3
0
Fork 0
mirror of https://github.com/Z3Prover/z3 synced 2026-03-07 22:04:53 +00:00

Add Phase 3: Nielsen transformation engine and equation solving

- New nseq_nielsen.h/cpp in src/ast/rewriter/: self-contained Nielsen
  transformation engine for word equations
  - simplify(): strip common prefix/suffix, empty elimination, variable
    stripping, single-var assignment detection
  - split(): case analysis for var vs constant, var vs var
  - is_conflict(): mismatch detection (different constants, one side
    has constants while other is empty)

- Wire Nielsen into theory_nseq:
  - solve_eqs()/solve_eq(): process word equations using Nielsen
    transformations with e-graph canonization
  - branch_eq()/branch_var_prefix(): binary empty/non-empty decisions
    and prefix enumeration (no fresh variable creation)
  - canonize(): rewrite equation sides using current e-graph equivalences
  - all_eqs_solved(): check if all equations are satisfied
  - mk_value(): basic model generation (walk e-class for string constants)

- Passes basic tests: simple equalities, concat equations, unsat detection

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This commit is contained in:
Nikolaj Bjorner 2026-02-27 18:01:08 -08:00
parent 58b57b2632
commit f48040d809
5 changed files with 743 additions and 1 deletions

View file

@ -41,6 +41,7 @@ z3_add_component(rewriter
seq_eq_solver.cpp
seq_rewriter.cpp
seq_skolem.cpp
nseq_nielsen.cpp
th_rewriter.cpp
value_sweep.cpp
var_subst.cpp

