src/theory/strings/theory_strings.h

   1 /*********************                                                        */
   2 /*! \file theory_strings.h
   3  ** \verbatim
   4  ** Original author: Tianyi Liang
   5  ** Major contributors: Andrew Reynolds
   6  ** Minor contributors (to current version): Martin Brain <>, Morgan Deters
   7  ** This file is part of the CVC4 project.
   8  ** Copyright (c) 2009-2014  New York University and The University of Iowa
   9  ** See the file COPYING in the top-level source directory for licensing
  10  ** information.\endverbatim
  11  **
  12  ** \brief Theory of strings
  13  **
  14  ** Theory of strings.
  15  **/
  16
  17 #include "cvc4_private.h"
  18
  19 #ifndef __CVC4__THEORY__STRINGS__THEORY_STRINGS_H
  20 #define __CVC4__THEORY__STRINGS__THEORY_STRINGS_H
  21
  22 #include "theory/theory.h"
  23 #include "theory/uf/equality_engine.h"
  24 #include "theory/strings/theory_strings_preprocess.h"
  25 #include "theory/strings/regexp_operation.h"
  26
  27 #include "context/cdchunk_list.h"
  28 #include "context/cdhashset.h"
  29
  30 #include <climits>
  31
  32 namespace CVC4 {
  33 namespace theory {
  34 namespace strings {
  35
  36 /**
  37  * Decision procedure for strings.
  38  *
  39  */
  40
  41 class TheoryStrings : public Theory {
  42   typedef context::CDChunkList<Node> NodeList;
  43   typedef context::CDHashMap<Node, NodeList*, NodeHashFunction> NodeListMap;
  44   typedef context::CDHashMap<Node, bool, NodeHashFunction> NodeBoolMap;
  45   typedef context::CDHashMap<Node, int, NodeHashFunction> NodeIntMap;
  46   typedef context::CDHashMap<Node, Node, NodeHashFunction> NodeNodeMap;
  47   typedef context::CDHashSet<Node, NodeHashFunction> NodeSet;
  48
  49 public:
  50   TheoryStrings(context::Context* c, context::UserContext* u, OutputChannel& out, Valuation valuation, const LogicInfo& logicInfo);
  51   ~TheoryStrings();
  52
  53   void setMasterEqualityEngine(eq::EqualityEngine* eq);
  54
  55   std::string identify() const { return std::string("TheoryStrings"); }
  56
  57 public:
  58   void propagate(Effort e);
  59   bool propagate(TNode literal);
  60   void explain( TNode literal, std::vector<TNode>& assumptions );
  61   Node explain( TNode literal );
  62
  63
  64   // NotifyClass for equality engine
  65   class NotifyClass : public eq::EqualityEngineNotify {
  66     TheoryStrings& d_str;
  67   public:
  68     NotifyClass(TheoryStrings& t_str): d_str(t_str) {}
  69     bool eqNotifyTriggerEquality(TNode equality, bool value) {
  70       Debug("strings") << "NotifyClass::eqNotifyTriggerEquality(" << equality << ", " << (value ? "true" : "false" )<< ")" << std::endl;
  71       if (value) {
  72         return d_str.propagate(equality);
  73       } else {
  74         // We use only literal triggers so taking not is safe
  75         return d_str.propagate(equality.notNode());
  76       }
  77     }
  78     bool eqNotifyTriggerPredicate(TNode predicate, bool value) {
  79       Debug("strings") << "NotifyClass::eqNotifyTriggerPredicate(" << predicate << ", " << (value ? "true" : "false") << ")" << std::endl;
  80       if (value) {
  81         return d_str.propagate(predicate);
  82       } else {
  83          return d_str.propagate(predicate.notNode());
  84       }
  85     }
  86     bool eqNotifyTriggerTermEquality(TheoryId tag, TNode t1, TNode t2, bool value) {
  87       Debug("strings") << "NotifyClass::eqNotifyTriggerTermMerge(" << tag << ", " << t1 << ", " << t2 << ")" << std::endl;
  88       if (value) {
  89       return d_str.propagate(t1.eqNode(t2));
  90       } else {
  91       return d_str.propagate(t1.eqNode(t2).notNode());
  92       }
  93     }
  94     void eqNotifyConstantTermMerge(TNode t1, TNode t2) {
  95       Debug("strings") << "NotifyClass::eqNotifyConstantTermMerge(" << t1 << ", " << t2 << ")" << std::endl;
  96       d_str.conflict(t1, t2);
  97     }
  98     void eqNotifyNewClass(TNode t) {
  99       Debug("strings") << "NotifyClass::eqNotifyNewClass(" << t << std::endl;
 100       d_str.eqNotifyNewClass(t);
 101     }
 102     void eqNotifyPreMerge(TNode t1, TNode t2) {
 103       Debug("strings") << "NotifyClass::eqNotifyPreMerge(" << t1 << ", " << t2 << std::endl;
 104       d_str.eqNotifyPreMerge(t1, t2);
 105     }
 106     void eqNotifyPostMerge(TNode t1, TNode t2) {
 107       Debug("strings") << "NotifyClass::eqNotifyPostMerge(" << t1 << ", " << t2 << std::endl;
 108       d_str.eqNotifyPostMerge(t1, t2);
 109     }
 110     void eqNotifyDisequal(TNode t1, TNode t2, TNode reason) {
 111       Debug("strings") << "NotifyClass::eqNotifyDisequal(" << t1 << ", " << t2 << ", " << reason << std::endl;
 112       d_str.eqNotifyDisequal(t1, t2, reason);
 113     }
 114   };/* class TheoryStrings::NotifyClass */
 115
 116 private:
 117   /**
 118    * Function symbol used to implement uninterpreted undefined string
 119    * semantics.  Needed to deal with partial charat/substr function.
