideCard ) ) );
3842 lit = Rewriter::rewrite( lit );
3843 d_cardinality_lits[decideCard] = lit;
3844 Node lem = NodeManager::currentNM()->mkNode( kind::OR, lit, lit.negate() );
3845 Trace("strings-fmf") << "Strings::FMF: Add decision lemma " << lem << ", decideCard = " << decideCard << std::endl;
3846 d_out->lemma( lem );
3847 d_out->requirePhase( lit, true );
3848 }
3849 Node lit = d_cardinality_lits[ decideCard ];
3850 Trace("strings-fmf") << "Strings::FMF: Decide positive on " << lit << std::endl;
3851 return lit;
3852 }
3853 }
3854 }
3855 return Node::null();
3856 }
3857
3858 Node TheoryStrings::ppRewrite(TNode atom) {
3859 Trace("strings-ppr") << "TheoryStrings::ppRewrite " << atom << std::endl;
3860 if( !options::stringLazyPreproc() ){
3861 //eager preprocess here
3862 std::vector< Node > new_nodes;
3863 Node ret = d_preproc.processAssertion( atom, new_nodes );
3864 if( ret!=atom ){
3865 Trace("strings-ppr") << " rewrote " << atom << " -> " << ret << ", with " << new_nodes.size() << " lemmas." << std::endl;
3866 for( unsigned i=0; i<new_nodes.size(); i++ ){
3867 Trace("strings-ppr") << " lemma : " << new_nodes[i] << std::endl;
3868 d_out->lemma( new_nodes[i] );
3869 }
3870 return ret;
3871 }else{
3872 Assert( new_nodes.empty() );
3873 }
3874 }
3875 return atom;
3876 }
3877
3878 void TheoryStrings::collectExtendedFuncTerms( Node n, std::map< Node, bool >& visited ) {
3879 if( visited.find( n )==visited.end() ){
3880 visited[n] = true;
3881 d_extt->registerTerm( n );
3882 for( unsigned i=0; i<n.getNumChildren(); i++ ){
3883 collectExtendedFuncTerms( n[i], visited );
3884 }
3885 }
3886 }
3887
3888 // Stats
3889 TheoryStrings::Statistics::Statistics():
3890 d_splits("TheoryStrings::NumOfSplitOnDemands", 0),
3891 d_eq_splits("TheoryStrings::NumOfEqSplits", 0),
3892 d_deq_splits("TheoryStrings::NumOfDiseqSplits", 0),
3893 d_loop_lemmas("TheoryStrings::NumOfLoops", 0),
3894 d_new_skolems("TheoryStrings::NumOfNewSkolems", 0)
3895 {
3896 smtStatisticsRegistry()->registerStat(&d_splits);
3897 smtStatisticsRegistry()->registerStat(&d_eq_splits);
3898 smtStatisticsRegistry()->registerStat(&d_deq_splits);
3899 smtStatisticsRegistry()->registerStat(&d_loop_lemmas);
3900 smtStatisticsRegistry()->registerStat(&d_new_skolems);
3901 }
3902
3903 TheoryStrings::Statistics::~Statistics(){
3904 smtStatisticsRegistry()->unregisterStat(&d_splits);
3905 smtStatisticsRegistry()->unregisterStat(&d_eq_splits);
3906 smtStatisticsRegistry()->unregisterStat(&d_deq_splits);
3907 smtStatisticsRegistry()->unregisterStat(&d_loop_lemmas);
3908 smtStatisticsRegistry()->unregisterStat(&d_new_skolems);
3909 }
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930 //// Regular Expressions
3931
3932
3933 unsigned TheoryStrings::getNumMemberships( Node n, bool isPos ) {
3934 if( isPos ){
3935 NodeIntMap::const_iterator it = d_pos_memberships.find( n );
3936 if( it!=d_pos_memberships.end() ){
3937 return (*it).second;
3938 }
3939 }else{
3940 NodeIntMap::const_iterator it = d_neg_memberships.find( n );
3941 if( it!=d_neg_memberships.end() ){
3942 return (*it).second;
3943 }
3944 }
3945 return 0;
3946 }
3947
3948 Node TheoryStrings::getMembership( Node n, bool isPos, unsigned i ) {
3949 return isPos ? d_pos_memberships_data[n][i] : d_neg_memberships_data[n][i];
3950 }
3951
3952 Node TheoryStrings::mkRegExpAntec(Node atom, Node ant) {
3953 if(d_regexp_ant.find(atom) == d_regexp_ant.end()) {
3954 return Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, ant, atom) );
3955 } else {
3956 Node n = d_regexp_ant[atom];
3957 return Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, ant, n) );
3958 }
3959 }
3960
3961 Node TheoryStrings::normalizeRegexp(Node r) {
3962 Node nf_r = r;
3963 if(d_nf_regexps.find(r) != d_nf_regexps.end()) {
3964 nf_r = d_nf_regexps[r];
3965 } else {
3966 std::vector< Node > nf_exp;
3967 if(!d_regexp_opr.checkConstRegExp(r)) {
3968 switch( r.getKind() ) {
3969 case kind::REGEXP_EMPTY:
3970 case kind::REGEXP_SIGMA: {
3971 break;
3972 }
3973 case kind::STRING_TO_REGEXP: {
3974 if(r[0].isConst()) {
3975 break;
3976 } else {
3977 if(d_normal_forms.find( r[0] ) != d_normal_forms.end()) {
3978 nf_r = mkConcat( d_normal_forms[r[0]] );
3979 Debug("regexp-nf") << "Term: " << r[0] << " has a normal form " << nf_r << std::endl;
3980 nf_exp.insert(nf_exp.end(), d_normal_forms_exp[r[0]].begin(), d_normal_forms_exp[r[0]].end());
3981 nf_r = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::STRING_TO_REGEXP, nf_r) );