View file

@ -0,0 +1,366 @@
/*++
Copyright (c) 2025 Microsoft Corporation
Module Name:
nseq_nielsen.cpp
Abstract:
Nielsen transformation-based word equation solver.
Author:
Clemens Eisenhofer
Nikolaj Bjorner (nbjorner) 2025-2-28
--*/
#include "ast/ast_pp.h"
#include "ast/ast_ll_pp.h"
#include "ast/rewriter/nseq_nielsen.h"
namespace seq {
nielsen::nielsen(ast_manager& m, seq_rewriter& rw)
: m(m), m_util(m), m_autil(m), m_rw(rw), m_lhs(m), m_rhs(m) {
}
bool nielsen::is_var(expr* e) const {
return m_util.is_seq(e) &&
!m_util.str.is_concat(e) &&
!m_util.str.is_unit(e) &&
!m_util.str.is_empty(e) &&
!m_util.str.is_string(e);
}
bool nielsen::is_unit(expr* e) const {
return m_util.str.is_unit(e);
}
bool nielsen::is_empty(expr* e) const {
return m_util.str.is_empty(e);
}
bool nielsen::has_var(expr_ref_vector const& es) const {
for (expr* e : es)
if (is_var(e))
return true;
return false;
}
// -------------------------------------------------------
// Strip matching constants/units from equation sides
// -------------------------------------------------------
bool nielsen::strip_common_prefix(expr_ref_vector& lhs, expr_ref_vector& rhs) {
unsigned i = 0;
unsigned min_sz = std::min(lhs.size(), rhs.size());
while (i < min_sz) {
expr* l = lhs.get(i);
expr* r = rhs.get(i);
// Both must be ground/unit and equal
if (l == r && (is_unit(l) || m_util.str.is_string(l))) {
i++;
continue;
}
// Check if both are string constants with matching prefix
zstring s1, s2;
if (m_util.str.is_string(l, s1) && m_util.str.is_string(r, s2)) {
if (s1 == s2) { i++; continue; }
}
break;
}
if (i == 0) return false;
expr_ref_vector new_lhs(m), new_rhs(m);
new_lhs.append(lhs.size() - i, lhs.data() + i);
new_rhs.append(rhs.size() - i, rhs.data() + i);
lhs.swap(new_lhs);
rhs.swap(new_rhs);
return true;
}
bool nielsen::strip_common_suffix(expr_ref_vector& lhs, expr_ref_vector& rhs) {
unsigned li = lhs.size();
unsigned ri = rhs.size();
unsigned stripped = 0;
while (li > 0 && ri > 0) {
expr* l = lhs.get(li - 1);
expr* r = rhs.get(ri - 1);
if (l == r && (is_unit(l) || m_util.str.is_string(l))) {
li--; ri--; stripped++;
continue;
}
zstring s1, s2;
if (m_util.str.is_string(l, s1) && m_util.str.is_string(r, s2)) {
if (s1 == s2) { li--; ri--; stripped++; continue; }
}
break;
}
if (stripped == 0) return false;
lhs.resize(li);
rhs.resize(ri);
return true;
}
// -------------------------------------------------------
// Main simplification (no case splitting)
// -------------------------------------------------------
nielsen_result nielsen::simplify(expr_ref_vector& lhs, expr_ref_vector& rhs) {
bool changed = false;
// Remove empty strings from both sides
unsigned j = 0;
for (unsigned i = 0; i < lhs.size(); ++i)
if (!is_empty(lhs.get(i)))
lhs[j++] = lhs.get(i);
lhs.resize(j);
j = 0;
for (unsigned i = 0; i < rhs.size(); ++i)
if (!is_empty(rhs.get(i)))
rhs[j++] = rhs.get(i);
rhs.resize(j);
// Check trivial cases
if (lhs.empty() && rhs.empty())
return nielsen_result::solved;
// Strip common prefix and suffix
changed |= strip_common_prefix(lhs, rhs);
changed |= strip_common_suffix(lhs, rhs);
if (lhs.empty() && rhs.empty())
return nielsen_result::solved;
// Check for conflict: both sides start with different constants
if (is_conflict(lhs, rhs))
return nielsen_result::conflict;
// Variable = empty: if one side is empty and other has single var
if (lhs.empty() && rhs.size() == 1 && is_var(rhs.get(0)))
return nielsen_result::solved; // x = ε is a solution
if (rhs.empty() && lhs.size() == 1 && is_var(lhs.get(0)))
return nielsen_result::solved; // x = ε is a solution
// Single variable = single term (x = t): a direct assignment, solved
if (lhs.size() == 1 && is_var(lhs.get(0)) && !has_var(rhs))
return nielsen_result::solved;
if (rhs.size() == 1 && is_var(rhs.get(0)) && !has_var(lhs))
return nielsen_result::solved;
// Both sides start with the same variable: strip it
if (!lhs.empty() && !rhs.empty() && lhs.get(0) == rhs.get(0) && is_var(lhs.get(0))) {
expr_ref_vector new_lhs(m), new_rhs(m);
new_lhs.append(lhs.size() - 1, lhs.data() + 1);
new_rhs.append(rhs.size() - 1, rhs.data() + 1);
lhs.swap(new_lhs);
rhs.swap(new_rhs);
changed = true;
}
// Both sides end with the same variable: strip it
if (!lhs.empty() && !rhs.empty() &&
lhs.back() == rhs.back() && is_var(lhs.back())) {
lhs.pop_back();
rhs.pop_back();
changed = true;
}
if (changed && lhs.empty() && rhs.empty())
return nielsen_result::solved;
if (changed)
return nielsen_result::reduced;
return nielsen_result::unchanged;
}
// -------------------------------------------------------
// Check for conflicts
// -------------------------------------------------------
bool nielsen::is_conflict(expr_ref_vector const& lhs, expr_ref_vector const& rhs) const {
