Implement proofs for arith BRAB lemmas (#5784)
[cvc5.git] / src / theory / theory.h
1 /********************* */
2 /*! \file theory.h
3 ** \verbatim
4 ** Top contributors (to current version):
5 ** Andrew Reynolds, Morgan Deters, Dejan Jovanovic
6 ** This file is part of the CVC4 project.
7 ** Copyright (c) 2009-2020 by the authors listed in the file AUTHORS
8 ** in the top-level source directory and their institutional affiliations.
9 ** All rights reserved. See the file COPYING in the top-level source
10 ** directory for licensing information.\endverbatim
11 **
12 ** \brief Base of the theory interface.
13 **
14 ** Base of the theory interface.
15 **/
16
17 #include "cvc4_private.h"
18
19 #ifndef CVC4__THEORY__THEORY_H
20 #define CVC4__THEORY__THEORY_H
21
22 #include <iosfwd>
23 #include <map>
24 #include <set>
25 #include <string>
26 #include <unordered_set>
27
28 #include "context/cdhashset.h"
29 #include "context/cdlist.h"
30 #include "context/cdo.h"
31 #include "context/context.h"
32 #include "expr/node.h"
33 #include "options/options.h"
34 #include "options/theory_options.h"
35 #include "smt/logic_request.h"
36 #include "theory/assertion.h"
37 #include "theory/care_graph.h"
38 #include "theory/decision_manager.h"
39 #include "theory/ee_setup_info.h"
40 #include "theory/logic_info.h"
41 #include "theory/output_channel.h"
42 #include "theory/theory_id.h"
43 #include "theory/theory_inference_manager.h"
44 #include "theory/theory_rewriter.h"
45 #include "theory/theory_state.h"
46 #include "theory/trust_node.h"
47 #include "theory/trust_substitutions.h"
48 #include "theory/valuation.h"
49 #include "util/statistics_registry.h"
50
51 namespace CVC4 {
52
53 class TheoryEngine;
54 class ProofNodeManager;
55
56 namespace theory {
57
58 class QuantifiersEngine;
59 class TheoryModel;
60 class SubstitutionMap;
61 class TheoryRewriter;
62
63 namespace rrinst {
64 class CandidateGenerator;
65 }/* CVC4::theory::rrinst namespace */
66
67 namespace eq {
68 class EqualityEngine;
69 }/* CVC4::theory::eq namespace */
70
71 /**
72 * Base class for T-solvers. Abstract DPLL(T).
73 *
74 * This is essentially an interface class. The TheoryEngine has
75 * pointers to Theory. Note that only one specific Theory type (e.g.,
76 * TheoryUF) can exist per NodeManager, because of how the
77 * RegisteredAttr works. (If you need multiple instances of the same
78 * theory, you'll have to write a multiplexed theory that dispatches
79 * all calls to them.)
80 *
81 * NOTE: A Theory has a special way of being initialized. The owner of a Theory
82 * is either:
83 *
84 * (A) Using Theory as a standalone object, not associated with a TheoryEngine.
85 * In this case, simply call the public initialization method
86 * (Theory::finishInitStandalone).
87 *
88 * (B) TheoryEngine, which determines how the Theory acts in accordance with
89 * its theory combination policy. We require the following steps in order:
90 * (B.1) Get information about whether the theory wishes to use an equality
91 * eninge, and more specifically which equality engine notifications the Theory
92 * would like to be notified of (Theory::needsEqualityEngine).
93 * (B.2) Set the equality engine of the theory (Theory::setEqualityEngine),
94 * which we refer to as the "official equality engine" of this Theory. The
95 * equality engine passed to the theory must respect the contract(s) specified
96 * by the equality engine setup information (EeSetupInfo) returned in the
97 * previous step.
98 * (B.3) Set the other required utilities including setQuantifiersEngine and
99 * setDecisionManager.
100 * (B.4) Call the private initialization method (Theory::finishInit).
101 *
102 * Initialization of the second form happens during TheoryEngine::finishInit,
103 * after the quantifiers engine and model objects have been set up.
104 */
105 class Theory {
106 friend class ::CVC4::TheoryEngine;
107
108 private:
109 // Disallow default construction, copy, assignment.