3982 }
3983 }
3984 }
3985 case kind::REGEXP_CONCAT:
3986 case kind::REGEXP_UNION:
3987 case kind::REGEXP_INTER: {
3988 bool flag = false;
3989 std::vector< Node > vec_nodes;
3990 for(unsigned i=0; i<r.getNumChildren(); ++i) {
3991 Node rtmp = normalizeRegexp(r[i]);
3992 vec_nodes.push_back(rtmp);
3993 if(rtmp != r[i]) {
3994 flag = true;
3995 }
3996 }
3997 if(flag) {
3998 Node rtmp = vec_nodes.size()==1 ? vec_nodes[0] : NodeManager::currentNM()->mkNode(r.getKind(), vec_nodes);
3999 nf_r = Rewriter::rewrite( rtmp );
4000 }
4001 }
4002 case kind::REGEXP_STAR: {
4003 Node rtmp = normalizeRegexp(r[0]);
4004 if(rtmp != r[0]) {
4005 rtmp = NodeManager::currentNM()->mkNode(kind::REGEXP_STAR, rtmp);
4006 nf_r = Rewriter::rewrite( rtmp );
4007 }
4008 }
4009 default: {
4010 Unreachable();
4011 }
4012 }
4013 }
4014 d_nf_regexps[r] = nf_r;
4015 d_nf_regexps_exp[r] = nf_exp;
4016 }
4017 return nf_r;
4018 }
4019
4020 bool TheoryStrings::normalizePosMemberships(std::map< Node, std::vector< Node > > &memb_with_exps) {
4021 std::map< Node, std::vector< Node > > unprocessed_x_exps;
4022 std::map< Node, std::vector< Node > > unprocessed_memberships;
4023 std::map< Node, std::vector< Node > > unprocessed_memberships_bases;
4024 bool addLemma = false;
4025
4026 Trace("regexp-check") << "Normalizing Positive Memberships ... " << std::endl;
4027
4028 for( NodeIntMap::const_iterator itr_xr = d_pos_memberships.begin(); itr_xr != d_pos_memberships.end(); ++itr_xr ){
4029 Node x = (*itr_xr).first;
4030 Node nf_x = x;
4031 std::vector< Node > nf_x_exp;
4032 if(d_normal_forms.find( x ) != d_normal_forms.end()) {
4033 //nf_x = mkConcat( d_normal_forms[x] );
4034 nf_x_exp.insert(nf_x_exp.end(), d_normal_forms_exp[x].begin(), d_normal_forms_exp[x].end());
4035 //Debug("regexp-nf") << "Term: " << x << " has a normal form " << ret << std::endl;
4036 } else {
4037 Assert(false);
4038 }
4039 Trace("regexp-nf") << "Checking Memberships for N(" << x << ") = " << nf_x << " :" << std::endl;
4040
4041 std::vector< Node > vec_x;
4042 std::vector< Node > vec_r;
4043 unsigned n_pmem = (*itr_xr).second;
4044 Assert( getNumMemberships( x, true )==n_pmem );
4045 for( unsigned k=0; k<n_pmem; k++ ){
4046 Node r = getMembership( x, true, k );
4047 Node nf_r = normalizeRegexp( r ); //AJR: fixed (was normalizing mem #0 always)
4048 Node memb = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, nf_x, nf_r);
4049 if(d_processed_memberships.find(memb) == d_processed_memberships.end()) {
4050 if(d_regexp_opr.checkConstRegExp(nf_r)) {
4051 vec_x.push_back(x);
4052 vec_r.push_back(r);
4053 } else {
4054 Trace("regexp-nf") << "Handling Symbolic Regexp for N(" << r << ") = " << nf_r << std::endl;
4055 //TODO: handle symbolic ones
4056 addLemma = true;
4057 }
4058 d_processed_memberships.insert(memb);
4059 }
4060 }
4061 if(!vec_x.empty()) {
4062 if(unprocessed_x_exps.find(nf_x) == unprocessed_x_exps.end()) {
4063 unprocessed_x_exps[nf_x] = nf_x_exp;
4064 unprocessed_memberships[nf_x] = vec_r;
4065 unprocessed_memberships_bases[nf_x] = vec_x;
4066 } else {
4067 unprocessed_x_exps[nf_x].insert(unprocessed_x_exps[nf_x].end(), nf_x_exp.begin(), nf_x_exp.end());
4068 unprocessed_memberships[nf_x].insert(unprocessed_memberships[nf_x].end(), vec_r.begin(), vec_r.end());
4069 unprocessed_memberships_bases[nf_x].insert(unprocessed_memberships_bases[nf_x].end(), vec_x.begin(), vec_x.end());
4070 }
4071 }
4072 }
4073 //Intersection
4074 for(std::map< Node, std::vector< Node > >::const_iterator itr = unprocessed_memberships.begin();
4075 itr != unprocessed_memberships.end(); ++itr) {
4076 Node nf_x = itr->first;
4077 std::vector< Node > exp( unprocessed_x_exps[nf_x] );
4078 Node r = itr->second[0];
4079 //get nf_r
4080 Node inter_r = d_nf_regexps[r];
4081 exp.insert(exp.end(), d_nf_regexps_exp[r].begin(), d_nf_regexps_exp[r].end());
4082 Node x = unprocessed_memberships_bases[itr->first][0];
4083 Node memb = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x, r);
4084 exp.push_back(memb);
4085 for(std::size_t i=1; i < itr->second.size(); i++) {
4086 //exps
4087 Node r2 = itr->second[i];
4088 Node inter_r2 = d_nf_regexps[r2];
4089 exp.insert(exp.end(), d_nf_regexps_exp[r2].begin(), d_nf_regexps_exp[r2].end());
4090 Node x2 = unprocessed_memberships_bases[itr->first][i];
4091 memb = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x2, r2);
4092 exp.push_back(memb);
4093 //intersection
4094 bool spflag = false;