if (lhs.empty() != rhs.empty()) {
// One side empty, other side has constants
expr_ref_vector const& nonempty = lhs.empty() ? rhs : lhs;
for (unsigned i = 0; i < nonempty.size(); ++i) {
zstring s;
if (m_util.str.is_string(nonempty[i], s) && s.length() > 0)
return true;
if (is_unit(nonempty[i]))
return true;
}
return false;
}
if (lhs.empty() && rhs.empty())
return false;
// Both start with different non-variable ground terms
expr* l = lhs[0];
expr* r = rhs[0];
zstring s1, s2;
if (m_util.str.is_string(l, s1) && m_util.str.is_string(r, s2)) {
if (s1.length() > 0 && s2.length() > 0 && s1[0] != s2[0])
return true;
}
if (is_unit(l) && is_unit(r) && l != r) {
// Different unit terms
expr* c1 = to_app(l)->get_arg(0);
expr* c2 = to_app(r)->get_arg(0);
rational v1, v2;
if (m_autil.is_numeral(c1, v1) && m_autil.is_numeral(c2, v2) && v1 != v2)
return true;
}
return false;
}
bool nielsen::is_solved(expr_ref_vector const& lhs, expr_ref_vector const& rhs) const {
return lhs.empty() && rhs.empty();
}
// -------------------------------------------------------
// Case splitting
// -------------------------------------------------------
void nielsen::apply_subst(expr* var, expr* term, expr_ref_vector const& src, expr_ref_vector& dst) {
dst.reset();
for (unsigned i = 0; i < src.size(); ++i) {
if (src[i] == var) {
// Replace variable with its substitution
m_util.str.get_concat_units(term, dst);
}
else {
dst.push_back(src[i]);
}
}
}
bool nielsen::split(expr_ref_vector const& lhs, expr_ref_vector const& rhs,
vector<nielsen_branch>& branches) {
if (lhs.empty() || rhs.empty()) {
// One side empty: all variables on other side must be empty
expr_ref_vector const& nonempty = lhs.empty() ? rhs : lhs;
for (unsigned i = 0; i < nonempty.size(); ++i) {
if (is_var(nonempty[i])) {
nielsen_branch b(m);
b.var = nonempty[i];
b.term = m_util.str.mk_empty(nonempty[i]->get_sort());
// After substitution, just remove the empty variable
expr_ref_vector const& other = lhs.empty() ? lhs : rhs;
b.new_lhs.append(other);
for (unsigned j = 0; j < nonempty.size(); ++j)
if (j != i && !is_empty(nonempty[j]))
b.new_rhs.push_back(nonempty[j]);
if (lhs.empty()) b.new_lhs.swap(b.new_rhs);
branches.push_back(std::move(b));
return true;
}
}
return false;
}
expr* l0 = lhs[0];
expr* r0 = rhs[0];
// Case 1: Variable vs constant/unit
// x·α = c·β → branch: x = ε or x = c·x'
if (is_var(l0) && (is_unit(r0) || m_util.str.is_string(r0))) {
// Branch 1: x = ε
{
nielsen_branch b(m);
b.var = l0;
b.term = m_util.str.mk_empty(l0->get_sort());
apply_subst(l0, b.term, lhs, b.new_lhs);
b.new_rhs.append(rhs);
branches.push_back(std::move(b));
}
// Branch 2: x = r0 · x' (peel first character)
{
nielsen_branch b(m);
b.var = l0;
expr_ref x_prime(m.mk_fresh_const("x", l0->get_sort()), m);
b.term = m_util.str.mk_concat(r0, x_prime);
apply_subst(l0, b.term, lhs, b.new_lhs);
b.new_rhs.append(rhs);
branches.push_back(std::move(b));
}
return true;
}
// Symmetric: constant vs variable on left
if (is_var(r0) && (is_unit(l0) || m_util.str.is_string(l0))) {
// Branch 1: y = ε
{
nielsen_branch b(m);
b.var = r0;
b.term = m_util.str.mk_empty(r0->get_sort());
b.new_lhs.append(lhs);
apply_subst(r0, b.term, rhs, b.new_rhs);
branches.push_back(std::move(b));
}
// Branch 2: y = l0 · y'
{
nielsen_branch b(m);
b.var = r0;
expr_ref y_prime(m.mk_fresh_const("y", r0->get_sort()), m);
b.term = m_util.str.mk_concat(l0, y_prime);
b.new_lhs.append(lhs);
apply_subst(r0, b.term, rhs, b.new_rhs);
branches.push_back(std::move(b));
}
return true;
}
// Case 2: Variable vs variable
// x·α = y·β → branch: x = y (if same), x = y·z, or y = x·z
if (is_var(l0) && is_var(r0)) {
if (l0 == r0) {
// Same variable: strip and continue (should have been handled by simplify)
return false;
}
// Branch 1: x = ε
{
nielsen_branch b(m);
b.var = l0;
b.term = m_util.str.mk_empty(l0->get_sort());
apply_subst(l0, b.term, lhs, b.new_lhs);
b.new_rhs.append(rhs);
branches.push_back(std::move(b));
}
// Branch 2: y = ε
{
nielsen_branch b(m);
b.var = r0;
b.term = m_util.str.mk_empty(r0->get_sort());
b.new_lhs.append(lhs);
apply_subst(r0, b.term, rhs, b.new_rhs);
branches.push_back(std::move(b));
}
// Branch 3: x = y · z (x is longer)
{
nielsen_branch b(m);
b.var = l0;
expr_ref z(m.mk_fresh_const("z", l0->get_sort()), m);
b.term = m_util.str.mk_concat(r0, z);
apply_subst(l0, b.term, lhs, b.new_lhs);
b.new_rhs.append(rhs);
branches.push_back(std::move(b));
}
// Branch 4: y = x · z (y is longer)
{
nielsen_branch b(m);
b.var = r0;
expr_ref z(m.mk_fresh_const("z", r0->get_sort()), m);
b.term = m_util.str.mk_concat(l0, z);
b.new_lhs.append(lhs);
apply_subst(r0, b.term, rhs, b.new_rhs);
branches.push_back(std::move(b));
}
return true;
}
return false;
}
}

