1 /********************* */
2 /*! \file theory_strings.cpp
4 ** Top contributors (to current version):
5 ** Andrew Reynolds, Tianyi Liang, Morgan Deters
6 ** This file is part of the CVC4 project.
7 ** Copyright (c) 2009-2018 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
12 ** \brief Implementation of the theory of strings.
14 ** Implementation of the theory of strings.
17 #include "theory/strings/theory_strings.h"
21 #include "expr/kind.h"
22 #include "options/strings_options.h"
23 #include "smt/command.h"
24 #include "smt/logic_exception.h"
25 #include "smt/smt_statistics_registry.h"
26 #include "theory/ext_theory.h"
27 #include "theory/quantifiers/term_database.h"
28 #include "theory/rewriter.h"
29 #include "theory/strings/theory_strings_rewriter.h"
30 #include "theory/strings/type_enumerator.h"
31 #include "theory/theory_model.h"
32 #include "theory/valuation.h"
35 using namespace CVC4::context
;
36 using namespace CVC4::kind
;
42 std::ostream
& operator<<(std::ostream
& out
, Inference i
)
46 case INFER_SSPLIT_CST_PROP
: out
<< "S-Split(CST-P)-prop"; break;
47 case INFER_SSPLIT_VAR_PROP
: out
<< "S-Split(VAR)-prop"; break;
48 case INFER_LEN_SPLIT
: out
<< "Len-Split(Len)"; break;
49 case INFER_LEN_SPLIT_EMP
: out
<< "Len-Split(Emp)"; break;
50 case INFER_SSPLIT_CST_BINARY
: out
<< "S-Split(CST-P)-binary"; break;
51 case INFER_SSPLIT_CST
: out
<< "S-Split(CST-P)"; break;
52 case INFER_SSPLIT_VAR
: out
<< "S-Split(VAR)"; break;
53 case INFER_FLOOP
: out
<< "F-Loop"; break;
54 default: out
<< "?"; break;
59 std::ostream
& operator<<(std::ostream
& out
, InferStep s
)
63 case BREAK
: out
<< "break"; break;
64 case CHECK_INIT
: out
<< "check_init"; break;
65 case CHECK_CONST_EQC
: out
<< "check_const_eqc"; break;
66 case CHECK_EXTF_EVAL
: out
<< "check_extf_eval"; break;
67 case CHECK_CYCLES
: out
<< "check_cycles"; break;
68 case CHECK_FLAT_FORMS
: out
<< "check_flat_forms"; break;
69 case CHECK_NORMAL_FORMS_EQ
: out
<< "check_normal_forms_eq"; break;
70 case CHECK_NORMAL_FORMS_DEQ
: out
<< "check_normal_forms_deq"; break;
71 case CHECK_CODES
: out
<< "check_codes"; break;
72 case CHECK_LENGTH_EQC
: out
<< "check_length_eqc"; break;
73 case CHECK_EXTF_REDUCTION
: out
<< "check_extf_reduction"; break;
74 case CHECK_MEMBERSHIP
: out
<< "check_membership"; break;
75 case CHECK_CARDINALITY
: out
<< "check_cardinality"; break;
76 default: out
<< "?"; break;
81 Node
TheoryStrings::TermIndex::add( TNode n
, unsigned index
, TheoryStrings
* t
, Node er
, std::vector
< Node
>& c
) {
82 if( index
==n
.getNumChildren() ){
83 if( d_data
.isNull() ){
88 Assert( index
<n
.getNumChildren() );
89 TNode nir
= t
->getRepresentative( n
[index
] );
90 //if it is empty, and doing CONCAT, ignore
91 if( nir
==er
&& n
.getKind()==kind::STRING_CONCAT
){
92 return add( n
, index
+1, t
, er
, c
);
95 return d_children
[nir
].add( n
, index
+1, t
, er
, c
);
100 TheoryStrings::TheoryStrings(context::Context
* c
,
101 context::UserContext
* u
,
104 const LogicInfo
& logicInfo
)
105 : Theory(THEORY_STRINGS
, c
, u
, out
, valuation
, logicInfo
),
107 d_equalityEngine(d_notify
, c
, "theory::strings", true),
108 d_conflict(c
, false),
112 d_pregistered_terms_cache(u
),
113 d_registered_terms_cache(u
),
114 d_length_lemma_terms_cache(u
),
115 d_skolem_ne_reg_cache(u
),
118 d_extf_infer_cache(c
),
119 d_extf_infer_cache_u(u
),
120 d_ee_disequalities(c
),
123 d_proxy_var_to_length(u
),
125 d_has_extf(c
, false),
126 d_has_str_code(false),
127 d_regexp_memberships(c
),
130 d_pos_memberships(c
),
131 d_neg_memberships(c
),
134 d_processed_memberships(c
),
138 d_cardinality_lits(u
),
139 d_curr_cardinality(c
, 0),
140 d_strategy_init(false)
143 getExtTheory()->addFunctionKind(kind::STRING_SUBSTR
);
144 getExtTheory()->addFunctionKind(kind::STRING_STRIDOF
);
145 getExtTheory()->addFunctionKind(kind::STRING_ITOS
);
146 getExtTheory()->addFunctionKind(kind::STRING_STOI
);
147 getExtTheory()->addFunctionKind(kind::STRING_STRREPL
);
148 getExtTheory()->addFunctionKind(kind::STRING_STRCTN
);
149 getExtTheory()->addFunctionKind(kind::STRING_IN_REGEXP
);
150 getExtTheory()->addFunctionKind(kind::STRING_LEQ
);
151 getExtTheory()->addFunctionKind(kind::STRING_CODE
);
153 // The kinds we are treating as function application in congruence
154 d_equalityEngine
.addFunctionKind(kind::STRING_LENGTH
);
155 d_equalityEngine
.addFunctionKind(kind::STRING_CONCAT
);
156 d_equalityEngine
.addFunctionKind(kind::STRING_IN_REGEXP
);
157 d_equalityEngine
.addFunctionKind(kind::STRING_CODE
);
158 if( options::stringLazyPreproc() ){
159 d_equalityEngine
.addFunctionKind(kind::STRING_STRCTN
);
160 d_equalityEngine
.addFunctionKind(kind::STRING_LEQ
);
161 d_equalityEngine
.addFunctionKind(kind::STRING_SUBSTR
);
162 d_equalityEngine
.addFunctionKind(kind::STRING_ITOS
);
163 d_equalityEngine
.addFunctionKind(kind::STRING_STOI
);
164 d_equalityEngine
.addFunctionKind(kind::STRING_STRIDOF
);
165 d_equalityEngine
.addFunctionKind(kind::STRING_STRREPL
);
168 d_zero
= NodeManager::currentNM()->mkConst( Rational( 0 ) );
169 d_one
= NodeManager::currentNM()->mkConst( Rational( 1 ) );
170 d_neg_one
= NodeManager::currentNM()->mkConst(Rational(-1));
171 d_emptyString
= NodeManager::currentNM()->mkConst( ::CVC4::String("") );
172 std::vector
< Node
> nvec
;
173 d_emptyRegexp
= NodeManager::currentNM()->mkNode( kind::REGEXP_EMPTY
, nvec
);
174 d_true
= NodeManager::currentNM()->mkConst( true );
175 d_false
= NodeManager::currentNM()->mkConst( false );
177 d_card_size
= TheoryStringsRewriter::getAlphabetCardinality();
180 TheoryStrings::~TheoryStrings() {
181 for( std::map
< Node
, EqcInfo
* >::iterator it
= d_eqc_info
.begin(); it
!= d_eqc_info
.end(); ++it
){
186 Node
TheoryStrings::getRepresentative( Node t
) {
187 if( d_equalityEngine
.hasTerm( t
) ){
188 return d_equalityEngine
.getRepresentative( t
);
194 bool TheoryStrings::hasTerm( Node a
){
195 return d_equalityEngine
.hasTerm( a
);
198 bool TheoryStrings::areEqual( Node a
, Node b
){
201 }else if( hasTerm( a
) && hasTerm( b
) ){
202 return d_equalityEngine
.areEqual( a
, b
);
208 bool TheoryStrings::areDisequal( Node a
, Node b
){
212 if( hasTerm( a
) && hasTerm( b
) ) {
213 Node ar
= d_equalityEngine
.getRepresentative( a
);
214 Node br
= d_equalityEngine
.getRepresentative( b
);
215 return ( ar
!=br
&& ar
.isConst() && br
.isConst() ) || d_equalityEngine
.areDisequal( ar
, br
, false );
217 Node ar
= getRepresentative( a
);
218 Node br
= getRepresentative( b
);
219 return ar
!=br
&& ar
.isConst() && br
.isConst();
224 bool TheoryStrings::areCareDisequal( TNode x
, TNode y
) {
225 Assert( d_equalityEngine
.hasTerm(x
) );
226 Assert( d_equalityEngine
.hasTerm(y
) );
227 if( d_equalityEngine
.isTriggerTerm(x
, THEORY_STRINGS
) && d_equalityEngine
.isTriggerTerm(y
, THEORY_STRINGS
) ){
228 TNode x_shared
= d_equalityEngine
.getTriggerTermRepresentative(x
, THEORY_STRINGS
);
229 TNode y_shared
= d_equalityEngine
.getTriggerTermRepresentative(y
, THEORY_STRINGS
);
230 EqualityStatus eqStatus
= d_valuation
.getEqualityStatus(x_shared
, y_shared
);
231 if( eqStatus
==EQUALITY_FALSE_AND_PROPAGATED
|| eqStatus
==EQUALITY_FALSE
|| eqStatus
==EQUALITY_FALSE_IN_MODEL
){
238 Node
TheoryStrings::getLengthExp( Node t
, std::vector
< Node
>& exp
, Node te
){
239 Assert( areEqual( t
, te
) );
240 Node lt
= mkLength( te
);
242 // use own length if it exists, leads to shorter explanation
245 EqcInfo
* ei
= getOrMakeEqcInfo( t
, false );
246 Node length_term
= ei
? ei
->d_length_term
: Node::null();
247 if( length_term
.isNull() ){
248 //typically shouldnt be necessary
251 Debug("strings") << "TheoryStrings::getLengthTerm " << t
<< " is " << length_term
<< std::endl
;
252 addToExplanation( length_term
, te
, exp
);
253 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, length_term
) );
257 Node
TheoryStrings::getLength( Node t
, std::vector
< Node
>& exp
) {
258 return getLengthExp( t
, exp
, t
);
261 void TheoryStrings::setMasterEqualityEngine(eq::EqualityEngine
* eq
) {
262 d_equalityEngine
.setMasterEqualityEngine(eq
);
265 void TheoryStrings::addSharedTerm(TNode t
) {
266 Debug("strings") << "TheoryStrings::addSharedTerm(): "
267 << t
<< " " << t
.getType().isBoolean() << endl
;
268 d_equalityEngine
.addTriggerTerm(t
, THEORY_STRINGS
);
269 Debug("strings") << "TheoryStrings::addSharedTerm() finished" << std::endl
;
272 EqualityStatus
TheoryStrings::getEqualityStatus(TNode a
, TNode b
) {
273 if( d_equalityEngine
.hasTerm(a
) && d_equalityEngine
.hasTerm(b
) ){
274 if (d_equalityEngine
.areEqual(a
, b
)) {
275 // The terms are implied to be equal
276 return EQUALITY_TRUE
;
278 if (d_equalityEngine
.areDisequal(a
, b
, false)) {
279 // The terms are implied to be dis-equal
280 return EQUALITY_FALSE
;
283 return EQUALITY_UNKNOWN
;
286 void TheoryStrings::propagate(Effort e
) {
287 // direct propagation now
290 bool TheoryStrings::propagate(TNode literal
) {
291 Debug("strings-propagate") << "TheoryStrings::propagate(" << literal
<< ")" << std::endl
;
292 // If already in conflict, no more propagation
294 Debug("strings-propagate") << "TheoryStrings::propagate(" << literal
<< "): already in conflict" << std::endl
;
298 bool ok
= d_out
->propagate(literal
);
306 void TheoryStrings::explain(TNode literal
, std::vector
<TNode
>& assumptions
) {
307 Debug("strings-explain") << "Explain " << literal
<< " " << d_conflict
<< std::endl
;
308 bool polarity
= literal
.getKind() != kind::NOT
;
309 TNode atom
= polarity
? literal
: literal
[0];
310 unsigned ps
= assumptions
.size();
311 std::vector
< TNode
> tassumptions
;
312 if (atom
.getKind() == kind::EQUAL
) {
313 if( atom
[0]!=atom
[1] ){
314 Assert( hasTerm( atom
[0] ) );
315 Assert( hasTerm( atom
[1] ) );
316 d_equalityEngine
.explainEquality(atom
[0], atom
[1], polarity
, tassumptions
);
319 d_equalityEngine
.explainPredicate(atom
, polarity
, tassumptions
);
321 for( unsigned i
=0; i
<tassumptions
.size(); i
++ ){
322 if( std::find( assumptions
.begin(), assumptions
.end(), tassumptions
[i
] )==assumptions
.end() ){
323 assumptions
.push_back( tassumptions
[i
] );
326 if (Debug
.isOn("strings-explain-debug"))
328 Debug("strings-explain-debug") << "Explanation for " << literal
<< " was "
330 for (unsigned i
= ps
; i
< assumptions
.size(); i
++)
332 Debug("strings-explain-debug") << " " << assumptions
[i
] << std::endl
;
337 Node
TheoryStrings::explain( TNode literal
){
338 Debug("strings-explain") << "explain called on " << literal
<< std::endl
;
339 std::vector
< TNode
> assumptions
;
340 explain( literal
, assumptions
);
341 if( assumptions
.empty() ){
343 }else if( assumptions
.size()==1 ){
344 return assumptions
[0];
346 return NodeManager::currentNM()->mkNode( kind::AND
, assumptions
);
350 bool TheoryStrings::getCurrentSubstitution( int effort
, std::vector
< Node
>& vars
,
351 std::vector
< Node
>& subs
, std::map
< Node
, std::vector
< Node
> >& exp
) {
352 Trace("strings-subs") << "getCurrentSubstitution, effort = " << effort
<< std::endl
;
353 for( unsigned i
=0; i
<vars
.size(); i
++ ){
355 Trace("strings-subs") << " get subs for " << n
<< "..." << std::endl
;
358 Node mv
= d_valuation
.getModel()->getRepresentative( n
);
359 Trace("strings-subs") << " model val : " << mv
<< std::endl
;
360 subs
.push_back( mv
);
362 Node nr
= getRepresentative( n
);
363 std::map
< Node
, Node
>::iterator itc
= d_eqc_to_const
.find( nr
);
364 if( itc
!=d_eqc_to_const
.end() ){
365 //constant equivalence classes
366 Trace("strings-subs") << " constant eqc : " << d_eqc_to_const_exp
[nr
] << " " << d_eqc_to_const_base
[nr
] << " " << nr
<< std::endl
;
367 subs
.push_back( itc
->second
);
368 if( !d_eqc_to_const_exp
[nr
].isNull() ){
369 exp
[n
].push_back( d_eqc_to_const_exp
[nr
] );
371 if( !d_eqc_to_const_base
[nr
].isNull() ){
372 addToExplanation( n
, d_eqc_to_const_base
[nr
], exp
[n
] );
374 }else if( effort
>=1 && effort
<3 && n
.getType().isString() ){
376 Node ns
= getNormalString( d_normal_forms_base
[nr
], exp
[n
] );
377 subs
.push_back( ns
);
378 Trace("strings-subs") << " normal eqc : " << ns
<< " " << d_normal_forms_base
[nr
] << " " << nr
<< std::endl
;
379 if( !d_normal_forms_base
[nr
].isNull() ) {
380 addToExplanation( n
, d_normal_forms_base
[nr
], exp
[n
] );
384 //Trace("strings-subs") << " representative : " << nr << std::endl;
385 //addToExplanation( n, nr, exp[n] );
386 //subs.push_back( nr );
394 int TheoryStrings::getReduction( int effort
, Node n
, Node
& nr
) {
395 //determine the effort level to process the extf at
396 // 0 - at assertion time, 1+ - after no other reduction is applicable
397 Assert( d_extf_info_tmp
.find( n
)!=d_extf_info_tmp
.end() );
398 if( d_extf_info_tmp
[n
].d_model_active
){
400 int pol
= d_extf_info_tmp
[n
].d_pol
;
401 if( n
.getKind()==kind::STRING_STRCTN
){
408 std::vector
< Node
> lexp
;
409 Node lenx
= getLength( x
, lexp
);
410 Node lens
= getLength( s
, lexp
);
411 if( areEqual( lenx
, lens
) ){
412 Trace("strings-extf-debug") << " resolve extf : " << n
<< " based on equal lengths disequality." << std::endl
;
413 //we can reduce to disequality when lengths are equal
414 if( !areDisequal( x
, s
) ){
415 lexp
.push_back( lenx
.eqNode(lens
) );
416 lexp
.push_back( n
.negate() );
417 Node xneqs
= x
.eqNode(s
).negate();
418 sendInference( lexp
, xneqs
, "NEG-CTN-EQL", true );
427 if( options::stringLazyPreproc() ){
428 if( n
.getKind()==kind::STRING_SUBSTR
){
430 }else if( n
.getKind()!=kind::STRING_IN_REGEXP
){
435 if( effort
==r_effort
){
436 Node c_n
= pol
==-1 ? n
.negate() : n
;
437 if( d_preproc_cache
.find( c_n
)==d_preproc_cache
.end() ){
438 d_preproc_cache
[ c_n
] = true;
439 Trace("strings-process-debug") << "Process reduction for " << n
<< ", pol = " << pol
<< std::endl
;
440 Kind k
= n
.getKind();
441 if (k
== kind::STRING_STRCTN
&& pol
== 1)
445 //positive contains reduces to a equality
446 Node sk1
= mkSkolemCached( x
, s
, sk_id_ctn_pre
, "sc1" );
447 Node sk2
= mkSkolemCached( x
, s
, sk_id_ctn_post
, "sc2" );
448 Node eq
= Rewriter::rewrite( x
.eqNode( mkConcat( sk1
, s
, sk2
) ) );
449 std::vector
< Node
> exp_vec
;
450 exp_vec
.push_back( n
);
451 sendInference( d_empty_vec
, exp_vec
, eq
, "POS-CTN", true );
452 //we've reduced this n
453 Trace("strings-extf-debug") << " resolve extf : " << n
<< " based on positive contain reduction." << std::endl
;
456 else if (k
!= kind::STRING_CODE
)
458 Assert(k
== STRING_SUBSTR
|| k
== STRING_STRCTN
|| k
== STRING_STRIDOF
461 || k
== STRING_STRREPL
463 std::vector
< Node
> new_nodes
;
464 Node res
= d_preproc
.simplify( n
, new_nodes
);
466 new_nodes
.push_back( NodeManager::currentNM()->mkNode( kind::EQUAL
, res
, n
) );
467 Node nnlem
= new_nodes
.size()==1 ? new_nodes
[0] : NodeManager::currentNM()->mkNode( kind::AND
, new_nodes
);
468 nnlem
= Rewriter::rewrite( nnlem
);
469 Trace("strings-red-lemma") << "Reduction_" << effort
<< " lemma : " << nnlem
<< std::endl
;
470 Trace("strings-red-lemma") << "...from " << n
<< std::endl
;
471 sendInference( d_empty_vec
, nnlem
, "Reduction", true );
472 //we've reduced this n
473 Trace("strings-extf-debug") << " resolve extf : " << n
<< " based on reduction." << std::endl
;
484 /////////////////////////////////////////////////////////////////////////////
486 /////////////////////////////////////////////////////////////////////////////
489 void TheoryStrings::presolve() {
490 Debug("strings-presolve") << "TheoryStrings::Presolving : get fmf options " << (options::stringFMF() ? "true" : "false") << std::endl
;
491 initializeStrategy();
495 /////////////////////////////////////////////////////////////////////////////
497 /////////////////////////////////////////////////////////////////////////////
499 bool TheoryStrings::collectModelInfo(TheoryModel
* m
)
501 Trace("strings-model") << "TheoryStrings : Collect model info" << std::endl
;
502 Trace("strings-model") << "TheoryStrings : assertEqualityEngine." << std::endl
;
504 //AJR : no use doing this since we cannot preregister terms with finite types that don't belong to strings.
505 // change this if we generalize to sequences.
507 // Compute terms appearing in assertions and shared terms
508 //computeRelevantTerms(termSet);
509 //m->assertEqualityEngine( &d_equalityEngine, &termSet );
511 if (!m
->assertEqualityEngine(&d_equalityEngine
))
516 NodeManager
* nm
= NodeManager::currentNM();
518 std::vector
< Node
> nodes
;
519 getEquivalenceClasses( nodes
);
520 std::map
< Node
, Node
> processed
;
521 std::vector
< std::vector
< Node
> > col
;
522 std::vector
< Node
> lts
;
523 separateByLength( nodes
, col
, lts
);
524 //step 1 : get all values for known lengths
525 std::vector
< Node
> lts_values
;
526 std::map
< unsigned, bool > values_used
;
527 for( unsigned i
=0; i
<col
.size(); i
++ ) {
528 Trace("strings-model") << "Checking length for {";
529 for( unsigned j
=0; j
<col
[i
].size(); j
++ ) {
531 Trace("strings-model") << ", ";
533 Trace("strings-model") << col
[i
][j
];
535 Trace("strings-model") << " } (length is " << lts
[i
] << ")" << std::endl
;
536 if( lts
[i
].isConst() ) {
537 lts_values
.push_back( lts
[i
] );
538 Assert(lts
[i
].getConst
<Rational
>() <= Rational(String::maxSize()),
539 "Exceeded UINT32_MAX in string model");
540 unsigned lvalue
= lts
[i
].getConst
<Rational
>().getNumerator().toUnsignedInt();
541 values_used
[ lvalue
] = true;
543 //get value for lts[i];
544 if( !lts
[i
].isNull() ){
545 Node v
= d_valuation
.getModelValue(lts
[i
]);
546 Trace("strings-model") << "Model value for " << lts
[i
] << " is " << v
<< std::endl
;
547 lts_values
.push_back( v
);
548 Assert(v
.getConst
<Rational
>() <= Rational(String::maxSize()),
549 "Exceeded UINT32_MAX in string model");
550 unsigned lvalue
= v
.getConst
<Rational
>().getNumerator().toUnsignedInt();
551 values_used
[ lvalue
] = true;
553 //Trace("strings-model-warn") << "No length for eqc " << col[i][0] << std::endl;
555 lts_values
.push_back( Node::null() );
559 ////step 2 : assign arbitrary values for unknown lengths?
560 // confirmed by calculus invariant, see paper
561 Trace("strings-model") << "Assign to equivalence classes..." << std::endl
;
562 std::map
<Node
, Node
> pure_eq_assign
;
563 //step 3 : assign values to equivalence classes that are pure variables
564 for( unsigned i
=0; i
<col
.size(); i
++ ){
565 std::vector
< Node
> pure_eq
;
566 Trace("strings-model") << "The equivalence classes ";
567 for (const Node
& eqc
: col
[i
])
569 Trace("strings-model") << eqc
<< " ";
570 //check if col[i][j] has only variables
573 Assert(d_normal_forms
.find(eqc
) != d_normal_forms
.end());
574 if (d_normal_forms
[eqc
].size() == 1)
576 // does it have a code and the length of these equivalence classes are
578 if (d_has_str_code
&& lts_values
[i
] == d_one
)
580 EqcInfo
* eip
= getOrMakeEqcInfo(eqc
, false);
581 if (eip
&& !eip
->d_code_term
.get().isNull())
583 // its value must be equal to its code
584 Node ct
= nm
->mkNode(kind::STRING_CODE
, eip
->d_code_term
.get());
585 Node ctv
= d_valuation
.getModelValue(ct
);
587 ctv
.getConst
<Rational
>().getNumerator().toUnsignedInt();
588 Trace("strings-model") << "(code: " << cvalue
<< ") ";
589 std::vector
<unsigned> vec
;
590 vec
.push_back(String::convertCodeToUnsignedInt(cvalue
));
591 Node mv
= nm
->mkConst(String(vec
));
592 pure_eq_assign
[eqc
] = mv
;
593 m
->getEqualityEngine()->addTerm(mv
);
596 pure_eq
.push_back(eqc
);
601 processed
[eqc
] = eqc
;
604 Trace("strings-model") << "have length " << lts_values
[i
] << std::endl
;
606 //assign a new length if necessary
607 if( !pure_eq
.empty() ){
608 if( lts_values
[i
].isNull() ){
609 // start with length two (other lengths have special precendence)
611 while( values_used
.find( lvalue
)!=values_used
.end() ){
614 Trace("strings-model") << "*** Decide to make length of " << lvalue
<< std::endl
;
615 lts_values
[i
] = nm
->mkConst(Rational(lvalue
));
616 values_used
[ lvalue
] = true;
618 Trace("strings-model") << "Need to assign values of length " << lts_values
[i
] << " to equivalence classes ";
619 for( unsigned j
=0; j
<pure_eq
.size(); j
++ ){
620 Trace("strings-model") << pure_eq
[j
] << " ";
622 Trace("strings-model") << std::endl
;
624 //use type enumerator
625 Assert(lts_values
[i
].getConst
<Rational
>() <= Rational(String::maxSize()),
626 "Exceeded UINT32_MAX in string model");
627 StringEnumeratorLength
sel(lts_values
[i
].getConst
<Rational
>().getNumerator().toUnsignedInt());
628 for (const Node
& eqc
: pure_eq
)
631 std::map
<Node
, Node
>::iterator itp
= pure_eq_assign
.find(eqc
);
632 if (itp
== pure_eq_assign
.end())
634 Assert( !sel
.isFinished() );
636 while (m
->hasTerm(c
))
639 Assert(!sel
.isFinished());
648 Trace("strings-model") << "*** Assigned constant " << c
<< " for "
651 if (!m
->assertEquality(eqc
, c
, true))
658 Trace("strings-model") << "String Model : Pure Assigned." << std::endl
;
659 //step 4 : assign constants to all other equivalence classes
660 for( unsigned i
=0; i
<nodes
.size(); i
++ ){
661 if( processed
.find( nodes
[i
] )==processed
.end() ){
662 Assert( d_normal_forms
.find( nodes
[i
] )!=d_normal_forms
.end() );
663 Trace("strings-model") << "Construct model for " << nodes
[i
] << " based on normal form ";
664 for( unsigned j
=0; j
<d_normal_forms
[nodes
[i
]].size(); j
++ ) {
665 if( j
>0 ) Trace("strings-model") << " ++ ";
666 Trace("strings-model") << d_normal_forms
[nodes
[i
]][j
];
667 Node r
= getRepresentative( d_normal_forms
[nodes
[i
]][j
] );
668 if( !r
.isConst() && processed
.find( r
)==processed
.end() ){
669 Trace("strings-model") << "(UNPROCESSED)";
672 Trace("strings-model") << std::endl
;
673 std::vector
< Node
> nc
;
674 for( unsigned j
=0; j
<d_normal_forms
[nodes
[i
]].size(); j
++ ) {
675 Node r
= getRepresentative( d_normal_forms
[nodes
[i
]][j
] );
676 Assert( r
.isConst() || processed
.find( r
)!=processed
.end() );
677 nc
.push_back(r
.isConst() ? r
: processed
[r
]);
679 Node cc
= mkConcat( nc
);
680 Assert( cc
.getKind()==kind::CONST_STRING
);
681 Trace("strings-model") << "*** Determined constant " << cc
<< " for " << nodes
[i
] << std::endl
;
682 processed
[nodes
[i
]] = cc
;
683 if (!m
->assertEquality(nodes
[i
], cc
, true))
689 //Trace("strings-model") << "String Model : Assigned." << std::endl;
690 Trace("strings-model") << "String Model : Finished." << std::endl
;
694 /////////////////////////////////////////////////////////////////////////////
696 /////////////////////////////////////////////////////////////////////////////
699 void TheoryStrings::preRegisterTerm(TNode n
) {
700 if( d_pregistered_terms_cache
.find(n
) == d_pregistered_terms_cache
.end() ) {
701 d_pregistered_terms_cache
.insert(n
);
702 Trace("strings-preregister")
703 << "TheoryString::preregister : " << n
<< std::endl
;
704 //check for logic exceptions
705 Kind k
= n
.getKind();
706 if( !options::stringExp() ){
707 if (k
== kind::STRING_STRIDOF
|| k
== kind::STRING_ITOS
708 || k
== kind::STRING_STOI
709 || k
== kind::STRING_STRREPL
710 || k
== kind::STRING_STRCTN
713 std::stringstream ss
;
714 ss
<< "Term of kind " << k
715 << " not supported in default mode, try --strings-exp";
716 throw LogicException(ss
.str());
722 d_equalityEngine
.addTriggerEquality(n
);
725 case kind::STRING_IN_REGEXP
: {
726 d_out
->requirePhase(n
, true);
727 d_equalityEngine
.addTriggerPredicate(n
);
728 d_equalityEngine
.addTerm(n
[0]);
729 d_equalityEngine
.addTerm(n
[1]);
734 TypeNode tn
= n
.getType();
735 if (tn
.isRegExp() && n
.isVar())
737 std::stringstream ss
;
738 ss
<< "Regular expression variables are not supported.";
739 throw LogicException(ss
.str());
741 if( tn
.isString() ) {
742 // all characters of constants should fall in the alphabet
745 std::vector
<unsigned> vec
= n
.getConst
<String
>().getVec();
746 for (unsigned u
: vec
)
748 if (u
>= d_card_size
)
750 std::stringstream ss
;
751 ss
<< "Characters in string \"" << n
752 << "\" are outside of the given alphabet.";
753 throw LogicException(ss
.str());
757 // if finite model finding is enabled,
758 // then we minimize the length of this term if it is a variable
759 // but not an internally generated Skolem, or a term that does
760 // not belong to this theory.