4095 inter_r = d_regexp_opr.intersect(inter_r, inter_r2, spflag);
4096 if(inter_r == d_emptyRegexp) {
4097 //conflict
4098 Node conc;
4099 sendInference( d_empty_vec, exp, conc, "INTERSECT CONFLICT", true );
4100 addLemma = true;
4101 break;
4102 }
4103 }
4104 //infer
4105 if(!d_conflict) {
4106 memb = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, nf_x, inter_r) );
4107 memb_with_exps[memb] = exp;
4108 } else {
4109 break;
4110 }
4111 }
4112
4113 return addLemma;
4114 }
4115
4116 bool TheoryStrings::applyRConsume( CVC4::String &s, Node &r) {
4117 Trace("regexp-derivative") << "TheoryStrings::derivative: s=" << s << ", r= " << r << std::endl;
4118 Assert( d_regexp_opr.checkConstRegExp(r) );
4119
4120 if( !s.isEmptyString() ) {
4121 Node dc = r;
4122
4123 for(unsigned i=0; i<s.size(); ++i) {
4124 CVC4::String c = s.substr(i, 1);
4125 Node dc2;
4126 int rt = d_regexp_opr.derivativeS(dc, c, dc2);
4127 dc = dc2;
4128 if(rt == 0) {
4129 Unreachable();
4130 } else if(rt == 2) {
4131 return false;
4132 }
4133 }
4134 r = dc;
4135 }
4136
4137 return true;
4138 }
4139
4140 Node TheoryStrings::applyRSplit(Node s1, Node s2, Node r) {
4141 Assert(d_regexp_opr.checkConstRegExp(r));
4142
4143 std::vector< std::pair< Node, Node > > vec_can;
4144 d_regexp_opr.splitRegExp(r, vec_can);
4145 //TODO: lazy cache or eager?
4146 std::vector< Node > vec_or;
4147
4148 for(unsigned int i=0; i<vec_can.size(); i++) {
4149 Node m1 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s1, vec_can[i].first);
4150 Node m2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s2, vec_can[i].second);
4151 Node c = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, m1, m2) );
4152 vec_or.push_back( c );
4153 }
4154 Node conc = vec_or.size()==0? Node::null() : vec_or.size()==1 ? vec_or[0] : Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::OR, vec_or) );
4155 return conc;
4156 }
4157
4158 bool TheoryStrings::applyRLen(std::map< Node, std::vector< Node > > &XinR_with_exps) {
4159 if(XinR_with_exps.size() > 0) {
4160 //TODO: get vector, var, store.
4161 return true;
4162 } else {
4163 return false;
4164 }
4165 }
4166
4167 bool TheoryStrings::checkMembershipsWithoutLength(
4168 std::map< Node, std::vector< Node > > &memb_with_exps,
4169 std::map< Node, std::vector< Node > > &XinR_with_exps) {
4170 for(std::map< Node, std::vector< Node > >::iterator itr = memb_with_exps.begin(); itr != memb_with_exps.end(); ++itr) {
4171 Node memb = itr->first;
4172 Node s = memb[0];
4173 Node r = memb[1];
4174 if(s.isConst()) {
4175 memb = Rewriter::rewrite( memb );
4176 if(memb == d_false) {
4177 Node conc;
4178 sendInference(d_empty_vec, itr->second, conc, "MEMBERSHIP CONFLICT", true);
4179 //addLemma = true;
4180 return true;
4181 } else {
4182 Assert(memb == d_true);
4183 }
4184 } else if(s.getKind() == kind::VARIABLE) {
4185 //add to XinR
4186 XinR_with_exps[itr->first] = itr->second;
4187 } else {
4188 Assert(s.getKind() == kind::STRING_CONCAT);
4189 Node conc;
4190 for( unsigned i=0; i<s.getNumChildren(); i++ ) {
4191 if(s[i].isConst()) {
4192 CVC4::String str( s[0].getConst< String >() );
4193 //R-Consume, see Tianyi's thesis
4194 if(!applyRConsume(str, r)) {
4195 sendInference(d_empty_vec, itr->second, conc, "R-Consume CONFLICT", true);
4196 //addLemma = true;
4197 return true;
4198 }
4199 } else {
4200 //R-Split, see Tianyi's thesis
4201 if(i == s.getNumChildren() - 1) {
4202 //add to XinR
4203 Node memb2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s[i], r);
4204 XinR_with_exps[itr->first] = itr->second;
4205 } else {
4206 Node s1 = s[i];
4207 std::vector< Node > vec_s2;
4208 for( unsigned j=i+1; j<s.getNumChildren(); j++ ) {
4209 vec_s2.push_back(s[j]);
4210 }
4211 Node s2 = mkConcat(vec_s2);
4212 conc = applyRSplit(s1, s2, r);
4213 if(conc == d_true) {
4214 break;
4215 } else if(conc.isNull() || conc == d_false) {
4216 conc = Node::null();
4217 sendInference(d_empty_vec, itr->second, conc, "R-Split Conflict", true);
4218 //addLemma = true;
4219 return true;
4220 } else {
4221 sendInference(d_empty_vec, itr->second, conc, "R-Split", true);
4222 //addLemma = true;
4223 return true;
4224 }
4225 }
4226 }
4227 }
4228 }
4229 }
4230 return false;
4231 }
4232
4233 bool TheoryStrings::checkMemberships2() {
4234 bool addedLemma = false;
4235 d_nf_regexps.clear();
4236 d_nf_regexps_exp.clear();
4237 std::map< Node, std::vector< Node > > memb_with_exps;
4238 std::map< Node, std::vector< Node > > XinR_with_exps;