110 Theory() = delete;
111 Theory(const Theory&) = delete;
112 Theory& operator=(const Theory&) = delete;
113
114 /** An integer identifying the type of the theory. */
115 TheoryId d_id;
116
117 /** The SAT search context for the Theory. */
118 context::Context* d_satContext;
119
120 /** The user level assertion context for the Theory. */
121 context::UserContext* d_userContext;
122
123 /** Information about the logic we're operating within. */
124 const LogicInfo& d_logicInfo;
125
126 /**
127 * The assertFact() queue.
128 *
129 * These can not be TNodes as some atoms (such as equalities) are sent
130 * across theories without being stored in a global map.
131 */
132 context::CDList<Assertion> d_facts;
133
134 /** Index into the head of the facts list */
135 context::CDO<unsigned> d_factsHead;
136
137 /** Indices for splitting on the shared terms. */
138 context::CDO<unsigned> d_sharedTermsIndex;
139
140 /** The care graph the theory will use during combination. */
141 CareGraph* d_careGraph;
142
143 /**
144 * Pointer to the quantifiers engine (or NULL, if quantifiers are not
145 * supported or not enabled). Not owned by the theory.
146 */
147 QuantifiersEngine* d_quantEngine;
148
149 /** Pointer to the decision manager. */
150 DecisionManager* d_decManager;
151
152 protected:
153 /** Name of this theory instance. Along with the TheoryId this should provide
154 * an unique string identifier for each instance of a Theory class. We need
155 * this to ensure unique statistics names over multiple theory instances. */
156 std::string d_instanceName;
157
158 // === STATISTICS ===
159 /** time spent in check calls */
160 TimerStat d_checkTime;
161 /** time spent in theory combination */
162 TimerStat d_computeCareGraphTime;
163
164 /**
165 * The only method to add suff to the care graph.
166 */
167 void addCarePair(TNode t1, TNode t2);
168
169 /**
170 * The function should compute the care graph over the shared terms.
171 * The default function returns all the pairs among the shared variables.
172 */
173 virtual void computeCareGraph();
174
175 /**
176 * A list of shared terms that the theory has.
177 */
178 context::CDList<TNode> d_sharedTerms;
179
180 /**
181 * Construct a Theory.
182 *
183 * The pair <id, instance> is assumed to uniquely identify this Theory
184 * w.r.t. the SmtEngine.
185 */
186 Theory(TheoryId id,
187 context::Context* satContext,
188 context::UserContext* userContext,
189 OutputChannel& out,
190 Valuation valuation,
191 const LogicInfo& logicInfo,
192 ProofNodeManager* pnm,
193 std::string instance = ""); // taking : No default.
194
195 /**
196 * This is called at shutdown time by the TheoryEngine, just before
197 * destruction. It is important because there are destruction
198 * ordering issues between PropEngine and Theory (based on what
199 * hard-links to Nodes are outstanding). As the fact queue might be
200 * nonempty, we ensure here that it's clear. If you overload this,
201 * you must make an explicit call here to this->Theory::shutdown()
202 * too.
203 */
204 virtual void shutdown() { }
205
206 /**
207 * The output channel for the Theory.
208 */
209 OutputChannel* d_out;
210
211 /**
212 * The valuation proxy for the Theory to communicate back with the
213 * theory engine (and other theories).
214 */
215 Valuation d_valuation;
216 /**
217 * Pointer to the official equality engine of this theory, which is owned by
218 * the equality engine manager of TheoryEngine.
219 */
220 eq::EqualityEngine* d_equalityEngine;
221 /**
222 * The official equality engine, if we allocated it.
223 */
224 std::unique_ptr<eq::EqualityEngine> d_allocEqualityEngine;
225 /**
226 * The theory state, which contains contexts, valuation, and equality engine.
227 * Notice the theory is responsible for memory management of this class.
228 */
229 TheoryState* d_theoryState;
230 /**
231 * The theory inference manager. This is a wrapper around the equality
232 * engine and the output channel. It ensures that the output channel and
233 * the equality engine are used properly.
234 */
235 TheoryInferenceManager* d_inferManager;
236
237 /** Pointer to proof node manager */
238 ProofNodeManager* d_pnm;
239
240 /**
241 * Are proofs enabled?
242 *
243 * They are considered enabled if the ProofNodeManager is non-null.