761 if (options::stringFMF()
762 && (n
.isVar() ? d_all_skolems
.find(n
) == d_all_skolems
.end()
763 : kindToTheoryId(k
) != THEORY_STRINGS
))
765 d_input_vars
.insert(n
);
767 d_equalityEngine
.addTerm(n
);
768 } else if (tn
.isBoolean()) {
769 // Get triggered for both equal and dis-equal
770 d_equalityEngine
.addTriggerPredicate(n
);
772 // Function applications/predicates
773 d_equalityEngine
.addTerm(n
);
774 if( options::stringExp() ){
775 //collect extended functions here: some may not be asserted to strings (such as those with return type Int),
776 // but we need to record them so they are treated properly
777 getExtTheory()->registerTermRec( n
);
780 //concat terms do not contribute to theory combination? TODO: verify
781 if (n
.hasOperator() && kindToTheoryId(k
) == THEORY_STRINGS
782 && k
!= kind::STRING_CONCAT
)
784 d_functionsTerms
.push_back( n
);
791 Node
TheoryStrings::expandDefinition(LogicRequest
&logicRequest
, Node node
) {
792 Trace("strings-exp-def") << "TheoryStrings::expandDefinition : " << node
<< std::endl
;
797 void TheoryStrings::check(Effort e
) {
798 if (done() && e
<EFFORT_FULL
) {
802 TimerStat::CodeTimer
checkTimer(d_checkTime
);
807 if( !done() && !hasTerm( d_emptyString
) ) {
808 preRegisterTerm( d_emptyString
);
811 // Trace("strings-process") << "Theory of strings, check : " << e << std::endl;
812 Trace("strings-check") << "Theory of strings, check : " << e
<< std::endl
;
813 while ( !done() && !d_conflict
) {
814 // Get all the assertions
815 Assertion assertion
= get();
816 TNode fact
= assertion
.assertion
;
818 Trace("strings-assertion") << "get assertion: " << fact
<< endl
;
819 polarity
= fact
.getKind() != kind::NOT
;
820 atom
= polarity
? fact
: fact
[0];
822 //assert pending fact
823 assertPendingFact( atom
, polarity
, fact
);
827 Assert(d_strategy_init
);
828 std::map
<Effort
, std::pair
<unsigned, unsigned> >::iterator itsr
=
829 d_strat_steps
.find(e
);
830 if (!d_conflict
&& !d_valuation
.needCheck() && itsr
!= d_strat_steps
.end())
832 Trace("strings-check") << "Theory of strings " << e
<< " effort check "
834 if(Trace
.isOn("strings-eqc")) {
835 for( unsigned t
=0; t
<2; t
++ ) {
836 eq::EqClassesIterator eqcs2_i
= eq::EqClassesIterator( &d_equalityEngine
);
837 Trace("strings-eqc") << (t
==0 ? "STRINGS:" : "OTHER:") << std::endl
;
838 while( !eqcs2_i
.isFinished() ){
839 Node eqc
= (*eqcs2_i
);
840 bool print
= (t
==0 && eqc
.getType().isString() ) || (t
==1 && !eqc
.getType().isString() );
842 eq::EqClassIterator eqc2_i
= eq::EqClassIterator( eqc
, &d_equalityEngine
);
843 Trace("strings-eqc") << "Eqc( " << eqc
<< " ) : { ";
844 while( !eqc2_i
.isFinished() ) {
845 if( (*eqc2_i
)!=eqc
&& (*eqc2_i
).getKind()!=kind::EQUAL
){
846 Trace("strings-eqc") << (*eqc2_i
) << " ";
850 Trace("strings-eqc") << " } " << std::endl
;
851 EqcInfo
* ei
= getOrMakeEqcInfo( eqc
, false );
853 Trace("strings-eqc-debug") << "* Length term : " << ei
->d_length_term
.get() << std::endl
;
854 Trace("strings-eqc-debug") << "* Cardinality lemma k : " << ei
->d_cardinality_lem_k
.get() << std::endl
;
855 Trace("strings-eqc-debug") << "* Normalization length lemma : " << ei
->d_normalized_length
.get() << std::endl
;
860 Trace("strings-eqc") << std::endl
;
862 Trace("strings-eqc") << std::endl
;
864 unsigned sbegin
= itsr
->second
.first
;
865 unsigned send
= itsr
->second
.second
;
866 bool addedLemma
= false;
869 runStrategy(sbegin
, send
);
871 addedFact
= !d_pending
.empty();
872 addedLemma
= !d_lemma_cache
.empty();
875 // repeat if we did not add a lemma or conflict
876 }while( !d_conflict
&& !addedLemma
&& addedFact
);
878 Trace("strings-check") << "Theory of strings done full effort check " << addedLemma
<< " " << d_conflict
<< std::endl
;
880 Trace("strings-check") << "Theory of strings, done check : " << e
<< std::endl
;
881 Assert( d_pending
.empty() );
882 Assert( d_lemma_cache
.empty() );
885 bool TheoryStrings::needsCheckLastEffort() {
886 if( options::stringGuessModel() ){
887 return d_has_extf
.get();
893 void TheoryStrings::checkExtfReductions( int effort
) {
895 //std::vector< Node > nred;
896 //getExtTheory()->doReductions( effort, nred, false );
898 std::vector
< Node
> extf
= getExtTheory()->getActive();
899 Trace("strings-process") << " checking " << extf
.size() << " active extf"
901 for( unsigned i
=0; i
<extf
.size(); i
++ ){
903 Trace("strings-process") << " check " << n
<< ", active in model="
904 << d_extf_info_tmp
[n
].d_model_active
<< std::endl
;
906 int ret
= getReduction( effort
, n
, nr
);
907 Assert( nr
.isNull() );
909 getExtTheory()->markReduced( extf
[i
] );
918 TheoryStrings::EqcInfo::EqcInfo(context::Context
* c
)
921 d_cardinality_lem_k(c
),
922 d_normalized_length(c
)
926 TheoryStrings::EqcInfo
* TheoryStrings::getOrMakeEqcInfo( Node eqc
, bool doMake
) {
927 std::map
< Node
, EqcInfo
* >::iterator eqc_i
= d_eqc_info
.find( eqc
);
928 if( eqc_i
==d_eqc_info
.end() ){
930 EqcInfo
* ei
= new EqcInfo( getSatContext() );
931 d_eqc_info
[eqc
] = ei
;
937 return (*eqc_i
).second
;
942 /** Conflict when merging two constants */
943 void TheoryStrings::conflict(TNode a
, TNode b
){
945 Debug("strings-conflict") << "Making conflict..." << std::endl
;
948 conflictNode
= explain( a
.eqNode(b
) );
949 Trace("strings-conflict") << "CONFLICT: Eq engine conflict : " << conflictNode
<< std::endl
;
950 d_out
->conflict( conflictNode
);
954 /** called when a new equivalance class is created */
955 void TheoryStrings::eqNotifyNewClass(TNode t
){
956 Kind k
= t
.getKind();
957 if (k
== kind::STRING_LENGTH
|| k
== kind::STRING_CODE
)
959 Trace("strings-debug") << "New length eqc : " << t
<< std::endl
;
960 Node r
= d_equalityEngine
.getRepresentative(t
[0]);
961 EqcInfo
* ei
= getOrMakeEqcInfo( r
, true );
962 if (k
== kind::STRING_LENGTH
)
964 ei
->d_length_term
= t
[0];
968 ei
->d_code_term
= t
[0];
970 //we care about the length of this string
971 registerTerm( t
[0], 1 );
973 //getExtTheory()->registerTerm( t );
977 /** called when two equivalance classes will merge */
978 void TheoryStrings::eqNotifyPreMerge(TNode t1
, TNode t2
){
979 EqcInfo
* e2
= getOrMakeEqcInfo(t2
, false);
981 EqcInfo
* e1
= getOrMakeEqcInfo( t1
);
982 //add information from e2 to e1
983 if( !e2
->d_length_term
.get().isNull() ){
984 e1
->d_length_term
.set( e2
->d_length_term
);
986 if (!e2
->d_code_term
.get().isNull())
988 e1
->d_code_term
.set(e2
->d_code_term
);
990 if( e2
->d_cardinality_lem_k
.get()>e1
->d_cardinality_lem_k
.get() ) {
991 e1
->d_cardinality_lem_k
.set( e2
->d_cardinality_lem_k
);
993 if( !e2
->d_normalized_length
.get().isNull() ){
994 e1
->d_normalized_length
.set( e2
->d_normalized_length
);
999 /** called when two equivalance classes have merged */
1000 void TheoryStrings::eqNotifyPostMerge(TNode t1
, TNode t2
) {
1004 /** called when two equivalance classes are disequal */
1005 void TheoryStrings::eqNotifyDisequal(TNode t1
, TNode t2
, TNode reason
) {
1006 if( t1
.getType().isString() ){
1007 //store disequalities between strings, may need to check if their lengths are equal/disequal
1008 d_ee_disequalities
.push_back( t1
.eqNode( t2
) );
1012 void TheoryStrings::addCarePairs( quantifiers::TermArgTrie
* t1
, quantifiers::TermArgTrie
* t2
, unsigned arity
, unsigned depth
) {
1015 Node f1
= t1
->getNodeData();
1016 Node f2
= t2
->getNodeData();
1017 if( !d_equalityEngine
.areEqual( f1
, f2
) ){
1018 Trace("strings-cg-debug") << "TheoryStrings::computeCareGraph(): checking function " << f1
<< " and " << f2
<< std::endl
;
1019 vector
< pair
<TNode
, TNode
> > currentPairs
;
1020 for (unsigned k
= 0; k
< f1
.getNumChildren(); ++ k
) {
1023 Assert( d_equalityEngine
.hasTerm(x
) );
1024 Assert( d_equalityEngine
.hasTerm(y
) );
1025 Assert( !d_equalityEngine
.areDisequal( x
, y
, false ) );
1026 Assert( !areCareDisequal( x
, y
) );
1027 if( !d_equalityEngine
.areEqual( x
, y
) ){
1028 if( d_equalityEngine
.isTriggerTerm(x
, THEORY_STRINGS
) && d_equalityEngine
.isTriggerTerm(y
, THEORY_STRINGS
) ){
1029 TNode x_shared
= d_equalityEngine
.getTriggerTermRepresentative(x
, THEORY_STRINGS
);
1030 TNode y_shared
= d_equalityEngine
.getTriggerTermRepresentative(y
, THEORY_STRINGS
);
1031 currentPairs
.push_back(make_pair(x_shared
, y_shared
));
1035 for (unsigned c
= 0; c
< currentPairs
.size(); ++ c
) {
1036 Trace("strings-cg-pair") << "TheoryStrings::computeCareGraph(): pair : " << currentPairs
[c
].first
<< " " << currentPairs
[c
].second
<< std::endl
;
1037 addCarePair(currentPairs
[c
].first
, currentPairs
[c
].second
);
1043 if( depth
<(arity
-1) ){
1044 //add care pairs internal to each child
1045 for( std::map
< TNode
, quantifiers::TermArgTrie
>::iterator it
= t1
->d_data
.begin(); it
!= t1
->d_data
.end(); ++it
){
1046 addCarePairs( &it
->second
, NULL
, arity
, depth
+1 );
1049 //add care pairs based on each pair of non-disequal arguments
1050 for( std::map
< TNode
, quantifiers::TermArgTrie
>::iterator it
= t1
->d_data
.begin(); it
!= t1
->d_data
.end(); ++it
){
1051 std::map
< TNode
, quantifiers::TermArgTrie
>::iterator it2
= it
;
1053 for( ; it2
!= t1
->d_data
.end(); ++it2
){
1054 if( !d_equalityEngine
.areDisequal(it
->first
, it2
->first
, false) ){
1055 if( !areCareDisequal(it
->first
, it2
->first
) ){
1056 addCarePairs( &it
->second
, &it2
->second
, arity
, depth
+1 );
1062 //add care pairs based on product of indices, non-disequal arguments
1063 for( std::map
< TNode
, quantifiers::TermArgTrie
>::iterator it
= t1
->d_data
.begin(); it
!= t1
->d_data
.end(); ++it
){
1064 for( std::map
< TNode
, quantifiers::TermArgTrie
>::iterator it2
= t2
->d_data
.begin(); it2
!= t2
->d_data
.end(); ++it2
){
1065 if( !d_equalityEngine
.areDisequal(it
->first
, it2
->first
, false) ){
1066 if( !areCareDisequal(it
->first
, it2
->first
) ){
1067 addCarePairs( &it
->second
, &it2
->second
, arity
, depth
+1 );
1076 void TheoryStrings::computeCareGraph(){
1077 //computing the care graph here is probably still necessary, due to operators that take non-string arguments TODO: verify
1078 Trace("strings-cg") << "TheoryStrings::computeCareGraph(): Build term indices..." << std::endl
;
1079 std::map
< Node
, quantifiers::TermArgTrie
> index
;
1080 std::map
< Node
, unsigned > arity
;
1081 unsigned functionTerms
= d_functionsTerms
.size();
1082 for (unsigned i
= 0; i
< functionTerms
; ++ i
) {
1083 TNode f1
= d_functionsTerms
[i
];
1084 Trace("strings-cg") << "...build for " << f1
<< std::endl
;
1085 Node op
= f1
.getOperator();
1086 std::vector
< TNode
> reps
;
1087 bool has_trigger_arg
= false;
1088 for( unsigned j
=0; j
<f1
.getNumChildren(); j
++ ){
1089 reps
.push_back( d_equalityEngine
.getRepresentative( f1
[j
] ) );
1090 if( d_equalityEngine
.isTriggerTerm( f1
[j
], THEORY_STRINGS
) ){
1091 has_trigger_arg
= true;
1094 if( has_trigger_arg
){
1095 index
[op
].addTerm( f1
, reps
);
1096 arity
[op
] = reps
.size();
1100 for( std::map
< Node
, quantifiers::TermArgTrie
>::iterator itii
= index
.begin(); itii
!= index
.end(); ++itii
){
1101 Trace("strings-cg") << "TheoryStrings::computeCareGraph(): Process index " << itii
->first
<< "..." << std::endl
;
1102 addCarePairs( &itii
->second
, NULL
, arity
[ itii
->first
], 0 );
1106 void TheoryStrings::assertPendingFact(Node atom
, bool polarity
, Node exp
) {
1107 Trace("strings-pending") << "Assert pending fact : " << atom
<< " " << polarity
<< " from " << exp
<< std::endl
;
1108 Assert(atom
.getKind() != kind::OR
, "Infer error: a split.");
1109 if( atom
.getKind()==kind::EQUAL
){
1110 Trace("strings-pending-debug") << " Register term" << std::endl
;
1111 for( unsigned j
=0; j
<2; j
++ ) {
1112 if( !d_equalityEngine
.hasTerm( atom
[j
] ) && atom
[j
].getType().isString() ) {
1113 registerTerm( atom
[j
], 0 );
1116 Trace("strings-pending-debug") << " Now assert equality" << std::endl
;
1117 d_equalityEngine
.assertEquality( atom
, polarity
, exp
);
1118 Trace("strings-pending-debug") << " Finished assert equality" << std::endl
;
1120 d_equalityEngine
.assertPredicate( atom
, polarity
, exp
);
1122 if( atom
.getKind()==kind::STRING_IN_REGEXP
){
1123 if( polarity
&& atom
[1].getKind()==kind::REGEXP_RANGE
){
1124 if( d_extf_infer_cache_u
.find( atom
)==d_extf_infer_cache_u
.end() ){
1125 d_extf_infer_cache_u
.insert( atom
);
1126 //length of first argument is one
1127 Node conc
= d_one
.eqNode( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, atom
[0] ) );
1128 Node lem
= NodeManager::currentNM()->mkNode( kind::OR
, atom
.negate(), conc
);
1129 Trace("strings-lemma") << "Strings::Lemma RE-Range-Len : " << lem
<< std::endl
;
1130 d_out
->lemma( lem
);
1134 //register the atom here, since it may not create a new equivalence class
1135 //getExtTheory()->registerTerm( atom );
1137 Trace("strings-pending-debug") << " Now collect terms" << std::endl
;
1138 //collect extended function terms in the atom
1139 getExtTheory()->registerTermRec( atom
);
1140 Trace("strings-pending-debug") << " Finished collect terms" << std::endl
;
1143 void TheoryStrings::doPendingFacts() {
1145 while( !d_conflict
&& i
<d_pending
.size() ) {
1146 Node fact
= d_pending
[i
];
1147 Node exp
= d_pending_exp
[ fact
];
1148 if(fact
.getKind() == kind::AND
) {
1149 for(size_t j
=0; j
<fact
.getNumChildren(); j
++) {
1150 bool polarity
= fact
[j
].getKind() != kind::NOT
;
1151 TNode atom
= polarity
? fact
[j
] : fact
[j
][0];
1152 assertPendingFact(atom
, polarity
, exp
);
1155 bool polarity
= fact
.getKind() != kind::NOT
;
1156 TNode atom
= polarity
? fact
: fact
[0];
1157 assertPendingFact(atom
, polarity
, exp
);
1162 d_pending_exp
.clear();
1165 void TheoryStrings::doPendingLemmas() {
1166 if( !d_conflict
&& !d_lemma_cache
.empty() ){
1167 for( unsigned i
=0; i
<d_lemma_cache
.size(); i
++ ){
1168 Trace("strings-pending") << "Process pending lemma : " << d_lemma_cache
[i
] << std::endl
;
1169 d_out
->lemma( d_lemma_cache
[i
] );
1171 for( std::map
< Node
, bool >::iterator it
= d_pending_req_phase
.begin(); it
!= d_pending_req_phase
.end(); ++it
){
1172 Trace("strings-pending") << "Require phase : " << it
->first
<< ", polarity = " << it
->second
<< std::endl
;
1173 d_out
->requirePhase( it
->first
, it
->second
);
1176 d_lemma_cache
.clear();
1177 d_pending_req_phase
.clear();
1180 bool TheoryStrings::hasProcessed() {
1181 return d_conflict
|| !d_lemma_cache
.empty() || !d_pending
.empty();
1184 void TheoryStrings::addToExplanation( Node a
, Node b
, std::vector
< Node
>& exp
) {
1186 Debug("strings-explain") << "Add to explanation : " << a
<< " == " << b
<< std::endl
;
1187 Assert( areEqual( a
, b
) );
1188 exp
.push_back( a
.eqNode( b
) );
1192 void TheoryStrings::addToExplanation( Node lit
, std::vector
< Node
>& exp
) {
1193 if( !lit
.isNull() ){
1194 exp
.push_back( lit
);
1198 void TheoryStrings::checkInit() {
1200 d_eqc_to_const
.clear();
1201 d_eqc_to_const_base
.clear();
1202 d_eqc_to_const_exp
.clear();
1203 d_eqc_to_len_term
.clear();
1204 d_term_index
.clear();
1205 d_strings_eqc
.clear();
1207 std::map
< Kind
, unsigned > ncongruent
;
1208 std::map
< Kind
, unsigned > congruent
;
1209 d_emptyString_r
= getRepresentative( d_emptyString
);
1210 eq::EqClassesIterator eqcs_i
= eq::EqClassesIterator( &d_equalityEngine
);
1211 while( !eqcs_i
.isFinished() ){
1212 Node eqc
= (*eqcs_i
);
1213 TypeNode tn
= eqc
.getType();
1214 if( !tn
.isRegExp() ){
1215 if( tn
.isString() ){
1216 d_strings_eqc
.push_back( eqc
);
1219 eq::EqClassIterator eqc_i
= eq::EqClassIterator( eqc
, &d_equalityEngine
);
1220 while( !eqc_i
.isFinished() ) {
1223 d_eqc_to_const
[eqc
] = n
;
1224 d_eqc_to_const_base
[eqc
] = n
;
1225 d_eqc_to_const_exp
[eqc
] = Node::null();
1226 }else if( tn
.isInteger() ){
1227 if( n
.getKind()==kind::STRING_LENGTH
){
1228 Node nr
= getRepresentative( n
[0] );
1229 d_eqc_to_len_term
[nr
] = n
[0];
1231 }else if( n
.getNumChildren()>0 ){
1232 Kind k
= n
.getKind();
1233 if( k
!=kind::EQUAL
){
1234 if( d_congruent
.find( n
)==d_congruent
.end() ){
1235 std::vector
< Node
> c
;
1236 Node nc
= d_term_index
[k
].add( n
, 0, this, d_emptyString_r
, c
);
1238 //check if we have inferred a new equality by removal of empty components
1239 if( n
.getKind()==kind::STRING_CONCAT
&& !areEqual( nc
, n
) ){
1240 std::vector
< Node
> exp
;
1241 unsigned count
[2] = { 0, 0 };
1242 while( count
[0]<nc
.getNumChildren() || count
[1]<n
.getNumChildren() ){
1243 //explain empty prefixes
1244 for( unsigned t
=0; t
<2; t
++ ){
1245 Node nn
= t
==0 ? nc
: n
;
1246 while( count
[t
]<nn
.getNumChildren() &&
1247 ( nn
[count
[t
]]==d_emptyString
|| areEqual( nn
[count
[t
]], d_emptyString
) ) ){
1248 if( nn
[count
[t
]]!=d_emptyString
){
1249 exp
.push_back( nn
[count
[t
]].eqNode( d_emptyString
) );
1254 //explain equal components
1255 if( count
[0]<nc
.getNumChildren() ){
1256 Assert( count
[1]<n
.getNumChildren() );
1257 if( nc
[count
[0]]!=n
[count
[1]] ){
1258 exp
.push_back( nc
[count
[0]].eqNode( n
[count
[1]] ) );
1264 //infer the equality
1265 sendInference( exp
, n
.eqNode( nc
), "I_Norm" );
1266 }else if( getExtTheory()->hasFunctionKind( n
.getKind() ) ){
1267 //mark as congruent : only process if neither has been reduced
1268 getExtTheory()->markCongruent( nc
, n
);
1270 //this node is congruent to another one, we can ignore it
1271 Trace("strings-process-debug") << " congruent term : " << n
<< std::endl
;
1272 d_congruent
.insert( n
);
1274 }else if( k
==kind::STRING_CONCAT
&& c
.size()==1 ){
1275 Trace("strings-process-debug") << " congruent term by singular : " << n
<< " " << c
[0] << std::endl
;
1277 if( !areEqual( c
[0], n
) ){
1278 std::vector
< Node
> exp
;
1279 //explain empty components
1280 bool foundNEmpty
= false;
1281 for( unsigned i
=0; i
<n
.getNumChildren(); i
++ ){
1282 if( areEqual( n
[i
], d_emptyString
) ){
1283 if( n
[i
]!=d_emptyString
){
1284 exp
.push_back( n
[i
].eqNode( d_emptyString
) );
1287 Assert( !foundNEmpty
);
1289 exp
.push_back( n
[i
].eqNode( c
[0] ) );
1294 AlwaysAssert( foundNEmpty
);
1295 //infer the equality
1296 sendInference( exp
, n
.eqNode( c
[0] ), "I_Norm_S" );
1298 d_congruent
.insert( n
);
1308 if( d_congruent
.find( n
)==d_congruent
.end() ){
1312 Trace("strings-process-debug") << " congruent variable : " << n
<< std::endl
;
1313 d_congruent
.insert( n
);
1322 if( Trace
.isOn("strings-process") ){
1323 for( std::map
< Kind
, TermIndex
>::iterator it
= d_term_index
.begin(); it
!= d_term_index
.end(); ++it
){
1324 Trace("strings-process") << " Terms[" << it
->first
<< "] = " << ncongruent
[it
->first
] << "/" << (congruent
[it
->first
]+ncongruent
[it
->first
]) << std::endl
;
1329 void TheoryStrings::checkConstantEquivalenceClasses()
1333 std::vector
<Node
> vecc
;
