summaryrefslogtreecommitdiff
path: root/src/theory
diff options
context:
space:
mode:
authorAndrew Reynolds <andrew.j.reynolds@gmail.com>2020-10-01 17:19:08 -0500
committerGitHub <noreply@github.com>2020-10-01 17:19:08 -0500
commit75d47fd5b65c3fe6dff3ef591d8331f737ca1cea (patch)
treef8b5d619f3cfe0b9d3f1e9e22c9209606b2c53e3 /src/theory
parent776ee02237b06eb3130e56af4d98d9ff36667d8b (diff)
(proof-new) Make arrays proof producing (#5112)
This includes adding a basic inference manager to arrays that ensures that the correct applications of PfRule are given to the inference manager. Note that some calls to lemma are yet to be converted. Also note that some minor simplifications were made for unnecessary parts of the code.
Diffstat (limited to 'src/theory')
-rw-r--r--src/theory/arrays/inference_manager.cpp128
-rw-r--r--src/theory/arrays/inference_manager.h77
-rw-r--r--src/theory/arrays/theory_arrays.cpp123
-rw-r--r--src/theory/arrays/theory_arrays.h6
4 files changed, 267 insertions, 67 deletions
diff --git a/src/theory/arrays/inference_manager.cpp b/src/theory/arrays/inference_manager.cpp
new file mode 100644
index 000000000..f429140be
--- /dev/null
+++ b/src/theory/arrays/inference_manager.cpp
@@ -0,0 +1,128 @@
+/********************* */
+/*! \file inference_manager.cpp
+ ** \verbatim
+ ** Top contributors (to current version):
+ ** Andrew Reynolds
+ ** This file is part of the CVC4 project.
+ ** Copyright (c) 2009-2020 by the authors listed in the file AUTHORS
+ ** in the top-level source directory) and their institutional affiliations.
+ ** All rights reserved. See the file COPYING in the top-level source
+ ** directory for licensing information.\endverbatim
+ **
+ ** \brief Arrays inference manager
+ **/
+
+#include "theory/arrays/inference_manager.h"
+
+#include "options/smt_options.h"
+#include "theory/theory.h"
+#include "theory/uf/equality_engine.h"
+
+using namespace CVC4::kind;
+
+namespace CVC4 {
+namespace theory {
+namespace arrays {
+
+InferenceManager::InferenceManager(Theory& t,
+ TheoryState& state,
+ ProofNodeManager* pnm)
+ : TheoryInferenceManager(t, state, pnm),
+ d_lemmaPg(pnm ? new EagerProofGenerator(pnm,
+ state.getUserContext(),
+ "ArrayLemmaProofGenerator")
+ : nullptr)
+{
+}
+
+bool InferenceManager::assertInference(TNode atom,
+ bool polarity,
+ TNode reason,
+ PfRule id)
+{
+ Trace("arrays-infer") << "TheoryArrays::assertInference: "
+ << (polarity ? Node(atom) : atom.notNode()) << " by "
+ << reason << "; " << id << std::endl;
+ Assert(atom.getKind() == EQUAL);
+ // if proofs are enabled, we determine which proof rule to add, otherwise
+ // we simply assert the internal fact
+ if (isProofEnabled())
+ {
+ Node fact = polarity ? Node(atom) : atom.notNode();
+ std::vector<Node> children;
+ std::vector<Node> args;
+ // convert to proof rule application
+ convert(id, fact, reason, children, args);
+ return assertInternalFact(atom, polarity, id, children, args);
+ }
+ return assertInternalFact(atom, polarity, reason);
+}
+
+bool InferenceManager::arrayLemma(
+ Node conc, Node exp, PfRule id, LemmaProperty p, bool doCache)
+{
+ Trace("arrays-infer") << "TheoryArrays::arrayLemma: " << conc << " by " << exp
+ << "; " << id << std::endl;
+ NodeManager* nm = NodeManager::currentNM();
+ if (isProofEnabled())
+ {
+ std::vector<Node> children;
+ std::vector<Node> args;
+ // convert to proof rule application
+ convert(id, conc, exp, children, args);
+ // make the trusted lemma based on the eager proof generator and send
+ TrustNode tlem = d_lemmaPg->mkTrustNode(conc, id, children, args);
+ return trustedLemma(tlem, p, doCache);
+ }
+ // send lemma without proofs
+ Node lem = nm->mkNode(IMPLIES, exp, conc);
+ return lemma(lem, p, doCache);
+}
+
+void InferenceManager::convert(PfRule& id,
+ Node conc,
+ Node exp,
+ std::vector<Node>& children,
+ std::vector<Node>& args)
+{
+ // note that children must contain something equivalent to exp,
+ // regardless of the PfRule.