244 */
245 bool proofsEnabled() const;
246
247 /**
248 * Returns the next assertion in the assertFact() queue.
249 *
250 * @return the next assertion in the assertFact() queue
251 */
252 inline Assertion get();
253
254 const LogicInfo& getLogicInfo() const {
255 return d_logicInfo;
256 }
257
258 /**
259 * Set separation logic heap. This is called when the location and data
260 * types for separation logic are determined. This should be called at
261 * most once, before solving.
262 *
263 * This currently should be overridden by the separation logic theory only.
264 */
265 virtual void declareSepHeap(TypeNode locT, TypeNode dataT) {}
266
267 /**
268 * The theory that owns the uninterpreted sort.
269 */
270 static TheoryId s_uninterpretedSortOwner;
271
272 void printFacts(std::ostream& os) const;
273 void debugPrintFacts() const;
274
275 /** is legal elimination
276 *
277 * Returns true if x -> val is a legal elimination of variable x. This is
278 * useful for ppAssert, when x = val is an entailed equality. This function
279 * determines whether indeed x can be eliminated from the problem via the
280 * substituion x -> val.
281 *
282 * The following criteria imply that x -> val is *not* a legal elimination:
283 * (1) If x is contained in val,
284 * (2) If the type of val is not a subtype of the type of x,
285 * (3) If val contains an operator that cannot be evaluated, and produceModels
286 * is true. For example, x -> sqrt(2) is not a legal elimination if we
287 * are producing models. This is because we care about the value of x, and
288 * its value must be computed (approximated) by the non-linear solver.
289 */
290 bool isLegalElimination(TNode x, TNode val);
291 //--------------------------------- private initialization
292 /**
293 * Called to set the official equality engine. This should be done by
294 * TheoryEngine only.
295 */
296 void setEqualityEngine(eq::EqualityEngine* ee);
297 /** Called to set the quantifiers engine. */
298 void setQuantifiersEngine(QuantifiersEngine* qe);
299 /** Called to set the decision manager. */
300 void setDecisionManager(DecisionManager* dm);
301 /**
302 * Finish theory initialization. At this point, options and the logic
303 * setting are final, the master equality engine and quantifiers
304 * engine (if any) are initialized, and the official equality engine of this
305 * theory has been assigned. This base class implementation
306 * does nothing. This should be called by TheoryEngine only.
307 */
308 virtual void finishInit() {}
309 //--------------------------------- end private initialization
310
311 /**
312 * This method is called to notify a theory that the node n should
313 * be considered a "shared term" by this theory. This does anything
314 * theory-specific concerning the fact that n is now marked as a shared
315 * term, which is done in addition to explicitly storing n as a shared
316 * term and adding it as a trigger term in the equality engine of this
317 * class (see addSharedTerm).
318 */
319 virtual void notifySharedTerm(TNode n);
320
321 public:
322 //--------------------------------- initialization
323 /**
324 * @return The theory rewriter associated with this theory.
325 */
326 virtual TheoryRewriter* getTheoryRewriter() = 0;
327 /**
328 * Returns true if this theory needs an equality engine for checking
329 * satisfiability.
330 *
331 * If this method returns true, then the equality engine manager will
332 * initialize its equality engine field via setEqualityEngine above during
333 * TheoryEngine::finishInit, prior to calling finishInit for this theory.
334 *
335 * Additionally, if this method returns true, then this method is required to
336 * update the argument esi with instructions for initializing and setting up
337 * notifications from its equality engine, which is commonly done with
338 * a notifications class (eq::EqualityEngineNotify).
339 */
340 virtual bool needsEqualityEngine(EeSetupInfo& esi);
341 /**
342 * Finish theory initialization, standalone version. This is used to
343 * initialize this class if it is not associated with a theory engine.
344 * This allocates the official equality engine of this Theory and then
345 * calls the finishInit method above.
346 */
347 void finishInitStandalone();
348 //--------------------------------- end initialization
349
350 /**
351 * Return the ID of the theory responsible for the given type.