 120    */
 121   Node d_ufSubstr;
 122
 123   // Constants
 124   Node d_emptyString;
 125   Node d_emptyRegexp;
 126   Node d_true;
 127   Node d_false;
 128   Node d_zero;
 129   Node d_one;
 130   CVC4::Rational RMAXINT;
 131   // Options
 132   bool d_opt_fmf;
 133   bool d_opt_regexp_gcd;
 134   // Helper functions
 135   Node getRepresentative( Node t );
 136   bool hasTerm( Node a );
 137   bool areEqual( Node a, Node b );
 138   bool areDisequal( Node a, Node b );
 139   Node getLengthTerm( Node t );
 140   Node getLength( Node t );
 141
 142 private:
 143   /** The notify class */
 144   NotifyClass d_notify;
 145   /** Equaltity engine */
 146   eq::EqualityEngine d_equalityEngine;
 147   /** Are we in conflict */
 148   context::CDO<bool> d_conflict;
 149   //list of pairs of nodes to merge
 150   std::map< Node, Node > d_pending_exp;
 151   std::vector< Node > d_pending;
 152   std::vector< Node > d_lemma_cache;
 153   std::map< Node, bool > d_pending_req_phase;
 154   /** inferences */
 155   NodeList d_infer;
 156   NodeList d_infer_exp;
 157   /** normal forms */
 158   std::map< Node, Node > d_normal_forms_base;
 159   std::map< Node, std::vector< Node > > d_normal_forms;
 160   std::map< Node, std::vector< Node > > d_normal_forms_exp;
 161   //map of pairs of terms that have the same normal form
 162   NodeListMap d_nf_pairs;
 163   void addNormalFormPair( Node n1, Node n2 );
 164   bool isNormalFormPair( Node n1, Node n2 );
 165   bool isNormalFormPair2( Node n1, Node n2 );
 166   // loop ant
 167   NodeSet d_loop_antec;
 168   NodeSet d_length_intro_vars;
 169   // preReg cache
 170   NodeSet d_registed_terms_cache;
 171   // term cache
 172   std::vector< Node > d_terms_cache;
 173   void collectTerm( Node n );
 174   void appendTermLemma();
 175
 176   /////////////////////////////////////////////////////////////////////////////
 177   // MODEL GENERATION
 178   /////////////////////////////////////////////////////////////////////////////
 179 public:
 180   void collectModelInfo(TheoryModel* m, bool fullModel);
 181
 182   /////////////////////////////////////////////////////////////////////////////
 183   // NOTIFICATIONS
 184   /////////////////////////////////////////////////////////////////////////////
 185 public:
 186   void presolve();
 187   void shutdown() { }
 188
 189   /////////////////////////////////////////////////////////////////////////////
 190   // MAIN SOLVER
 191   /////////////////////////////////////////////////////////////////////////////
 192 private:
 193   void addSharedTerm(TNode n);
 194   EqualityStatus getEqualityStatus(TNode a, TNode b);
 195
 196 private:
 197   class EqcInfo {
 198   public:
 199     EqcInfo( context::Context* c );
 200     ~EqcInfo(){}
 201     //constant in this eqc
 202     context::CDO< Node > d_const_term;
 203     context::CDO< Node > d_length_term;
 204     context::CDO< unsigned > d_cardinality_lem_k;
 205     // 1 = added length lemma
 206     context::CDO< Node > d_normalized_length;
 207   };
 208   /** map from representatives to information necessary for equivalence classes */
 209   std::map< Node, EqcInfo* > d_eqc_info;
 210   EqcInfo * getOrMakeEqcInfo( Node eqc, bool doMake = true );
 211   //maintain which concat terms have the length lemma instantiated
 212   NodeNodeMap d_length_inst;
 213 private:
 214   void mergeCstVec(std::vector< Node > &vec_strings);
 215   bool getNormalForms(Node &eqc, std::vector< Node > & visited, std::vector< Node > & nf,
 216         std::vector< std::vector< Node > > &normal_forms,
 217         std::vector< std::vector< Node > > &normal_forms_exp,
 218         std::vector< Node > &normal_form_src);
 219   bool detectLoop(std::vector< std::vector< Node > > &normal_forms,