+ switch (id)
+ {
+ case PfRule::MACRO_SR_PRED_INTRO:
+ Assert(exp.isConst());
+ args.push_back(conc);
+ break;
+ case PfRule::ARRAYS_READ_OVER_WRITE:
+ if (exp.isConst())
+ {
+ // Premise can be shown by rewriting, use standard predicate intro rule.
+ // This is the case where we have 2 constant indices.
+ id = PfRule::MACRO_SR_PRED_INTRO;
+ args.push_back(conc);
+ }
+ else
+ {
+ children.push_back(exp);
+ args.push_back(conc[0]);
+ }
+ break;
+ case PfRule::ARRAYS_READ_OVER_WRITE_CONTRA: children.push_back(exp); break;
+ case PfRule::ARRAYS_READ_OVER_WRITE_1:
+ Assert(exp.isConst());
+ args.push_back(conc[0]);
+ break;
+ case PfRule::ARRAYS_EXT: children.push_back(exp); break;
+ default:
+ // unknown rule, should never happen
+ Assert(false);
+ children.push_back(exp);
+ args.push_back(conc);
+ id = PfRule::ARRAYS_TRUST;
+ break;
+ }
+}
+
+} // namespace arrays
+} // namespace theory
+} // namespace CVC4
diff --git a/src/theory/arrays/inference_manager.h b/src/theory/arrays/inference_manager.h
new file mode 100644
index 000000000..2073fac6a
--- /dev/null
+++ b/src/theory/arrays/inference_manager.h
@@ -0,0 +1,77 @@
+/********************* */
+/*! \file inference_manager.h
+ ** \verbatim
+ ** Top contributors (to current version):
+ ** Andrew Reynolds
+ ** This file is part of the CVC4 project.
+ ** Copyright (c) 2009-2020 by the authors listed in the file AUTHORS
+ ** in the top-level source directory) and their institutional affiliations.
+ ** All rights reserved. See the file COPYING in the top-level source
+ ** directory for licensing information.\endverbatim
+ **
+ ** \brief Arrays inference manager
+ **/
+
+#include "cvc4_private.h"
+
+#ifndef CVC4__THEORY__ARRAYS__INFERENCE_MANAGER_H
+#define CVC4__THEORY__ARRAYS__INFERENCE_MANAGER_H
+
+#include "expr/node.h"
+#include "expr/proof_rule.h"
+#include "theory/eager_proof_generator.h"
+#include "theory/theory_inference_manager.h"
+
+namespace CVC4 {
+namespace theory {
+namespace arrays {
+
+/**
+ * The arrays inference manager.
+ */
+class InferenceManager : public TheoryInferenceManager
+{
+ public:
+ InferenceManager(Theory& t, TheoryState& state, ProofNodeManager* pnm);
+ ~InferenceManager() {}
+
+ /**
+ * Assert inference. This sends an internal fact to the equality engine
+ * immediately, possibly with proof support. The identifier id which
+ * rule to apply when proofs are enabled. The order of children
+ * and arguments to use in the proof step are determined internally in
+ * this method.
+ *
+ * @return true if the fact was successfully asserted, and false if the
+ * fact was redundant.
+ */
+ bool assertInference(TNode atom, bool polarity, TNode reason, PfRule id);
+ /**
+ * Send lemma (exp => conc) based on proof rule id with properties p. Cache
+ * the lemma if doCache is true.
+ */
+ bool arrayLemma(Node conc,
+ Node exp,
+ PfRule id,
+ LemmaProperty p = LemmaProperty::NONE,
+ bool doCache = false);
+
+ private:
+ /**
+ * Converts a conclusion, explanation and proof rule id used by the array
+ * theory to the set of arguments required for a proof rule application.