4239
4240 addedLemma = normalizePosMemberships( memb_with_exps );
4241 if(!d_conflict) {
4242 // main procedure
4243 addedLemma |= checkMembershipsWithoutLength( memb_with_exps, XinR_with_exps );
4244 //TODO: check addlemma
4245 if (!addedLemma && !d_conflict) {
4246 for(std::map< Node, std::vector< Node > >::const_iterator itr = XinR_with_exps.begin();
4247 itr != XinR_with_exps.end(); ++itr) {
4248 std::vector<Node> vec_or;
4249 d_regexp_opr.disjunctRegExp( itr->first, vec_or );
4250 Node tmp = NodeManager::currentNM()->mkNode(kind::REGEXP_UNION, vec_or);
4251 Trace("regexp-process") << "Got r: " << itr->first << " to " << tmp << std::endl;
4252 /*
4253 if(r.getKind() == kind::REGEXP_STAR) {
4254 //TODO: apply R-Len
4255 addedLemma = applyRLen(XinR_with_exps);
4256 } else {
4257 //TODO: split
4258 }
4259 */
4260 }
4261 Assert(false); //TODO:tmp
4262 }
4263 }
4264
4265 return addedLemma;
4266 }
4267
4268 void TheoryStrings::checkMemberships() {
4269 //add the memberships
4270 std::vector< Node > mems;
4271 d_extt->getActive( mems, kind::STRING_IN_REGEXP );
4272 for( unsigned i=0; i<mems.size(); i++ ){
4273 Node n = mems[i];
4274 Assert( d_extf_info_tmp.find( n )!=d_extf_info_tmp.end() );
4275 Assert( d_extf_info_tmp[n].d_pol==1 || d_extf_info_tmp[n].d_pol==-1 );
4276 bool pol = d_extf_info_tmp[n].d_pol==1;
4277 Trace("strings-process-debug") << " add membership : " << n << ", pol = " << pol << std::endl;
4278 addMembership( pol ? n : n.negate() );
4279 }
4280
4281 bool addedLemma = false;
4282 bool changed = false;
4283 std::vector< Node > processed;
4284 std::vector< Node > cprocessed;
4285
4286 Trace("regexp-debug") << "Checking Memberships ... " << std::endl;
4287 //if(options::stringEIT()) {
4288 //TODO: Opt for normal forms
4289 for( NodeIntMap::const_iterator itr_xr = d_pos_memberships.begin(); itr_xr != d_pos_memberships.end(); ++itr_xr ){
4290 bool spflag = false;
4291 Node x = (*itr_xr).first;
4292 Trace("regexp-debug") << "Checking Memberships for " << x << std::endl;
4293 if(d_inter_index.find(x) == d_inter_index.end()) {
4294 d_inter_index[x] = 0;
4295 }
4296 int cur_inter_idx = d_inter_index[x];
4297 unsigned n_pmem = (*itr_xr).second;
4298 Assert( getNumMemberships( x, true )==n_pmem );
4299 if( cur_inter_idx != (int)n_pmem ) {
4300 if( n_pmem == 1) {
4301 d_inter_cache[x] = getMembership( x, true, 0 );
4302 d_inter_index[x] = 1;
4303 Trace("regexp-debug") << "... only one choice " << std::endl;
4304 } else if(n_pmem > 1) {
4305 Node r;
4306 if(d_inter_cache.find(x) != d_inter_cache.end()) {
4307 r = d_inter_cache[x];
4308 }
4309 if(r.isNull()) {
4310 r = getMembership( x, true, 0 );
4311 cur_inter_idx = 1;
4312 }
4313
4314 unsigned k_start = cur_inter_idx;
4315 Trace("regexp-debug") << "... staring from : " << cur_inter_idx << ", we have " << n_pmem << std::endl;
4316 for(unsigned k = k_start; k<n_pmem; k++) {
4317 Node r2 = getMembership( x, true, k );
4318 r = d_regexp_opr.intersect(r, r2, spflag);
4319 if(spflag) {
4320 break;
4321 } else if(r == d_emptyRegexp) {
4322 std::vector< Node > vec_nodes;
4323 for( unsigned kk=0; kk<=k; kk++ ){
4324 Node rr = getMembership( x, true, kk );
4325 Node n = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x, rr);
4326 vec_nodes.push_back( n );
4327 }
4328 Node conc;
4329 sendInference(vec_nodes, conc, "INTERSECT CONFLICT", true);
4330 addedLemma = true;
4331 break;
4332 }
4333 if(d_conflict) {
4334 break;
4335 }
4336 }
4337 //updates
4338 if(!d_conflict && !spflag) {
4339 d_inter_cache[x] = r;
4340 d_inter_index[x] = (int)n_pmem;
4341 }
4342 }
4343 }
4344 }
4345 //}
4346
4347 Trace("regexp-debug") << "... No Intersect Conflict in Memberships, addedLemma: " << addedLemma << std::endl;
4348 if(!addedLemma) {
4349 for( unsigned i=0; i<d_regexp_memberships.size(); i++ ) {
4350 //check regular expression membership
4351 Node assertion = d_regexp_memberships[i];
4352 if( d_regexp_ucached.find(assertion) == d_regexp_ucached.end()
4353 && d_regexp_ccached.find(assertion) == d_regexp_ccached.end() ) {
4354 Trace("strings-regexp") << "We have regular expression assertion : " << assertion << std::endl;
4355 Node atom = assertion.getKind()==kind::NOT ? assertion[0] : assertion;
4356 bool polarity = assertion.getKind()!=kind::NOT;
4357 bool flag = true;
4358 Node x = atom[0];
4359 Node r = atom[1];
4360 std::vector< Node > rnfexp;
4361
4362 //if(options::stringOpt1()) {
4363 if(true){
4364 if(!x.isConst()) {