352 */
353 static inline TheoryId theoryOf(TypeNode typeNode) {
354 Trace("theory::internal") << "theoryOf(" << typeNode << ")" << std::endl;
355 TheoryId id;
356 if (typeNode.getKind() == kind::TYPE_CONSTANT) {
357 id = typeConstantToTheoryId(typeNode.getConst<TypeConstant>());
358 } else {
359 id = kindToTheoryId(typeNode.getKind());
360 }
361 if (id == THEORY_BUILTIN) {
362 Trace("theory::internal") << "theoryOf(" << typeNode << ") == " << s_uninterpretedSortOwner << std::endl;
363 return s_uninterpretedSortOwner;
364 }
365 return id;
366 }
367
368 /**
369 * Returns the ID of the theory responsible for the given node.
370 */
371 static TheoryId theoryOf(options::TheoryOfMode mode, TNode node);
372
373 /**
374 * Returns the ID of the theory responsible for the given node.
375 */
376 static inline TheoryId theoryOf(TNode node) {
377 return theoryOf(options::theoryOfMode(), node);
378 }
379
380 /**
381 * Set the owner of the uninterpreted sort.
382 */
383 static void setUninterpretedSortOwner(TheoryId theory) {
384 s_uninterpretedSortOwner = theory;
385 }
386
387 /**
388 * Get the owner of the uninterpreted sort.
389 */
390 static TheoryId getUninterpretedSortOwner() {
391 return s_uninterpretedSortOwner;
392 }
393
394 /**
395 * Checks if the node is a leaf node of this theory
396 */
397 inline bool isLeaf(TNode node) const {
398 return node.getNumChildren() == 0 || theoryOf(node) != d_id;
399 }
400
401 /**
402 * Checks if the node is a leaf node of a theory.
403 */
404 inline static bool isLeafOf(TNode node, TheoryId theoryId) {
405 return node.getNumChildren() == 0 || theoryOf(node) != theoryId;
406 }
407
408 /** Returns true if the assertFact queue is empty*/
409 bool done() const { return d_factsHead == d_facts.size(); }
410 /**
411 * Destructs a Theory.
412 */
413 virtual ~Theory();
414
415 /**
416 * Subclasses of Theory may add additional efforts. DO NOT CHECK
417 * equality with one of these values (e.g. if STANDARD xxx) but
418 * rather use range checks (or use the helper functions below).
419 * Normally we call QUICK_CHECK or STANDARD; at the leaves we call
420 * with FULL_EFFORT.
421 */
422 enum Effort
423 {
424 /**
425 * Standard effort where theory need not do anything
426 */
427 EFFORT_STANDARD = 50,
428 /**
429 * Full effort requires the theory make sure its assertions are satisfiable
430 * or not
431 */
432 EFFORT_FULL = 100,
433 /**
434 * Last call effort, called after theory combination has completed with
435 * no lemmas and a model is available.
436 */
437 EFFORT_LAST_CALL = 200
438 }; /* enum Effort */
439
440 static inline bool standardEffortOrMore(Effort e) CVC4_CONST_FUNCTION
441 { return e >= EFFORT_STANDARD; }
442 static inline bool standardEffortOnly(Effort e) CVC4_CONST_FUNCTION
443 { return e >= EFFORT_STANDARD && e < EFFORT_FULL; }
444 static inline bool fullEffort(Effort e) CVC4_CONST_FUNCTION
445 { return e == EFFORT_FULL; }
446
447 /**
448 * Get the id for this Theory.
449 */
450 TheoryId getId() const {
451 return d_id;
452 }
453
454 /**
455 * Get the SAT context associated to this Theory.
456 */
457 context::Context* getSatContext() const {
458 return d_satContext;
459 }
460
461 /**
462 * Get the context associated to this Theory.
463 */
464 context::UserContext* getUserContext() const {
465 return d_userContext;
466 }
467
468 /**
469 * Set the output channel associated to this theory.
470 */
471 void setOutputChannel(OutputChannel& out) {
472 d_out = &out;
473 }
474
475 /**
476 * Get the output channel associated to this theory.
477 */
478 OutputChannel& getOutputChannel() {
479 return *d_out;
480 }
481
482 /**
483 * Get the valuation associated to this theory.
484 */
485 Valuation& getValuation() {
486 return d_valuation;
487 }
488
489 /** Get the equality engine being used by this theory. */
490 eq::EqualityEngine* getEqualityEngine();
491
492 /**
493 * Get the quantifiers engine associated to this theory.
