mirrors/z3
3
0
Fork 0
mirror of https://github.com/Z3Prover/z3 synced 2025-04-08 18:31:49 +00:00
Code Activity

add tc and trc functionals for binary relations

Browse source 
Signed-off-by: Nikolaj Bjorner <nbjorner@microsoft.com>
This commit is contained in:
Nikolaj Bjorner 2019-04-10 00:57:50 +02:00
parent 182039eb44
commit 9c9cd5ebf7
3 changed files with 71 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
src/ast/special_relations_decl_plugin.cpp
View file

@ -45,6 +45,9 @@ func_decl * special_relations_decl_plugin::mk_func_decl(
if (!range) {
range = m_manager->mk_bool_sort();
}
if (!m_manager->is_bool(range) && k != OP_SPECIAL_RELATION_NEXT) {
m_manager->raise_exception("range type is expected to be Boolean for special relations");
}
func_decl_info info(m_family_id, k, num_parameters, parameters);
symbol name;
switch(k) {
@ -67,6 +70,8 @@ void special_relations_decl_plugin::get_op_names(svector<builtin_name> & op_name
op_names.push_back(builtin_name(m_to.bare_str(), OP_SPECIAL_RELATION_TO));
op_names.push_back(builtin_name(m_tc.bare_str(), OP_SPECIAL_RELATION_TC));
op_names.push_back(builtin_name(m_trc.bare_str(), OP_SPECIAL_RELATION_TRC));
// next is an internal skolem function used for unfolding a relation R to satisfy a relation that
// is asserted for the transitive closure of R.
}
}

74
src/smt/theory_special_relations.cpp
View file

@ -97,9 +97,41 @@ namespace smt {
return alloc(theory_special_relations, new_ctx->get_manager());
}
/**
\brief for term := next(next(a,b),c) for relation f
assert f(term,c) or term != c
assert f(term,c) or term != next(a,b)
assert f(term,c) or term != b
assert f(term,c) or term != a
*/
bool theory_special_relations::internalize_term(app * term) {
if (m_util.is_next(term)) {
return false;
}
mk_var(term);
context& ctx = get_context();
ast_manager& m = get_manager();
func_decl* f = to_func_decl(term->get_decl()->get_parameter(0).get_ast());
expr* src = term->get_arg(0);
expr* dst = term->get_arg(1);
expr_ref f_rel(m.mk_app(f, src, dst), m);
literal f_lit = ctx.get_literal(f_rel);
src = term;
while (m_util.is_next(src)) {
dst = to_app(src)->get_arg(1);
src = to_app(src)->get_arg(0);
ctx.mk_th_axiom(get_id(), f_lit, ~mk_eq(term, src, false));
ctx.mk_th_axiom(get_id(), f_lit, ~mk_eq(term, dst, false));
}
return true;
}
bool theory_special_relations::internalize_atom(app * atm, bool gate_ctx) {
SASSERT(m_util.is_special_relation(atm));
relation* r = 0;
if (m_util.is_next(atm)) {
return internalize_term(atm);
}
ast_manager& m = get_manager();
if (!m_relations.find(atm->get_decl(), r)) {
r = alloc(relation, m_util.get_property(atm), atm->get_decl());
@ -267,28 +299,27 @@ namespace smt {
// Ensure that TC(R)xy -> Rxz1 Rz1z2 .. Rzky
//
// if not Rxy and no path in graph:
// Find z reachable from x and u that an creach y, such that not Ruz is not forced
// introduce Ruy
// If no such element for infinite domains, then introduce z
// Rxz & Rzy
// For finite domains extract conflict.
// TBD:
// - infinite/large domain version is naive.
// It could be made fair by finding cut-points between x and y
// quantified axioms could prevent
// Introduce next(x,y), such that
// TC(R)(x,y) => R(x,y) or TR(R)(next(x,y),y) & R(x,next(x,y))
//
// next(x,y) is fresh unless R(x,y) is true:
// R(x,y) or x != next(x,y)
// R(x,y) or y != next(x,y)
//
unsigned sz = r.m_asserted_atoms.size();
for (unsigned i = 0; i < sz; ++i) {
atom& a = *r.m_asserted_atoms[i];
if (a.phase()) {
bool_var bv = a.var();
TRACE("special_relations", tout << bv << " is positive\n";);
expr* arg1 = get_expr(a.v1());
expr* arg2 = get_expr(a.v2());
// we need reachability in the R graph not R* graph
theory_var r1 = get_representative(a.v1());
theory_var r2 = get_representative(a.v2());
if (r.m_property == sr_trc && r1 == r2) {
continue;
}
if (r_graph.can_reach(r1, r2)) {
TRACE("special_relations",
tout << a.v1() << ": " << mk_pp(arg1, m) << " -> "
@ -307,7 +338,18 @@ namespace smt {
// whenever f(n1, n2) is asserted.
break;
case l_false: {
//
// Add the axioms:
// TC(R)(x,y) => R(x,y) or TC(R)(next(x,y),y)
// TC(R)(x,y) => R(x,y) or R(x,next(x,y))
// R(x,y) or next(x,y) != x
// R(x,y) or next(x,y) != y,
// and recursively on all next subterms of x.
// Add the literal R(next(x,y),y) - set case split preference to true.
//
// TBD: perhaps replace by recursion unfolding similar to theory_rec_fun
//
expr_ref next(m.mk_app(m_util.mk_next(f), arg1, arg2), m);
expr_ref a2next(m.mk_app(f, arg1, next), m);
expr_ref next2b(m.mk_app(tcf, next, arg2), m);
@ -1072,6 +1114,10 @@ namespace smt {
}
}
bool theory_special_relations::has_quantifiers() {
return get_context().has_quantifiers();
}
void theory_special_relations::init_model(model_generator & m) {
for (auto const& kv : m_relations) {
switch (kv.m_value->m_property) {
@ -1088,8 +1134,16 @@ namespace smt {
init_model_po(*kv.m_value, m, true);
break;
case sr_tc:
if (has_quantifiers()) {
// model construction for transitive relation is reserved for quantified formulas where
// TC(R) may be needed to perform MBQI
init_model_po(*kv.m_value, m, true);
}
break;
case sr_trc:
if (has_quantifiers()) {
init_model_po(*kv.m_value, m, true);
}
break;
default:
// other 28 combinations of 0x1F

3
src/smt/theory_special_relations.h
View file

@ -172,6 +172,7 @@ namespace smt {
void collect_asserted_po_atoms(vector< std::pair<bool_var,bool> >& atoms) const;
void display_atom(std::ostream & out, atom& a) const;
bool has_quantifiers();
public:
theory_special_relations(ast_manager& m);
@ -179,7 +180,7 @@ namespace smt {
theory * mk_fresh(context * new_ctx) override;
bool internalize_atom(app * atom, bool gate_ctx) override;
bool internalize_term(app * term) override { return false; }
bool internalize_term(app * term) override;
void new_eq_eh(theory_var v1, theory_var v2) override;
void new_diseq_eh(theory_var v1, theory_var v2) override {}
bool use_diseqs() const override { return false; }