4365 x = getNormalString( x, rnfexp);
4366 changed = true;
4367 }
4368 if(!d_regexp_opr.checkConstRegExp(r)) {
4369 r = getNormalSymRegExp(r, rnfexp);
4370 changed = true;
4371 }
4372 Trace("strings-regexp-nf") << "Term " << atom << " is normalized to " << x << " IN " << r << std::endl;
4373 if(changed) {
4374 Node tmp = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x, r) );
4375 if(!polarity) {
4376 tmp = tmp.negate();
4377 }
4378 if(tmp == d_true) {
4379 d_regexp_ccached.insert(assertion);
4380 continue;
4381 } else if(tmp == d_false) {
4382 Node antec = mkRegExpAntec(assertion, mkExplain(rnfexp));
4383 Node conc = Node::null();
4384 sendLemma(antec, conc, "REGEXP NF Conflict");
4385 addedLemma = true;
4386 break;
4387 }
4388 }
4389 }
4390
4391 if( polarity ) {
4392 flag = checkPDerivative(x, r, atom, addedLemma, processed, cprocessed, rnfexp);
4393 if(options::stringOpt2() && flag) {
4394 if(d_regexp_opr.checkConstRegExp(r) && x.getKind()==kind::STRING_CONCAT) {
4395 std::vector< std::pair< Node, Node > > vec_can;
4396 d_regexp_opr.splitRegExp(r, vec_can);
4397 //TODO: lazy cache or eager?
4398 std::vector< Node > vec_or;
4399 std::vector< Node > vec_s2;
4400 for(unsigned int s2i=1; s2i<x.getNumChildren(); s2i++) {
4401 vec_s2.push_back(x[s2i]);
4402 }
4403 Node s1 = x[0];
4404 Node s2 = mkConcat(vec_s2);
4405 for(unsigned int i=0; i<vec_can.size(); i++) {
4406 Node m1 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s1, vec_can[i].first);
4407 Node m2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s2, vec_can[i].second);
4408 Node c = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, m1, m2) );
4409 vec_or.push_back( c );
4410 }
4411 Node conc = vec_or.size()==1 ? vec_or[0] : Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::OR, vec_or) );
4412 //Trace("regexp-split") << "R " << r << " to " << conc << std::endl;
4413 Node antec = mkRegExpAntec(atom, mkExplain(rnfexp));
4414 if(conc == d_true) {
4415 if(changed) {
4416 cprocessed.push_back( assertion );
4417 } else {
4418 processed.push_back( assertion );
4419 }
4420 } else {
4421 sendLemma(antec, conc, "RegExp-CST-SP");
4422 }
4423 addedLemma = true;
4424 flag = false;
4425 }
4426 }
4427 } else {
4428 if(! options::stringExp()) {
4429 throw LogicException("Strings Incomplete (due to Negative Membership) by default, try --strings-exp option.");
4430 }
4431 }
4432 if(flag) {
4433 //check if the term is atomic
4434 Node xr = getRepresentative( x );
4435 //Trace("strings-regexp") << xr << " is rep of " << x << std::endl;
4436 //Assert( d_normal_forms.find( xr )!=d_normal_forms.end() );
4437 //TODO
4438 if( true || r.getKind()!=kind::REGEXP_STAR || ( d_normal_forms[xr].size()==1 && x.getKind()!=kind::STRING_CONCAT ) ){
4439 Trace("strings-regexp") << "Unroll/simplify membership of atomic term " << xr << std::endl;
4440 //if so, do simple unrolling
4441 std::vector< Node > nvec;
4442
4443 /*if(xr.isConst()) {
4444 Node tmp = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, xr, r) );
4445 if(tmp==d_true || tmp==d_false) {
4446 if(!polarity) {
4447 tmp = tmp==d_true? d_false : d_true;
4448 }
4449 nvec.push_back( tmp );
4450 }
4451 }*/
4452
4453 if(nvec.empty()) {
4454 d_regexp_opr.simplify(atom, nvec, polarity);
4455 }
4456 Node antec = assertion;
4457 if(d_regexp_ant.find(assertion) != d_regexp_ant.end()) {
4458 antec = d_regexp_ant[assertion];
4459 for(std::vector< Node >::const_iterator itr=nvec.begin(); itr<nvec.end(); itr++) {
4460 if(itr->getKind() == kind::STRING_IN_REGEXP) {
4461 if(d_regexp_ant.find( *itr ) == d_regexp_ant.end()) {
4462 d_regexp_ant[ *itr ] = antec;
4463 }
4464 }
4465 }
4466 }
4467 antec = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::AND, antec, mkExplain(rnfexp)) );
4468 Node conc = nvec.size()==1 ? nvec[0] : NodeManager::currentNM()->mkNode(kind::AND, nvec);
4469 conc = Rewriter::rewrite(conc);
4470 sendLemma( antec, conc, "REGEXP_Unfold" );
4471 addedLemma = true;
4472 if(changed) {
4473 cprocessed.push_back( assertion );
4474 } else {
4475 processed.push_back( assertion );
4476 }
4477 //d_regexp_ucached[assertion] = true;
4478 }else{
4479 Trace("strings-regexp") << "Unroll/simplify membership of non-atomic term " << xr << " = ";
4480 for( unsigned j=0; j<d_normal_forms[xr].size(); j++ ){
4481 Trace("strings-regexp") << d_normal_forms[xr][j] << " ";
4482 }
4483 Trace("strings-regexp") << ", polarity = " << polarity << std::endl;
4484 //otherwise, distribute unrolling over parts