494 */
495 QuantifiersEngine* getQuantifiersEngine() {
496 return d_quantEngine;
497 }
498
499 /** Get the decision manager associated to this theory. */
500 DecisionManager* getDecisionManager() { return d_decManager; }
501
502 /**
503 * @return The theory state associated with this theory.
504 */
505 TheoryState* getTheoryState() { return d_theoryState; }
506
507 /**
508 * @return The theory inference manager associated with this theory.
509 */
510 TheoryInferenceManager* getInferenceManager() { return d_inferManager; }
511
512 /**
513 * Expand definitions in the term node. This returns a term that is
514 * equivalent to node. It wraps this term in a TrustNode of kind
515 * TrustNodeKind::REWRITE. If node is unchanged by this method, the
516 * null TrustNode may be returned. This is an optimization to avoid
517 * constructing the trivial equality (= node node) internally within
518 * TrustNode.
519 *
520 * The purpose of this method is typically to eliminate the operators in node
521 * that are syntax sugar that cannot otherwise be eliminated during rewriting.
522 * For example, division relies on the introduction of an uninterpreted
523 * function for the divide-by-zero case, which we do not introduce with
524 * the rewriter, since this function may be cached in a non-global fashion.
525 *
526 * Some theories have kinds that are effectively definitions and should be
527 * expanded before they are handled. Definitions allow a much wider range of
528 * actions than the normal forms given by the rewriter. However no
529 * assumptions can be made about subterms having been expanded or rewritten.
530 * Where possible rewrite rules should be used, definitions should only be
531 * used when rewrites are not possible, for example in handling
532 * under-specified operations using partially defined functions.
533 *
534 * Some theories like sets use expandDefinition as a "context
535 * independent preRegisterTerm". This is required for cases where
536 * a theory wants to be notified about a term before preprocessing
537 * and simplification but doesn't necessarily want to rewrite it.
538 */
539 virtual TrustNode expandDefinition(Node node)
540 {
541 // by default, do nothing
542 return TrustNode::null();
543 }
544
545 /**
546 * Pre-register a term. Done one time for a Node per SAT context level.
547 */
548 virtual void preRegisterTerm(TNode);
549
550 /**
551 * Assert a fact in the current context.
552 */
553 void assertFact(TNode assertion, bool isPreregistered) {
554 Trace("theory") << "Theory<" << getId() << ">::assertFact["
555 << d_satContext->getLevel() << "](" << assertion << ", "
556 << (isPreregistered ? "true" : "false") << ")" << std::endl;
557 d_facts.push_back(Assertion(assertion, isPreregistered));
558 }
559
560 /** Add shared term to the theory. */
561 void addSharedTerm(TNode node);
562
563 /**
564 * Return the current theory care graph. Theories should overload
565 * computeCareGraph to do the actual computation, and use addCarePair to add
566 * pairs to the care graph.
567 */
568 void getCareGraph(CareGraph* careGraph);
569
570 /**
571 * Return the status of two terms in the current context. Should be
572 * implemented in sub-theories to enable more efficient theory-combination.
573 */
574 virtual EqualityStatus getEqualityStatus(TNode a, TNode b);
575
576 /**
577 * Return the model value of the give shared term (or null if not available).
578 *
579 * TODO (project #39): this method is likely to become deprecated.
580 */
581 virtual Node getModelValue(TNode var) { return Node::null(); }
582
583 /** T-propagate new literal assignments in the current context. */
584 virtual void propagate(Effort level = EFFORT_FULL) {}
585
586 /**
587 * Return an explanation for the literal represented by parameter n
588 * (which was previously propagated by this theory).
589 */
590 virtual TrustNode explain(TNode n)
591 {
592 Unimplemented() << "Theory " << identify()
593 << " propagated a node but doesn't implement the "
594 "Theory::explain() interface!";
595 return TrustNode::null();
596 }
597
598 //--------------------------------- check
599 /**
600 * Does this theory wish to be called to check at last call effort? This is
601 * the case for any theory that wishes to run when a model is available.
602 */
603 virtual bool needsCheckLastEffort() { return false; }
604 /**
605 * Check the current assignment's consistency.
606 *
607 * An implementation of check() is required to either:
608 * - return a conflict on the output channel,
609 * - be interrupted,
610 * - throw an exception
611 * - or call get() until done() is true.
