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