1337 Trace("strings-process-debug") << "Check constant equivalence classes..."
1339 prevSize
= d_eqc_to_const
.size();
1340 checkConstantEquivalenceClasses(&d_term_index
[kind::STRING_CONCAT
], vecc
);
1341 } while (!hasProcessed() && d_eqc_to_const
.size() > prevSize
);
1344 void TheoryStrings::checkConstantEquivalenceClasses( TermIndex
* ti
, std::vector
< Node
>& vecc
) {
1345 Node n
= ti
->d_data
;
1347 //construct the constant
1348 Node c
= mkConcat( vecc
);
1349 if( !areEqual( n
, c
) ){
1350 Trace("strings-debug") << "Constant eqc : " << c
<< " for " << n
<< std::endl
;
1351 Trace("strings-debug") << " ";
1352 for( unsigned i
=0; i
<vecc
.size(); i
++ ){
1353 Trace("strings-debug") << vecc
[i
] << " ";
1355 Trace("strings-debug") << std::endl
;
1357 unsigned countc
= 0;
1358 std::vector
< Node
> exp
;
1359 while( count
<n
.getNumChildren() ){
1360 while( count
<n
.getNumChildren() && areEqual( n
[count
], d_emptyString
) ){
1361 addToExplanation( n
[count
], d_emptyString
, exp
);
1364 if( count
<n
.getNumChildren() ){
1365 Trace("strings-debug") << "...explain " << n
[count
] << " " << vecc
[countc
] << std::endl
;
1366 if( !areEqual( n
[count
], vecc
[countc
] ) ){
1367 Node nrr
= getRepresentative( n
[count
] );
1368 Assert( !d_eqc_to_const_exp
[nrr
].isNull() );
1369 addToExplanation( n
[count
], d_eqc_to_const_base
[nrr
], exp
);
1370 exp
.push_back( d_eqc_to_const_exp
[nrr
] );
1372 addToExplanation( n
[count
], vecc
[countc
], exp
);
1378 //exp contains an explanation of n==c
1379 Assert( countc
==vecc
.size() );
1381 sendInference( exp
, n
.eqNode( c
), "I_CONST_MERGE" );
1383 }else if( !hasProcessed() ){
1384 Node nr
= getRepresentative( n
);
1385 std::map
< Node
, Node
>::iterator it
= d_eqc_to_const
.find( nr
);
1386 if( it
==d_eqc_to_const
.end() ){
1387 Trace("strings-debug") << "Set eqc const " << n
<< " to " << c
<< std::endl
;
1388 d_eqc_to_const
[nr
] = c
;
1389 d_eqc_to_const_base
[nr
] = n
;
1390 d_eqc_to_const_exp
[nr
] = mkAnd( exp
);
1391 }else if( c
!=it
->second
){
1393 Trace("strings-debug") << "Conflict, other constant was " << it
->second
<< ", this constant was " << c
<< std::endl
;
1394 if( d_eqc_to_const_exp
[nr
].isNull() ){
1395 // n==c ^ n == c' => false
1396 addToExplanation( n
, it
->second
, exp
);
1398 // n==c ^ n == d_eqc_to_const_base[nr] == c' => false
1399 exp
.push_back( d_eqc_to_const_exp
[nr
] );
1400 addToExplanation( n
, d_eqc_to_const_base
[nr
], exp
);
1402 sendInference( exp
, d_false
, "I_CONST_CONFLICT" );
1405 Trace("strings-debug") << "Duplicate constant." << std::endl
;
1410 for( std::map
< TNode
, TermIndex
>::iterator it
= ti
->d_children
.begin(); it
!= ti
->d_children
.end(); ++it
){
1411 std::map
< Node
, Node
>::iterator itc
= d_eqc_to_const
.find( it
->first
);
1412 if( itc
!=d_eqc_to_const
.end() ){
1413 vecc
.push_back( itc
->second
);
1414 checkConstantEquivalenceClasses( &it
->second
, vecc
);
1416 if( hasProcessed() ){
1423 void TheoryStrings::checkExtfEval( int effort
) {
1424 Trace("strings-extf-list") << "Active extended functions, effort=" << effort
<< " : " << std::endl
;
1425 d_extf_info_tmp
.clear();
1426 bool has_nreduce
= false;
1427 std::vector
< Node
> terms
= getExtTheory()->getActive();
1428 std::vector
< Node
> sterms
;
1429 std::vector
< std::vector
< Node
> > exp
;
1430 getExtTheory()->getSubstitutedTerms( effort
, terms
, sterms
, exp
);
1431 for( unsigned i
=0; i
<terms
.size(); i
++ ){
1433 Node sn
= sterms
[i
];
1434 //setup information about extf
1435 d_extf_info_tmp
[n
].init();
1436 std::map
< Node
, ExtfInfoTmp
>::iterator itit
= d_extf_info_tmp
.find( n
);
1437 if( n
.getType().isBoolean() ){
1438 if( areEqual( n
, d_true
) ){
1439 itit
->second
.d_pol
= 1;
1440 }else if( areEqual( n
, d_false
) ){
1441 itit
->second
.d_pol
= -1;
1444 Trace("strings-extf-debug") << "Check extf " << n
<< " == " << sn
<< ", pol = " << itit
->second
.d_pol
<< ", effort=" << effort
<< "..." << std::endl
;
1448 itit
->second
.d_exp
.insert( itit
->second
.d_exp
.end(), exp
[i
].begin(), exp
[i
].end() );
1449 // inference is rewriting the substituted node
1450 Node nrc
= Rewriter::rewrite( sn
);
1451 //if rewrites to a constant, then do the inference and mark as reduced
1452 if( nrc
.isConst() ){
1454 getExtTheory()->markReduced( n
);
1455 Trace("strings-extf-debug") << " resolvable by evaluation..." << std::endl
;
1456 std::vector
< Node
> exps
;
1457 // The following optimization gets the "symbolic definition" of
1458 // an extended term. The symbolic definition of a term t is a term
1459 // t' where constants are replaced by their corresponding proxy
1461 // For example, if lsym is a proxy variable for "", then
1462 // str.replace( lsym, lsym, lsym ) is the symbolic definition for
1463 // str.replace( "", "", "" ). It is generally better to use symbolic
1464 // definitions when doing cd-rewriting for the purpose of minimizing
1465 // clauses, e.g. we infer the unit equality:
1466 // str.replace( lsym, lsym, lsym ) == ""
1467 // instead of making this inference multiple times:
1468 // x = "" => str.replace( x, x, x ) == ""
1469 // y = "" => str.replace( y, y, y ) == ""
1470 Trace("strings-extf-debug") << " get symbolic definition..." << std::endl
;
1471 Node nrs
= getSymbolicDefinition( sn
, exps
);
1472 if( !nrs
.isNull() ){
1473 Trace("strings-extf-debug") << " rewrite " << nrs
<< "..." << std::endl
;
1474 Node nrsr
= Rewriter::rewrite(nrs
);
1475 // ensure the symbolic form is not rewritable
1478 // we cannot use the symbolic definition if it rewrites
1479 Trace("strings-extf-debug") << " symbolic definition is trivial..." << std::endl
;
1483 Trace("strings-extf-debug") << " could not infer symbolic definition." << std::endl
;
1486 if( !nrs
.isNull() ){
1487 Trace("strings-extf-debug") << " symbolic def : " << nrs
<< std::endl
;
1488 if( !areEqual( nrs
, nrc
) ){
1489 //infer symbolic unit
1490 if( n
.getType().isBoolean() ){
1491 conc
= nrc
==d_true
? nrs
: nrs
.negate();
1493 conc
= nrs
.eqNode( nrc
);
1495 itit
->second
.d_exp
.clear();
1498 if( !areEqual( n
, nrc
) ){
1499 if( n
.getType().isBoolean() ){
1500 if( areEqual( n
, nrc
==d_true
? d_false
: d_true
) ){
1501 itit
->second
.d_exp
.push_back( nrc
==d_true
? n
.negate() : n
);
1504 conc
= nrc
==d_true
? n
: n
.negate();
1507 conc
= n
.eqNode( nrc
);
1511 if( !conc
.isNull() ){
1512 Trace("strings-extf") << " resolve extf : " << sn
<< " -> " << nrc
<< std::endl
;
1513 sendInference( itit
->second
.d_exp
, conc
, effort
==0 ? "EXTF" : "EXTF-N", true );
1515 Trace("strings-extf-debug") << " conflict, return." << std::endl
;
1520 //check if it is already equal, if so, mark as reduced. Otherwise, do nothing.
1521 if( areEqual( n
, nrc
) ){
1522 Trace("strings-extf") << " resolved extf, since satisfied by model: " << n
<< std::endl
;
1523 itit
->second
.d_model_active
= false;
1526 //if it reduces to a conjunction, infer each and reduce
1527 }else if( ( nrc
.getKind()==kind::OR
&& itit
->second
.d_pol
==-1 ) || ( nrc
.getKind()==kind::AND
&& itit
->second
.d_pol
==1 ) ){
1529 getExtTheory()->markReduced( n
);
1530 itit
->second
.d_exp
.push_back( itit
->second
.d_pol
==-1 ? n
.negate() : n
);
1531 Trace("strings-extf-debug") << " decomposable..." << std::endl
;
1532 Trace("strings-extf") << " resolve extf : " << sn
<< " -> " << nrc
<< ", pol = " << itit
->second
.d_pol
<< std::endl
;
1533 for( unsigned i
=0; i
<nrc
.getNumChildren(); i
++ ){
1534 sendInference( itit
->second
.d_exp
, itit
->second
.d_pol
==-1 ? nrc
[i
].negate() : nrc
[i
], effort
==0 ? "EXTF_d" : "EXTF_d-N" );
1540 to_reduce
= sterms
[i
];
1543 if( !to_reduce
.isNull() ){
1546 Trace("strings-extf") << " cannot rewrite extf : " << to_reduce
<< std::endl
;
1548 checkExtfInference( n
, to_reduce
, itit
->second
, effort
);
1549 if( Trace
.isOn("strings-extf-list") ){
1550 Trace("strings-extf-list") << " * " << to_reduce
;
1551 if( itit
->second
.d_pol
!=0 ){
1552 Trace("strings-extf-list") << ", pol = " << itit
->second
.d_pol
;
1555 Trace("strings-extf-list") << ", from " << n
;
1557 Trace("strings-extf-list") << std::endl
;
1559 if( getExtTheory()->isActive( n
) && itit
->second
.d_model_active
){
1564 d_has_extf
= has_nreduce
;
1567 void TheoryStrings::checkExtfInference( Node n
, Node nr
, ExtfInfoTmp
& in
, int effort
){
1568 //make additional inferences that do not contribute to the reduction of n, but may help show a refutation
1570 //add original to explanation
1571 in
.d_exp
.push_back( in
.d_pol
==1 ? n
: n
.negate() );
1573 //d_extf_infer_cache stores whether we have made the inferences associated with a node n,
1574 // this may need to be generalized if multiple inferences apply
1576 if( nr
.getKind()==kind::STRING_STRCTN
){
1577 if( ( in
.d_pol
==1 && nr
[1].getKind()==kind::STRING_CONCAT
) || ( in
.d_pol
==-1 && nr
[0].getKind()==kind::STRING_CONCAT
) ){
1578 if( d_extf_infer_cache
.find( nr
)==d_extf_infer_cache
.end() ){
1579 d_extf_infer_cache
.insert( nr
);
1581 //one argument does (not) contain each of the components of the other argument
1582 int index
= in
.d_pol
==1 ? 1 : 0;
1583 std::vector
< Node
> children
;
1584 children
.push_back( nr
[0] );
1585 children
.push_back( nr
[1] );
1586 //Node exp_n = mkAnd( exp );
1587 for( unsigned i
=0; i
<nr
[index
].getNumChildren(); i
++ ){
1588 children
[index
] = nr
[index
][i
];
1589 Node conc
= NodeManager::currentNM()->mkNode( kind::STRING_STRCTN
, children
);
1590 conc
= Rewriter::rewrite(in
.d_pol
== 1 ? conc
: conc
.negate());
1591 // check if it already (does not) hold
1594 if (areEqual(conc
, d_false
))
1596 // should be a conflict
1597 sendInference(in
.d_exp
, conc
, "CTN_Decompose");
1599 else if (getExtTheory()->hasFunctionKind(conc
.getKind()))
1601 // can mark as reduced, since model for n => model for conc
1602 getExtTheory()->markReduced(conc
);
1609 //store this (reduced) assertion
1610 //Assert( effort==0 || nr[0]==getRepresentative( nr[0] ) );
1611 bool pol
= in
.d_pol
==1;
1612 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() ){
1613 Trace("strings-extf-debug") << " store contains info : " << nr
[0] << " " << pol
<< " " << nr
[1] << std::endl
;
1614 d_extf_info_tmp
[nr
[0]].d_ctn
[pol
].push_back( nr
[1] );
1615 d_extf_info_tmp
[nr
[0]].d_ctn_from
[pol
].push_back( n
);
1616 //transitive closure for contains
1618 for( unsigned i
=0; i
<d_extf_info_tmp
[nr
[0]].d_ctn
[opol
].size(); i
++ ){
1619 Node onr
= d_extf_info_tmp
[nr
[0]].d_ctn
[opol
][i
];
1620 Node conc
= NodeManager::currentNM()->mkNode( kind::STRING_STRCTN
, pol
? nr
[1] : onr
, pol
? onr
: nr
[1] );
1621 conc
= Rewriter::rewrite( conc
);
1622 bool do_infer
= false;
1623 if( conc
.getKind()==kind::EQUAL
){
1624 do_infer
= !areDisequal( conc
[0], conc
[1] );
1626 do_infer
= !areEqual( conc
, d_false
);
1629 conc
= conc
.negate();
1630 std::vector
< Node
> exp_c
;
1631 exp_c
.insert( exp_c
.end(), in
.d_exp
.begin(), in
.d_exp
.end() );
1632 Node ofrom
= d_extf_info_tmp
[nr
[0]].d_ctn_from
[opol
][i
];
1633 Assert( d_extf_info_tmp
.find( ofrom
)!=d_extf_info_tmp
.end() );
1634 exp_c
.insert( exp_c
.end(), d_extf_info_tmp
[ofrom
].d_exp
.begin(), d_extf_info_tmp
[ofrom
].d_exp
.end() );
1635 sendInference( exp_c
, conc
, "CTN_Trans" );
1639 Trace("strings-extf-debug") << " redundant." << std::endl
;
1640 getExtTheory()->markReduced( n
);
1647 Node
TheoryStrings::getSymbolicDefinition( Node n
, std::vector
< Node
>& exp
) {
1648 if( n
.getNumChildren()==0 ){
1649 NodeNodeMap::const_iterator it
= d_proxy_var
.find( n
);
1650 if( it
==d_proxy_var
.end() ){
1651 return Node::null();
1653 Node eq
= n
.eqNode( (*it
).second
);
1654 eq
= Rewriter::rewrite( eq
);
1655 if( std::find( exp
.begin(), exp
.end(), eq
)==exp
.end() ){
1656 exp
.push_back( eq
);
1658 return (*it
).second
;
1661 std::vector
< Node
> children
;
1662 if (n
.getMetaKind() == kind::metakind::PARAMETERIZED
) {
1663 children
.push_back( n
.getOperator() );
1665 for( unsigned i
=0; i
<n
.getNumChildren(); i
++ ){
1666 if( n
.getKind()==kind::STRING_IN_REGEXP
&& i
==1 ){
1667 children
.push_back( n
[i
] );
1669 Node ns
= getSymbolicDefinition( n
[i
], exp
);
1671 return Node::null();
1673 children
.push_back( ns
);
1677 return NodeManager::currentNM()->mkNode( n
.getKind(), children
);
1681 Node
TheoryStrings::getConstantEqc( Node eqc
) {
1682 std::map
< Node
, Node
>::iterator it
= d_eqc_to_const
.find( eqc
);
1683 if( it
!=d_eqc_to_const
.end() ){
1686 return Node::null();
1690 void TheoryStrings::debugPrintFlatForms( const char * tc
){
1691 for( unsigned k
=0; k
<d_strings_eqc
.size(); k
++ ){
1692 Node eqc
= d_strings_eqc
[k
];
1693 if( d_eqc
[eqc
].size()>1 ){
1694 Trace( tc
) << "EQC [" << eqc
<< "]" << std::endl
;
1696 Trace( tc
) << "eqc [" << eqc
<< "]";
1698 std::map
< Node
, Node
>::iterator itc
= d_eqc_to_const
.find( eqc
);
1699 if( itc
!=d_eqc_to_const
.end() ){
1700 Trace( tc
) << " C: " << itc
->second
;
1701 if( d_eqc
[eqc
].size()>1 ){
1702 Trace( tc
) << std::endl
;
1705 if( d_eqc
[eqc
].size()>1 ){
1706 for( unsigned i
=0; i
<d_eqc
[eqc
].size(); i
++ ){
1707 Node n
= d_eqc
[eqc
][i
];
1709 for( unsigned j
=0; j
<d_flat_form
[n
].size(); j
++ ){
1710 Node fc
= d_flat_form
[n
][j
];
1711 itc
= d_eqc_to_const
.find( fc
);
1713 if( itc
!=d_eqc_to_const
.end() ){
1714 Trace( tc
) << itc
->second
;
1720 Trace( tc
) << ", from " << n
;
1722 Trace( tc
) << std::endl
;
1725 Trace( tc
) << std::endl
;
1728 Trace( tc
) << std::endl
;
1731 void TheoryStrings::debugPrintNormalForms( const char * tc
) {
1734 struct sortConstLength
{
1735 std::map
< Node
, unsigned > d_const_length
;
1736 bool operator() (Node i
, Node j
) {
1737 std::map
< Node
, unsigned >::iterator it_i
= d_const_length
.find( i
);
1738 std::map
< Node
, unsigned >::iterator it_j
= d_const_length
.find( j
);
1739 if( it_i
==d_const_length
.end() ){
1740 if( it_j
==d_const_length
.end() ){
1746 if( it_j
==d_const_length
.end() ){
1749 return it_i
->second
<it_j
->second
;
1755 void TheoryStrings::checkCycles()
1757 // first check for cycles, while building ordering of equivalence classes
1758 d_flat_form
.clear();
1759 d_flat_form_index
.clear();
1761 //rebuild strings eqc based on acyclic ordering
1762 std::vector
< Node
> eqc
;
1763 eqc
.insert( eqc
.end(), d_strings_eqc
.begin(), d_strings_eqc
.end() );
1764 d_strings_eqc
.clear();
1765 if( options::stringBinaryCsp() ){
1766 //sort: process smallest constants first (necessary if doing binary splits)
1767 sortConstLength scl
;
1768 for( unsigned i
=0; i
<eqc
.size(); i
++ ){
1769 std::map
< Node
, Node
>::iterator itc
= d_eqc_to_const
.find( eqc
[i
] );
1770 if( itc
!=d_eqc_to_const
.end() ){
1771 scl
.d_const_length
[eqc
[i
]] = itc
->second
.getConst
<String
>().size();
1774 std::sort( eqc
.begin(), eqc
.end(), scl
);
1776 for( unsigned i
=0; i
<eqc
.size(); i
++ ){
1777 std::vector
< Node
> curr
;
1778 std::vector
< Node
> exp
;
1779 checkCycles( eqc
[i
], curr
, exp
);
1780 if( hasProcessed() ){
1786 void TheoryStrings::checkFlatForms()
1788 // debug print flat forms
1789 if (Trace
.isOn("strings-ff"))
1791 Trace("strings-ff") << "Flat forms : " << std::endl
;
1792 debugPrintFlatForms("strings-ff");
1795 // inferences without recursively expanding flat forms
1797 //(1) approximate equality by containment, infer conflicts
1798 for (const Node
& eqc
: d_strings_eqc
)
1800 Node c
= getConstantEqc(eqc
);
1803 // if equivalence class is constant, all component constants in flat forms
1804 // must be contained in it, in order
1805 std::map
<Node
, std::vector
<Node
> >::iterator it
= d_eqc
.find(eqc
);
1806 if (it
!= d_eqc
.end())
1808 for (const Node
& n
: it
->second
)
1811 if (!TheoryStringsRewriter::canConstantContainList(
1812 c
, d_flat_form
[n
], firstc
, lastc
))
1814 Trace("strings-ff-debug") << "Flat form for " << n
1815 << " cannot be contained in constant "
1817 Trace("strings-ff-debug") << " indices = " << firstc
<< "/"
1818 << lastc
<< std::endl
;
1819 // conflict, explanation is n = base ^ base = c ^ relevant portion
1821 std::vector
<Node
> exp
;
1822 Assert(d_eqc_to_const_base
.find(eqc
) != d_eqc_to_const_base
.end());
1823 addToExplanation(n
, d_eqc_to_const_base
[eqc
], exp
);
1824 Assert(d_eqc_to_const_exp
.find(eqc
) != d_eqc_to_const_exp
.end());
1825 if (!d_eqc_to_const_exp
[eqc
].isNull())
1827 exp
.push_back(d_eqc_to_const_exp
[eqc
]);
1829 for (int e
= firstc
; e
<= lastc
; e
++)
1831 if (d_flat_form
[n
][e
].isConst())
1833 Assert(e
>= 0 && e
< (int)d_flat_form_index
[n
].size());
1834 Assert(d_flat_form_index
[n
][e
] >= 0
1835 && d_flat_form_index
[n
][e
] < (int)n
.getNumChildren());
1837 d_flat_form
[n
][e
], n
[d_flat_form_index
[n
][e
]], exp
);
1840 Node conc
= d_false
;
1841 sendInference(exp
, conc
, "F_NCTN");
1849 //(2) scan lists, unification to infer conflicts and equalities
1850 for (const Node
& eqc
: d_strings_eqc
)
1852 std::map
<Node
, std::vector
<Node
> >::iterator it
= d_eqc
.find(eqc
);
1853 if (it
== d_eqc
.end() || it
->second
.size() <= 1)
1857 // iterate over start index
1858 for (unsigned start
= 0; start
< it
->second
.size() - 1; start
++)
1860 for (unsigned r
= 0; r
< 2; r
++)
1862 bool isRev
= r
== 1;
1863 checkFlatForm(it
->second
, start
, isRev
);
1873 void TheoryStrings::checkFlatForm(std::vector
<Node
>& eqc
,
1878 std::vector
<Node
> inelig
;
1879 for (unsigned i
= 0; i
<= start
; i
++)
1881 inelig
.push_back(eqc
[start
]);
1883 Node a
= eqc
[start
];
1887 std::vector
<Node
> exp
;
1890 unsigned eqc_size
= eqc
.size();
1891 unsigned asize
= d_flat_form
[a
].size();
1894 for (unsigned i
= start
+ 1; i
< eqc_size
; i
++)
1897 if (std::find(inelig
.begin(), inelig
.end(), b
) == inelig
.end())
1899 unsigned bsize
= d_flat_form
[b
].size();
1903 std::vector
<Node
> conc_c
;
1904 for (unsigned j
= count
; j
< bsize
; j
++)
1907 b
[d_flat_form_index
[b
][j
]].eqNode(d_emptyString
));
1909 Assert(!conc_c
.empty());
1910 conc
= mkAnd(conc_c
);
1913 // swap, will enforce is empty past current
1919 inelig
.push_back(eqc
[i
]);
1925 Node curr
= d_flat_form
[a
][count
];
1926 Node curr_c
= getConstantEqc(curr
);
1927 Node ac
= a
[d_flat_form_index
[a
][count
]];
1928 std::vector
<Node
> lexp
;
1929 Node lcurr
= getLength(ac
, lexp
);
1930 for (unsigned i
= 1; i
< eqc_size
; i
++)
1933 if (std::find(inelig
.begin(), inelig
.end(), b
) == inelig
.end())
1935 if (count
== d_flat_form
[b
].size())
1937 inelig
.push_back(b
);
1939 std::vector
<Node
> conc_c
;
1940 for (unsigned j
= count
; j
< asize
; j
++)
1943 a
[d_flat_form_index
[a
][j
]].eqNode(d_emptyString
));
1945 Assert(!conc_c
.empty());
1946 conc
= mkAnd(conc_c
);
1954 Node cc
= d_flat_form
[b
][count
];
1957 Node bc
= b
[d_flat_form_index
[b
][count
]];
1958 inelig
.push_back(b
);
1959 Assert(!areEqual(curr
, cc
));
1960 Node cc_c
= getConstantEqc(cc
);
1961 if (!curr_c
.isNull() && !cc_c
.isNull())
1963 // check for constant conflict
1965 Node s
= TheoryStringsRewriter::splitConstant(
1966 cc_c
, curr_c
, index
, isRev
);
1969 addToExplanation(ac
, d_eqc_to_const_base
[curr
], exp
);
1970 addToExplanation(d_eqc_to_const_exp
[curr
], exp
);
1971 addToExplanation(bc
, d_eqc_to_const_base
[cc
], exp
);
1972 addToExplanation(d_eqc_to_const_exp
[cc
], exp
);
1978 else if ((d_flat_form
[a
].size() - 1) == count
1979 && (d_flat_form
[b
].size() - 1) == count
)
1981 conc
= ac
.eqNode(bc
);
1987 // if lengths are the same, apply LengthEq
1988 std::vector
<Node
> lexp2
;
1989 Node lcc
= getLength(bc
, lexp2
);
1990 if (areEqual(lcurr
, lcc
))
1992 Trace("strings-ff-debug") << "Infer " << ac
<< " == " << bc
1993 << " since " << lcurr
1994 << " == " << lcc
<< std::endl
;
1995 // exp_n.push_back( getLength( curr, true ).eqNode(
1996 // getLength( cc, true ) ) );
1997 Trace("strings-ff-debug") << "Explanation for " << lcurr
1999 for (unsigned j
= 0; j
< lexp
.size(); j
++)
2001 Trace("strings-ff-debug") << lexp
[j
] << std::endl
;
2003 Trace("strings-ff-debug") << "Explanation for " << lcc
2005 for (unsigned j
= 0; j
< lexp2
.size(); j
++)
2007 Trace("strings-ff-debug") << lexp2
[j
] << std::endl
;
2009 exp
.insert(exp
.end(), lexp
.begin(), lexp
.end());
2010 exp
.insert(exp
.end(), lexp2
.begin(), lexp2
.end());
2011 addToExplanation(lcurr
, lcc
, exp
);
2012 conc
= ac
.eqNode(bc
);
2024 Trace("strings-ff-debug")
2025 << "Found inference : " << conc
<< " based on equality " << a
2026 << " == " << b
<< ", " << isRev
<< " " << inf_type
<< std::endl
;
2027 addToExplanation(a
, b
, exp
);
2028 // explain why prefixes up to now were the same
2029 for (unsigned j
= 0; j
< count
; j
++)
2031 Trace("strings-ff-debug") << "Add at " << d_flat_form_index
[a
][j
] << " "
2032 << d_flat_form_index
[b
][j
] << std::endl
;
2034 a
[d_flat_form_index
[a
][j
]], b
[d_flat_form_index
[b
][j
]], exp
);
2036 // explain why other components up to now are empty
2037 for (unsigned t
= 0; t
< 2; t
++)
2039 Node c
= t
== 0 ? a
: b
;
2041 if (inf_type
== 3 || (t
== 1 && inf_type
== 2))
2043 // explain all the empty components for F_EndpointEq, all for
2044 // the short end for F_EndpointEmp
2045 jj
= isRev
? -1 : c
.getNumChildren();
2049 jj
= t
== 0 ? d_flat_form_index
[a
][count
]
2050 : d_flat_form_index
[b
][count
];
2052 int startj
= isRev
? jj
+ 1 : 0;
2053 int endj
= isRev
? c
.getNumChildren() : jj
;
2054 for (int j
= startj
; j
< endj
; j
++)
2056 if (areEqual(c
[j
], d_emptyString
))
2058 addToExplanation(c
[j
], d_emptyString
, exp
);
2062 // notice that F_EndpointEmp is not typically applied, since
2063 // strict prefix equality ( a.b = a ) where a,b non-empty
2064 // is conflicting by arithmetic len(a.b)=len(a)+len(b)!=len(a)
2071 : (inf_type
== 1 ? "F_Unify" : (inf_type