612 *
613 * The standard method for check consists of a loop that processes the entire
614 * fact queue when preCheck returns false. It makes four theory-specific
615 * callbacks, (preCheck, postCheck, preNotifyFact, notifyFact) as described
616 * below. It asserts each fact to the official equality engine when
617 * preNotifyFact returns false.
618 *
619 * Theories that use this check method must use an official theory
620 * state object (d_theoryState).
621 */
622 void check(Effort level = EFFORT_FULL);
623 /**
624 * Pre-check, called before the fact queue of the theory is processed.
625 * If this method returns false, then the theory will process its fact
626 * queue. If this method returns true, then the theory has indicated
627 * its check method should finish immediately.
628 */
629 virtual bool preCheck(Effort level = EFFORT_FULL);
630 /**
631 * Post-check, called after the fact queue of the theory is processed.
632 */
633 virtual void postCheck(Effort level = EFFORT_FULL);
634 /**
635 * Pre-notify fact, return true if the theory processed it. If this
636 * method returns false, then the atom will be added to the equality engine
637 * of the theory and notifyFact will be called with isInternal=false.
638 *
639 * Theories that implement check but do not use official equality
640 * engines should always return true for this method.
641 *
642 * @param atom The atom
643 * @param polarity Its polarity
644 * @param fact The original literal that was asserted
645 * @param isPrereg Whether the assertion is preregistered
646 * @param isInternal Whether the origin of the fact was internal. If this
647 * is false, the fact was asserted via the fact queue of the theory.
648 * @return true if the theory completely processed this fact, i.e. it does
649 * not need to assert the fact to its equality engine.
650 */
651 virtual bool preNotifyFact(
652 TNode atom, bool pol, TNode fact, bool isPrereg, bool isInternal);
653 /**
654 * Notify fact, called immediately after the fact was pushed into the
655 * equality engine.
656 *
657 * @param atom The atom
658 * @param polarity Its polarity
659 * @param fact The original literal that was asserted.
660 * @param isInternal Whether the origin of the fact was internal. If this
661 * is false, the fact was asserted via the fact queue of the theory.
662 */
663 virtual void notifyFact(TNode atom, bool pol, TNode fact, bool isInternal);
664 //--------------------------------- end check
665
666 //--------------------------------- collect model info
667 /**
668 * Get all relevant information in this theory regarding the current
669 * model. This should be called after a call to check( FULL_EFFORT )
670 * for all theories with no conflicts and no lemmas added.
671 *
672 * This method returns true if and only if the equality engine of m is
673 * consistent as a result of this call.
674 *
675 * The standard method for collectModelInfo computes the relevant terms,
676 * asserts the theory's equality engine to the model (if necessary) and
677 * then calls computeModelValues.
678 *
679 * TODO (project #39): this method should be non-virtual, once all theories
680 * conform to the new standard, delete, move to model manager distributed.
681 */
682 virtual bool collectModelInfo(TheoryModel* m, const std::set<Node>& termSet);
683 /**
684 * Compute terms that are not necessarily part of the assertions or
685 * shared terms that should be considered relevant, add them to termSet.
686 */
687 virtual void computeRelevantTerms(std::set<Node>& termSet);
688 /**
689 * Collect model values, after equality information is added to the model.
690 * The argument termSet is the set of relevant terms returned by
691 * computeRelevantTerms.
692 */
693 virtual bool collectModelValues(TheoryModel* m,
694 const std::set<Node>& termSet);
695 /** if theories want to do something with model after building, do it here */
696 virtual void postProcessModel( TheoryModel* m ){ }
697 //--------------------------------- end collect model info
698
699 //--------------------------------- preprocessing
700 /**
701 * Statically learn from assertion "in," which has been asserted
702 * true at the top level. The theory should only add (via
703 * ::operator<< or ::append()) to the "learned" builder---it should
704 * *never* clear it. It is a conjunction to add to the formula at
705 * the top-level and may contain other theories' contributions.
706 */
707 virtual void ppStaticLearn(TNode in, NodeBuilder<>& learned) { }
708
709 enum PPAssertStatus {
710 /** Atom has been solved */
711 PP_ASSERT_STATUS_SOLVED,
712 /** Atom has not been solved */
713 PP_ASSERT_STATUS_UNSOLVED,
714 /** Atom is inconsistent */
715 PP_ASSERT_STATUS_CONFLICT
716 };
717
718 /**
719 * Given a literal and its proof generator (encapsulated by trust node tin),
720 * add the solved substitutions to the map, if any. The method should return
721 * true if the literal can be safely removed from the input problem.