View file

@ -0,0 +1,106 @@
/*++
Copyright (c) 2025 Microsoft Corporation
Module Name:
nseq_nielsen.h
Abstract:
Nielsen transformation-based word equation solver.
Self-contained rewriter utility for decomposing and solving
word equations using Nielsen transformations.
Given a word equation lhs_1·...·lhs_n = rhs_1·...·rhs_m,
the solver applies:
- Constant matching: strip matching constants from both sides
- Empty elimination: x = ε
- Variable splitting: x·α = c·β x = c·x' or x = ε
- Length-based reasoning for pruning
Author:
Clemens Eisenhofer
Nikolaj Bjorner (nbjorner) 2025-2-28
--*/
#pragma once
#include "ast/seq_decl_plugin.h"
#include "ast/arith_decl_plugin.h"
#include "ast/rewriter/seq_rewriter.h"
namespace seq {
// Result of a Nielsen transformation step
enum class nielsen_result {
solved, // equation is trivially satisfied
conflict, // equation is unsatisfiable
reduced, // equation was simplified
split, // case split needed
unchanged // no progress
};
// A case split produced by Nielsen transformation.
// Represents a substitution x -> t and the resulting simplified equation.
struct nielsen_branch {
expr_ref var; // variable being assigned
expr_ref term; // value assigned (ε, constant prefix + fresh var, etc.)
expr_ref_vector new_lhs; // resulting LHS after substitution
expr_ref_vector new_rhs; // resulting RHS after substitution
nielsen_branch(ast_manager& m) : var(m), term(m), new_lhs(m), new_rhs(m) {}
};
// Nielsen transformation engine for word equations.
// Self-contained, no dependency on SMT context.
class nielsen {
ast_manager& m;
seq_util m_util;
arith_util m_autil;
seq_rewriter& m_rw;
// Scratch space
expr_ref_vector m_lhs, m_rhs;
// Strip matching constants/units from both sides of an equation.
// Returns true if anything was stripped.
bool strip_common_prefix(expr_ref_vector& lhs, expr_ref_vector& rhs);
bool strip_common_suffix(expr_ref_vector& lhs, expr_ref_vector& rhs);
// Check if an expression is a unit (single character)
bool is_unit(expr* e) const;
// Check if an expression is the empty string
bool is_empty(expr* e) const;
// Check if an expression vector contains any variables
bool has_var(expr_ref_vector const& es) const;
// Apply substitution x -> t in an expression vector
void apply_subst(expr* var, expr* term, expr_ref_vector const& src, expr_ref_vector& dst);
public:
nielsen(ast_manager& m, seq_rewriter& rw);
// Check if an expression is a string variable (not unit, not constant)
bool is_var(expr* e) const;
// Main simplification: reduce a word equation as far as possible
// without case splitting. Returns the result status.
// lhs, rhs: in/out - the equation sides (decomposed into concat components)
nielsen_result simplify(expr_ref_vector& lhs, expr_ref_vector& rhs);
// Generate case split branches for a word equation that cannot
// be further simplified. The equation should be in simplified form.
// branches: output - possible branches to explore
// Returns false if no splits are possible (stuck).
bool split(expr_ref_vector const& lhs, expr_ref_vector const& rhs,
vector<nielsen_branch>& branches);
// Check if an equation is trivially solved (both sides empty)
bool is_solved(expr_ref_vector const& lhs, expr_ref_vector const& rhs) const;
// Check for conflict (mismatched constants)
bool is_conflict(expr_ref_vector const& lhs, expr_ref_vector const& rhs) const;
};
}