 220         int i, int j, int index_i, int index_j,
 221         int &loop_in_i, int &loop_in_j);
 222   bool processLoop(std::vector< Node > &antec,
 223         std::vector< std::vector< Node > > &normal_forms,
 224         std::vector< Node > &normal_form_src,
 225         int i, int j, int loop_n_index, int other_n_index,
 226         int loop_index, int index, int other_index);
 227   bool processNEqc(std::vector< std::vector< Node > > &normal_forms,
 228         std::vector< std::vector< Node > > &normal_forms_exp,
 229         std::vector< Node > &normal_form_src);
 230   bool processReverseNEq(std::vector< std::vector< Node > > &normal_forms,
 231         std::vector< Node > &normal_form_src, std::vector< Node > &curr_exp, unsigned i, unsigned j );
 232   bool processSimpleNEq( std::vector< std::vector< Node > > &normal_forms,
 233         std::vector< Node > &normal_form_src, std::vector< Node > &curr_exp, unsigned i, unsigned j,
 234         unsigned& index_i, unsigned& index_j, bool isRev );
 235   bool normalizeEquivalenceClass( Node n, std::vector< Node > & visited, std::vector< Node > & nf, std::vector< Node > & nf_exp );
 236   bool processDeq( Node n1, Node n2 );
 237   int processReverseDeq( std::vector< Node >& nfi, std::vector< Node >& nfj, Node ni, Node nj );
 238   int processSimpleDeq( std::vector< Node >& nfi, std::vector< Node >& nfj, Node ni, Node nj, unsigned& index, bool isRev );
 239   //bool unrollStar( Node atom );
 240   Node mkRegExpAntec(Node atom, Node ant);
 241
 242   //bool checkSimple();
 243   bool checkNormalForms();
 244   void checkDeqNF();
 245   bool checkLengthsEqc();
 246   bool checkCardinality();
 247   bool checkInductiveEquations();
 248   //check membership constraints
 249   Node normalizeRegexp(Node r);
 250   bool normalizePosMemberships(std::map< Node, std::vector< Node > > &memb_with_exps);
 251   bool applyRConsume( CVC4::String &s, Node &r);
 252   Node applyRSplit(Node s1, Node s2, Node r);
 253   bool applyRLen(std::map< Node, std::vector< Node > > &XinR_with_exps);
 254   bool checkMembershipsWithoutLength(
 255     std::map< Node, std::vector< Node > > &memb_with_exps,
 256     std::map< Node, std::vector< Node > > &XinR_with_exps);
 257   bool checkMemberships();
 258   //temp
 259   bool checkMemberships2();
 260   bool checkPDerivative(Node x, Node r, Node atom, bool &addedLemma,
 261     std::vector< Node > &processed, std::vector< Node > &cprocessed,
 262     std::vector< Node > &nf_exp);
 263   bool checkContains();
 264   bool checkPosContains();
 265   bool checkNegContains();
 266
 267 public:
 268   void preRegisterTerm(TNode n);
 269   Node expandDefinition(LogicRequest &logicRequest, Node n);
 270   void check(Effort e);
 271
 272   /** Conflict when merging two constants */
 273   void conflict(TNode a, TNode b);
 274   /** called when a new equivalence class is created */
 275   void eqNotifyNewClass(TNode t);
 276   /** called when two equivalence classes will merge */
 277   void eqNotifyPreMerge(TNode t1, TNode t2);
 278   /** called when two equivalence classes have merged */
 279   void eqNotifyPostMerge(TNode t1, TNode t2);
 280   /** called when two equivalence classes are made disequal */
 281   void eqNotifyDisequal(TNode t1, TNode t2, TNode reason);
 282 protected:
 283   /** compute care graph */
 284   void computeCareGraph();
 285
 286   //do pending merges
 287   void assertPendingFact(Node fact, Node exp);
 288   void doPendingFacts();
 289   void doPendingLemmas();
 290
 291   //register term
 292   bool registerTerm( Node n );
 293   //send lemma
 294   void sendLemma( Node ant, Node conc, const char * c );