722 *
723 * Note that tin has trude node kind LEMMA. Its proof generator should be
724 * take into account when adding a substitution to outSubstitutions when
725 * proofs are enabled.
726 */
727 virtual PPAssertStatus ppAssert(TrustNode tin,
728 TrustSubstitutionMap& outSubstitutions);
729
730 /**
731 * Given a term of the theory coming from the input formula or
732 * from a lemma generated during solving, this method can be overridden in a
733 * theory implementation to rewrite the term into an equivalent form.
734 *
735 * This method returns a TrustNode of kind TrustNodeKind::REWRITE, which
736 * carries information about the proof generator for the rewrite, which can
737 * be the null TrustNode if n is unchanged.
738 *
739 * Notice this method is used both in the "theory rewrite equalities"
740 * preprocessing pass, where n is an equality from the input formula,
741 * and in theory preprocessing, where n is a (non-equality) term occurring
742 * in the input or generated in a lemma.
743 */
744 virtual TrustNode ppRewrite(TNode n) { return TrustNode::null(); }
745
746 /**
747 * Notify preprocessed assertions. Called on new assertions after
748 * preprocessing before they are asserted to theory engine.
749 */
750 virtual void ppNotifyAssertions(const std::vector<Node>& assertions) {}
751 //--------------------------------- end preprocessing
752
753 /**
754 * A Theory is called with presolve exactly one time per user
755 * check-sat. presolve() is called after preregistration,
756 * rewriting, and Boolean propagation, (other theories'
757 * propagation?), but the notified Theory has not yet had its
758 * check() or propagate() method called. A Theory may empty its
759 * assertFact() queue using get(). A Theory can raise conflicts,
760 * add lemmas, and propagate literals during presolve().
761 *
762 * NOTE: The presolve property must be added to the kinds file for
763 * the theory.
764 */
765 virtual void presolve() { }
766
767 /**
768 * A Theory is called with postsolve exactly one time per user
769 * check-sat. postsolve() is called after the query has completed
770 * (regardless of whether sat, unsat, or unknown), and after any
771 * model-querying related to the query has been performed.
772 * After this call, the theory will not get another check() or
773 * propagate() call until presolve() is called again. A Theory
774 * cannot raise conflicts, add lemmas, or propagate literals during
775 * postsolve().
776 */
777 virtual void postsolve() { }
778
779 /**
780 * Notification sent to the theory wheneven the search restarts.
781 * Serves as a good time to do some clean-up work, and you can
782 * assume you're at DL 0 for the purposes of Contexts. This function
783 * should not use the output channel.
784 */
785 virtual void notifyRestart() { }
786
787 /**
788 * Identify this theory (for debugging, dynamic configuration,
789 * etc..)
790 */
791 virtual std::string identify() const = 0;
792
793 /** Set user attribute
794 * This function is called when an attribute is set by a user. In SMT-LIBv2 this is done
795 * via the syntax (! n :attr)
796 */
797 virtual void setUserAttribute(const std::string& attr, Node n, std::vector<Node> node_values, std::string str_value) {
798 Unimplemented() << "Theory " << identify()
799 << " doesn't support Theory::setUserAttribute interface";
800 }
801
802 typedef context::CDList<Assertion>::const_iterator assertions_iterator;
803
804 /**
805 * Provides access to the facts queue, primarily intended for theory
806 * debugging purposes.
807 *
808 * @return the iterator to the beginning of the fact queue
809 */
810 assertions_iterator facts_begin() const {
811 return d_facts.begin();
812 }
813
814 /**
815 * Provides access to the facts queue, primarily intended for theory
816 * debugging purposes.
817 *
818 * @return the iterator to the end of the fact queue
819 */
820 assertions_iterator facts_end() const {
821 return d_facts.end();
822 }
823 /**
824 * Whether facts have been asserted to this theory.
825 *
826 * @return true iff facts have been asserted to this theory.
827 */
828 bool hasFacts() { return !d_facts.empty(); }
829
830 /** Return total number of facts asserted to this theory */
831 size_t numAssertions() {
832 return d_facts.size();
833 }
834
835 typedef context::CDList<TNode>::const_iterator shared_terms_iterator;
836
837 /**
838 * Provides access to the shared terms, primarily intended for theory
839 * debugging purposes.