View file

@ -33,6 +33,7 @@ namespace smt {
m_sk(ctx.get_manager(), m_rewrite),
m_arith_value(ctx.get_manager()),
m_state(ctx.get_manager(), m_util),
m_nielsen(ctx.get_manager(), m_seq_rewrite),
m_find(*this),
m_has_seq(false),
m_new_propagation(false) {
@ -53,6 +54,7 @@ namespace smt {
if (!m_has_seq)
return FC_DONE;
m_new_propagation = false;
TRACE(seq, display(tout << "final_check level=" << ctx.get_scope_level() << "\n"););
// Process pending axioms
@ -65,9 +67,16 @@ namespace smt {
return FC_CONTINUE;
}
// TODO: implement Nielsen transformation-based solving
// Solve word equations using Nielsen transformations
if (solve_eqs())
return FC_CONTINUE;
// TODO: implement regex membership checking
// TODO: implement length/Parikh reasoning
if (all_eqs_solved() && m_state.mems().empty())
return FC_DONE;
return FC_GIVEUP;
}
@ -370,6 +379,226 @@ namespace smt {
return expr_ref(m_util.str.mk_concat(es.size(), es.data(), s), m);
}
// Canonize an equation side using current e-graph equivalences.
// Replaces each element with its canonical representative, decomposing
// concatenations and string constants as needed.
bool theory_nseq::canonize(expr_ref_vector const& src, expr_ref_vector& dst,
nseq_dependency*& dep) {
dst.reset();
for (expr* e : src) {
if (!ctx.e_internalized(e)) {
dst.push_back(e);
continue;
}
enode* n = ctx.get_enode(e);
enode* r = n->get_root();
expr* re = r->get_expr();
if (re != e) {
// Track the dependency for the equality
dep = m_state.mk_join(dep, m_state.mk_dep(n, r));
}
// Decompose the canonical representative into concat components
m_util.str.get_concat_units(re, dst);
}
return true;
}
// Check if all equations are satisfied in the current e-graph.
bool theory_nseq::all_eqs_solved() {
for (auto const& eq : m_state.eqs()) {
expr_ref_vector lhs(m), rhs(m);
nseq_dependency* dep = eq.dep();
if (!canonize(eq.lhs(), lhs, dep) || !canonize(eq.rhs(), rhs, dep))
return false;
seq::nielsen_result result = m_nielsen.simplify(lhs, rhs);
TRACE(seq, tout << "all_eqs_solved [" << eq.id() << "]: ";
for (auto* e : lhs) tout << mk_bounded_pp(e, m, 2) << " ";
tout << "= ";
for (auto* e : rhs) tout << mk_bounded_pp(e, m, 2) << " ";
tout << " -> " << (int)result << "\n";);
if (result != seq::nielsen_result::solved)
return false;
}
return true;
}
// -------------------------------------------------------
// Nielsen equation solving
// -------------------------------------------------------
bool theory_nseq::solve_eqs() {
auto const& eqs = m_state.eqs();
for (unsigned i = 0; !ctx.inconsistent() && i < eqs.size(); ++i)
solve_eq(eqs[i]);
return m_new_propagation || ctx.inconsistent();
}
bool theory_nseq::solve_eq(nseq_eq const& eq) {
expr_ref_vector lhs(m), rhs(m);
nseq_dependency* dep = eq.dep();
// Canonize using current e-graph equivalences
if (!canonize(eq.lhs(), lhs, dep) || !canonize(eq.rhs(), rhs, dep))
return false;
TRACE(seq, tout << "solve_eq [" << eq.id() << "]: ";
for (auto* e : lhs) tout << mk_bounded_pp(e, m, 2) << " ";
tout << "= ";
for (auto* e : rhs) tout << mk_bounded_pp(e, m, 2) << " ";
tout << "\n";);
// Apply Nielsen simplification
seq::nielsen_result result = m_nielsen.simplify(lhs, rhs);
switch (result) {
case seq::nielsen_result::solved: {
// Propagate solved form: either both empty, var = empty, or var = ground term
bool changed = false;
if (lhs.size() == 1 && m_nielsen.is_var(lhs.get(0)) && !rhs.empty()) {
sort* s = lhs.get(0)->get_sort();
expr_ref rhs_concat = mk_concat(rhs, s);
changed = propagate_eq(dep, lhs.get(0), rhs_concat);
}
else if (rhs.size() == 1 && m_nielsen.is_var(rhs.get(0)) && !lhs.empty()) {
sort* s = rhs.get(0)->get_sort();
expr_ref lhs_concat = mk_concat(lhs, s);
changed = propagate_eq(dep, rhs.get(0), lhs_concat);
}
else {
// All remaining vars must be empty
for (unsigned i = 0; i < lhs.size(); ++i)
if (m_nielsen.is_var(lhs.get(i)))
changed |= propagate_eq(dep, lhs.get(i), expr_ref(m_util.str.mk_empty(lhs.get(i)->get_sort()), m));
for (unsigned i = 0; i < rhs.size(); ++i)
if (m_nielsen.is_var(rhs.get(i)))
changed |= propagate_eq(dep, rhs.get(i), expr_ref(m_util.str.mk_empty(rhs.get(i)->get_sort()), m));
}
TRACE(seq, tout << "solved" << (changed ? " (propagated)" : " (no change)") << "\n";);
return changed;
}
case seq::nielsen_result::conflict:
TRACE(seq, tout << "conflict\n";);
set_conflict(dep);
return true;
case seq::nielsen_result::reduced: {
if (lhs.empty() && rhs.empty())
return false; // already equal
bool changed = false;