== 2 ? "F_EndpointEmp"
2072 : "F_EndpointEq")));
2080 } while (inelig
.size() < eqc
.size());
2082 for (const Node
& n
: eqc
)
2084 std::reverse(d_flat_form
[n
].begin(), d_flat_form
[n
].end());
2085 std::reverse(d_flat_form_index
[n
].begin(), d_flat_form_index
[n
].end());
2089 Node
TheoryStrings::checkCycles( Node eqc
, std::vector
< Node
>& curr
, std::vector
< Node
>& exp
){
2090 if( std::find( curr
.begin(), curr
.end(), eqc
)!=curr
.end() ){
2093 }else if( std::find( d_strings_eqc
.begin(), d_strings_eqc
.end(), eqc
)==d_strings_eqc
.end() ){
2094 curr
.push_back( eqc
);
2095 //look at all terms in this equivalence class
2096 eq::EqClassIterator eqc_i
= eq::EqClassIterator( eqc
, &d_equalityEngine
);
2097 while( !eqc_i
.isFinished() ) {
2099 if( d_congruent
.find( n
)==d_congruent
.end() ){
2100 if( n
.getKind() == kind::STRING_CONCAT
){
2101 Trace("strings-cycle") << eqc
<< " check term : " << n
<< " in " << eqc
<< std::endl
;
2102 if( eqc
!=d_emptyString_r
){
2103 d_eqc
[eqc
].push_back( n
);
2105 for( unsigned i
=0; i
<n
.getNumChildren(); i
++ ){
2106 Node nr
= getRepresentative( n
[i
] );
2107 if( eqc
==d_emptyString_r
){
2108 //for empty eqc, ensure all components are empty
2109 if( nr
!=d_emptyString_r
){
2110 std::vector
< Node
> exp
;
2111 exp
.push_back( n
.eqNode( d_emptyString
) );
2112 sendInference( exp
, n
[i
].eqNode( d_emptyString
), "I_CYCLE_E" );
2113 return Node::null();
2116 if( nr
!=d_emptyString_r
){
2117 d_flat_form
[n
].push_back( nr
);
2118 d_flat_form_index
[n
].push_back( i
);
2120 //for non-empty eqc, recurse and see if we find a loop
2121 Node ncy
= checkCycles( nr
, curr
, exp
);
2122 if( !ncy
.isNull() ){
2123 Trace("strings-cycle") << eqc
<< " cycle: " << ncy
<< " at " << n
<< "[" << i
<< "] : " << n
[i
] << std::endl
;
2124 addToExplanation( n
, eqc
, exp
);
2125 addToExplanation( nr
, n
[i
], exp
);
2127 //can infer all other components must be empty
2128 for( unsigned j
=0; j
<n
.getNumChildren(); j
++ ){
2129 //take first non-empty
2130 if( j
!=i
&& !areEqual( n
[j
], d_emptyString
) ){
2131 sendInference( exp
, n
[j
].eqNode( d_emptyString
), "I_CYCLE" );
2132 return Node::null();
2135 Trace("strings-error") << "Looping term should be congruent : " << n
<< " " << eqc
<< " " << ncy
<< std::endl
;
2136 //should find a non-empty component, otherwise would have been singular congruent (I_Norm_S)
2142 if( hasProcessed() ){
2143 return Node::null();
2153 //now we can add it to the list of equivalence classes
2154 d_strings_eqc
.push_back( eqc
);
2158 return Node::null();
2161 void TheoryStrings::checkNormalFormsEq()
2163 if( !options::stringEagerLen() ){
2164 for( unsigned i
=0; i
<d_strings_eqc
.size(); i
++ ) {
2165 Node eqc
= d_strings_eqc
[i
];
2166 eq::EqClassIterator eqc_i
= eq::EqClassIterator( eqc
, &d_equalityEngine
);
2167 while( !eqc_i
.isFinished() ) {
2169 if( d_congruent
.find( n
)==d_congruent
.end() ){
2170 registerTerm( n
, 2 );
2181 // calculate normal forms for each equivalence class, possibly adding
2183 d_normal_forms
.clear();
2184 d_normal_forms_exp
.clear();
2185 std::map
<Node
, Node
> nf_to_eqc
;
2186 std::map
<Node
, Node
> eqc_to_nf
;
2187 std::map
<Node
, Node
> eqc_to_exp
;
2188 for (const Node
& eqc
: d_strings_eqc
)
2190 Trace("strings-process-debug") << "- Verify normal forms are the same for "
2191 << eqc
<< std::endl
;
2192 normalizeEquivalenceClass(eqc
);
2193 Trace("strings-debug") << "Finished normalizing eqc..." << std::endl
;
2198 Node nf_term
= mkConcat(d_normal_forms
[eqc
]);
2199 std::map
<Node
, Node
>::iterator itn
= nf_to_eqc
.find(nf_term
);
2200 if (itn
!= nf_to_eqc
.end())
2202 // two equivalence classes have same normal form, merge
2203 std::vector
<Node
> nf_exp
;
2204 nf_exp
.push_back(mkAnd(d_normal_forms_exp
[eqc
]));
2205 nf_exp
.push_back(eqc_to_exp
[itn
->second
]);
2207 d_normal_forms_base
[eqc
].eqNode(d_normal_forms_base
[itn
->second
]);
2208 sendInference(nf_exp
, eq
, "Normal_Form");
2209 if( hasProcessed() ){
2215 nf_to_eqc
[nf_term
] = eqc
;
2216 eqc_to_nf
[eqc
] = nf_term
;
2217 eqc_to_exp
[eqc
] = mkAnd(d_normal_forms_exp
[eqc
]);
2219 Trace("strings-process-debug")
2220 << "Done verifying normal forms are the same for " << eqc
<< std::endl
;
2222 if (Trace
.isOn("strings-nf"))
2224 Trace("strings-nf") << "**** Normal forms are : " << std::endl
;
2225 for (std::map
<Node
, Node
>::iterator it
= eqc_to_exp
.begin();
2226 it
!= eqc_to_exp
.end();
2229 Trace("strings-nf") << " N[" << it
->first
<< "] (base "
2230 << d_normal_forms_base
[it
->first
]
2231 << ") = " << eqc_to_nf
[it
->first
] << std::endl
;
2232 Trace("strings-nf") << " exp: " << it
->second
<< std::endl
;
2234 Trace("strings-nf") << std::endl
;
2238 void TheoryStrings::checkCodes()
2240 // ensure that lemmas regarding str.code been added for each constant string
2244 NodeManager
* nm
= NodeManager::currentNM();
2245 // str.code applied to the code term for each equivalence class that has a
2246 // code term but is not a constant
2247 std::vector
<Node
> nconst_codes
;
2248 // str.code applied to the proxy variables for each equivalence classes that
2249 // are constants of size one
2250 std::vector
<Node
> const_codes
;
2251 for (const Node
& eqc
: d_strings_eqc
)
2253 if (d_normal_forms
[eqc
].size() == 1 && d_normal_forms
[eqc
][0].isConst())
2255 Node c
= d_normal_forms
[eqc
][0];
2256 Trace("strings-code-debug") << "Get proxy variable for " << c
2258 Node cc
= nm
->mkNode(kind::STRING_CODE
, c
);
2259 cc
= Rewriter::rewrite(cc
);
2260 Assert(cc
.isConst());
2261 NodeNodeMap::const_iterator it
= d_proxy_var
.find(c
);
2262 AlwaysAssert(it
!= d_proxy_var
.end());
2263 Node vc
= nm
->mkNode(kind::STRING_CODE
, (*it
).second
);
2264 if (!areEqual(cc
, vc
))
2266 sendInference(d_empty_vec
, cc
.eqNode(vc
), "Code_Proxy");
2268 const_codes
.push_back(vc
);
2272 EqcInfo
* ei
= getOrMakeEqcInfo(eqc
, false);
2273 if (ei
&& !ei
->d_code_term
.get().isNull())
2275 Node vc
= nm
->mkNode(kind::STRING_CODE
, ei
->d_code_term
.get());
2276 nconst_codes
.push_back(vc
);
2284 // now, ensure that str.code is injective
2285 std::vector
<Node
> cmps
;
2286 cmps
.insert(cmps
.end(), const_codes
.rbegin(), const_codes
.rend());
2287 cmps
.insert(cmps
.end(), nconst_codes
.rbegin(), nconst_codes
.rend());
2288 for (unsigned i
= 0, num_ncc
= nconst_codes
.size(); i
< num_ncc
; i
++)
2290 Node c1
= nconst_codes
[i
];
2292 for (const Node
& c2
: cmps
)
2294 Trace("strings-code-debug")
2295 << "Compare codes : " << c1
<< " " << c2
<< std::endl
;
2296 if (!areDisequal(c1
, c2
) && !areEqual(c1
, d_neg_one
))
2298 Node eq_no
= c1
.eqNode(d_neg_one
);
2299 Node deq
= c1
.eqNode(c2
).negate();
2300 Node eqn
= c1
[0].eqNode(c2
[0]);
2301 // str.code(x)==-1 V str.code(x)!=str.code(y) V x==y
2302 Node inj_lem
= nm
->mkNode(kind::OR
, eq_no
, deq
, eqn
);
2303 sendInference(d_empty_vec
, inj_lem
, "Code_Inj");
2310 //compute d_normal_forms_(base,exp,exp_depend)[eqc]
2311 void TheoryStrings::normalizeEquivalenceClass( Node eqc
) {
2312 Trace("strings-process-debug") << "Process equivalence class " << eqc
<< std::endl
;
2313 if( areEqual( eqc
, d_emptyString
) ) {
2314 #ifdef CVC4_ASSERTIONS
2315 for( unsigned j
=0; j
<d_eqc
[eqc
].size(); j
++ ){
2316 Node n
= d_eqc
[eqc
][j
];
2317 for( unsigned i
=0; i
<n
.getNumChildren(); i
++ ){
2318 Assert( areEqual( n
[i
], d_emptyString
) );
2323 Trace("strings-process-debug") << "Return process equivalence class " << eqc
<< " : empty." << std::endl
;
2324 d_normal_forms_base
[eqc
] = d_emptyString
;
2325 d_normal_forms
[eqc
].clear();
2326 d_normal_forms_exp
[eqc
].clear();
2328 Assert( d_normal_forms
.find(eqc
)==d_normal_forms
.end() );
2329 //phi => t = s1 * ... * sn
2330 // normal form for each non-variable term in this eqc (s1...sn)
2331 std::vector
< std::vector
< Node
> > normal_forms
;
2332 // explanation for each normal form (phi)
2333 std::vector
< std::vector
< Node
> > normal_forms_exp
;
2334 // dependency information
2335 std::vector
< std::map
< Node
, std::map
< bool, int > > > normal_forms_exp_depend
;
2336 // record terms for each normal form (t)
2337 std::vector
< Node
> normal_form_src
;
2339 getNormalForms(eqc
, normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
);
2340 if( hasProcessed() ){
2343 // process the normal forms
2344 processNEqc( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
);
2345 if( hasProcessed() ){
2348 //debugPrintNormalForms( "strings-solve", eqc, normal_forms, normal_form_src, normal_forms_exp, normal_forms_exp_depend );
2350 //construct the normal form
2351 Assert( !normal_forms
.empty() );
2354 std::vector
< Node
>::iterator itn
= std::find( normal_form_src
.begin(), normal_form_src
.end(), eqc
);
2355 if( itn
!=normal_form_src
.end() ){
2356 nf_index
= itn
- normal_form_src
.begin();
2357 Trace("strings-solve-debug2") << "take normal form " << nf_index
<< std::endl
;
2358 Assert( normal_form_src
[nf_index
]==eqc
);
2360 //just take the first normal form
2361 Trace("strings-solve-debug2") << "take the first normal form" << std::endl
;
2363 d_normal_forms
[eqc
].insert( d_normal_forms
[eqc
].end(), normal_forms
[nf_index
].begin(), normal_forms
[nf_index
].end() );
2364 d_normal_forms_exp
[eqc
].insert( d_normal_forms_exp
[eqc
].end(), normal_forms_exp
[nf_index
].begin(), normal_forms_exp
[nf_index
].end() );
2365 Trace("strings-solve-debug2") << "take normal form ... done" << std::endl
;
2366 d_normal_forms_base
[eqc
] = normal_form_src
[nf_index
];
2367 //track dependencies
2368 for( unsigned i
=0; i
<normal_forms_exp
[nf_index
].size(); i
++ ){
2369 Node exp
= normal_forms_exp
[nf_index
][i
];
2370 for( unsigned r
=0; r
<2; r
++ ){
2371 d_normal_forms_exp_depend
[eqc
][exp
][r
==0] = normal_forms_exp_depend
[nf_index
][exp
][r
==0];
2374 Trace("strings-process-debug") << "Return process equivalence class " << eqc
<< " : returned, size = " << d_normal_forms
[eqc
].size() << std::endl
;
2378 void trackNfExpDependency( std::vector
< Node
>& nf_exp_n
, std::map
< Node
, std::map
< bool, int > >& nf_exp_depend_n
, Node exp
, int new_val
, int new_rev_val
){
2379 if( std::find( nf_exp_n
.begin(), nf_exp_n
.end(), exp
)==nf_exp_n
.end() ){
2380 nf_exp_n
.push_back( exp
);
2382 for( unsigned k
=0; k
<2; k
++ ){
2383 int val
= k
==0 ? new_val
: new_rev_val
;
2384 std::map
< bool, int >::iterator itned
= nf_exp_depend_n
[exp
].find( k
==1 );
2385 if( itned
==nf_exp_depend_n
[exp
].end() ){
2386 Trace("strings-process-debug") << "Deps : set dependency on " << exp
<< " to " << val
<< " isRev=" << (k
==0) << std::endl
;
2387 nf_exp_depend_n
[exp
][k
==1] = val
;
2389 Trace("strings-process-debug") << "Deps : Multiple dependencies on " << exp
<< " : " << itned
->second
<< " " << val
<< " isRev=" << (k
==0) << std::endl
;
2390 //if we already have a dependency (in the case of non-linear string equalities), it is min/max
2391 bool cmp
= val
> itned
->second
;
2393 nf_exp_depend_n
[exp
][k
==1] = val
;
2399 void TheoryStrings::getNormalForms( Node
&eqc
, std::vector
< std::vector
< Node
> > &normal_forms
, std::vector
< Node
> &normal_form_src
,
2400 std::vector
< std::vector
< Node
> > &normal_forms_exp
, std::vector
< std::map
< Node
, std::map
< bool, int > > >& normal_forms_exp_depend
) {
2401 //constant for equivalence class
2402 Node eqc_non_c
= eqc
;
2403 Trace("strings-process-debug") << "Get normal forms " << eqc
<< std::endl
;
2404 eq::EqClassIterator eqc_i
= eq::EqClassIterator( eqc
, &d_equalityEngine
);
2405 while( !eqc_i
.isFinished() ){
2407 if( d_congruent
.find( n
)==d_congruent
.end() ){
2408 if( n
.getKind() == kind::CONST_STRING
|| n
.getKind() == kind::STRING_CONCAT
){
2409 Trace("strings-process-debug") << "Get Normal Form : Process term " << n
<< " in eqc " << eqc
<< std::endl
;
2410 std::vector
< Node
> nf_n
;
2411 std::vector
< Node
> nf_exp_n
;
2412 std::map
< Node
, std::map
< bool, int > > nf_exp_depend_n
;
2413 if( n
.getKind()==kind::CONST_STRING
){
2414 if( n
!=d_emptyString
) {
2415 nf_n
.push_back( n
);
2417 }else if( n
.getKind()==kind::STRING_CONCAT
){
2418 for( unsigned i
=0; i
<n
.getNumChildren(); i
++ ) {
2419 Node nr
= d_equalityEngine
.getRepresentative( n
[i
] );
2420 Trace("strings-process-debug") << "Normalizing subterm " << n
[i
] << " = " << nr
<< std::endl
;
2421 Assert( d_normal_forms
.find( nr
)!=d_normal_forms
.end() );
2422 unsigned orig_size
= nf_n
.size();
2423 unsigned add_size
= d_normal_forms
[nr
].size();
2424 //if not the empty string, add to current normal form
2425 if( !d_normal_forms
[nr
].empty() ){
2426 for( unsigned r
=0; r
<d_normal_forms
[nr
].size(); r
++ ) {
2427 if( Trace
.isOn("strings-error") ) {
2428 if( d_normal_forms
[nr
][r
].getKind()==kind::STRING_CONCAT
){
2429 Trace("strings-error") << "Strings::Error: From eqc = " << eqc
<< ", " << n
<< " index " << i
<< ", bad normal form : ";
2430 for( unsigned rr
=0; rr
<d_normal_forms
[nr
].size(); rr
++ ) {
2431 Trace("strings-error") << d_normal_forms
[nr
][rr
] << " ";
2433 Trace("strings-error") << std::endl
;
2436 Assert( d_normal_forms
[nr
][r
].getKind()!=kind::STRING_CONCAT
);
2438 nf_n
.insert( nf_n
.end(), d_normal_forms
[nr
].begin(), d_normal_forms
[nr
].end() );
2441 for( unsigned j
=0; j
<d_normal_forms_exp
[nr
].size(); j
++ ){
2442 Node exp
= d_normal_forms_exp
[nr
][j
];
2444 trackNfExpDependency( nf_exp_n
, nf_exp_depend_n
, exp
,
2445 orig_size
+ d_normal_forms_exp_depend
[nr
][exp
][false],
2446 orig_size
+ ( add_size
- d_normal_forms_exp_depend
[nr
][exp
][true] ) );
2448 if( d_normal_forms_base
[nr
]!=n
[i
] ){
2449 Assert( d_normal_forms_base
.find( nr
)!=d_normal_forms_base
.end() );
2450 Node eq
= n
[i
].eqNode( d_normal_forms_base
[nr
] );
2451 //track depends : entire current segment is dependent upon base equality
2452 trackNfExpDependency( nf_exp_n
, nf_exp_depend_n
, eq
, orig_size
, orig_size
+ add_size
);
2455 //convert forward indices to reverse indices
2456 int total_size
= nf_n
.size();
2457 for( std::map
< Node
, std::map
< bool, int > >::iterator it
= nf_exp_depend_n
.begin(); it
!= nf_exp_depend_n
.end(); ++it
){
2458 it
->second
[true] = total_size
- it
->second
[true];
2459 Assert( it
->second
[true]>=0 );
2462 //if not equal to self
2463 if( nf_n
.size()>1 || ( nf_n
.size()==1 && nf_n
[0].getKind()==kind::CONST_STRING
) ){
2464 if( nf_n
.size()>1 ) {
2465 for( unsigned i
=0; i
<nf_n
.size(); i
++ ){
2466 if( Trace
.isOn("strings-error") ){
2467 Trace("strings-error") << "Cycle for normal form ";
2468 printConcat(nf_n
,"strings-error");
2469 Trace("strings-error") << "..." << nf_n
[i
] << std::endl
;
2471 Assert( !areEqual( nf_n
[i
], n
) );
2474 normal_forms
.push_back(nf_n
);
2475 normal_form_src
.push_back(n
);
2476 normal_forms_exp
.push_back(nf_exp_n
);
2477 normal_forms_exp_depend
.push_back(nf_exp_depend_n
);
2479 //this was redundant: combination of self + empty string(s)
2480 Node nn
= nf_n
.size()==0 ? d_emptyString
: nf_n
[0];
2481 Assert( areEqual( nn
, eqc
) );
2490 if( normal_forms
.empty() ) {
2491 Trace("strings-solve-debug2") << "construct the normal form" << std::endl
;
2492 //do not choose a concat here use "eqc_non_c" (in this case they have non-trivial explanation why they normalize to self)
2493 std::vector
< Node
> eqc_non_c_nf
;
2494 getConcatVec( eqc_non_c
, eqc_non_c_nf
);
2495 normal_forms
.push_back( eqc_non_c_nf
);
2496 normal_form_src
.push_back( eqc_non_c
);
2497 normal_forms_exp
.push_back( std::vector
< Node
>() );
2498 normal_forms_exp_depend
.push_back( std::map
< Node
, std::map
< bool, int > >() );
2500 if(Trace
.isOn("strings-solve")) {
2501 Trace("strings-solve") << "--- Normal forms for equivalance class " << eqc
<< " : " << std::endl
;
2502 for( unsigned i
=0; i
<normal_forms
.size(); i
++ ) {
2503 Trace("strings-solve") << "#" << i
<< " (from " << normal_form_src
[i
] << ") : ";
2504 for( unsigned j
=0; j
<normal_forms
[i
].size(); j
++ ) {
2506 Trace("strings-solve") << ", ";
2508 Trace("strings-solve") << normal_forms
[i
][j
];
2510 Trace("strings-solve") << std::endl
;
2511 Trace("strings-solve") << " Explanation is : ";
2512 if(normal_forms_exp
[i
].size() == 0) {
2513 Trace("strings-solve") << "NONE";
2515 for( unsigned j
=0; j
<normal_forms_exp
[i
].size(); j
++ ) {
2517 Trace("strings-solve") << " AND ";
2519 Trace("strings-solve") << normal_forms_exp
[i
][j
];
2521 Trace("strings-solve") << std::endl
;
2522 Trace("strings-solve") << "WITH DEPENDENCIES : " << std::endl
;
2523 for( unsigned j
=0; j
<normal_forms_exp
[i
].size(); j
++ ) {
2524 Trace("strings-solve") << " " << normal_forms_exp
[i
][j
] << " -> ";
2525 Trace("strings-solve") << normal_forms_exp_depend
[i
][normal_forms_exp
[i
][j
]][false] << ",";
2526 Trace("strings-solve") << normal_forms_exp_depend
[i
][normal_forms_exp
[i
][j
]][true] << std::endl
;
2529 Trace("strings-solve") << std::endl
;
2533 Trace("strings-solve") << "--- Single normal form for equivalence class " << eqc
<< std::endl
;
2536 //if equivalence class is constant, approximate as containment, infer conflicts
2537 Node c
= getConstantEqc( eqc
);
2539 Trace("strings-solve") << "Eqc is constant " << c
<< std::endl
;
2540 for( unsigned i
=0; i
<normal_forms
.size(); i
++ ) {
2542 if( !TheoryStringsRewriter::canConstantContainList( c
, normal_forms
[i
], firstc
, lastc
) ){
2543 Node n
= normal_form_src
[i
];
2545 Trace("strings-solve") << "Normal form for " << n
<< " cannot be contained in constant " << c
<< std::endl
;
2546 //conflict, explanation is n = base ^ base = c ^ relevant porition of ( n = N[n] )
2547 std::vector
< Node
> exp
;
2548 Assert( d_eqc_to_const_base
.find( eqc
)!=d_eqc_to_const_base
.end() );
2549 addToExplanation( n
, d_eqc_to_const_base
[eqc
], exp
);
2550 Assert( d_eqc_to_const_exp
.find( eqc
)!=d_eqc_to_const_exp
.end() );
2551 if( !d_eqc_to_const_exp
[eqc
].isNull() ){
2552 exp
.push_back( d_eqc_to_const_exp
[eqc
] );
2554 //TODO: this can be minimized based on firstc/lastc, normal_forms_exp_depend
2555 exp
.insert( exp
.end(), normal_forms_exp
[i
].begin(), normal_forms_exp
[i
].end() );
2556 Node conc
= d_false
;
2557 sendInference( exp
, conc
, "N_NCTN" );
2564 void TheoryStrings::getExplanationVectorForPrefix( std::vector
< std::vector
< Node
> > &normal_forms_exp
, std::vector
< std::map
< Node
, std::map
< bool, int > > >& normal_forms_exp_depend
,
2565 unsigned i
, int index
, bool isRev
, std::vector
< Node
>& curr_exp
) {
2566 if( index
==-1 || !options::stringMinPrefixExplain() ){
2567 curr_exp
.insert(curr_exp
.end(), normal_forms_exp
[i
].begin(), normal_forms_exp
[i
].end() );
2569 for( unsigned k
=0; k
<normal_forms_exp
[i
].size(); k
++ ){
2570 Node exp
= normal_forms_exp
[i
][k
];
2571 int dep
= normal_forms_exp_depend
[i
][exp
][isRev
];
2573 curr_exp
.push_back( exp
);
2574 Trace("strings-explain-prefix-debug") << " include : " << exp
<< std::endl
;
2576 Trace("strings-explain-prefix-debug") << " exclude : " << exp
<< std::endl
;
2582 void TheoryStrings::getExplanationVectorForPrefixEq( std::vector
< std::vector
< Node
> > &normal_forms
, std::vector
< Node
> &normal_form_src
,
2583 std::vector
< std::vector
< Node
> > &normal_forms_exp
, std::vector
< std::map
< Node
, std::map
< bool, int > > >& normal_forms_exp_depend
,
2584 unsigned i
, unsigned j
, int index_i
, int index_j
, bool isRev
, std::vector
< Node
>& curr_exp
) {
2585 Trace("strings-explain-prefix") << "Get explanation for prefix " << index_i
<< ", " << index_j
<< " of normal forms " << i
<< " and " << j
<< ", reverse = " << isRev
<< std::endl
;
2586 for( unsigned r
=0; r
<2; r
++ ){
2587 getExplanationVectorForPrefix( normal_forms_exp
, normal_forms_exp_depend
, r
==0 ? i
: j
, r
==0 ? index_i
: index_j
, isRev
, curr_exp
);
2589 Trace("strings-explain-prefix") << "Included " << curr_exp
.size() << " / " << ( normal_forms_exp
[i
].size() + normal_forms_exp
[j
].size() ) << std::endl
;
2590 addToExplanation( normal_form_src
[i
], normal_form_src
[j
], curr_exp
);
2594 void TheoryStrings::processNEqc( std::vector
< std::vector
< Node
> > &normal_forms
, std::vector
< Node
> &normal_form_src
,
2595 std::vector
< std::vector
< Node
> > &normal_forms_exp
, std::vector
< std::map
< Node
, std::map
< bool, int > > >& normal_forms_exp_depend
){
2596 //the possible inferences
2597 std::vector
< InferInfo
> pinfer
;
2598 // loop over all pairs
2599 for(unsigned i
=0; i
<normal_forms
.size()-1; i
++) {
2600 //unify each normalform[j] with normal_forms[i]
2601 for(unsigned j
=i
+1; j
<normal_forms
.size(); j
++ ) {
2602 //ensure that normal_forms[i] and normal_forms[j] are the same modulo equality, add to pinfer if not
2603 Trace("strings-solve") << "Strings: Process normal form #" << i
<< " against #" << j
<< "..." << std::endl
;
2604 if( isNormalFormPair( normal_form_src
[i
], normal_form_src
[j
] ) ) {
2605 Trace("strings-solve") << "Strings: Already cached." << std::endl
;
2607 //process the reverse direction first (check for easy conflicts and inferences)
2608 unsigned rindex
= 0;
2609 processReverseNEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, rindex
, 0, pinfer
);
2610 if( hasProcessed() ){
2612 }else if( !pinfer
.empty() && pinfer
.back().d_id
==1 ){
2615 //AJR: for less aggressive endpoint inference
2619 processSimpleNEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, index
, false, rindex
, pinfer
);
2620 if( hasProcessed() ){
2622 }else if( !pinfer
.empty() && pinfer
.back().d_id
==1 ){
2628 if( !pinfer
.empty() ){
2629 //now, determine which of the possible inferences we want to add
2630 unsigned use_index
= 0;
2631 bool set_use_index
= false;
2632 Trace("strings-solve") << "Possible inferences (" << pinfer
.size() << ") : " << std::endl