4485 Node p1;
4486 Node p2;
4487 if( d_normal_forms[xr].size()>1 ){
4488 p1 = d_normal_forms[xr][0];
4489 std::vector< Node > cc;
4490 cc.insert( cc.begin(), d_normal_forms[xr].begin() + 1, d_normal_forms[xr].end() );
4491 p2 = mkConcat( cc );
4492 }
4493
4494 Trace("strings-regexp-debug") << "Construct antecedant..." << std::endl;
4495 std::vector< Node > antec;
4496 std::vector< Node > antecn;
4497 antec.insert( antec.begin(), d_normal_forms_exp[xr].begin(), d_normal_forms_exp[xr].end() );
4498 if( x!=xr ){
4499 antec.push_back( x.eqNode( xr ) );
4500 }
4501 antecn.push_back( assertion );
4502 Node ant = mkExplain( antec, antecn );
4503 Trace("strings-regexp-debug") << "Construct conclusion..." << std::endl;
4504 Node conc;
4505 if( polarity ){
4506 if( d_normal_forms[xr].size()==0 ){
4507 conc = d_true;
4508 }else if( d_normal_forms[xr].size()==1 ){
4509 Trace("strings-regexp-debug") << "Case 1\n";
4510 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, d_normal_forms[xr][0], r);
4511 }else{
4512 Trace("strings-regexp-debug") << "Case 2\n";
4513 std::vector< Node > conc_c;
4514 Node s11 = mkSkolemS( "s11" );
4515 Node s12 = mkSkolemS( "s12" );
4516 Node s21 = mkSkolemS( "s21" );
4517 Node s22 = mkSkolemS( "s22" );
4518 conc = p1.eqNode( mkConcat(s11, s12) );
4519 conc_c.push_back(conc);
4520 conc = p2.eqNode( mkConcat(s21, s22) );
4521 conc_c.push_back(conc);
4522 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s11, r);
4523 conc_c.push_back(conc);
4524 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, mkConcat(s12, s21), r[0]);
4525 conc_c.push_back(conc);
4526 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s22, r);
4527 conc_c.push_back(conc);
4528 conc = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, conc_c));
4529 Node eqz = Rewriter::rewrite(x.eqNode(d_emptyString));
4530 conc = NodeManager::currentNM()->mkNode(kind::OR, eqz, conc);
4531 d_pending_req_phase[eqz] = true;
4532 }
4533 }else{
4534 if( d_normal_forms[xr].size()==0 ){
4535 conc = d_false;
4536 }else if( d_normal_forms[xr].size()==1 ){
4537 Trace("strings-regexp-debug") << "Case 3\n";
4538 conc = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, d_normal_forms[xr][0], r).negate();
4539 }else{
4540 Trace("strings-regexp-debug") << "Case 4\n";
4541 Node len1 = NodeManager::currentNM()->mkNode(kind::STRING_LENGTH, p1);
4542 Node len2 = NodeManager::currentNM()->mkNode(kind::STRING_LENGTH, p2);
4543 Node bi = NodeManager::currentNM()->mkBoundVar(NodeManager::currentNM()->integerType());
4544 Node bj = NodeManager::currentNM()->mkBoundVar(NodeManager::currentNM()->integerType());
4545 Node b1v = NodeManager::currentNM()->mkNode(kind::BOUND_VAR_LIST, bi, bj);
4546 Node g1 = NodeManager::currentNM()->mkNode(kind::AND,
4547 NodeManager::currentNM()->mkNode(kind::GEQ, bi, d_zero),
4548 NodeManager::currentNM()->mkNode(kind::GEQ, len1, bi),
4549 NodeManager::currentNM()->mkNode(kind::GEQ, bj, d_zero),
4550 NodeManager::currentNM()->mkNode(kind::GEQ, len2, bj));
4551 Node s11 = NodeManager::currentNM()->mkNode(kind::STRING_SUBSTR, p1, d_zero, bi);
4552 Node s12 = NodeManager::currentNM()->mkNode(kind::STRING_SUBSTR, p1, bi, NodeManager::currentNM()->mkNode(kind::MINUS, len1, bi));
4553 Node s21 = NodeManager::currentNM()->mkNode(kind::STRING_SUBSTR, p2, d_zero, bj);
4554 Node s22 = NodeManager::currentNM()->mkNode(kind::STRING_SUBSTR, p2, bj, NodeManager::currentNM()->mkNode(kind::MINUS, len2, bj));
4555 Node cc1 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s11, r).negate();
4556 Node cc2 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, mkConcat(s12, s21), r[0]).negate();
4557 Node cc3 = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, s22, r).negate();
4558 conc = NodeManager::currentNM()->mkNode(kind::OR, cc1, cc2, cc3);
4559 conc = NodeManager::currentNM()->mkNode(kind::IMPLIES, g1, conc);
4560 conc = NodeManager::currentNM()->mkNode(kind::FORALL, b1v, conc);
4561 conc = NodeManager::currentNM()->mkNode(kind::AND, x.eqNode(d_emptyString).negate(), conc);
4562 }
4563 }
4564 if( conc!=d_true ){
4565 ant = mkRegExpAntec(assertion, ant);
4566 sendLemma(ant, conc, "REGEXP CSTAR");
4567 addedLemma = true;
4568 if( conc==d_false ){
4569 d_regexp_ccached.insert( assertion );
4570 }else{
4571 cprocessed.push_back( assertion );
4572 }
4573 }else{
4574 d_regexp_ccached.insert(assertion);
4575 }
4576 }
4577 }
4578 }
4579 if(d_conflict) {
4580 break;