+ */
+ void convert(PfRule& id,
+ Node conc,
+ Node exp,
+ std::vector<Node>& children,
+ std::vector<Node>& args);
+ /** Eager proof generator for lemmas from the above method */
+ std::unique_ptr<EagerProofGenerator> d_lemmaPg;
+};
+
+} // namespace arrays
+} // namespace theory
+} // namespace CVC4
+
+#endif
diff --git a/src/theory/arrays/theory_arrays.cpp b/src/theory/arrays/theory_arrays.cpp
index c0a3af8f3..5dee75a6c 100644
--- a/src/theory/arrays/theory_arrays.cpp
+++ b/src/theory/arrays/theory_arrays.cpp
@@ -21,10 +21,12 @@
#include "expr/kind.h"
#include "expr/node_algorithm.h"
+#include "expr/proof_checker.h"
#include "options/arrays_options.h"
#include "options/smt_options.h"
#include "smt/logic_exception.h"
#include "smt/smt_statistics_registry.h"
+#include "theory/arrays/skolem_cache.h"
#include "theory/arrays/theory_arrays_rewriter.h"
#include "theory/rewriter.h"
#include "theory/theory_model.h"
@@ -78,6 +80,7 @@ TheoryArrays::TheoryArrays(context::Context* c,
d_ppEqualityEngine(u, name + "theory::arrays::pp", true),
d_ppFacts(u),
d_state(c, u, valuation),
+ d_im(*this, d_state, pnm),
d_literalsToPropagate(c),
d_literalsToPropagateIndex(c, 0),
d_isPreRegistered(c),
@@ -131,9 +134,10 @@ TheoryArrays::TheoryArrays(context::Context* c,
{
d_pchecker.registerTo(pc);
}
-
- // indicate we are using the default theory state object
+ // indicate we are using the default theory state object, and the arrays
+ // inference manager
d_theoryState = &d_state;
+ d_inferManager = &d_im;
}
TheoryArrays::~TheoryArrays() {
@@ -758,10 +762,9 @@ void TheoryArrays::preRegisterTermInternal(TNode node)
{
preRegisterTermInternal(ni);
}
-
// Apply RIntro1 Rule
- d_equalityEngine->assertEquality(
- ni.eqNode(v), true, d_true, theory::eq::MERGED_THROUGH_ROW1);
+ d_im.assertInference(
+ ni.eqNode(v), true, d_true, PfRule::ARRAYS_READ_OVER_WRITE_1);
d_infoMap.addStore(node, node);
d_infoMap.addInStore(a, node);
@@ -837,9 +840,7 @@ void TheoryArrays::explain(TNode literal, Node& explanation)
TrustNode TheoryArrays::explain(TNode literal)
{
- Node explanation;
- explain(literal, explanation);
- return TrustNode::mkTrustPropExp(literal, explanation, nullptr);
+ return d_im.explainLit(literal);
}
/////////////////////////////////////////////////////////////////////////////
@@ -1179,38 +1180,26 @@ void TheoryArrays::presolve()
// MAIN SOLVER
/////////////////////////////////////////////////////////////////////////////
-
-Node TheoryArrays::getSkolem(TNode ref, const string& name, const TypeNode& type, const string& comment, bool makeEqual)
+Node TheoryArrays::getSkolem(TNode ref)
{
+ // the call to SkolemCache::getExtIndexSkolem should be deterministic, but use
+ // cache anyways for now
Node skolem;
std::unordered_map<Node, Node, NodeHashFunction>::iterator it = d_skolemCache.find(ref);
if (it == d_skolemCache.end()) {
- NodeManager* nm = NodeManager::currentNM();
- skolem = nm->mkSkolem(name, type, comment);
+ Assert(ref.getKind() == kind::NOT && ref[0].getKind() == kind::EQUAL);
+ // make the skolem using the skolem cache utility
+ skolem = SkolemCache::getExtIndexSkolem(ref);
d_skolemCache[ref] = skolem;
}
else {