;
2633 unsigned min_id
= 9;
2634 unsigned max_index
= 0;
2635 for (unsigned i
= 0, size
= pinfer
.size(); i
< size
; i
++)
2637 Trace("strings-solve") << "From " << pinfer
[i
].d_i
<< " / " << pinfer
[i
].d_j
<< " (rev=" << pinfer
[i
].d_rev
<< ") : ";
2638 Trace("strings-solve")
2639 << pinfer
[i
].d_conc
<< " by " << pinfer
[i
].d_id
<< std::endl
;
2640 if (!set_use_index
|| pinfer
[i
].d_id
< min_id
2641 || (pinfer
[i
].d_id
== min_id
&& pinfer
[i
].d_index
> max_index
))
2643 min_id
= pinfer
[i
].d_id
;
2644 max_index
= pinfer
[i
].d_index
;
2646 set_use_index
= true;
2649 //send the inference
2650 if( !pinfer
[use_index
].d_nf_pair
[0].isNull() ){
2651 Assert( !pinfer
[use_index
].d_nf_pair
[1].isNull() );
2652 addNormalFormPair( pinfer
[use_index
].d_nf_pair
[0], pinfer
[use_index
].d_nf_pair
[1] );
2654 std::stringstream ssi
;
2655 ssi
<< pinfer
[use_index
].d_id
;
2656 sendInference(pinfer
[use_index
].d_ant
,
2657 pinfer
[use_index
].d_antn
,
2658 pinfer
[use_index
].d_conc
,
2660 pinfer
[use_index
].sendAsLemma());
2661 for( std::map
< int, std::vector
< Node
> >::iterator it
= pinfer
[use_index
].d_new_skolem
.begin(); it
!= pinfer
[use_index
].d_new_skolem
.end(); ++it
){
2662 for( unsigned i
=0; i
<it
->second
.size(); i
++ ){
2664 sendLengthLemma( it
->second
[i
] );
2665 }else if( it
->first
==1 ){
2666 registerNonEmptySkolem( it
->second
[i
] );
2673 bool TheoryStrings::InferInfo::sendAsLemma() {
2677 void TheoryStrings::processReverseNEq( std::vector
< std::vector
< Node
> > &normal_forms
, std::vector
< Node
> &normal_form_src
,
2678 std::vector
< std::vector
< Node
> > &normal_forms_exp
, std::vector
< std::map
< Node
, std::map
< bool, int > > >& normal_forms_exp_depend
,
2679 unsigned i
, unsigned j
, unsigned& index
, unsigned rproc
, std::vector
< InferInfo
>& pinfer
) {
2680 //reverse normal form of i, j
2681 std::reverse( normal_forms
[i
].begin(), normal_forms
[i
].end() );
2682 std::reverse( normal_forms
[j
].begin(), normal_forms
[j
].end() );
2684 processSimpleNEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, index
, true, rproc
, pinfer
);
2686 //reverse normal form of i, j
2687 std::reverse( normal_forms
[i
].begin(), normal_forms
[i
].end() );
2688 std::reverse( normal_forms
[j
].begin(), normal_forms
[j
].end() );
2691 //rproc is the # is the size of suffix that is identical
2692 void TheoryStrings::processSimpleNEq( std::vector
< std::vector
< Node
> > &normal_forms
, std::vector
< Node
> &normal_form_src
,
2693 std::vector
< std::vector
< Node
> > &normal_forms_exp
, std::vector
< std::map
< Node
, std::map
< bool, int > > >& normal_forms_exp_depend
,
2694 unsigned i
, unsigned j
, unsigned& index
, bool isRev
, unsigned rproc
, std::vector
< InferInfo
>& pinfer
) {
2695 Assert( rproc
<=normal_forms
[i
].size() && rproc
<=normal_forms
[j
].size() );
2699 //if we are at the end
2700 if( index
==(normal_forms
[i
].size()-rproc
) || index
==(normal_forms
[j
].size()-rproc
) ){
2701 if( index
==(normal_forms
[i
].size()-rproc
) && index
==(normal_forms
[j
].size()-rproc
) ){
2704 //the remainder must be empty
2705 unsigned k
= index
==(normal_forms
[i
].size()-rproc
) ? j
: i
;
2706 unsigned index_k
= index
;
2707 //Node eq_exp = mkAnd( curr_exp );
2708 std::vector
< Node
> curr_exp
;
2709 getExplanationVectorForPrefixEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, -1, -1, isRev
, curr_exp
);
2710 while( !d_conflict
&& index_k
<(normal_forms
[k
].size()-rproc
) ){
2711 //can infer that this string must be empty
2712 Node eq
= normal_forms
[k
][index_k
].eqNode( d_emptyString
);
2713 //Trace("strings-lemma") << "Strings: Infer " << eq << " from " << eq_exp << std::endl;
2714 Assert( !areEqual( d_emptyString
, normal_forms
[k
][index_k
] ) );
2715 sendInference( curr_exp
, eq
, "N_EndpointEmp" );
2720 Trace("strings-solve-debug") << "Process " << normal_forms
[i
][index
] << " ... " << normal_forms
[j
][index
] << std::endl
;
2721 if( normal_forms
[i
][index
]==normal_forms
[j
][index
] ){
2722 Trace("strings-solve-debug") << "Simple Case 1 : strings are equal" << std::endl
;
2726 Assert( !areEqual(normal_forms
[i
][index
], normal_forms
[j
][index
]) );
2727 std::vector
< Node
> temp_exp
;
2728 Node length_term_i
= getLength( normal_forms
[i
][index
], temp_exp
);
2729 Node length_term_j
= getLength( normal_forms
[j
][index
], temp_exp
);
2730 //check length(normal_forms[i][index]) == length(normal_forms[j][index])
2731 if( areEqual( length_term_i
, length_term_j
) ){
2732 Trace("strings-solve-debug") << "Simple Case 2 : string lengths are equal" << std::endl
;
2733 Node eq
= normal_forms
[i
][index
].eqNode( normal_forms
[j
][index
] );
2734 //eq = Rewriter::rewrite( eq );
2735 Node length_eq
= length_term_i
.eqNode( length_term_j
);
2736 //temp_exp.insert(temp_exp.end(), curr_exp.begin(), curr_exp.end() );
2737 getExplanationVectorForPrefixEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, index
, index
, isRev
, temp_exp
);
2738 temp_exp
.push_back(length_eq
);
2739 sendInference( temp_exp
, eq
, "N_Unify" );
2741 }else if( ( normal_forms
[i
][index
].getKind()!=kind::CONST_STRING
&& index
==normal_forms
[i
].size()-rproc
-1 ) ||
2742 ( normal_forms
[j
][index
].getKind()!=kind::CONST_STRING
&& index
==normal_forms
[j
].size()-rproc
-1 ) ){
2743 Trace("strings-solve-debug") << "Simple Case 3 : at endpoint" << std::endl
;
2744 std::vector
< Node
> antec
;
2745 //antec.insert(antec.end(), curr_exp.begin(), curr_exp.end() );
2746 getExplanationVectorForPrefixEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, -1, -1, isRev
, antec
);
2747 std::vector
< Node
> eqn
;
2748 for( unsigned r
=0; r
<2; r
++ ) {
2749 int index_k
= index
;
2750 int k
= r
==0 ? i
: j
;
2751 std::vector
< Node
> eqnc
;
2752 for( unsigned index_l
=index_k
; index_l
<(normal_forms
[k
].size()-rproc
); index_l
++ ) {
2754 eqnc
.insert(eqnc
.begin(), normal_forms
[k
][index_l
] );
2756 eqnc
.push_back( normal_forms
[k
][index_l
] );
2759 eqn
.push_back( mkConcat( eqnc
) );
2761 if( !areEqual( eqn
[0], eqn
[1] ) ){
2762 sendInference( antec
, eqn
[0].eqNode( eqn
[1] ), "N_EndpointEq", true );
2765 Assert( normal_forms
[i
].size()==normal_forms
[j
].size() );
2766 index
= normal_forms
[i
].size()-rproc
;
2768 }else if( normal_forms
[i
][index
].isConst() && normal_forms
[j
][index
].isConst() ){
2769 Node const_str
= normal_forms
[i
][index
];
2770 Node other_str
= normal_forms
[j
][index
];
2771 Trace("strings-solve-debug") << "Simple Case 3 : Const Split : " << const_str
<< " vs " << other_str
<< " at index " << index
<< ", isRev = " << isRev
<< std::endl
;
2772 unsigned len_short
= const_str
.getConst
<String
>().size() <= other_str
.getConst
<String
>().size() ? const_str
.getConst
<String
>().size() : other_str
.getConst
<String
>().size();
2773 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
);
2775 //same prefix/suffix
2776 //k is the index of the string that is shorter
2777 int k
= const_str
.getConst
<String
>().size()<other_str
.getConst
<String
>().size() ? i
: j
;
2778 int l
= const_str
.getConst
<String
>().size()<other_str
.getConst
<String
>().size() ? j
: i
;
2779 //update the nf exp dependencies
2780 //notice this is not critical for soundness: not doing the below incrementing will only lead to overapproximating when antecedants are required in explanations
2781 for( std::map
< Node
, std::map
< bool, int > >::iterator itnd
= normal_forms_exp_depend
[l
].begin(); itnd
!= normal_forms_exp_depend
[l
].end(); ++itnd
){
2782 for( std::map
< bool, int >::iterator itnd2
= itnd
->second
.begin(); itnd2
!= itnd
->second
.end(); ++itnd2
){
2783 //see if this can be incremented: it can if it is not relevant to the current index
2784 Assert( itnd2
->second
>=0 && itnd2
->second
<=(int)normal_forms
[l
].size() );
2785 bool increment
= (itnd2
->first
==isRev
) ? itnd2
->second
>(int)index
: ( (int)normal_forms
[l
].size()-1-itnd2
->second
)<(int)index
;
2787 normal_forms_exp_depend
[l
][itnd
->first
][itnd2
->first
] = itnd2
->second
+ 1;
2792 int new_len
= normal_forms
[l
][index
].getConst
<String
>().size() - len_short
;
2793 Node remainderStr
= NodeManager::currentNM()->mkConst( normal_forms
[l
][index
].getConst
<String
>().substr(0, new_len
) );
2794 Trace("strings-solve-debug-test") << "Break normal form of " << normal_forms
[l
][index
] << " into " << normal_forms
[k
][index
] << ", " << remainderStr
<< std::endl
;
2795 normal_forms
[l
].insert( normal_forms
[l
].begin()+index
+ 1, remainderStr
);
2797 Node remainderStr
= NodeManager::currentNM()->mkConst(normal_forms
[l
][index
].getConst
<String
>().substr(len_short
));
2798 Trace("strings-solve-debug-test") << "Break normal form of " << normal_forms
[l
][index
] << " into " << normal_forms
[k
][index
] << ", " << remainderStr
<< std::endl
;
2799 normal_forms
[l
].insert( normal_forms
[l
].begin()+index
+ 1, remainderStr
);
2801 normal_forms
[l
][index
] = normal_forms
[k
][index
];
2806 std::vector
< Node
> antec
;
2807 getExplanationVectorForPrefixEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, index
, index
, isRev
, antec
);
2808 sendInference( antec
, d_false
, "N_Const", true );
2812 //construct the candidate inference "info"
2814 info
.d_index
= index
;
2819 bool info_valid
= false;
2820 Assert( index
<normal_forms
[i
].size()-rproc
&& index
<normal_forms
[j
].size()-rproc
);
2821 std::vector
< Node
> lexp
;
2822 Node length_term_i
= getLength( normal_forms
[i
][index
], lexp
);
2823 Node length_term_j
= getLength( normal_forms
[j
][index
], lexp
);
2824 //split on equality between string lengths (note that splitting on equality between strings is worse since it is harder to process)
2825 if( !areDisequal( length_term_i
, length_term_j
) && !areEqual( length_term_i
, length_term_j
) &&
2826 normal_forms
[i
][index
].getKind()!=kind::CONST_STRING
&& normal_forms
[j
][index
].getKind()!=kind::CONST_STRING
){ //AJR: remove the latter 2 conditions?
2827 Trace("strings-solve-debug") << "Non-simple Case 1 : string lengths neither equal nor disequal" << std::endl
;
2828 //try to make the lengths equal via splitting on demand
2829 Node length_eq
= NodeManager::currentNM()->mkNode( kind::EQUAL
, length_term_i
, length_term_j
);
2830 length_eq
= Rewriter::rewrite( length_eq
);
2832 info
.d_conc
= NodeManager::currentNM()->mkNode( kind::OR
, length_eq
, length_eq
.negate() );
2833 info
.d_pending_phase
[ length_eq
] = true;
2834 info
.d_id
= INFER_LEN_SPLIT
;
2837 Trace("strings-solve-debug") << "Non-simple Case 2 : must compare strings" << std::endl
;
2840 if( detectLoop( normal_forms
, i
, j
, index
, loop_in_i
, loop_in_j
, rproc
) ){
2841 if( !isRev
){ //FIXME
2842 getExplanationVectorForPrefixEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, -1, -1, isRev
, info
.d_ant
);
2844 if( processLoop( normal_forms
, normal_form_src
, i
, j
, loop_in_i
!=-1 ? i
: j
, loop_in_i
!=-1 ? j
: i
, loop_in_i
!=-1 ? loop_in_i
: loop_in_j
, index
, info
) ){
2849 //AJR: length entailment here?
2850 if( normal_forms
[i
][index
].getKind() == kind::CONST_STRING
|| normal_forms
[j
][index
].getKind() == kind::CONST_STRING
){
2851 unsigned const_k
= normal_forms
[i
][index
].getKind() == kind::CONST_STRING
? i
: j
;
2852 unsigned nconst_k
= normal_forms
[i
][index
].getKind() == kind::CONST_STRING
? j
: i
;
2853 Node other_str
= normal_forms
[nconst_k
][index
];
2854 Assert( other_str
.getKind()!=kind::CONST_STRING
, "Other string is not constant." );
2855 Assert( other_str
.getKind()!=kind::STRING_CONCAT
, "Other string is not CONCAT." );
2856 if( !d_equalityEngine
.areDisequal( other_str
, d_emptyString
, true ) ){
2857 Node eq
= other_str
.eqNode( d_emptyString
);
2859 info
.d_conc
= NodeManager::currentNM()->mkNode( kind::OR
, eq
, eq
.negate() );
2860 info
.d_id
= INFER_LEN_SPLIT_EMP
;
2863 if( !isRev
){ //FIXME
2864 Node xnz
= other_str
.eqNode( d_emptyString
).negate();
2865 unsigned index_nc_k
= index
+1;
2866 //Node next_const_str = TheoryStringsRewriter::collectConstantStringAt( normal_forms[nconst_k], index_nc_k, false );
2867 unsigned start_index_nc_k
= index
+1;
2868 Node next_const_str
= TheoryStringsRewriter::getNextConstantAt( normal_forms
[nconst_k
], start_index_nc_k
, index_nc_k
, false );
2869 if( !next_const_str
.isNull() ) {
2870 unsigned index_c_k
= index
;
2871 Node const_str
= TheoryStringsRewriter::collectConstantStringAt( normal_forms
[const_k
], index_c_k
, false );
2872 Assert( !const_str
.isNull() );
2873 CVC4::String stra
= const_str
.getConst
<String
>();
2874 CVC4::String strb
= next_const_str
.getConst
<String
>();
2875 //since non-empty, we start with charecter #1
2878 CVC4::String stra1
= stra
.prefix( stra
.size()-1 );
2879 p
= stra
.size() - stra1
.roverlap(strb
);
2880 Trace("strings-csp-debug") << "Compute roverlap : " << const_str
<< " " << next_const_str
<< std::endl
;
2881 size_t p2
= stra1
.rfind(strb
);
2882 p
= p2
==std::string::npos
? p
: ( p
>p2
+1? p2
+1 : p
);
2883 Trace("strings-csp-debug") << "overlap : " << stra1
<< " " << strb
<< " returned " << p
<< " " << p2
<< " " << (p2
==std::string::npos
) << std::endl
;
2885 CVC4::String stra1
= stra
.substr( 1 );
2886 p
= stra
.size() - stra1
.overlap(strb
);
2887 Trace("strings-csp-debug") << "Compute overlap : " << const_str
<< " " << next_const_str
<< std::endl
;
2888 size_t p2
= stra1
.find(strb
);
2889 p
= p2
==std::string::npos
? p
: ( p
>p2
+1? p2
+1 : p
);
2890 Trace("strings-csp-debug") << "overlap : " << stra1
<< " " << strb
<< " returned " << p
<< " " << p2
<< " " << (p2
==std::string::npos
) << std::endl
;
2893 if( start_index_nc_k
==index
+1 ){
2894 info
.d_ant
.push_back( xnz
);
2895 getExplanationVectorForPrefixEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
,
2896 const_k
, nconst_k
, index_c_k
, index_nc_k
, isRev
, info
.d_ant
);
2897 Node prea
= p
==stra
.size() ? const_str
: NodeManager::currentNM()->mkConst( isRev
? stra
.suffix( p
) : stra
.prefix( p
) );
2898 Node sk
= mkSkolemCached( other_str
, prea
, isRev
? sk_id_c_spt_rev
: sk_id_c_spt
, "c_spt", -1 );
2899 Trace("strings-csp") << "Const Split: " << prea
<< " is removed from " << stra
<< " due to " << strb
<< ", p=" << p
<< std::endl
;
2901 info
.d_conc
= other_str
.eqNode( isRev
? mkConcat( sk
, prea
) : mkConcat(prea
, sk
) );
2902 info
.d_new_skolem
[0].push_back( sk
);
2903 info
.d_id
= INFER_SSPLIT_CST_PROP
;
2906 /* FIXME for isRev, speculative
2907 else if( options::stringLenPropCsp() ){
2908 //propagate length constraint
2909 std::vector< Node > cc;
2910 for( unsigned i=index; i<start_index_nc_k; i++ ){
2911 cc.push_back( normal_forms[nconst_k][i] );
2913 Node lt = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH, mkConcat( cc ) );
2914 conc = NodeManager::currentNM()->mkNode( kind::GEQ, lt, NodeManager::currentNM()->mkConst( Rational(p) ) );
2915 sendInference( ant, conc, "S-Split(CSP-P)-lprop", true );
2921 info
.d_ant
.push_back( xnz
);
2922 Node const_str
= normal_forms
[const_k
][index
];
2923 getExplanationVectorForPrefixEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, index
, index
, isRev
, info
.d_ant
);
2924 CVC4::String stra
= const_str
.getConst
<String
>();
2925 if( options::stringBinaryCsp() && stra
.size()>3 ){
2926 //split string in half
2927 Node c_firstHalf
= NodeManager::currentNM()->mkConst( isRev
? stra
.substr( stra
.size()/2 ) : stra
.substr(0, stra
.size()/2 ) );
2928 Node sk
= mkSkolemCached( other_str
, c_firstHalf
, isRev
? sk_id_vc_bin_spt_rev
: sk_id_vc_bin_spt
, "cb_spt", -1 );
2929 Trace("strings-csp") << "Const Split: " << c_firstHalf
<< " is removed from " << const_str
<< " (binary) " << std::endl
;
2930 info
.d_conc
= NodeManager::currentNM()->mkNode( kind::OR
, other_str
.eqNode( isRev
? mkConcat( sk
, c_firstHalf
) : mkConcat( c_firstHalf
, sk
) ),
2931 NodeManager::currentNM()->mkNode( kind::AND
,
2932 sk
.eqNode( d_emptyString
).negate(),
2933 c_firstHalf
.eqNode( isRev
? mkConcat( sk
, other_str
) : mkConcat( other_str
, sk
) ) ) );
2934 info
.d_new_skolem
[0].push_back( sk
);
2935 info
.d_id
= INFER_SSPLIT_CST_BINARY
;
2939 Node firstChar
= stra
.size() == 1 ? const_str
: NodeManager::currentNM()->mkConst( isRev
? stra
.suffix( 1 ) : stra
.prefix( 1 ) );
2940 Node sk
= mkSkolemCached( other_str
, firstChar
, isRev
? sk_id_vc_spt_rev
: sk_id_vc_spt
, "c_spt", -1 );
2941 Trace("strings-csp") << "Const Split: " << firstChar
<< " is removed from " << const_str
<< " (serial) " << std::endl
;
2942 info
.d_conc
= other_str
.eqNode( isRev
? mkConcat( sk
, firstChar
) : mkConcat(firstChar
, sk
) );
2943 info
.d_new_skolem
[0].push_back( sk
);
2944 info
.d_id
= INFER_SSPLIT_CST
;
2951 int lentTestSuccess
= -1;
2953 if( options::stringCheckEntailLen() ){
2955 for( unsigned e
=0; e
<2; e
++ ){
2956 Node t
= e
==0 ? normal_forms
[i
][index
] : normal_forms
[j
][index
];
2957 //do not infer constants are larger than variables
2958 if( t
.getKind()!=kind::CONST_STRING
){
2959 Node lt1
= e
==0 ? length_term_i
: length_term_j
;
2960 Node lt2
= e
==0 ? length_term_j
: length_term_i
;
2961 Node ent_lit
= Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::GT
, lt1
, lt2
) );
2962 std::pair
<bool, Node
> et
= d_valuation
.entailmentCheck( THEORY_OF_TYPE_BASED
, ent_lit
);
2964 Trace("strings-entail") << "Strings entailment : " << ent_lit
<< " is entailed in the current context." << std::endl
;
2965 Trace("strings-entail") << " explanation was : " << et
.second
<< std::endl
;
2966 lentTestSuccess
= e
;
2967 lentTestExp
= et
.second
;
2974 getExplanationVectorForPrefixEq( normal_forms
, normal_form_src
, normal_forms_exp
, normal_forms_exp_depend
, i
, j
, index
, index
, isRev
, info
.d_ant
);
2976 for(unsigned xory
=0; xory
<2; xory
++) {
2977 Node x
= xory
==0 ? normal_forms
[i
][index
] : normal_forms
[j
][index
];
2978 Node xgtz
= x
.eqNode( d_emptyString
).negate();
2979 if( d_equalityEngine
.areDisequal( x
, d_emptyString
, true ) ) {
2980 info
.d_ant
.push_back( xgtz
);
2982 info
.d_antn
.push_back( xgtz
);
2985 Node sk
= mkSkolemCached( normal_forms
[i
][index
], normal_forms
[j
][index
], isRev
? sk_id_v_spt_rev
: sk_id_v_spt
, "v_spt", -1 );
2986 //must add length requirement
2987 info
.d_new_skolem
[1].push_back( sk
);
2988 Node eq1
= normal_forms
[i
][index
].eqNode( isRev
? mkConcat(sk
, normal_forms
[j
][index
]) : mkConcat(normal_forms
[j
][index
], sk
) );
2989 Node eq2
= normal_forms
[j
][index
].eqNode( isRev
? mkConcat(sk
, normal_forms
[i
][index
]) : mkConcat(normal_forms
[i
][index
], sk
) );
2991 if( lentTestSuccess
!=-1 ){
2992 info
.d_antn
.push_back( lentTestExp
);
2993 info
.d_conc
= lentTestSuccess
==0 ? eq1
: eq2
;
2994 info
.d_id
= INFER_SSPLIT_VAR_PROP
;
2997 Node ldeq
= NodeManager::currentNM()->mkNode( kind::EQUAL
, length_term_i
, length_term_j
).negate();
2998 if( d_equalityEngine
.areDisequal( length_term_i
, length_term_j
, true ) ){
2999 info
.d_ant
.push_back( ldeq
);
3001 info
.d_antn
.push_back(ldeq
);
3004 info
.d_conc
= NodeManager::currentNM()->mkNode( kind::OR
, eq1
, eq2
);
3005 info
.d_id
= INFER_SSPLIT_VAR
;
3012 pinfer
.push_back( info
);
3021 bool TheoryStrings::detectLoop( std::vector
< std::vector
< Node
> > &normal_forms
, int i
, int j
, int index
, int &loop_in_i
, int &loop_in_j
, unsigned rproc
){
3022 int has_loop
[2] = { -1, -1 };
3023 if( options::stringLB() != 2 ) {
3024 for( unsigned r
=0; r
<2; r
++ ) {
3025 int n_index
= (r
==0 ? i
: j
);
3026 int other_n_index
= (r
==0 ? j
: i
);
3027 if( normal_forms
[other_n_index
][index
].getKind() != kind::CONST_STRING
) {
3028 for( unsigned lp
= index
+1; lp
<normal_forms
[n_index
].size()-rproc
; lp
++ ){
3029 if( normal_forms
[n_index
][lp
]==normal_forms
[other_n_index
][index
] ){
3037 if( has_loop
[0]!=-1 || has_loop
[1]!=-1 ) {
3038 loop_in_i
= has_loop
[0];
3039 loop_in_j
= has_loop
[1];
3042 Trace("strings-solve-debug") << "No loops detected." << std::endl
;
3048 bool TheoryStrings::processLoop( std::vector
< std::vector
< Node
> > &normal_forms
, std::vector
< Node
> &normal_form_src
,
3049 int i
, int j
, int loop_n_index
, int other_n_index
, int loop_index
, int index
, InferInfo
& info
){
3050 if( options::stringAbortLoop() ){
3051 std::stringstream ss
;
3052 ss
<< "Looping word equation encountered." << std::endl
;
3053 throw LogicException(ss
.str());
3055 NodeManager
* nm
= NodeManager::currentNM();
3057 Trace("strings-loop") << "Detected possible loop for "
3058 << normal_forms
[loop_n_index
][loop_index
] << std::endl
;
3059 Trace("strings-loop") << " ... (X)= " << normal_forms
[other_n_index
][index
]
3062 Trace("strings-loop") << " ... T(Y.Z)= ";
3063 std::vector
<Node
>& veci
= normal_forms
[loop_n_index
];
3064 std::vector
<Node
> vec_t(veci
.begin() + index
, veci
.begin() + loop_index
);
3065 Node t_yz
= mkConcat(vec_t
);
3066 Trace("strings-loop") << " (" << t_yz
<< ")" << std::endl
;
3067 Trace("strings-loop") << " ... S(Z.Y)= ";
3068 std::vector
<Node
>& vecoi
= normal_forms
[other_n_index
];
3069 std::vector
<Node
> vec_s(vecoi
.begin() + index
+ 1, vecoi
.end());
3070 Node s_zy
= mkConcat(vec_s
);
3071 Trace("strings-loop") << s_zy
<< std::endl
;
3072 Trace("strings-loop") << " ... R= ";
3073 std::vector
<Node
> vec_r(veci
.begin() + loop_index
+ 1, veci
.end());
3074 Node r
= mkConcat(vec_r
);
3075 Trace("strings-loop") << r
<< std::endl
;
3077 if (s_zy
.isConst() && r
.isConst() && r
!= d_emptyString
)
3081 if (s_zy
.getConst
<String
>().tailcmp(r
.getConst
<String
>(), c
))
3085 s_zy
= nm
->mkConst(s_zy
.getConst
<String
>().substr(0, c
));
3088 Trace("strings-loop") << "Strings::Loop: Refactor S(Z.Y)= " << s_zy
3089 << ", c=" << c
<< std::endl
;