if (!lhs.empty() && !rhs.empty()) {
sort* s = lhs[0]->get_sort();
expr_ref l = mk_concat(lhs, s);
expr_ref r = mk_concat(rhs, s);
changed = propagate_eq(dep, l, r);
}
else {
expr_ref_vector& nonempty = lhs.empty() ? rhs : lhs;
for (expr* e : nonempty) {
if (m_util.is_seq(e)) {
expr_ref emp(m_util.str.mk_empty(e->get_sort()), m);
changed |= propagate_eq(dep, e, emp);
}
}
}
TRACE(seq, tout << "reduced" << (changed ? " (propagated)" : " (no change)") << "\n";);
return changed;
}
case seq::nielsen_result::split:
case seq::nielsen_result::unchanged:
break;
}
// Try branching: find a variable and decide x = "" or x ≠ ""
return branch_eq(lhs, rhs, dep);
}
bool theory_nseq::branch_eq(expr_ref_vector const& lhs, expr_ref_vector const& rhs,
nseq_dependency* dep) {
// Try branching on variables from the LHS first, then RHS
for (unsigned side = 0; side < 2; ++side) {
expr_ref_vector const& es = (side == 0) ? lhs : rhs;
for (expr* e : es) {
if (!m_nielsen.is_var(e))
continue;
// Check if this variable is already known to be empty
enode* n = ctx.get_enode(e);
expr_ref emp(m_util.str.mk_empty(e->get_sort()), m);
if (ctx.e_internalized(emp)) {
enode* n_emp = ctx.get_enode(emp);
if (n->get_root() == n_emp->get_root())
continue; // already equal to empty, skip
}
// Decide: x = "" or x ≠ ""
literal eq_empty = mk_eq(e, emp, false);
switch (ctx.get_assignment(eq_empty)) {
case l_undef:
// Force the empty branch first
TRACE(seq, tout << "branch " << mk_bounded_pp(e, m) << " = \"\"\n";);
ctx.force_phase(eq_empty);
ctx.mark_as_relevant(eq_empty);
m_new_propagation = true;
m_state.stats().m_num_splits++;
return true;
case l_true:
// Variable assigned to empty but EQ not yet merged
// Propagate the equality
propagate_eq(dep, e, emp);
return true;
case l_false:
// x ≠ "": try to find a prefix from the other side
break;
}
}
}
// If all variables on one side are decided non-empty,
// try to match against the other side using prefix enumeration
return branch_eq_prefix(lhs, rhs, dep);
}
bool theory_nseq::branch_eq_prefix(expr_ref_vector const& lhs, expr_ref_vector const& rhs,
nseq_dependency* dep) {
// Find a leading variable on either side
if (!lhs.empty() && m_nielsen.is_var(lhs[0]) && !rhs.empty())
return branch_var_prefix(lhs[0], rhs, dep);
if (!rhs.empty() && m_nielsen.is_var(rhs[0]) && !lhs.empty())
return branch_var_prefix(rhs[0], lhs, dep);
return false;
}
bool theory_nseq::branch_var_prefix(expr* x, expr_ref_vector const& other,
nseq_dependency* dep) {
// For x starting one side, try x = prefix of other side
// x = "" was already tried (assigned false)
// Now enumerate: x = other[0], x = other[0]·other[1], ...
expr_ref candidate(m);
for (unsigned len = 1; len <= other.size(); ++len) {
if (len == 1)
candidate = other[0];
else
candidate = expr_ref(m_util.str.mk_concat(len, other.data(), x->get_sort()), m);
literal eq = mk_eq(x, candidate, false);
switch (ctx.get_assignment(eq)) {
case l_undef:
TRACE(seq, tout << "branch " << mk_bounded_pp(x, m) << " = " << mk_bounded_pp(candidate, m) << "\n";);
ctx.force_phase(eq);
ctx.mark_as_relevant(eq);
m_new_propagation = true;
m_state.stats().m_num_splits++;
return true;
case l_true:
propagate_eq(dep, x, candidate);
return true;
case l_false:
continue;
}
}
return false;
}
// -------------------------------------------------------
// Display and statistics
// -------------------------------------------------------
@ -396,6 +625,35 @@ namespace smt {
// -------------------------------------------------------
model_value_proc* theory_nseq::mk_value(enode* n, model_generator& mg) {
app* e = n->get_expr();
TRACE(seq, tout << "mk_value: " << mk_bounded_pp(e, m) << "\n";);
if (m_util.is_seq(e)) {
// Walk the equivalence class looking for a concrete string value
enode* root = n->get_root();
enode* curr = root;
do {
expr* ce = curr->get_expr();
zstring s;
if (m_util.str.is_string(ce, s))
return alloc(expr_wrapper_proc, to_app(ce));
if (m_util.str.is_empty(ce))
return alloc(expr_wrapper_proc, to_app(ce));
curr = curr->get_next();
} while (curr != root);
// No concrete value found: use seq_factory to get a fresh value
expr_ref val(m);
val = mg.get_model().get_fresh_value(e->get_sort());
if (val)
return alloc(expr_wrapper_proc, to_app(val));
// Fallback: empty string
return alloc(expr_wrapper_proc, to_app(m_util.str.mk_empty(e->get_sort())));
}
if (m_util.is_re(e)) {
return alloc(expr_wrapper_proc, to_app(m_util.re.mk_full_seq(e->get_sort())));
}
UNREACHABLE();
return nullptr;
}