skolem = (*it).second;
- if (d_equalityEngine->hasTerm(ref) && d_equalityEngine->hasTerm(skolem)
- && d_equalityEngine->areEqual(ref, skolem))
- {
- makeEqual = false;
- }
}
Debug("pf::array") << "Pregistering a Skolem" << std::endl;
preRegisterTermInternal(skolem);
Debug("pf::array") << "Pregistering a Skolem DONE" << std::endl;
- if (makeEqual) {
- Node d = skolem.eqNode(ref);
- Debug("arrays-model-based") << "Asserting skolem equality " << d << endl;
- d_equalityEngine->assertEquality(d, true, d_true);
- Assert(!d_state.isInConflict());
- d_skolemAssertions.push_back(d);
- d_skolemIndex = d_skolemIndex + 1;
- }
-
Debug("pf::array") << "getSkolem DONE" << std::endl;
return skolem;
}
@@ -1294,7 +1283,8 @@ void TheoryArrays::postCheck(Effort level)
weakEquivBuildCond(r2[0], r[1], conjunctions);
lemma = mkAnd(conjunctions, true);
// LSH FIXME: which kind of arrays lemma is this
- Trace("arrays-lem") << "Arrays::addExtLemma " << lemma <<"\n";
+ Trace("arrays-lem")
+ << "Arrays::addExtLemma (weak-eq) " << lemma << "\n";
d_out->lemma(lemma, LemmaProperty::SEND_ATOMS);
d_readTableContext->pop();
Trace("arrays") << spaces(getSatContext()->getLevel()) << "Arrays::check(): done" << endl;
@@ -1325,7 +1315,7 @@ void TheoryArrays::postCheck(Effort level)
bool TheoryArrays::preNotifyFact(
TNode atom, bool pol, TNode fact, bool isPrereg, bool isInternal)
{
- if (!isPrereg)
+ if (!isInternal && !isPrereg)
{
if (atom.getKind() == kind::EQUAL)
{
@@ -1347,13 +1337,14 @@ bool TheoryArrays::preNotifyFact(
void TheoryArrays::notifyFact(TNode atom, bool pol, TNode fact, bool isInternal)
{
// if a disequality
- if (atom.getKind() == kind::EQUAL && !pol)
+ if (atom.getKind() == kind::EQUAL && !pol && !isInternal)
{
+ // Notice that this should be an external assertion, since we do not
+ // internally infer disequalities.
// Apply ArrDiseq Rule if diseq is between arrays
if (fact[0][0].getType().isArray() && !d_state.isInConflict())
{
NodeManager* nm = NodeManager::currentNM();
- TypeNode indexType = fact[0][0].getType()[0];
TNode k;
// k is the skolem for this disequality.
@@ -1361,12 +1352,7 @@ void TheoryArrays::notifyFact(TNode atom, bool pol, TNode fact, bool isInternal)
<< std::endl;
// If not in replay mode, generate a fresh skolem variable
- k = getSkolem(
- fact,
- "array_ext_index",
- indexType,
- "an extensional lemma index variable from the theory of arrays",
- false);
+ k = getSkolem(fact);
Node ak = nm->mkNode(kind::SELECT, fact[0][0], k);
Node bk = nm->mkNode(kind::SELECT, fact[0][1], k);
@@ -1377,19 +1363,17 @@ void TheoryArrays::notifyFact(TNode atom, bool pol, TNode fact, bool isInternal)
&& d_equalityEngine->hasTerm(bk))
{
// Propagate witness disequality - might produce a conflict
- d_permRef.push_back(lemma);
Debug("pf::array") << "Asserting to the equality engine:" << std::endl
<< "\teq = " << eq << std::endl
<< "\treason = " << fact << std::endl;
-
- d_equalityEngine->assertEquality(eq, false, fact);
+ d_im.assertInference(eq, false, fact, PfRule::ARRAYS_EXT);
++d_numProp;
}
// If this is the solution pass, generate the lemma. Otherwise, don't
// generate it - as this is the lemma that we're reproving...