3095 Trace("strings-loop") << "Strings::Loop: tails are different."
3097 sendInference(info
.d_ant
, conc
, "Loop Conflict", true);
3103 for (unsigned r
= 0; r
< 2; r
++)
3105 Node t
= r
== 0 ? normal_forms
[loop_n_index
][loop_index
] : t_yz
;
3106 split_eq
= t
.eqNode(d_emptyString
);
3107 Node split_eqr
= Rewriter::rewrite(split_eq
);
3108 // the equality could rewrite to false
3109 if (!split_eqr
.isConst())
3111 if (!areDisequal(t
, d_emptyString
))
3113 // try to make t equal to empty to avoid loop
3114 info
.d_conc
= nm
->mkNode(kind::OR
, split_eq
, split_eq
.negate());
3115 info
.d_id
= INFER_LEN_SPLIT_EMP
;
3120 info
.d_ant
.push_back(split_eq
.negate());
3125 Assert(!split_eqr
.getConst
<bool>());
3129 Node ant
= mkExplain(info
.d_ant
);
3131 info
.d_antn
.push_back(ant
);
3134 if (s_zy
== t_yz
&& r
== d_emptyString
&& s_zy
.isConst()
3135 && s_zy
.getConst
<String
>().isRepeated())
3137 Node rep_c
= nm
->mkConst(s_zy
.getConst
<String
>().substr(0, 1));
3138 Trace("strings-loop") << "Special case (X)="
3139 << normal_forms
[other_n_index
][index
] << " "
3141 Trace("strings-loop") << "... (C)=" << rep_c
<< " " << std::endl
;
3144 nm
->mkNode(kind::STRING_IN_REGEXP
,
3145 normal_forms
[other_n_index
][index
],
3146 nm
->mkNode(kind::REGEXP_STAR
,
3147 nm
->mkNode(kind::STRING_TO_REGEXP
, rep_c
)));
3150 else if (t_yz
.isConst())
3152 Trace("strings-loop") << "Strings::Loop: Const Normal Breaking."
3154 CVC4::String s
= t_yz
.getConst
<CVC4::String
>();
3155 unsigned size
= s
.size();
3156 std::vector
<Node
> vconc
;
3157 for (unsigned len
= 1; len
<= size
; len
++)
3159 Node y
= nm
->mkConst(s
.substr(0, len
));
3160 Node z
= nm
->mkConst(s
.substr(len
, size
- len
));
3163 if (r
!= d_emptyString
)
3165 std::vector
<Node
> v2(vec_r
);
3166 v2
.insert(v2
.begin(), y
);
3167 v2
.insert(v2
.begin(), z
);
3168 restr
= mkConcat(z
, y
);
3169 cc
= Rewriter::rewrite(s_zy
.eqNode(mkConcat(v2
)));
3173 cc
= Rewriter::rewrite(s_zy
.eqNode(mkConcat(z
, y
)));
3179 Node conc2
= nm
->mkNode(
3180 kind::STRING_IN_REGEXP
,
3181 normal_forms
[other_n_index
][index
],
3182 nm
->mkNode(kind::REGEXP_CONCAT
,
3183 nm
->mkNode(kind::STRING_TO_REGEXP
, y
),
3184 nm
->mkNode(kind::REGEXP_STAR
,
3185 nm
->mkNode(kind::STRING_TO_REGEXP
, restr
))));
3186 cc
= cc
== d_true
? conc2
: nm
->mkNode(kind::AND
, cc
, conc2
);
3187 d_regexp_ant
[conc2
] = ant
;
3188 vconc
.push_back(cc
);
3190 conc
= vconc
.size() == 0 ? Node::null() : vconc
.size() == 1
3192 : nm
->mkNode(kind::OR
, vconc
);
3196 Trace("strings-loop") << "Strings::Loop: Normal Loop Breaking."
3199 Node sk_w
= mkSkolemS("w_loop");
3200 Node sk_y
= mkSkolemS("y_loop", 1);
3201 Node sk_z
= mkSkolemS("z_loop");
3202 // t1 * ... * tn = y * z
3203 Node conc1
= t_yz
.eqNode(mkConcat(sk_y
, sk_z
));
3204 // s1 * ... * sk = z * y * r
3205 vec_r
.insert(vec_r
.begin(), sk_y
);
3206 vec_r
.insert(vec_r
.begin(), sk_z
);
3207 Node conc2
= s_zy
.eqNode(mkConcat(vec_r
));
3209 normal_forms
[other_n_index
][index
].eqNode(mkConcat(sk_y
, sk_w
));
3210 Node restr
= r
== d_emptyString
? s_zy
: mkConcat(sk_z
, sk_y
);
3212 nm
->mkNode(kind::STRING_IN_REGEXP
,
3214 nm
->mkNode(kind::REGEXP_STAR
,
3215 nm
->mkNode(kind::STRING_TO_REGEXP
, restr
)));
3217 std::vector
<Node
> vec_conc
;
3218 vec_conc
.push_back(conc1
);
3219 vec_conc
.push_back(conc2
);
3220 vec_conc
.push_back(conc3
);
3221 vec_conc
.push_back(str_in_re
);
3222 // vec_conc.push_back(sk_y.eqNode(d_emptyString).negate());//by mkskolems
3223 conc
= nm
->mkNode(kind::AND
, vec_conc
);
3226 // set its antecedant to ant, to say when it is relevant
3227 if (!str_in_re
.isNull())
3229 d_regexp_ant
[str_in_re
] = ant
;
3232 if (options::stringProcessLoop())
3235 info
.d_id
= INFER_FLOOP
;
3236 info
.d_nf_pair
[0] = normal_form_src
[i
];
3237 info
.d_nf_pair
[1] = normal_form_src
[j
];
3240 d_out
->setIncomplete();
3244 //return true for lemma, false if we succeed
3245 void TheoryStrings::processDeq( Node ni
, Node nj
) {
3246 //Assert( areDisequal( ni, nj ) );
3247 if( d_normal_forms
[ni
].size()>1 || d_normal_forms
[nj
].size()>1 ){
3248 std::vector
< Node
> nfi
;
3249 nfi
.insert( nfi
.end(), d_normal_forms
[ni
].begin(), d_normal_forms
[ni
].end() );
3250 std::vector
< Node
> nfj
;
3251 nfj
.insert( nfj
.end(), d_normal_forms
[nj
].begin(), d_normal_forms
[nj
].end() );
3253 int revRet
= processReverseDeq( nfi
, nfj
, ni
, nj
);
3259 nfi
.insert( nfi
.end(), d_normal_forms
[ni
].begin(), d_normal_forms
[ni
].end() );
3261 nfj
.insert( nfj
.end(), d_normal_forms
[nj
].begin(), d_normal_forms
[nj
].end() );
3264 while( index
<nfi
.size() || index
<nfj
.size() ){
3265 int ret
= processSimpleDeq( nfi
, nfj
, ni
, nj
, index
, false );
3269 Assert( index
<nfi
.size() && index
<nfj
.size() );
3270 Node i
= nfi
[index
];
3271 Node j
= nfj
[index
];
3272 Trace("strings-solve-debug") << "...Processing(DEQ) " << i
<< " " << j
<< std::endl
;
3273 if( !areEqual( i
, j
) ){
3274 Assert( i
.getKind()!=kind::CONST_STRING
|| j
.getKind()!=kind::CONST_STRING
);
3275 std::vector
< Node
> lexp
;
3276 Node li
= getLength( i
, lexp
);
3277 Node lj
= getLength( j
, lexp
);
3278 if( areDisequal( li
, lj
) ){
3279 if( i
.getKind()==kind::CONST_STRING
|| j
.getKind()==kind::CONST_STRING
){
3281 Node const_k
= i
.getKind() == kind::CONST_STRING
? i
: j
;
3282 Node nconst_k
= i
.getKind() == kind::CONST_STRING
? j
: i
;
3283 Node lnck
= i
.getKind() == kind::CONST_STRING
? lj
: li
;
3284 if( !d_equalityEngine
.areDisequal( nconst_k
, d_emptyString
, true ) ){
3285 Node eq
= nconst_k
.eqNode( d_emptyString
);
3286 Node conc
= NodeManager::currentNM()->mkNode( kind::OR
, eq
, eq
.negate() );
3287 sendInference( d_empty_vec
, conc
, "D-DISL-Emp-Split" );
3290 //split on first character
3291 CVC4::String str
= const_k
.getConst
<String
>();
3292 Node firstChar
= str
.size() == 1 ? const_k
: NodeManager::currentNM()->mkConst( str
.prefix( 1 ) );
3293 if( areEqual( lnck
, d_one
) ){
3294 if( areDisequal( firstChar
, nconst_k
) ){
3296 }else if( !areEqual( firstChar
, nconst_k
) ){
3297 //splitting on demand : try to make them disequal
3299 firstChar
, nconst_k
, "S-Split(DEQL-Const)", false))
3305 Node sk
= mkSkolemCached( nconst_k
, firstChar
, sk_id_dc_spt
, "dc_spt", 2 );
3306 Node skr
= mkSkolemCached( nconst_k
, firstChar
, sk_id_dc_spt_rem
, "dc_spt_rem" );
3307 Node eq1
= nconst_k
.eqNode( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT
, sk
, skr
) );
3308 eq1
= Rewriter::rewrite( eq1
);
3309 Node eq2
= nconst_k
.eqNode( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT
, firstChar
, skr
) );
3310 std::vector
< Node
> antec
;
3311 antec
.insert( antec
.end(), d_normal_forms_exp
[ni
].begin(), d_normal_forms_exp
[ni
].end() );
3312 antec
.insert( antec
.end(), d_normal_forms_exp
[nj
].begin(), d_normal_forms_exp
[nj
].end() );
3313 antec
.push_back( nconst_k
.eqNode( d_emptyString
).negate() );
3314 sendInference( antec
, NodeManager::currentNM()->mkNode( kind::OR
,
3315 NodeManager::currentNM()->mkNode( kind::AND
, eq1
, sk
.eqNode( firstChar
).negate() ), eq2
), "D-DISL-CSplit" );
3316 d_pending_req_phase
[ eq1
] = true;
3321 Trace("strings-solve") << "Non-Simple Case 1 : add lemma " << std::endl
;
3323 std::vector
< Node
> antec
;
3324 std::vector
< Node
> antec_new_lits
;
3325 antec
.insert( antec
.end(), d_normal_forms_exp
[ni
].begin(), d_normal_forms_exp
[ni
].end() );
3326 antec
.insert( antec
.end(), d_normal_forms_exp
[nj
].begin(), d_normal_forms_exp
[nj
].end() );
3328 if( areDisequal( ni
, nj
) ){
3329 antec
.push_back( ni
.eqNode( nj
).negate() );
3331 antec_new_lits
.push_back( ni
.eqNode( nj
).negate() );
3333 antec_new_lits
.push_back( li
.eqNode( lj
).negate() );
3334 std::vector
< Node
> conc
;
3335 Node sk1
= mkSkolemCached( i
, j
, sk_id_deq_x
, "x_dsplit" );
3336 Node sk2
= mkSkolemCached( i
, j
, sk_id_deq_y
, "y_dsplit" );
3337 Node sk3
= mkSkolemCached( i
, j
, sk_id_deq_z
, "z_dsplit", 1 );
3338 //Node nemp = sk3.eqNode(d_emptyString).negate();
3339 //conc.push_back(nemp);
3340 Node lsk1
= mkLength( sk1
);
3341 conc
.push_back( lsk1
.eqNode( li
) );
3342 Node lsk2
= mkLength( sk2
);
3343 conc
.push_back( lsk2
.eqNode( lj
) );
3344 conc
.push_back( NodeManager::currentNM()->mkNode( kind::OR
, j
.eqNode( mkConcat( sk1
, sk3
) ), i
.eqNode( mkConcat( sk2
, sk3
) ) ) );
3345 sendInference( antec
, antec_new_lits
, NodeManager::currentNM()->mkNode( kind::AND
, conc
), "D-DISL-Split" );
3346 ++(d_statistics
.d_deq_splits
);
3349 }else if( areEqual( li
, lj
) ){
3350 Assert( !areDisequal( i
, j
) );
3351 //splitting on demand : try to make them disequal
3352 if (sendSplit(i
, j
, "S-Split(DEQL)", false))
3357 //splitting on demand : try to make lengths equal
3358 if (sendSplit(li
, lj
, "D-Split"))
3371 int TheoryStrings::processReverseDeq( std::vector
< Node
>& nfi
, std::vector
< Node
>& nfj
, Node ni
, Node nj
) {
3372 //reverse normal form of i, j
3373 std::reverse( nfi
.begin(), nfi
.end() );
3374 std::reverse( nfj
.begin(), nfj
.end() );
3377 int ret
= processSimpleDeq( nfi
, nfj
, ni
, nj
, index
, true );
3379 //reverse normal form of i, j
3380 std::reverse( nfi
.begin(), nfi
.end() );
3381 std::reverse( nfj
.begin(), nfj
.end() );
3386 int TheoryStrings::processSimpleDeq( std::vector
< Node
>& nfi
, std::vector
< Node
>& nfj
, Node ni
, Node nj
, unsigned& index
, bool isRev
){
3387 // See if one side is constant, if so, the disequality ni != nj is satisfied
3388 // since ni does not contain nj or vice versa.
3389 // This is only valid when isRev is false, since when isRev=true, the contents
3390 // of normal form vectors nfi and nfj are reversed.
3393 for (unsigned i
= 0; i
< 2; i
++)
3395 Node c
= getConstantEqc(i
== 0 ? ni
: nj
);
3399 if (!TheoryStringsRewriter::canConstantContainList(
3400 c
, i
== 0 ? nfj
: nfi
, findex
, lindex
))
3402 Trace("strings-solve-debug")
3403 << "Disequality: constant cannot contain list" << std::endl
;
3409 while( index
<nfi
.size() || index
<nfj
.size() ) {
3410 if( index
>=nfi
.size() || index
>=nfj
.size() ){
3411 Trace("strings-solve-debug") << "Disequality normalize empty" << std::endl
;
3412 std::vector
< Node
> ant
;
3413 //we have a conflict : because the lengths are equal, the remainder needs to be empty, which will lead to a conflict
3414 Node lni
= getLengthExp( ni
, ant
, d_normal_forms_base
[ni
] );
3415 Node lnj
= getLengthExp( nj
, ant
, d_normal_forms_base
[nj
] );
3416 ant
.push_back( lni
.eqNode( lnj
) );
3417 ant
.insert( ant
.end(), d_normal_forms_exp
[ni
].begin(), d_normal_forms_exp
[ni
].end() );
3418 ant
.insert( ant
.end(), d_normal_forms_exp
[nj
].begin(), d_normal_forms_exp
[nj
].end() );
3419 std::vector
< Node
> cc
;
3420 std::vector
< Node
>& nfk
= index
>=nfi
.size() ? nfj
: nfi
;
3421 for( unsigned index_k
=index
; index_k
<nfk
.size(); index_k
++ ){
3422 cc
.push_back( nfk
[index_k
].eqNode( d_emptyString
) );
3424 Node conc
= cc
.size()==1 ? cc
[0] : NodeManager::currentNM()->mkNode( kind::AND
, cc
);
3425 conc
= Rewriter::rewrite( conc
);
3426 sendInference( ant
, conc
, "Disequality Normalize Empty", true);
3429 Node i
= nfi
[index
];
3430 Node j
= nfj
[index
];
3431 Trace("strings-solve-debug") << "...Processing(QED) " << i
<< " " << j
<< std::endl
;
3432 if( !areEqual( i
, j
) ) {
3433 if( i
.getKind()==kind::CONST_STRING
&& j
.getKind()==kind::CONST_STRING
) {
3434 unsigned int len_short
= i
.getConst
<String
>().size() < j
.getConst
<String
>().size() ? i
.getConst
<String
>().size() : j
.getConst
<String
>().size();
3435 bool isSameFix
= isRev
? i
.getConst
<String
>().rstrncmp(j
.getConst
<String
>(), len_short
): i
.getConst
<String
>().strncmp(j
.getConst
<String
>(), len_short
);
3437 //same prefix/suffix
3438 //k is the index of the string that is shorter
3439 Node nk
= i
.getConst
<String
>().size() < j
.getConst
<String
>().size() ? i
: j
;
3440 Node nl
= i
.getConst
<String
>().size() < j
.getConst
<String
>().size() ? j
: i
;
3443 int new_len
= nl
.getConst
<String
>().size() - len_short
;
3444 remainderStr
= NodeManager::currentNM()->mkConst( nl
.getConst
<String
>().substr(0, new_len
) );
3445 Trace("strings-solve-debug-test") << "Rev. Break normal form of " << nl
<< " into " << nk
<< ", " << remainderStr
<< std::endl
;
3447 remainderStr
= NodeManager::currentNM()->mkConst( nl
.getConst
<String
>().substr( len_short
) );
3448 Trace("strings-solve-debug-test") << "Break normal form of " << nl
<< " into " << nk
<< ", " << remainderStr
<< std::endl
;
3450 if( i
.getConst
<String
>().size() < j
.getConst
<String
>().size() ) {
3451 nfj
.insert( nfj
.begin() + index
+ 1, remainderStr
);
3452 nfj
[index
] = nfi
[index
];
3454 nfi
.insert( nfi
.begin() + index
+ 1, remainderStr
);
3455 nfi
[index
] = nfj
[index
];
3461 std::vector
< Node
> lexp
;
3462 Node li
= getLength( i
, lexp
);
3463 Node lj
= getLength( j
, lexp
);
3464 if( areEqual( li
, lj
) && areDisequal( i
, j
) ){
3465 Trace("strings-solve") << "Simple Case 2 : found equal length disequal sub strings " << i
<< " " << j
<< std::endl
;
3466 //we are done: D-Remove
3479 void TheoryStrings::addNormalFormPair( Node n1
, Node n2
){
3480 if( !isNormalFormPair( n1
, n2
) ){
3482 NodeIntMap::const_iterator it
= d_nf_pairs
.find( n1
);
3483 if( it
!=d_nf_pairs
.end() ){
3484 index
= (*it
).second
;
3486 d_nf_pairs
[n1
] = index
+ 1;
3487 if( index
<(int)d_nf_pairs_data
[n1
].size() ){
3488 d_nf_pairs_data
[n1
][index
] = n2
;
3490 d_nf_pairs_data
[n1
].push_back( n2
);
3492 Assert( isNormalFormPair( n1
, n2
) );
3494 Trace("strings-nf-debug") << "Already a normal form pair " << n1
<< " " << n2
<< std::endl
;
3498 bool TheoryStrings::isNormalFormPair( Node n1
, Node n2
) {
3499 //TODO: modulo equality?
3500 return isNormalFormPair2( n1
, n2
) || isNormalFormPair2( n2
, n1
);
3503 bool TheoryStrings::isNormalFormPair2( Node n1
, Node n2
) {
3504 //Trace("strings-debug") << "is normal form pair. " << n1 << " " << n2 << std::endl;
3505 NodeIntMap::const_iterator it
= d_nf_pairs
.find( n1
);
3506 if( it
!=d_nf_pairs
.end() ){
3507 Assert( d_nf_pairs_data
.find( n1
)!=d_nf_pairs_data
.end() );
3508 for( int i
=0; i
<(*it
).second
; i
++ ){
3509 Assert( i
<(int)d_nf_pairs_data
[n1
].size() );
3510 if( d_nf_pairs_data
[n1
][i
]==n2
){
3518 void TheoryStrings::registerTerm( Node n
, int effort
) {
3519 TypeNode tn
= n
.getType();
3520 bool do_register
= true;
3523 if (options::stringEagerLen())
3525 do_register
= effort
== 0;
3529 do_register
= effort
> 0 || n
.getKind() != kind::STRING_CONCAT
;
3533 if(d_registered_terms_cache
.find(n
) == d_registered_terms_cache
.end()) {
3534 d_registered_terms_cache
.insert(n
);
3535 Debug("strings-register") << "TheoryStrings::registerTerm() " << n
<< ", effort = " << effort
<< std::endl
;
3538 //register length information:
3539 // for variables, split on empty vs positive length
3540 // for concat/const/replace, introduce proxy var and state length relation
3542 bool processed
= false;
3543 if( n
.getKind()!=kind::STRING_CONCAT
&& n
.getKind()!=kind::CONST_STRING
) {
3544 if( d_length_lemma_terms_cache
.find( n
)==d_length_lemma_terms_cache
.end() ){
3545 Node lsumb
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, n
);
3546 lsum
= Rewriter::rewrite( lsumb
);
3547 // can register length term if it does not rewrite
3549 sendLengthLemma( n
);
3557 Node sk
= mkSkolemS( "lsym", -1 );
3558 StringsProxyVarAttribute spva
;
3559 sk
.setAttribute(spva
,true);
3560 Node eq
= Rewriter::rewrite( sk
.eqNode(n
) );
3561 Trace("strings-lemma") << "Strings::Lemma LENGTH Term : " << eq
<< std::endl
;
3562 d_proxy_var
[n
] = sk
;
3563 Trace("strings-assert") << "(assert " << eq
<< ")" << std::endl
;
3565 Node skl
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, sk
);
3566 if( n
.getKind()==kind::STRING_CONCAT
){
3567 std::vector
<Node
> node_vec
;
3568 for( unsigned i
=0; i
<n
.getNumChildren(); i
++ ) {
3569 if( n
[i
].getAttribute(StringsProxyVarAttribute()) ){
3570 Assert( d_proxy_var_to_length
.find( n
[i
] )!=d_proxy_var_to_length
.end() );
3571 node_vec
.push_back( d_proxy_var_to_length
[n
[i
]] );
3573 Node lni
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, n
[i
] );
3574 node_vec
.push_back(lni
);
3577 lsum
= NodeManager::currentNM()->mkNode( kind::PLUS
, node_vec
);
3578 lsum
= Rewriter::rewrite( lsum
);
3579 }else if( n
.getKind()==kind::CONST_STRING
){
3580 lsum
= NodeManager::currentNM()->mkConst( ::CVC4::Rational( n
.getConst
<String
>().size() ) );
3582 Assert( !lsum
.isNull() );
3583 d_proxy_var_to_length
[sk
] = lsum
;
3584 Node ceq
= Rewriter::rewrite( skl
.eqNode( lsum
) );
3585 Trace("strings-lemma") << "Strings::Lemma LENGTH : " << ceq
<< std::endl
;
3586 Trace("strings-lemma-debug") << " prerewrite : " << skl
.eqNode( lsum
) << std::endl
;
3587 Trace("strings-assert") << "(assert " << ceq
<< ")" << std::endl
;
3591 else if (n
.getKind() == kind::STRING_CODE
)
3593 d_has_str_code
= true;
3594 NodeManager
* nm
= NodeManager::currentNM();
3595 // ite( str.len(s)==1, 0 <= str.code(s) < num_codes, str.code(s)=-1 )
3596 Node code_len
= mkLength(n
[0]).eqNode(d_one
);
3597 Node code_eq_neg1
= n
.eqNode(d_neg_one
);
3598 Node code_range
= nm
->mkNode(
3600 nm
->mkNode(kind::GEQ
, n
, d_zero
),
3602 kind::LT
, n
, nm
->mkConst(Rational(CVC4::String::num_codes()))));
3603 Node lem
= nm
->mkNode(kind::ITE
, code_len
, code_range
, code_eq_neg1
);
3604 Trace("strings-lemma") << "Strings::Lemma CODE : " << lem
<< std::endl
;
3605 Trace("strings-assert") << "(assert " << lem
<< ")" << std::endl
;
3612 void TheoryStrings::sendInference( std::vector
< Node
>& exp
, std::vector
< Node
>& exp_n
, Node eq
, const char * c
, bool asLemma
) {
3613 eq
= eq
.isNull() ? d_false
: Rewriter::rewrite( eq
);
3615 if( Trace
.isOn("strings-infer-debug") ){
3616 Trace("strings-infer-debug") << "By " << c
<< ", infer : " << eq
<< " from: " << std::endl
;
3617 for( unsigned i
=0; i
<exp
.size(); i
++ ){
3618 Trace("strings-infer-debug") << " " << exp
[i
] << std::endl
;
3620 for( unsigned i
=0; i
<exp_n
.size(); i
++ ){
3621 Trace("strings-infer-debug") << " N:" << exp_n
[i
] << std::endl
;
3623 //Trace("strings-infer-debug") << "as lemma : " << asLemma << std::endl;
3625 //check if we should send a lemma or an inference
3626 if( asLemma
|| eq
==d_false
|| eq
.getKind()==kind::OR
|| !exp_n
.empty() || options::stringInferAsLemmas() ){
3628 if( options::stringRExplainLemmas() ){
3629 eq_exp
= mkExplain( exp
, exp_n
);
3632 eq_exp
= mkAnd( exp_n
);
3633 }else if( exp_n
.empty() ){
3634 eq_exp
= mkAnd( exp
);
3636 std::vector
< Node
> ev
;
3637 ev
.insert( ev
.end(), exp
.begin(), exp
.end() );
3638 ev
.insert( ev
.end(), exp_n
.begin(), exp_n
.end() );
3639 eq_exp
= NodeManager::currentNM()->mkNode( kind::AND
, ev
);
3642 // if we have unexplained literals, this lemma is not a conflict
3643 if (eq
== d_false
&& !exp_n
.empty())
3645 eq
= eq_exp
.negate();
3648 sendLemma( eq_exp
, eq
, c
);
3650 sendInfer( mkAnd( exp
), eq
, c
);
3655 void TheoryStrings::sendInference( std::vector
< Node
>& exp
, Node eq
, const char * c
, bool asLemma
) {
3656 std::vector
< Node
> exp_n
;
3657 sendInference( exp
, exp_n
, eq
, c
, asLemma
);
3660 void TheoryStrings::sendLemma( Node ant
, Node conc
, const char * c
) {
3661 if( conc
.isNull() || conc
== d_false
) {
3662 Trace("strings-conflict") << "Strings::Conflict : " << c
<< " : " << ant
<< std::endl
;
3663 Trace("strings-lemma") << "Strings::Conflict : " << c
<< " : " << ant
<< std::endl
;
3664 Trace("strings-assert") << "(assert (not " << ant
<< ")) ; conflict " << c
<< std::endl
;
3665 d_out
->conflict(ant
);
3669 if( ant
== d_true
) {
3672 lem
= NodeManager::currentNM()->mkNode( kind::IMPLIES
, ant
, conc
);
3674 Trace("strings-lemma") << "Strings::Lemma " << c
<< " : " << lem
<< std::endl
;
3675 Trace("strings-assert") << "(assert " << lem
<< ") ; lemma " << c
<< std::endl
;
3676 d_lemma_cache
.push_back( lem
);
3680 void TheoryStrings::sendInfer( Node eq_exp
, Node eq
, const char * c
) {
3681 if( options::stringInferSym() ){
3682 std::vector
< Node
> vars
;
3683 std::vector
< Node
> subs
;
3684 std::vector
< Node
> unproc
;
3685 inferSubstitutionProxyVars( eq_exp
, vars
, subs
, unproc
);
3686 if( unproc
.empty() ){
3687 Trace("strings-lemma-debug") << "Strings::Infer " << eq
<< " from " << eq_exp
<< " by " << c
<< std::endl
;
3688 Node eqs
= eq
.substitute( vars
.begin(), vars
.end(), subs
.begin(), subs
.end() );
3689 Trace("strings-lemma-debug") << "Strings::Infer Alternate : " << eqs
<< std::endl
;
3690 for( unsigned i
=0; i
<vars
.size(); i
++ ){
3691 Trace("strings-lemma-debug") << " " << vars
[i
] << " -> " << subs
[i
] << std::endl
;
3693 sendLemma( d_true
, eqs
, c
);
3696 for( unsigned i
=0; i
<unproc
.size(); i
++ ){
3697 Trace("strings-lemma-debug") << " non-trivial exp : " << unproc
[i
] << std::endl
;
3701 Trace("strings-lemma") << "Strings::Infer " << eq
<< " from " << eq_exp
<< " by " << c
<< std::endl
;
3702 Trace("strings-assert") << "(assert (=> " << eq_exp
<< " " << eq
<< ")) ; infer " << c
<< std::endl
;
3703 d_pending
.push_back( eq
);
3704 d_pending_exp
[eq
] = eq_exp
;
3705 d_infer
.push_back( eq
);
3706 d_infer_exp
.push_back( eq_exp
);
3709 bool TheoryStrings::sendSplit(Node a
, Node b
, const char* c
, bool preq
)
3711 Node eq
= a
.eqNode( b
);
3712 eq
= Rewriter::rewrite( eq
);
3715 Node neq
= NodeManager::currentNM()->mkNode(kind::NOT
, eq
);
3716 Node lemma_or
= NodeManager::currentNM()->mkNode(kind::OR
, eq
, neq
);
3717 Trace("strings-lemma") << "Strings::Lemma " << c
<< " SPLIT : " << lemma_or
3719 d_lemma_cache
.push_back(lemma_or
);
3720 d_pending_req_phase
[eq
] = preq
;
3721 ++(d_statistics
.d_splits
);
3728 void TheoryStrings::sendLengthLemma( Node n
){
3729 Node n_len
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, n
);
3730 if( options::stringSplitEmp() || !options::stringLenGeqZ() ){
3731 NodeManager
* nm
= NodeManager::currentNM();
3732 Node n_len_eq_z
= n_len
.eqNode( d_zero
);
3733 Node n_len_eq_z_2
= n
.eqNode( d_emptyString
);
3734 Node case_empty
= nm
->mkNode(AND
, n_len_eq_z
, n_len_eq_z_2
);
3735 case_empty
= Rewriter::rewrite(case_empty
);
3736 Node case_nempty
= nm
->mkNode(GT
, n_len
, d_zero
);
3737 if (!case_empty
.isConst())
3739 Node lem
= nm
->mkNode(OR
, case_empty
, case_nempty
);
3741 Trace("strings-lemma") << "Strings::Lemma LENGTH >= 0 : " << lem
3743 // prefer trying the empty case first
3744 // notice that requirePhase must only be called on rewritten literals that
3745 // occur in the CNF stream.