View file

@ -23,6 +23,7 @@ Author:
#include "ast/rewriter/seq_rewriter.h"
#include "ast/rewriter/seq_skolem.h"
#include "ast/rewriter/th_rewriter.h"
#include "ast/rewriter/nseq_nielsen.h"
#include "model/seq_factory.h"
#include "smt/smt_theory.h"
#include "smt/smt_arith_value.h"
@ -39,6 +40,7 @@ namespace smt {
seq::skolem m_sk;
arith_value m_arith_value;
nseq_state m_state;
seq::nielsen m_nielsen;
nseq_union_find m_find;
bool m_has_seq;
bool m_new_propagation;
@ -86,6 +88,15 @@ namespace smt {
expr_ref mk_len(expr* s);
expr_ref mk_concat(expr_ref_vector const& es, sort* s);
// Nielsen equation solving
bool solve_eqs();
bool solve_eq(nseq_eq const& eq);
bool branch_eq(expr_ref_vector const& lhs, expr_ref_vector const& rhs, nseq_dependency* dep);
bool branch_eq_prefix(expr_ref_vector const& lhs, expr_ref_vector const& rhs, nseq_dependency* dep);
bool branch_var_prefix(expr* x, expr_ref_vector const& other, nseq_dependency* dep);
bool canonize(expr_ref_vector const& src, expr_ref_vector& dst, nseq_dependency*& dep);
bool all_eqs_solved();
public:
theory_nseq(context& ctx);
~theory_nseq() override;