Trace("arrays-lem") << "Arrays::addExtLemma " << lemma << "\n";
- d_out->lemma(lemma);
+ d_im.arrayLemma(eq.notNode(), fact, PfRule::ARRAYS_EXT);
++d_numExt;
}
else
@@ -1669,7 +1653,8 @@ void TheoryArrays::checkRowForIndex(TNode i, TNode a)
{
preRegisterTermInternal(selConst);
}
- d_equalityEngine->assertEquality(selConst.eqNode(defValue), true, d_true);
+ d_im.assertInference(
+ selConst.eqNode(defValue), true, d_true, PfRule::ARRAYS_TRUST);
}
const CTNodeList* stores = d_infoMap.getStores(a);
@@ -1805,8 +1790,8 @@ void TheoryArrays::propagate(RowLemmaType lem)
if (!bjExists) {
preRegisterTermInternal(bj);
}
- d_equalityEngine->assertEquality(
- aj_eq_bj, true, reason, theory::eq::MERGED_THROUGH_ROW);
+ d_im.assertInference(
+ aj_eq_bj, true, reason, PfRule::ARRAYS_READ_OVER_WRITE);
++d_numProp;
return;
}
@@ -1815,10 +1800,9 @@ void TheoryArrays::propagate(RowLemmaType lem)
Trace("arrays-lem") << spaces(getSatContext()->getLevel()) <<"Arrays::queueRowLemma: propagating i = j ("<<i<<", "<<j<<")\n";
Node reason =
(aj.isConst() && bj.isConst()) ? d_true : aj.eqNode(bj).notNode();
- Node i_eq_j = i.eqNode(j);
- d_permRef.push_back(reason);
- d_equalityEngine->assertEquality(
- i_eq_j, true, reason, theory::eq::MERGED_THROUGH_ROW);
+ Node j_eq_i = j.eqNode(i);
+ d_im.assertInference(
+ j_eq_i, true, reason, PfRule::ARRAYS_READ_OVER_WRITE_CONTRA);
++d_numProp;
return;
}
@@ -1884,7 +1868,8 @@ void TheoryArrays::queueRowLemma(RowLemmaType lem)
{
preRegisterTermInternal(aj2);
}
- d_equalityEngine->assertEquality(aj.eqNode(aj2), true, d_true);
+ d_im.assertInference(
+ aj.eqNode(aj2), true, d_true, PfRule::MACRO_SR_PRED_INTRO);
}
Node bj2 = Rewriter::rewrite(bj);
if (bj != bj2) {
@@ -1895,7 +1880,8 @@ void TheoryArrays::queueRowLemma(RowLemmaType lem)
{
preRegisterTermInternal(bj2);
}
- d_equalityEngine->assertEquality(bj.eqNode(bj2), true, d_true);
+ d_im.assertInference(
+ bj.eqNode(bj2), true, d_true, PfRule::MACRO_SR_PRED_INTRO);
}
if (aj2 == bj2) {
return;
@@ -1913,22 +1899,24 @@ void TheoryArrays::queueRowLemma(RowLemmaType lem)
{
preRegisterTermInternal(bj2);
}
- d_equalityEngine->assertEquality(eq1, true, d_true);
+ d_im.assertInference(eq1, true, d_true, PfRule::MACRO_SR_PRED_INTRO);
return;
}
Node eq2 = i.eqNode(j);
Node eq2_r = Rewriter::rewrite(eq2);
if (eq2_r == d_true) {
- d_equalityEngine->assertEquality(eq2, true, d_true);
+ d_im.assertInference(eq2, true, d_true, PfRule::MACRO_SR_PRED_INTRO);
return;
}
Node lemma = nm->mkNode(kind::OR, eq2_r, eq1_r);
- Trace("arrays-lem")<<"Arrays::addRowLemma adding "<<lemma<<"\n";
+ Trace("arrays-lem") << "Arrays::addRowLemma (1) adding " << lemma << "\n";
d_RowAlreadyAdded.insert(lem);
- d_out->lemma(lemma);
+ // use non-rewritten nodes
+ d_im.arrayLemma(
+ aj.eqNode(bj), eq2.notNode(), PfRule::ARRAYS_READ_OVER_WRITE);
++d_numRow;
}
else {
@@ -1997,7 +1985,8 @@ bool TheoryArrays::dischargeLemmas()
{
preRegisterTermInternal(aj2);
}
- d_equalityEngine->assertEquality(aj.eqNode(aj2), true, d_true);
+ d_im.assertInference(
+ aj.eqNode(aj2), true, d_true, PfRule::MACRO_SR_PRED_INTRO);
}
Node bj2 = Rewriter::rewrite(bj);
if (bj != bj2) {
@@ -2008,7 +1997,8 @@ bool TheoryArrays::dischargeLemmas()