3746 n_len_eq_z
= Rewriter::rewrite(n_len_eq_z
);
3747 Assert(!n_len_eq_z
.isConst());
3748 d_out
->requirePhase(n_len_eq_z
, true);
3749 n_len_eq_z_2
= Rewriter::rewrite(n_len_eq_z_2
);
3750 Assert(!n_len_eq_z_2
.isConst());
3751 d_out
->requirePhase(n_len_eq_z_2
, true);
3753 else if (!case_empty
.getConst
<bool>())
3755 // the rewriter knows that n is non-empty
3756 Trace("strings-lemma")
3757 << "Strings::Lemma LENGTH > 0 (non-empty): " << case_nempty
3759 d_out
->lemma(case_nempty
);
3763 // If n = "" ---> true or len( n ) = 0 ----> true, then we expect that
3764 // n ---> "". Since this method is only called on non-constants n, it must
3765 // be that n = "" ^ len( n ) = 0 does not rewrite to true.
3769 //AJR: probably a good idea
3770 if( options::stringLenGeqZ() ){
3771 Node n_len_geq
= NodeManager::currentNM()->mkNode( kind::GEQ
, n_len
, d_zero
);
3772 n_len_geq
= Rewriter::rewrite( n_len_geq
);
3773 d_out
->lemma( n_len_geq
);
3777 void TheoryStrings::inferSubstitutionProxyVars( Node n
, std::vector
< Node
>& vars
, std::vector
< Node
>& subs
, std::vector
< Node
>& unproc
) {
3778 if( n
.getKind()==kind::AND
){
3779 for( unsigned i
=0; i
<n
.getNumChildren(); i
++ ){
3780 inferSubstitutionProxyVars( n
[i
], vars
, subs
, unproc
);
3783 }else if( n
.getKind()==kind::EQUAL
){
3784 Node ns
= n
.substitute( vars
.begin(), vars
.end(), subs
.begin(), subs
.end() );
3785 ns
= Rewriter::rewrite( ns
);
3786 if( ns
.getKind()==kind::EQUAL
){
3789 for( unsigned i
=0; i
<2; i
++ ){
3791 if( ns
[i
].getAttribute(StringsProxyVarAttribute()) ){
3793 }else if( ns
[i
].isConst() ){
3794 NodeNodeMap::const_iterator it
= d_proxy_var
.find( ns
[i
] );
3795 if( it
!=d_proxy_var
.end() ){
3801 if( v
.getNumChildren()==0 ){
3805 //both sides involved in proxy var
3816 subs
.push_back( s
);
3817 vars
.push_back( v
);
3825 unproc
.push_back( n
);
3830 Node
TheoryStrings::mkConcat( Node n1
, Node n2
) {
3831 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT
, n1
, n2
) );
3834 Node
TheoryStrings::mkConcat( Node n1
, Node n2
, Node n3
) {
3835 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_CONCAT
, n1
, n2
, n3
) );
3838 Node
TheoryStrings::mkConcat( const std::vector
< Node
>& c
) {
3839 return Rewriter::rewrite( c
.size()>1 ? NodeManager::currentNM()->mkNode( kind::STRING_CONCAT
, c
) : ( c
.size()==1 ? c
[0] : d_emptyString
) );
3842 Node
TheoryStrings::mkLength( Node t
) {
3843 return Rewriter::rewrite( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, t
) );
3846 Node
TheoryStrings::mkSkolemCached( Node a
, Node b
, int id
, const char * c
, int isLenSplit
){
3847 //return mkSkolemS( c, isLenSplit );
3848 std::map
< int, Node
>::iterator it
= d_skolem_cache
[a
][b
].find( id
);
3849 if( it
==d_skolem_cache
[a
][b
].end() ){
3850 Node sk
= mkSkolemS( c
, isLenSplit
);
3851 d_skolem_cache
[a
][b
][id
] = sk
;
3858 //isLenSplit: -1-ignore, 0-no restriction, 1-greater than one, 2-one
3859 Node
TheoryStrings::mkSkolemS( const char *c
, int isLenSplit
) {
3860 Node n
= NodeManager::currentNM()->mkSkolem( c
, NodeManager::currentNM()->stringType(), "string sko" );
3861 d_all_skolems
.insert(n
);
3862 d_length_lemma_terms_cache
.insert( n
);
3863 ++(d_statistics
.d_new_skolems
);
3864 if( isLenSplit
==0 ){
3865 sendLengthLemma( n
);
3866 } else if( isLenSplit
== 1 ){
3867 registerNonEmptySkolem( n
);
3868 }else if( isLenSplit
==2 ){
3869 Node len_one
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, n
).eqNode( d_one
);
3870 Trace("strings-lemma") << "Strings::Lemma SK-ONE : " << len_one
<< std::endl
;
3871 Trace("strings-assert") << "(assert " << len_one
<< ")" << std::endl
;
3872 d_out
->lemma( len_one
);
3877 void TheoryStrings::registerNonEmptySkolem( Node n
) {
3878 if( d_skolem_ne_reg_cache
.find( n
)==d_skolem_ne_reg_cache
.end() ){
3879 d_skolem_ne_reg_cache
.insert( n
);
3880 d_equalityEngine
.assertEquality(n
.eqNode(d_emptyString
), false, d_true
);
3881 Node len_n_gt_z
= NodeManager::currentNM()->mkNode(kind::GT
,
3882 NodeManager::currentNM()->mkNode(kind::STRING_LENGTH
, n
), d_zero
);
3883 Trace("strings-lemma") << "Strings::Lemma SK-NON-ZERO : " << len_n_gt_z
<< std::endl
;
3884 Trace("strings-assert") << "(assert " << len_n_gt_z
<< ")" << std::endl
;
3885 d_out
->lemma(len_n_gt_z
);
3889 Node
TheoryStrings::mkExplain( std::vector
< Node
>& a
) {
3890 std::vector
< Node
> an
;
3891 return mkExplain( a
, an
);
3894 Node
TheoryStrings::mkExplain( std::vector
< Node
>& a
, std::vector
< Node
>& an
) {
3895 std::vector
< TNode
> antec_exp
;
3896 for( unsigned i
=0; i
<a
.size(); i
++ ) {
3897 if( std::find( a
.begin(), a
.begin() + i
, a
[i
] )==a
.begin() + i
) {
3899 Debug("strings-explain") << "Ask for explanation of " << a
[i
] << std::endl
;
3901 if(a
[i
].getKind() == kind::EQUAL
) {
3902 //Assert( hasTerm(a[i][0]) );
3903 //Assert( hasTerm(a[i][1]) );
3904 Assert( areEqual(a
[i
][0], a
[i
][1]) );
3905 if( a
[i
][0]==a
[i
][1] ){
3908 } else if( a
[i
].getKind()==kind::NOT
&& a
[i
][0].getKind()==kind::EQUAL
) {
3909 Assert( hasTerm(a
[i
][0][0]) );
3910 Assert( hasTerm(a
[i
][0][1]) );
3911 AlwaysAssert( d_equalityEngine
.areDisequal(a
[i
][0][0], a
[i
][0][1], true) );
3912 }else if( a
[i
].getKind() == kind::AND
){
3913 for( unsigned j
=0; j
<a
[i
].getNumChildren(); j
++ ){
3914 a
.push_back( a
[i
][j
] );
3919 unsigned ps
= antec_exp
.size();
3920 explain(a
[i
], antec_exp
);
3921 Debug("strings-explain") << "Done, explanation was : " << std::endl
;
3922 for( unsigned j
=ps
; j
<antec_exp
.size(); j
++ ) {
3923 Debug("strings-explain") << " " << antec_exp
[j
] << std::endl
;
3925 Debug("strings-explain") << std::endl
;
3929 for( unsigned i
=0; i
<an
.size(); i
++ ) {
3930 if( std::find( an
.begin(), an
.begin() + i
, an
[i
] )==an
.begin() + i
){
3931 Debug("strings-explain") << "Add to explanation (new literal) " << an
[i
] << std::endl
;
3932 antec_exp
.push_back(an
[i
]);
3936 if( antec_exp
.empty() ) {
3938 } else if( antec_exp
.size()==1 ) {
3941 ant
= NodeManager::currentNM()->mkNode( kind::AND
, antec_exp
);
3943 //ant = Rewriter::rewrite( ant );
3947 Node
TheoryStrings::mkAnd( std::vector
< Node
>& a
) {
3948 std::vector
< Node
> au
;
3949 for( unsigned i
=0; i
<a
.size(); i
++ ){
3950 if( std::find( au
.begin(), au
.end(), a
[i
] )==au
.end() ){
3951 au
.push_back( a
[i
] );
3956 } else if( au
.size() == 1 ) {
3959 return NodeManager::currentNM()->mkNode( kind::AND
, au
);
3963 void TheoryStrings::getConcatVec( Node n
, std::vector
< Node
>& c
) {
3964 if( n
.getKind()==kind::STRING_CONCAT
) {
3965 for( unsigned i
=0; i
<n
.getNumChildren(); i
++ ) {
3966 if( !areEqual( n
[i
], d_emptyString
) ) {
3967 c
.push_back( n
[i
] );
3975 void TheoryStrings::checkNormalFormsDeq()
3977 std::vector
< std::vector
< Node
> > cols
;
3978 std::vector
< Node
> lts
;
3979 std::map
< Node
, std::map
< Node
, bool > > processed
;
3981 //for each pair of disequal strings, must determine whether their lengths are equal or disequal
3982 for( NodeList::const_iterator id
= d_ee_disequalities
.begin(); id
!= d_ee_disequalities
.end(); ++id
) {
3985 for( unsigned i
=0; i
<2; i
++ ){
3986 n
[i
] = d_equalityEngine
.getRepresentative( eq
[i
] );
3988 if( processed
[n
[0]].find( n
[1] )==processed
[n
[0]].end() ){
3989 processed
[n
[0]][n
[1]] = true;
3991 for( unsigned i
=0; i
<2; i
++ ){
3992 EqcInfo
* ei
= getOrMakeEqcInfo( n
[i
], false );
3993 lt
[i
] = ei
? ei
->d_length_term
: Node::null();
3994 if( lt
[i
].isNull() ){
3997 lt
[i
] = NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, lt
[i
] );
3999 if( !areEqual( lt
[0], lt
[1] ) && !areDisequal( lt
[0], lt
[1] ) ){
4000 sendSplit( lt
[0], lt
[1], "DEQ-LENGTH-SP" );
4005 if( !hasProcessed() ){
4006 separateByLength( d_strings_eqc
, cols
, lts
);
4007 for( unsigned i
=0; i
<cols
.size(); i
++ ){
4008 if( cols
[i
].size()>1 && d_lemma_cache
.empty() ){
4009 Trace("strings-solve") << "- Verify disequalities are processed for " << cols
[i
][0] << ", normal form : ";
4010 printConcat( d_normal_forms
[cols
[i
][0]], "strings-solve" );
4011 Trace("strings-solve") << "... #eql = " << cols
[i
].size() << std::endl
;
4012 //must ensure that normal forms are disequal
4013 for( unsigned j
=0; j
<cols
[i
].size(); j
++ ){
4014 for( unsigned k
=(j
+1); k
<cols
[i
].size(); k
++ ){
4015 //for strings that are disequal, but have the same length
4016 if( areDisequal( cols
[i
][j
], cols
[i
][k
] ) ){
4017 Assert( !d_conflict
);
4018 Trace("strings-solve") << "- Compare " << cols
[i
][j
] << " ";
4019 printConcat( d_normal_forms
[cols
[i
][j
]], "strings-solve" );
4020 Trace("strings-solve") << " against " << cols
[i
][k
] << " ";
4021 printConcat( d_normal_forms
[cols
[i
][k
]], "strings-solve" );
4022 Trace("strings-solve") << "..." << std::endl
;
4023 processDeq( cols
[i
][j
], cols
[i
][k
] );
4024 if( hasProcessed() ){
4035 void TheoryStrings::checkLengthsEqc() {
4036 if( options::stringLenNorm() ){
4037 for( unsigned i
=0; i
<d_strings_eqc
.size(); i
++ ){
4038 //if( d_normal_forms[nodes[i]].size()>1 ) {
4039 Trace("strings-process-debug") << "Process length constraints for " << d_strings_eqc
[i
] << std::endl
;
4040 //check if there is a length term for this equivalence class
4041 EqcInfo
* ei
= getOrMakeEqcInfo( d_strings_eqc
[i
], false );
4042 Node lt
= ei
? ei
->d_length_term
: Node::null();
4043 if( !lt
.isNull() ) {
4044 Node llt
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, lt
);
4045 //now, check if length normalization has occurred
4046 if( ei
->d_normalized_length
.get().isNull() ) {
4047 Node nf
= mkConcat( d_normal_forms
[d_strings_eqc
[i
]] );
4048 if( Trace
.isOn("strings-process-debug") ){
4049 Trace("strings-process-debug") << " normal form is " << nf
<< " from base " << d_normal_forms_base
[d_strings_eqc
[i
]] << std::endl
;
4050 Trace("strings-process-debug") << " normal form exp is: " << std::endl
;
4051 for( unsigned j
=0; j
<d_normal_forms_exp
[d_strings_eqc
[i
]].size(); j
++ ){
4052 Trace("strings-process-debug") << " " << d_normal_forms_exp
[d_strings_eqc
[i
]][j
] << std::endl
;
4056 //if not, add the lemma
4057 std::vector
< Node
> ant
;
4058 ant
.insert( ant
.end(), d_normal_forms_exp
[d_strings_eqc
[i
]].begin(), d_normal_forms_exp
[d_strings_eqc
[i
]].end() );
4059 ant
.push_back( d_normal_forms_base
[d_strings_eqc
[i
]].eqNode( lt
) );
4060 Node lc
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, nf
);
4061 Node lcr
= Rewriter::rewrite( lc
);
4062 Trace("strings-process-debug") << "Rewrote length " << lc
<< " to " << lcr
<< std::endl
;
4063 Node eq
= llt
.eqNode( lcr
);
4065 ei
->d_normalized_length
.set( eq
);
4066 sendInference( ant
, eq
, "LEN-NORM", true );
4070 Trace("strings-process-debug") << "No length term for eqc " << d_strings_eqc
[i
] << " " << d_eqc_to_len_term
[d_strings_eqc
[i
]] << std::endl
;
4071 if( !options::stringEagerLen() ){
4072 Node c
= mkConcat( d_normal_forms
[d_strings_eqc
[i
]] );
4073 registerTerm( c
, 3 );
4076 NodeNodeMap::const_iterator it = d_proxy_var.find( c );
4077 if( it!=d_proxy_var.end() ){
4078 Node pv = (*it).second;
4079 Assert( d_proxy_var_to_length.find( pv )!=d_proxy_var_to_length.end() );
4080 Node pvl = d_proxy_var_to_length[pv];
4081 Node ceq = Rewriter::rewrite( mkLength( pv ).eqNode( pvl ) );
4082 sendInference( d_empty_vec, ceq, "LEN-NORM-I", true );
4089 // Trace("strings-process-debug") << "Do not process length constraints for " << nodes[i] << " " << d_normal_forms[nodes[i]].size() << std::endl;
4095 void TheoryStrings::checkCardinality() {
4096 //int cardinality = options::stringCharCardinality();
4097 //Trace("strings-solve-debug2") << "get cardinality: " << cardinality << endl;
4099 //AJR: this will create a partition of eqc, where each collection has length that are pairwise propagated to be equal.
4100 // 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).
4101 // TODO: revisit this?