840 *
841 * @return the iterator to the beginning of the shared terms list
842 */
843 shared_terms_iterator shared_terms_begin() const {
844 return d_sharedTerms.begin();
845 }
846
847 /**
848 * Provides access to the facts queue, primarily intended for theory
849 * debugging purposes.
850 *
851 * @return the iterator to the end of the shared terms list
852 */
853 shared_terms_iterator shared_terms_end() const {
854 return d_sharedTerms.end();
855 }
856
857
858 /**
859 * This is a utility function for constructing a copy of the currently shared terms
860 * in a queriable form. As this is
861 */
862 std::unordered_set<TNode, TNodeHashFunction> currentlySharedTerms() const;
863
864 /**
865 * This allows the theory to be queried for whether a literal, lit, is
866 * entailed by the theory. This returns a pair of a Boolean and a node E.
867 *
868 * If the Boolean is true, then E is a formula that entails lit and E is propositionally
869 * entailed by the assertions to the theory.
870 *
871 * If the Boolean is false, it is "unknown" if lit is entailed and E may be
872 * any node.
873 *
874 * The literal lit is either an atom a or (not a), which must belong to the theory:
875 * There is some TheoryOfMode m s.t. Theory::theoryOf(m, a) == this->getId().
876 *
877 * There are NO assumptions that a or the subterms of a have been
878 * preprocessed in any form. This includes ppRewrite, rewriting,
879 * preregistering, registering, definition expansion or ITE removal!
880 *
881 * Theories are free to limit the amount of effort they use and so may
882 * always opt to return "unknown". Both "unknown" and "not entailed",
883 * may return for E a non-boolean Node (e.g. Node::null()). (There is no explicit output
884 * for the negation of lit is entailed.)
885 *
886 * If lit is theory valid, the return result may be the Boolean constant
887 * true for E.
888 *
889 * If lit is entailed by multiple assertions on the theory's getFact()
890 * queue, a_1, a_2, ... and a_k, this may return E=(and a_1 a_2 ... a_k) or
891 * another theory entailed explanation E=(and (and a_1 a_2) (and a3 a_4) ... a_k)
892 *
893 * If lit is entailed by a single assertion on the theory's getFact()
894 * queue, say a, this may return E=a.
895 *
896 * The theory may always return false!
897 *
898 * Theories may not touch their output stream during an entailment check.
899 *
900 * @param lit a literal belonging to the theory.
901 * @return a pair <b,E> s.t. if b is true, then a formula E such that
902 * E |= lit in the theory.
903 */
904 virtual std::pair<bool, Node> entailmentCheck(TNode lit);
905 };/* class Theory */
906
907 std::ostream& operator<<(std::ostream& os, theory::Theory::Effort level);
908
909
910 inline theory::Assertion Theory::get() {
911 Assert(!done()) << "Theory::get() called with assertion queue empty!";
912
913 // Get the assertion
914 Assertion fact = d_facts[d_factsHead];
915 d_factsHead = d_factsHead + 1;
916
917 Trace("theory") << "Theory::get() => " << fact << " (" << d_facts.size() - d_factsHead << " left)" << std::endl;
918
919 return fact;
920 }
921
922 inline std::ostream& operator<<(std::ostream& out,
923 const CVC4::theory::Theory& theory) {
924 return out << theory.identify();
925 }
926
927 inline std::ostream& operator << (std::ostream& out, theory::Theory::PPAssertStatus status) {
928 switch (status) {
929 case theory::Theory::PP_ASSERT_STATUS_SOLVED:
930 out << "SOLVE_STATUS_SOLVED"; break;
931 case theory::Theory::PP_ASSERT_STATUS_UNSOLVED:
932 out << "SOLVE_STATUS_UNSOLVED"; break;
933 case theory::Theory::PP_ASSERT_STATUS_CONFLICT:
934 out << "SOLVE_STATUS_CONFLICT"; break;
935 default:
936 Unhandled();
937 }
938 return out;
939 }
940
941 }/* CVC4::theory namespace */
942 }/* CVC4 namespace */
943
944 #endif /* CVC4__THEORY__THEORY_H */