4581 }
4582 }
4583 }
4584 if( addedLemma ) {
4585 if( !d_conflict ){
4586 for( unsigned i=0; i<processed.size(); i++ ) {
4587 d_regexp_ucached.insert(processed[i]);
4588 }
4589 for( unsigned i=0; i<cprocessed.size(); i++ ) {
4590 d_regexp_ccached.insert(cprocessed[i]);
4591 }
4592 }
4593 }
4594 }
4595
4596 bool TheoryStrings::checkPDerivative(Node x, Node r, Node atom, bool &addedLemma,
4597 std::vector< Node > &processed, std::vector< Node > &cprocessed, std::vector< Node > &nf_exp) {
4598
4599 Node antnf = mkExplain(nf_exp);
4600
4601 if(areEqual(x, d_emptyString)) {
4602 Node exp;
4603 switch(d_regexp_opr.delta(r, exp)) {
4604 case 0: {
4605 Node antec = mkRegExpAntec(atom, x.eqNode(d_emptyString));
4606 antec = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, antec, antnf));
4607 sendLemma(antec, exp, "RegExp Delta");
4608 addedLemma = true;
4609 d_regexp_ccached.insert(atom);
4610 return false;
4611 }
4612 case 1: {
4613 d_regexp_ccached.insert(atom);
4614 break;
4615 }
4616 case 2: {
4617 Node antec = mkRegExpAntec(atom, x.eqNode(d_emptyString));
4618 antec = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, antec, antnf));
4619 Node conc = Node::null();
4620 sendLemma(antec, conc, "RegExp Delta CONFLICT");
4621 addedLemma = true;
4622 d_regexp_ccached.insert(atom);
4623 return false;
4624 }
4625 default:
4626 //Impossible
4627 break;
4628 }
4629 } else {
4630 /*Node xr = getRepresentative( x );
4631 if(x != xr) {
4632 Node n = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, xr, r);
4633 Node nn = Rewriter::rewrite( n );
4634 if(nn == d_true) {
4635 d_regexp_ccached.insert(atom);
4636 return false;
4637 } else if(nn == d_false) {
4638 Node antec = mkRegExpAntec(atom, x.eqNode(xr));
4639 Node conc = Node::null();
4640 sendLemma(antec, conc, "RegExp Delta CONFLICT");
4641 addedLemma = true;
4642 d_regexp_ccached.insert(atom);
4643 return false;
4644 }
4645 }*/
4646 Node sREant = mkRegExpAntec(atom, d_true);
4647 sREant = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, sREant, antnf));
4648 if(deriveRegExp( x, r, sREant )) {
4649 addedLemma = true;
4650 processed.push_back( atom );
4651 return false;
4652 }
4653 }
4654 return true;
4655 }
4656
4657 CVC4::String TheoryStrings::getHeadConst( Node x ) {
4658 if( x.isConst() ) {
4659 return x.getConst< String >();
4660 } else if( x.getKind() == kind::STRING_CONCAT ) {
4661 if( x[0].isConst() ) {
4662 return x[0].getConst< String >();
4663 } else {
4664 return d_emptyString.getConst< String >();
4665 }
4666 } else {
4667 return d_emptyString.getConst< String >();
4668 }
4669 }
4670
4671 bool TheoryStrings::deriveRegExp( Node x, Node r, Node ant ) {
4672 // TODO cstr in vre
4673 Assert(x != d_emptyString);
4674 Trace("regexp-derive") << "TheoryStrings::deriveRegExp: x=" << x << ", r= " << r << std::endl;
4675 //if(x.isConst()) {
4676 // Node n = NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP, x, r );
4677 // Node r = Rewriter::rewrite( n );
4678 // if(n != r) {
4679 // sendLemma(ant, r, "REGEXP REWRITE");
4680 // return true;
4681 // }
4682 //}
4683 CVC4::String s = getHeadConst( x );
4684 if( !s.isEmptyString() && d_regexp_opr.checkConstRegExp( r ) ) {
4685 Node conc = Node::null();
4686 Node dc = r;
4687 bool flag = true;
4688 for(unsigned i=0; i<s.size(); ++i) {
4689 CVC4::String c = s.substr(i, 1);
4690 Node dc2;
4691 int rt = d_regexp_opr.derivativeS(dc, c, dc2);
4692 dc = dc2;
4693 if(rt == 0) {
4694 //TODO
4695 } else if(rt == 2) {
4696 // CONFLICT
4697 flag = false;
4698 break;
4699 }
4700 }
4701 // send lemma
4702 if(flag) {
4703 if(x.isConst()) {
4704 Assert(false, "Impossible: TheoryStrings::deriveRegExp: const string in const regular expression.");
4705 return false;
4706 } else {
4707 Assert( x.getKind() == kind::STRING_CONCAT );
4708 std::vector< Node > vec_nodes;
4709 for(unsigned int i=1; i<x.getNumChildren(); ++i ) {
4710 vec_nodes.push_back( x[i] );
4711 }
4712 Node left = mkConcat( vec_nodes );
4713 left = Rewriter::rewrite( left );
4714 conc = NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP, left, dc );
4715
4716 /*std::vector< Node > sdc;
4717 d_regexp_opr.simplify(conc, sdc, true);
4718 if(sdc.size() == 1) {
4719 conc = sdc[0];
4720 } else {
4721 conc = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, conc));
4722 }*/
4723 }
4724 }
4725 sendLemma(ant, conc, "RegExp-Derive");
4726 return true;
4727 } else {
4728 return false;
4729 }
4730 }
4731