4102 std::vector
< std::vector
< Node
> > cols
;
4103 std::vector
< Node
> lts
;
4104 separateByLength( d_strings_eqc
, cols
, lts
);
4106 for( unsigned i
= 0; i
<cols
.size(); ++i
) {
4108 Trace("strings-card") << "Number of strings with length equal to " << lr
<< " is " << cols
[i
].size() << std::endl
;
4109 if( cols
[i
].size() > 1 ) {
4111 unsigned card_need
= 1;
4112 double curr
= (double)cols
[i
].size();
4113 while( curr
>d_card_size
){
4114 curr
= curr
/(double)d_card_size
;
4117 Trace("strings-card") << "Need length " << card_need
<< " for this number of strings (where alphabet size is " << d_card_size
<< ")." << std::endl
;
4118 //check if we need to split
4119 bool needsSplit
= true;
4121 // if constant, compare
4122 Node cmp
= NodeManager::currentNM()->mkNode( kind::GEQ
, lr
, NodeManager::currentNM()->mkConst( Rational( card_need
) ) );
4123 cmp
= Rewriter::rewrite( cmp
);
4124 needsSplit
= cmp
!=d_true
;
4126 // find the minimimum constant that we are unknown to be disequal from, or otherwise stop if we increment such that cardinality does not apply
4128 bool success
= true;
4129 while( r
<card_need
&& success
){
4130 Node rr
= NodeManager::currentNM()->mkConst
<Rational
>( Rational(r
) );
4131 if( areDisequal( rr
, lr
) ){
4138 Trace("strings-card") << "Symbolic length " << lr
<< " must be at least " << r
<< " due to constant disequalities." << std::endl
;
4140 needsSplit
= r
<card_need
;
4144 unsigned int int_k
= (unsigned int)card_need
;
4145 for( std::vector
< Node
>::iterator itr1
= cols
[i
].begin();
4146 itr1
!= cols
[i
].end(); ++itr1
) {
4147 for( std::vector
< Node
>::iterator itr2
= itr1
+ 1;
4148 itr2
!= cols
[i
].end(); ++itr2
) {
4149 if(!areDisequal( *itr1
, *itr2
)) {
4151 if (sendSplit(*itr1
, *itr2
, "CARD-SP"))
4158 EqcInfo
* ei
= getOrMakeEqcInfo( lr
, true );
4159 Trace("strings-card") << "Previous cardinality used for " << lr
<< " is " << ((int)ei
->d_cardinality_lem_k
.get()-1) << std::endl
;
4160 if( int_k
+1 > ei
->d_cardinality_lem_k
.get() ){
4161 Node k_node
= NodeManager::currentNM()->mkConst( ::CVC4::Rational( int_k
) );
4162 //add cardinality lemma
4163 Node dist
= NodeManager::currentNM()->mkNode( kind::DISTINCT
, cols
[i
] );
4164 std::vector
< Node
> vec_node
;
4165 vec_node
.push_back( dist
);
4166 for( std::vector
< Node
>::iterator itr1
= cols
[i
].begin();
4167 itr1
!= cols
[i
].end(); ++itr1
) {
4168 Node len
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, *itr1
);
4170 Node len_eq_lr
= len
.eqNode(lr
);
4171 vec_node
.push_back( len_eq_lr
);
4174 Node len
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, cols
[i
][0] );
4175 Node cons
= NodeManager::currentNM()->mkNode( kind::GEQ
, len
, k_node
);
4176 cons
= Rewriter::rewrite( cons
);
4177 ei
->d_cardinality_lem_k
.set( int_k
+1 );
4179 sendInference( d_empty_vec
, vec_node
, cons
, "CARDINALITY", true );
4188 void TheoryStrings::getEquivalenceClasses( std::vector
< Node
>& eqcs
) {
4189 eq::EqClassesIterator eqcs_i
= eq::EqClassesIterator( &d_equalityEngine
);
4190 while( !eqcs_i
.isFinished() ) {
4191 Node eqc
= (*eqcs_i
);
4192 //if eqc.getType is string
4193 if (eqc
.getType().isString()) {
4194 eqcs
.push_back( eqc
);
4200 void TheoryStrings::separateByLength(std::vector
< Node
>& n
,
4201 std::vector
< std::vector
< Node
> >& cols
,
4202 std::vector
< Node
>& lts
) {
4203 unsigned leqc_counter
= 0;
4204 std::map
< Node
, unsigned > eqc_to_leqc
;
4205 std::map
< unsigned, Node
> leqc_to_eqc
;
4206 std::map
< unsigned, std::vector
< Node
> > eqc_to_strings
;
4207 for( unsigned i
=0; i
<n
.size(); i
++ ) {
4209 Assert( d_equalityEngine
.getRepresentative(eqc
)==eqc
);
4210 EqcInfo
* ei
= getOrMakeEqcInfo( eqc
, false );
4211 Node lt
= ei
? ei
->d_length_term
: Node::null();
4213 lt
= NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, lt
);
4214 Node r
= d_equalityEngine
.getRepresentative( lt
);
4215 if( eqc_to_leqc
.find( r
)==eqc_to_leqc
.end() ){
4216 eqc_to_leqc
[r
] = leqc_counter
;
4217 leqc_to_eqc
[leqc_counter
] = r
;
4220 eqc_to_strings
[ eqc_to_leqc
[r
] ].push_back( eqc
);
4222 eqc_to_strings
[leqc_counter
].push_back( eqc
);
4226 for( std::map
< unsigned, std::vector
< Node
> >::iterator it
= eqc_to_strings
.begin(); it
!= eqc_to_strings
.end(); ++it
){
4227 cols
.push_back( std::vector
< Node
>() );
4228 cols
.back().insert( cols
.back().end(), it
->second
.begin(), it
->second
.end() );
4229 lts
.push_back( leqc_to_eqc
[it
->first
] );
4233 void TheoryStrings::printConcat( std::vector
< Node
>& n
, const char * c
) {
4234 for( unsigned i
=0; i
<n
.size(); i
++ ){
4235 if( i
>0 ) Trace(c
) << " ++ ";
4242 //// Finite Model Finding
4244 Node
TheoryStrings::getNextDecisionRequest( unsigned& priority
) {
4245 if( options::stringFMF() && !d_conflict
){
4246 Node in_var_lsum
= d_input_var_lsum
.get();
4247 //Trace("strings-fmf-debug") << "Strings::FMF: Assertion Level = " << d_valuation.getAssertionLevel() << std::endl;
4248 //initialize the term we will minimize
4249 if( in_var_lsum
.isNull() && !d_input_vars
.empty() ){
4250 Trace("strings-fmf-debug") << "Input variables: ";
4251 std::vector
< Node
> ll
;
4252 for(NodeSet::key_iterator itr
= d_input_vars
.key_begin();
4253 itr
!= d_input_vars
.key_end(); ++itr
) {
4254 Trace("strings-fmf-debug") << " " << (*itr
) ;
4255 ll
.push_back( NodeManager::currentNM()->mkNode( kind::STRING_LENGTH
, *itr
) );
4257 Trace("strings-fmf-debug") << std::endl
;
4258 in_var_lsum
= ll
.size()==1 ? ll
[0] : NodeManager::currentNM()->mkNode( kind::PLUS
, ll
);
4259 in_var_lsum
= Rewriter::rewrite( in_var_lsum
);
4260 d_input_var_lsum
.set( in_var_lsum
);
4262 if( !in_var_lsum
.isNull() ){
4263 //Trace("strings-fmf") << "Get next decision request." << std::endl;
4264 //check if we need to decide on something
4265 int decideCard
= d_curr_cardinality
.get();
4266 if( d_cardinality_lits
.find( decideCard
)!=d_cardinality_lits
.end() ){
4268 Node cnode
= d_cardinality_lits
[ d_curr_cardinality
.get() ];
4269 if( d_valuation
.hasSatValue( cnode
, value
) ) {
4271 d_curr_cardinality
.set( d_curr_cardinality
.get() + 1 );
4272 decideCard
= d_curr_cardinality
.get();
4273 Trace("strings-fmf-debug") << "Has false SAT value, increment and decide." << std::endl
;
4276 Trace("strings-fmf-debug") << "Has true SAT value, do not decide." << std::endl
;
4279 Trace("strings-fmf-debug") << "No SAT value, decide." << std::endl
;
4282 if( decideCard
!=-1 ){
4283 if( d_cardinality_lits
.find( decideCard
)==d_cardinality_lits
.end() ){
4284 Node lit
= NodeManager::currentNM()->mkNode( kind::LEQ
, in_var_lsum
, NodeManager::currentNM()->mkConst( Rational( decideCard
) ) );
4285 lit
= Rewriter::rewrite( lit
);
4286 d_cardinality_lits
[decideCard
] = lit
;
4287 Node lem
= NodeManager::currentNM()->mkNode( kind::OR
, lit
, lit
.negate() );
4288 Trace("strings-fmf") << "Strings::FMF: Add decision lemma " << lem
<< ", decideCard = " << decideCard
<< std::endl
;
4289 d_out
->lemma( lem
);
4290 d_out
->requirePhase( lit
, true );
4292 Node lit
= d_cardinality_lits
[ decideCard
];
4293 Trace("strings-fmf") << "Strings::FMF: Decide positive on " << lit
<< std::endl
;
4299 return Node::null();
4302 Node
TheoryStrings::ppRewrite(TNode atom
) {
4303 Trace("strings-ppr") << "TheoryStrings::ppRewrite " << atom
<< std::endl
;
4304 if( !options::stringLazyPreproc() ){
4305 //eager preprocess here
4306 std::vector
< Node
> new_nodes
;
4307 Node ret
= d_preproc
.processAssertion( atom
, new_nodes
);
4309 Trace("strings-ppr") << " rewrote " << atom
<< " -> " << ret
<< ", with " << new_nodes
.size() << " lemmas." << std::endl
;
4310 for( unsigned i
=0; i
<new_nodes
.size(); i
++ ){
4311 Trace("strings-ppr") << " lemma : " << new_nodes
[i
] << std::endl
;
4312 d_out
->lemma( new_nodes
[i
] );
4316 Assert( new_nodes
.empty() );
4323 TheoryStrings::Statistics::Statistics():
4324 d_splits("theory::strings::NumOfSplitOnDemands", 0),
4325 d_eq_splits("theory::strings::NumOfEqSplits", 0),
4326 d_deq_splits("theory::strings::NumOfDiseqSplits", 0),
4327 d_loop_lemmas("theory::strings::NumOfLoops", 0),
4328 d_new_skolems("theory::strings::NumOfNewSkolems", 0)
4330 smtStatisticsRegistry()->registerStat(&d_splits
);
4331 smtStatisticsRegistry()->registerStat(&d_eq_splits
);
4332 smtStatisticsRegistry()->registerStat(&d_deq_splits
);
4333 smtStatisticsRegistry()->registerStat(&d_loop_lemmas
);
4334 smtStatisticsRegistry()->registerStat(&d_new_skolems
);
4337 TheoryStrings::Statistics::~Statistics(){
4338 smtStatisticsRegistry()->unregisterStat(&d_splits
);
4339 smtStatisticsRegistry()->unregisterStat(&d_eq_splits
);
4340 smtStatisticsRegistry()->unregisterStat(&d_deq_splits
);
4341 smtStatisticsRegistry()->unregisterStat(&d_loop_lemmas
);
4342 smtStatisticsRegistry()->unregisterStat(&d_new_skolems
);
4364 //// Regular Expressions
4367 unsigned TheoryStrings::getNumMemberships( Node n
, bool isPos
) {
4369 NodeIntMap::const_iterator it
= d_pos_memberships
.find( n
);
4370 if( it
!=d_pos_memberships
.end() ){
4371 return (*it
).second
;
4374 NodeIntMap::const_iterator it
= d_neg_memberships
.find( n
);
4375 if( it
!=d_neg_memberships
.end() ){
4376 return (*it
).second
;
4382 Node
TheoryStrings::getMembership( Node n
, bool isPos
, unsigned i
) {
4383 return isPos
? d_pos_memberships_data
[n
][i
] : d_neg_memberships_data
[n
][i
];
4386 Node
TheoryStrings::mkRegExpAntec(Node atom
, Node ant
) {
4387 if(d_regexp_ant
.find(atom
) == d_regexp_ant
.end()) {
4388 return NodeManager::currentNM()->mkNode(kind::AND
, ant
, atom
);
4390 Node n
= d_regexp_ant
[atom
];
4391 return NodeManager::currentNM()->mkNode(kind::AND
, ant
, n
);
4395 void TheoryStrings::checkMemberships() {
4396 //add the memberships
4397 std::vector
<Node
> mems
= getExtTheory()->getActive(kind::STRING_IN_REGEXP
);
4398 for (unsigned i
= 0; i
< mems
.size(); i
++) {
4400 Assert( d_extf_info_tmp
.find( n
)!=d_extf_info_tmp
.end() );
4401 if( d_extf_info_tmp
[n
].d_pol
==1 || d_extf_info_tmp
[n
].d_pol
==-1 ){
4402 bool pol
= d_extf_info_tmp
[n
].d_pol
==1;
4403 Trace("strings-process-debug") << " add membership : " << n
<< ", pol = " << pol
<< std::endl
;
4404 addMembership( pol
? n
: n
.negate() );
4406 Trace("strings-process-debug") << " irrelevant (non-asserted) membership : " << n
<< std::endl
;
4410 bool addedLemma
= false;
4411 bool changed
= false;
4412 std::vector
< Node
> processed
;
4413 std::vector
< Node
> cprocessed
;
4415 Trace("regexp-debug") << "Checking Memberships ... " << std::endl
;
4416 //if(options::stringEIT()) {
4417 //TODO: Opt for normal forms
4418 for( NodeIntMap::const_iterator itr_xr
= d_pos_memberships
.begin(); itr_xr
!= d_pos_memberships
.end(); ++itr_xr
){
4419 bool spflag
= false;
4420 Node x
= (*itr_xr
).first
;
4421 Trace("regexp-debug") << "Checking Memberships for " << x
<< std::endl
;
4422 if(d_inter_index
.find(x
) == d_inter_index
.end()) {
4423 d_inter_index
[x
] = 0;
4425 int cur_inter_idx
= d_inter_index
[x
];
4426 unsigned n_pmem
= (*itr_xr
).second
;
4427 Assert( getNumMemberships( x
, true )==n_pmem
);
4428 if( cur_inter_idx
!= (int)n_pmem
) {
4430 d_inter_cache
[x
] = getMembership( x
, true, 0 );
4431 d_inter_index
[x
] = 1;
4432 Trace("regexp-debug") << "... only one choice " << std::endl
;
4433 } else if(n_pmem
> 1) {
4435 if(d_inter_cache
.find(x
) != d_inter_cache
.end()) {
4436 r
= d_inter_cache
[x
];
4439 r
= getMembership( x
, true, 0 );
4443 unsigned k_start
= cur_inter_idx
;
4444 Trace("regexp-debug") << "... staring from : " << cur_inter_idx
<< ", we have " << n_pmem
<< std::endl
;
4445 for(unsigned k
= k_start
; k
<n_pmem
; k
++) {
4446 Node r2
= getMembership( x
, true, k
);
4447 r
= d_regexp_opr
.intersect(r
, r2
, spflag
);
4450 } else if(r
== d_emptyRegexp
) {
4451 std::vector
< Node
> vec_nodes
;
4452 for( unsigned kk
=0; kk
<=k
; kk
++ ){
4453 Node rr
= getMembership( x
, true, kk
);
4454 Node n
= NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP
, x
, rr
);
4455 vec_nodes
.push_back( n
);
4458 sendInference(vec_nodes
, conc
, "INTERSECT CONFLICT", true);
4467 if(!d_conflict
&& !spflag
) {
4468 d_inter_cache
[x
] = r
;
4469 d_inter_index
[x
] = (int)n_pmem
;
4476 Trace("regexp-debug") << "... No Intersect Conflict in Memberships, addedLemma: " << addedLemma
<< std::endl
;
4478 NodeManager
* nm
= NodeManager::currentNM();
4479 for( unsigned i
=0; i
<d_regexp_memberships
.size(); i
++ ) {
4480 //check regular expression membership
4481 Node assertion
= d_regexp_memberships
[i
];
4482 Trace("regexp-debug") << "Check : " << assertion
<< " " << (d_regexp_ucached
.find(assertion
) == d_regexp_ucached
.end()) << " " << (d_regexp_ccached
.find(assertion
) == d_regexp_ccached
.end()) << std::endl
;
4483 if( d_regexp_ucached
.find(assertion
) == d_regexp_ucached
.end()
4484 && d_regexp_ccached
.find(assertion
) == d_regexp_ccached
.end() ) {
4485 Trace("strings-regexp") << "We have regular expression assertion : " << assertion
<< std::endl
;
4486 Node atom
= assertion
.getKind()==kind::NOT
? assertion
[0] : assertion
;
4487 bool polarity
= assertion
.getKind()!=kind::NOT
;
4491 std::vector
< Node
> rnfexp
;
4495 x
= getNormalString(x
, rnfexp
);
4498 if (!d_regexp_opr
.checkConstRegExp(r
))
4500 r
= getNormalSymRegExp(r
, rnfexp
);
4503 Trace("strings-regexp-nf") << "Term " << atom
<< " is normalized to "
4504 << x
<< " IN " << r
<< std::endl
;
4508 Rewriter::rewrite(nm
->mkNode(kind::STRING_IN_REGEXP
, x
, r
));
4515 d_regexp_ccached
.insert(assertion
);
4518 else if (tmp
== d_false
)
4520 Node antec
= mkRegExpAntec(assertion
, mkExplain(rnfexp
));
4521 Node conc
= Node::null();
4522 sendLemma(antec
, conc
, "REGEXP NF Conflict");
4529 flag
= checkPDerivative(x
, r
, atom
, addedLemma
, rnfexp
);
4531 if(! options::stringExp()) {
4532 throw LogicException("Strings Incomplete (due to Negative Membership) by default, try --strings-exp option.");
4536 //check if the term is atomic
4537 Node xr
= getRepresentative( x
);
4538 //Trace("strings-regexp") << xr << " is rep of " << x << std::endl;
4539 //Assert( d_normal_forms.find( xr )!=d_normal_forms.end() );
4540 Trace("strings-regexp")
4541 << "Unroll/simplify membership of atomic term " << xr
4543 // if so, do simple unrolling
4544 std::vector
<Node
> nvec
;
4548 d_regexp_opr
.simplify(atom
, nvec
, polarity
);
4550 Node antec
= assertion
;
4551 if (d_regexp_ant
.find(assertion
) != d_regexp_ant
.end())
4553 antec
= d_regexp_ant
[assertion
];
4554 for (std::vector
<Node
>::const_iterator itr
= nvec
.begin();
4558 if (itr
->getKind() == kind::STRING_IN_REGEXP
)
4560 if (d_regexp_ant
.find(*itr
) == d_regexp_ant
.end())
4562 d_regexp_ant
[*itr
] = antec
;
4567 antec
= NodeManager::currentNM()->mkNode(
4568 kind::AND
, antec
, mkExplain(rnfexp
));
4569 Node conc
= nvec
.size() == 1
4571 : NodeManager::currentNM()->mkNode(kind::AND
, nvec
);
4572 conc
= Rewriter::rewrite(conc
);
4573 sendLemma(antec
, conc
, "REGEXP_Unfold");
4577 cprocessed
.push_back(assertion
);
4581 processed
.push_back(assertion
);
4583 // d_regexp_ucached[assertion] = true;
4593 for( unsigned i
=0; i
<processed
.size(); i
++ ) {
4594 Trace("strings-regexp") << "...add " << processed
[i
] << " to u-cache." << std::endl
;
4595 d_regexp_ucached
.insert(processed
[i
]);
4597 for( unsigned i
=0; i
<cprocessed
.size(); i
++ ) {
4598 Trace("strings-regexp") << "...add " << cprocessed
[i
] << " to c-cache." << std::endl
;
4599 d_regexp_ccached
.insert(cprocessed
[i
]);
4605 bool TheoryStrings::checkPDerivative( Node x
, Node r
, Node atom
, bool &addedLemma
, std::vector
< Node
> &nf_exp
) {
4607 Node antnf
= mkExplain(nf_exp
);
4609 if(areEqual(x
, d_emptyString
)) {
4611 switch(d_regexp_opr
.delta(r
, exp
)) {
4613 Node antec
= mkRegExpAntec(atom
, x
.eqNode(d_emptyString
));
4614 antec
= NodeManager::currentNM()->mkNode(kind::AND
, antec
, antnf
);
4615 sendLemma(antec
, exp
, "RegExp Delta");
4617 d_regexp_ccached
.insert(atom
);
4621 d_regexp_ccached
.insert(atom
);
4625 Node antec
= mkRegExpAntec(atom
, x
.eqNode(d_emptyString
));
4626 antec
= NodeManager::currentNM()->mkNode(kind::AND
, antec
, antnf
);
4627 Node conc
= Node::null();
4628 sendLemma(antec
, conc
, "RegExp Delta CONFLICT");
4630 d_regexp_ccached
.insert(atom
);
4638 /*Node xr = getRepresentative( x );
4640 Node n = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, xr, r);
4641 Node nn = Rewriter::rewrite( n );
4643 d_regexp_ccached.insert(atom);
4645 } else if(nn == d_false) {
4646 Node antec = mkRegExpAntec(atom, x.eqNode(xr));
4647 Node conc = Node::null();
4648 sendLemma(antec, conc, "RegExp Delta CONFLICT");
4650 d_regexp_ccached.insert(atom);
4654 Node sREant
= mkRegExpAntec(atom
, d_true
);
4655 sREant
= NodeManager::currentNM()->mkNode(kind::AND
, sREant
, antnf
);
4656 if(deriveRegExp( x
, r
, sREant
)) {
4658 d_regexp_ccached
.insert(atom
);
4665 CVC4::String
TheoryStrings::getHeadConst( Node x
) {
4667 return x
.getConst
< String
>();
4668 } else if( x
.getKind() == kind::STRING_CONCAT
) {
4669 if( x
[0].isConst() ) {
4670 return x
[0].getConst
< String
>();
4672 return d_emptyString
.getConst
< String
>();
4675 return d_emptyString
.getConst
< String
>();
4679 bool TheoryStrings::deriveRegExp( Node x
, Node r
, Node ant
) {
4681 Assert(x
!= d_emptyString
);
4682 Trace("regexp-derive") << "TheoryStrings::deriveRegExp: x=" << x
<< ", r= " << r
<< std::endl
;
4684 // Node n = NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP, x, r );
4685 // Node r = Rewriter::rewrite( n );
4687 // sendLemma(ant, r, "REGEXP REWRITE");
4691 CVC4::String s
= getHeadConst( x
);
4692 if( !s
.isEmptyString() && d_regexp_opr
.checkConstRegExp( r
) ) {
4693 Node conc
= Node::null();
4696 for(unsigned i
=0; i
<s
.size(); ++i
) {
4697 CVC4::String c
= s
.substr(i
, 1);
4699 int rt
= d_regexp_opr
.derivativeS(dc
, c
, dc2
);
4703 } else if(rt
== 2) {
4712 Assert(false, "Impossible: TheoryStrings::deriveRegExp: const string in const regular expression.");
4715 Assert( x
.getKind() == kind::STRING_CONCAT
);
4716 std::vector
< Node
> vec_nodes
;
4717 for(unsigned int i
=1; i
<x
.getNumChildren(); ++i
) {
4718 vec_nodes
.push_back( x
[i
] );
4720 Node left
= mkConcat( vec_nodes
);
4721 left
= Rewriter::rewrite( left
);
4722 conc
= NodeManager::currentNM()->mkNode( kind::STRING_IN_REGEXP
, left
, dc
);
4724 /*std::vector< Node > sdc;
4725 d_regexp_opr.simplify(conc, sdc, true);
4726 if(sdc.size() == 1) {
4729 conc = Rewriter::rewrite(NodeManager::currentNM()->mkNode(kind::AND, conc));
4733 sendLemma(ant
, conc
, "RegExp-Derive");
4740 void TheoryStrings::addMembership(Node assertion
) {
4741 bool polarity
= assertion
.getKind() != kind::NOT
;
4742 TNode atom
= polarity
? assertion
: assertion
[0];
4747 NodeIntMap::const_iterator it
= d_pos_memberships
.find( x
);
4748 if( it
!=d_nf_pairs
.end() ){
4749 index
= (*it
).second
;
4750 for( int k
=0; k
<index
; k
++ ){
4751 if( k
<(int)d_pos_memberships_data
[x
].size() ){
4752 if( d_pos_memberships_data
[x
][k
]==r
){
4760 d_pos_memberships
[x
] = index
+ 1;
4761 if( index
<(int)d_pos_memberships_data
[x
].size() ){
4762 d_pos_memberships_data
[x
][index
] = r
;
4764 d_pos_memberships_data
[x
].push_back( r
);
4766 } else if(!options::stringIgnNegMembership()) {
4767 /*if(options::stringEIT() && d_regexp_opr.checkConstRegExp(r)) {
4769 Node r2 = d_regexp_opr.complement(r, rt);
4770 Node a = NodeManager::currentNM()->mkNode(kind::STRING_IN_REGEXP, x, r2);
4773 NodeIntMap::const_iterator it
= d_neg_memberships
.find( x
);
4774 if( it
!=d_nf_pairs
.end() ){
4775 index
= (*it
).second
;
4776 for( int k
=0; k
<index
; k
++ ){
4777 if( k
<(int)d_neg_memberships_data
[x
].size() ){
4778 if( d_neg_memberships_data
[x
][k
]==r
){
4786 d_neg_memberships
[x
] = index
+ 1;
4787 if( index
<(int)d_neg_memberships_data
[x
].size() ){
4788 d_neg_memberships_data
[x
][index
] = r
;
4790 d_neg_memberships_data
[x
].push_back( r
);
4794 if(polarity
|| !options::stringIgnNegMembership()) {
4795 d_regexp_memberships
.push_back( assertion
);
4799 Node
TheoryStrings::getNormalString( Node x
, std::vector
< Node
>& nf_exp
){
4801 Node xr
= getRepresentative( x
);
4802 if( d_normal_forms
.find( xr
) != d_normal_forms
.end() ){
4803 Node ret
= mkConcat( d_normal_forms
[xr
] );
4804 nf_exp
.insert( nf_exp
.end(), d_normal_forms_exp
[xr
].begin(), d_normal_forms_exp
[xr
].end() );
4805 addToExplanation( x
, d_normal_forms_base
[xr
], nf_exp
);
4806 Trace("strings-debug") << "Term: " << x
<< " has a normal form " << ret
<< std::endl
;
4809 if(x
.getKind() == kind::STRING_CONCAT
) {
4810 std::vector
< Node
> vec_nodes
;
4811 for(unsigned i
=0; i
<x
.getNumChildren(); i
++) {
4812 Node nc
= getNormalString( x
[i
], nf_exp
);
4813 vec_nodes
.push_back( nc
);
4815 return mkConcat( vec_nodes
);
4822 Node
TheoryStrings::getNormalSymRegExp(Node r
, std::vector
<Node
> &nf_exp
) {
4824 switch( r
.getKind() ) {
4825 case kind::REGEXP_EMPTY
:
4826 case kind::REGEXP_SIGMA
:
4828 case kind::STRING_TO_REGEXP
: {
4829 if(!r
[0].isConst()) {
4830 Node tmp
= getNormalString( r
[0], nf_exp
);
4832 ret
= NodeManager::currentNM()->mkNode(kind::STRING_TO_REGEXP
, tmp
);
4837 case kind::REGEXP_CONCAT
:
4838 case kind::REGEXP_UNION
:
4839 case kind::REGEXP_INTER
:
4840 case kind::REGEXP_STAR
:
4842 std::vector
< Node
> vec_nodes
;
4843 for (const Node
& cr
: r
)
4845 vec_nodes
.push_back(getNormalSymRegExp(cr
, nf_exp
));
4847 ret
= Rewriter::rewrite(
4848 NodeManager::currentNM()->mkNode(r
.getKind(), vec_nodes
));
4851 //case kind::REGEXP_PLUS:
4852 //case kind::REGEXP_OPT:
4853 //case kind::REGEXP_RANGE:
4855 Trace("strings-error") << "Unsupported term: " << r
<< " in normalization SymRegExp." << std::endl
;
4857 //return Node::null();
4863 /** run the given inference step */
4864 void TheoryStrings::runInferStep(InferStep s
, int effort
)
4866 Trace("strings-process") << "Run " << s
;
4869 Trace("strings-process") << ", effort = " << effort
;
4871 Trace("strings-process") << "..." << std::endl
;
4874 case CHECK_INIT
: checkInit(); break;
4875 case CHECK_CONST_EQC
: checkConstantEquivalenceClasses(); break;
4876 case CHECK_EXTF_EVAL
: checkExtfEval(effort
); break;
4877 case CHECK_CYCLES
: checkCycles(); break;
4878 case CHECK_FLAT_FORMS
: checkFlatForms(); break;
4879 case CHECK_NORMAL_FORMS_EQ
: checkNormalFormsEq(); break;
4880 case CHECK_NORMAL_FORMS_DEQ
: checkNormalFormsDeq(); break;
4881 case CHECK_CODES
: checkCodes(); break;
4882 case CHECK_LENGTH_EQC
: checkLengthsEqc(); break;
4883 case CHECK_EXTF_REDUCTION
: checkExtfReductions(effort
); break;
4884 case CHECK_MEMBERSHIP
: checkMemberships(); break;
4885 case CHECK_CARDINALITY
: checkCardinality(); break;
4886 default: Unreachable(); break;
4888 Trace("strings-process") << "Done " << s
4889 << ", addedFact = " << !d_pending
.empty() << " "
4890 << !d_lemma_cache
.empty()
4891 << ", d_conflict = " << d_conflict
<< std::endl
;
4894 bool TheoryStrings::hasStrategyEffort(Effort e
) const
4896 return d_strat_steps
.find(e
) != d_strat_steps
.end();
4899 void TheoryStrings::addStrategyStep(InferStep s
, int effort
, bool addBreak
)
4901 // must run check init first
4902 Assert((s
== CHECK_INIT
)==d_infer_steps
.empty());
4903 // must use check cycles when using flat forms
4904 Assert(s
!= CHECK_FLAT_FORMS
4905 || std::find(d_infer_steps
.begin(), d_infer_steps
.end(), CHECK_CYCLES
)
4906 != d_infer_steps
.end());
4907 d_infer_steps
.push_back(s
);
4908 d_infer_step_effort
.push_back(effort
);
4911 d_infer_steps
.push_back(BREAK
);
4912 d_infer_step_effort
.push_back(0);
4916 void TheoryStrings::initializeStrategy()
4918 // initialize the strategy if not already done so
4919 if (!d_strategy_init
)
4921 std::map
<Effort
, unsigned> step_begin
;
4922 std::map
<Effort
, unsigned> step_end
;
4923 d_strategy_init
= true;
4924 // beginning indices
4925 step_begin
[EFFORT_FULL
] = 0;
4926 if (options::stringEager())
4928 step_begin
[EFFORT_STANDARD
] = 0;
4930 // add the inference steps
4931 addStrategyStep(CHECK_INIT
);
4932 addStrategyStep(CHECK_CONST_EQC
);
4933 addStrategyStep(CHECK_EXTF_EVAL
, 0);
4934 addStrategyStep(CHECK_CYCLES
);
4935 if (options::stringFlatForms())
4937 addStrategyStep(CHECK_FLAT_FORMS
);
4939 addStrategyStep(CHECK_EXTF_REDUCTION
, 1);
4940 if (options::stringEager())
4942 // do only the above inferences at standard effort, if applicable
4943 step_end
[EFFORT_STANDARD
] = d_infer_steps
.size() - 1;
4945 addStrategyStep(CHECK_NORMAL_FORMS_EQ
);
4946 addStrategyStep(CHECK_EXTF_EVAL
, 1);
4947 if (!options::stringEagerLen())
4949 addStrategyStep(CHECK_LENGTH_EQC
);
4951 addStrategyStep(CHECK_NORMAL_FORMS_DEQ
);
4952 addStrategyStep(CHECK_CODES
);
4953 if (options::stringEagerLen())
4955 addStrategyStep(CHECK_LENGTH_EQC
);
4957 if (options::stringExp() && !options::stringGuessModel())
4959 addStrategyStep(CHECK_EXTF_REDUCTION
, 2);
4961 addStrategyStep(CHECK_MEMBERSHIP
);
4962 addStrategyStep(CHECK_CARDINALITY
);
4963 step_end
[EFFORT_FULL
] = d_infer_steps
.size() - 1;
4964 if (options::stringExp() && options::stringGuessModel())
4966 step_begin
[EFFORT_LAST_CALL
] = d_infer_steps
.size();
4967 // these two steps are run in parallel
4968 addStrategyStep(CHECK_EXTF_REDUCTION
, 2, false);
4969 addStrategyStep(CHECK_EXTF_EVAL
, 3);
4970 step_end
[EFFORT_LAST_CALL
] = d_infer_steps
.size() - 1;
4972 // set the beginning/ending ranges
4973 for (const std::pair
<const Effort
, unsigned>& it_begin
: step_begin
)
4975 Effort e
= it_begin
.first
;
4976 std::map
<Effort
, unsigned>::iterator it_end
= step_end
.find(e
);
4977 Assert(it_end
!= step_end
.end());
4979 std::pair
<unsigned, unsigned>(it_begin
.second
, it_end
->second
);
4984 void TheoryStrings::runStrategy(unsigned sbegin
, unsigned send
)
4986 Trace("strings-process") << "----check, next round---" << std::endl
;
4987 for (unsigned i
= sbegin
; i
<= send
; i
++)
4989 InferStep curr
= d_infer_steps
[i
];
4999 runInferStep(curr
, d_infer_step_effort
[i
]);
5006 Trace("strings-process") << "----finished round---" << std::endl
;
5009 }/* CVC4::theory::strings namespace */
5010 }/* CVC4::theory namespace */
5011 }/* CVC4 namespace */