{
preRegisterTermInternal(bj2);
}
- d_equalityEngine->assertEquality(bj.eqNode(bj2), true, d_true);
+ d_im.assertInference(
+ bj.eqNode(bj2), true, d_true, PfRule::MACRO_SR_PRED_INTRO);
}
if (aj2 == bj2) {
continue;
@@ -2026,22 +2016,24 @@ bool TheoryArrays::dischargeLemmas()
{
preRegisterTermInternal(bj2);
}
- d_equalityEngine->assertEquality(eq1, true, d_true);
+ d_im.assertInference(eq1, true, d_true, PfRule::MACRO_SR_PRED_INTRO);
continue;
}
Node eq2 = i.eqNode(j);
Node eq2_r = Rewriter::rewrite(eq2);
if (eq2_r == d_true) {
- d_equalityEngine->assertEquality(eq2, true, d_true);
+ d_im.assertInference(eq2, true, d_true, PfRule::MACRO_SR_PRED_INTRO);
continue;
}
Node lem = nm->mkNode(kind::OR, eq2_r, eq1_r);
- Trace("arrays-lem")<<"Arrays::addRowLemma adding "<<lem<<"\n";
+ Trace("arrays-lem") << "Arrays::addRowLemma (2) adding " << lem << "\n";
d_RowAlreadyAdded.insert(l);
- d_out->lemma(lem);
+ // use non-rewritten nodes, theory preprocessing will rewrite
+ d_im.arrayLemma(
+ aj.eqNode(bj), eq2.notNode(), PfRule::ARRAYS_READ_OVER_WRITE);
++d_numRow;
lemmasAdded = true;
if (options::arraysReduceSharing()) {
@@ -2053,14 +2045,13 @@ bool TheoryArrays::dischargeLemmas()
void TheoryArrays::conflict(TNode a, TNode b) {
Debug("pf::array") << "TheoryArrays::Conflict called" << std::endl;
-
- explain(a.eqNode(b), d_conflictNode);
-
- if (!d_inCheckModel) {
- d_out->conflict(d_conflictNode);
+ if (d_inCheckModel)
+ {
+ // if in check model, don't send the conflict
+ d_state.notifyInConflict();
+ return;
}
-
- d_state.notifyInConflict();
+ d_im.conflictEqConstantMerge(a, b);
}
TheoryArrays::TheoryArraysDecisionStrategy::TheoryArraysDecisionStrategy(
diff --git a/src/theory/arrays/theory_arrays.h b/src/theory/arrays/theory_arrays.h
index 8c15a50bf..0dd95795b 100644
--- a/src/theory/arrays/theory_arrays.h
+++ b/src/theory/arrays/theory_arrays.h
@@ -26,10 +26,12 @@
#include "context/cdhashset.h"
#include "context/cdqueue.h"
#include "theory/arrays/array_info.h"
+#include "theory/arrays/inference_manager.h"
#include "theory/arrays/proof_checker.h"
#include "theory/arrays/theory_arrays_rewriter.h"
#include "theory/theory.h"
#include "theory/uf/equality_engine.h"
+#include "theory/uf/proof_equality_engine.h"
#include "util/statistics_registry.h"
namespace CVC4 {
@@ -187,6 +189,8 @@ class TheoryArrays : public Theory {
TheoryArraysRewriter d_rewriter;
/** A (default) theory state object */
TheoryState d_state;
+ /** The arrays inference manager */
+ InferenceManager d_im;
public:
PPAssertStatus ppAssert(TNode in, SubstitutionMap& outSubstitutions) override;
@@ -428,7 +432,7 @@ class TheoryArrays : public Theory {
context::CDList<Node> d_arrayMerges;
std::vector<CTNodeList*> d_readBucketAllocations;
- Node getSkolem(TNode ref, const std::string& name, const TypeNode& type, const std::string& comment, bool makeEqual = true);
+ Node getSkolem(TNode ref);
Node mkAnd(std::vector<TNode>& conjunctions, bool invert = false, unsigned startIndex = 0);
void setNonLinear(TNode a);
Node removeRepLoops(TNode a, TNode rep);
generated by cgit on debian on lair
contact matthew@masot.net with questions or feedback