4732 void TheoryStrings::addMembership(Node assertion) {
4733 bool polarity = assertion.getKind() != kind::NOT;
4734 TNode atom = polarity ? assertion : assertion[0];
4735 Node x = atom[0];
4736 Node r = atom[1];
4737 if(polarity) {
4738 int index = 0;
4739 NodeIntMap::const_iterator it = d_pos_memberships.find( x );
4740 if( it!=d_nf_pairs.end() ){
4741 index = (*it).second;
4742 for( int k=0; k<index; k++ ){
4743 if( k<(int)d_pos_memberships_data[x].size() ){
4744 if( d_pos_memberships_data[x][k]==r ){
4745 return;
4746 }
4747 }else{
4748 break;
4749 }
4750 }
4751 }
4752 d_pos_memberships[x] = index + 1;
4753 if( index<(int)d_pos_memberships_data[x].size() ){
4754 d_pos_memberships_data[x][index] = r;
4755 }else{
4756 d_pos_memberships_data[x].push_back( r );
4757 }
4758 } else if(!options::stringIgnNegMembership()) {
4759 /*if(options::stringEIT() && d_regexp_opr.checkConstRegExp(r)) {
4760 int rt;
4761 Node r2 = d_regexp_opr.complement(r, rt);
4762 Node a = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x, r2);
4763 }*/
4764 int index = 0;
4765 NodeIntMap::const_iterator it = d_neg_memberships.find( x );
4766 if( it!=d_nf_pairs.end() ){
4767 index = (*it).second;
4768 for( int k=0; k<index; k++ ){
4769 if( k<(int)d_neg_memberships_data[x].size() ){
4770 if( d_neg_memberships_data[x][k]==r ){
4771 return;
4772 }
4773 }else{
4774 break;
4775 }
4776 }
4777 }
4778 d_neg_memberships[x] = index + 1;
4779 if( index<(int)d_neg_memberships_data[x].size() ){
4780 d_neg_memberships_data[x][index] = r;
4781 }else{
4782 d_neg_memberships_data[x].push_back( r );
4783 }
4784 }
4785 // old
4786 if(polarity || !options::stringIgnNegMembership()) {
4787 d_regexp_memberships.push_back( assertion );
4788 }
4789 }
4790
4791 Node TheoryStrings::getNormalString( Node x, std::vector< Node >& nf_exp ){
4792 if( !x.isConst() ){
4793 Node xr = getRepresentative( x );
4794 if( d_normal_forms.find( xr ) != d_normal_forms.end() ){
4795 Node ret = mkConcat( d_normal_forms[xr] );
4796 nf_exp.insert( nf_exp.end(), d_normal_forms_exp[xr].begin(), d_normal_forms_exp[xr].end() );
4797 addToExplanation( x, d_normal_forms_base[xr], nf_exp );
4798 Trace("strings-debug") << "Term: " << x << " has a normal form " << ret << std::endl;
4799 return ret;
4800 } else {
4801 if(x.getKind() == kind::STRING_CONCAT) {
4802 std::vector< Node > vec_nodes;
4803 for(unsigned i=0; i<x.getNumChildren(); i++) {
4804 Node nc = getNormalString( x[i], nf_exp );
4805 vec_nodes.push_back( nc );
4806 }
4807 return mkConcat( vec_nodes );
4808 }
4809 }
4810 }
4811 return x;
4812 }
4813
4814 Node TheoryStrings::getNormalSymRegExp(Node r, std::vector<Node> &nf_exp) {
4815 Node ret = r;
4816 switch( r.getKind() ) {
4817 case kind::REGEXP_EMPTY:
4818 case kind::REGEXP_SIGMA:
4819 break;
4820 case kind::STRING_TO_REGEXP: {
4821 if(!r[0].isConst()) {
4822 Node tmp = getNormalString( r[0], nf_exp );
4823 if(tmp != r[0]) {
4824 ret = NodeManager::currentNM()->mkNode(kind::STRING_TO_REGEXP, tmp);
4825 }
4826 }
4827 break;
4828 }
4829 case kind::REGEXP_CONCAT: {
4830 std::vector< Node > vec_nodes;
4831 for(unsigned i=0; i<r.getNumChildren(); ++i) {
4832 vec_nodes.push_back( getNormalSymRegExp(r[i], nf_exp) );
4833 }
4834 ret = mkConcat(vec_nodes);
4835 break;
4836 }
4837 case kind::REGEXP_UNION: {
4838 std::vector< Node > vec_nodes;
4839 for(unsigned i=0; i<r.getNumChildren(); ++i) {
4840 vec_nodes.push_back( getNormalSymRegExp(r[i], nf_exp) );
4841 }
4842 ret = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::REGEXP_UNION, vec_nodes) );
4843 break;
4844 }
4845 case kind::REGEXP_INTER: {
4846 std::vector< Node > vec_nodes;
4847 for(unsigned i=0; i<r.getNumChildren(); ++i) {
4848 vec_nodes.push_back( getNormalSymRegExp(r[i], nf_exp) );
4849 }
4850 ret = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::REGEXP_INTER, vec_nodes) );
4851 break;
4852 }
4853 case kind::REGEXP_STAR: {
4854 ret = getNormalSymRegExp( r[0], nf_exp );
4855 ret = Rewriter::rewrite( NodeManager::currentNM()->mkNode(kind::REGEXP_STAR, ret) );
4856 break;
4857 }
4858 //case kind::REGEXP_PLUS:
4859 //case kind::REGEXP_OPT:
4860 //case kind::REGEXP_RANGE:
4861 default: {
4862 Trace("strings-error") << "Unsupported term: " << r << " in normalization SymRegExp." << std::endl;
4863 Assert( false );
4864 //return Node::null();
4865 }
4866 }
4867 return ret;
4868 }
4869
4870 }/* CVC4::theory::strings namespace */
4871 }/* CVC4::theory namespace */
4872 }/* CVC4 namespace */