 295   void sendInfer( Node eq_exp, Node eq, const char * c );
 296   void sendSplit( Node a, Node b, const char * c, bool preq = true );
 297   void sendLengthLemma( Node n );
 298   /** mkConcat **/
 299   inline Node mkConcat( Node n1, Node n2 );
 300   inline Node mkConcat( Node n1, Node n2, Node n3 );
 301   inline Node mkConcat( const std::vector< Node >& c );
 302   //mkSkolem
 303   inline Node mkSkolemS(const char * c, int isLenSplit = 0);
 304   //inline Node mkSkolemI(const char * c);
 305   /** mkExplain **/
 306   Node mkExplain( std::vector< Node >& a );
 307   Node mkExplain( std::vector< Node >& a, std::vector< Node >& an );
 308   /** mkAnd **/
 309   Node mkAnd( std::vector< Node >& a );
 310   /** get concat vector */
 311   void getConcatVec( Node n, std::vector< Node >& c );
 312
 313   //get equivalence classes
 314   void getEquivalenceClasses( std::vector< Node >& eqcs );
 315   //get final normal form
 316   void getFinalNormalForm( Node n, std::vector< Node >& nf, std::vector< Node >& exp );
 317
 318   //separate into collections with equal length
 319   void separateByLength( std::vector< Node >& n, std::vector< std::vector< Node > >& col, std::vector< Node >& lts );
 320   void printConcat( std::vector< Node >& n, const char * c );
 321
 322 private:
 323   Node mkSplitEq( const char * c, const char * info, Node lhs, Node rhs, bool lgtZero );
 324
 325   // Special String Functions
 326   NodeList d_str_pos_ctn;
 327   NodeList d_str_neg_ctn;
 328   NodeSet d_neg_ctn_eqlen;
 329   NodeSet d_neg_ctn_ulen;
 330   NodeSet d_pos_ctn_cached;
 331   NodeSet d_neg_ctn_cached;
 332
 333   // Symbolic Regular Expression
 334 private:
 335   // regular expression memberships
 336   NodeList d_regexp_memberships;
 337   NodeSet d_regexp_ucached;
 338   NodeSet d_regexp_ccached;
 339   // stored assertions
 340   NodeListMap d_pos_memberships;
 341   NodeListMap d_neg_memberships;
 342   // semi normal forms for symbolic expression
 343   std::map< Node, Node > d_nf_regexps;
 344   std::map< Node, std::vector< Node > > d_nf_regexps_exp;
 345   // intersection
 346   NodeNodeMap d_inter_cache;
 347   NodeIntMap d_inter_index;
 348   // processed memberships
 349   NodeSet d_processed_memberships;
 350   // antecedant for why regexp membership must be true
 351   NodeNodeMap d_regexp_ant;
 352   // membership length
 353   //std::map< Node, bool > d_membership_length;
 354   // regular expression operations
 355   RegExpOpr d_regexp_opr;
 356
 357   CVC4::String getHeadConst( Node x );
 358   bool deriveRegExp( Node x, Node r, Node ant );
 359   bool addMembershipLength(Node atom);
 360   void addMembership(Node assertion);
 361   Node getNormalString(Node x, std::vector<Node> &nf_exp);
 362   Node getNormalSymRegExp(Node r, std::vector<Node> &nf_exp);
 363
 364
 365   // Finite Model Finding
 366 private:
 367   NodeSet d_input_vars;
 368   context::CDO< Node > d_input_var_lsum;
 369   context::CDHashMap< int, Node > d_cardinality_lits;
 370   context::CDO< int > d_curr_cardinality;
 371 public:
 372   //for finite model finding
 373   Node getNextDecisionRequest();
 374   void assertNode( Node lit );
 375
 376 public:
 377 /** statistics class */
 378   class Statistics {
 379   public:
 380     IntStat d_splits;
 381     IntStat d_eq_splits;
 382     IntStat d_deq_splits;
 383     IntStat d_loop_lemmas;
 384     IntStat d_new_skolems;
 385     Statistics();
 386     ~Statistics();
 387   };/* class TheoryStrings::Statistics */
 388   Statistics d_statistics;
 389 };/* class TheoryStrings */
 390
 391 }/* CVC4::theory::strings namespace */
 392 }/* CVC4::theory namespace */
 393 }/* CVC4 namespace */
 394
 395 #endif /* __CVC4__THEORY__STRINGS__THEORY_STRINGS_H */