3
0
Fork 0
mirror of https://github.com/Z3Prover/z3 synced 2026-03-19 03:23:10 +00:00
z3/src/smt/nseq_regex.h
copilot-swe-agent[bot] 0bdec633d7 Implement ZIPT string solver skeleton (theory_nseq)
Add theory_nseq, a Nielsen-graph-based string solver plugin for Z3.

## New files
- src/smt/nseq_state.h/.cpp: constraint store bridging SMT context to
  Nielsen graph with manual push/pop backtracking
- src/smt/nseq_regex.h/.cpp: regex membership handling via Brzozowski
  derivatives (stub delegates to sgraph::brzozowski_deriv)
- src/smt/nseq_model.h/.cpp: model generation stub
- src/smt/theory_nseq.h/.cpp: main theory class implementing smt::theory
  with its own private egraph/sgraph, returns FC_GIVEUP as skeleton
- src/test/nseq_basic.cpp: unit tests covering instantiation, parameter
  validation, trivial-equality SAT, and node simplification

## Extensions to seq_nielsen.h/.cpp
- Add search_result enum and solve() iterative-deepening DFS entry point
- Add search_dfs() recursive DFS driver
- Add simplify_node(), generate_extensions(), collect_conflict_deps()
- Add nielsen_node::simplify_and_init(): trivial removal, empty
  propagation, prefix matching, symbol clash detection
- Add nielsen_node::is_satisfied(), is_subsumed_by()
- Implement Det, Const Nielsen, and Eq-split modifiers in
  generate_extensions()

## Integration
- smt_params.cpp: accept 'nseq' as valid string_solver value
- smt_params_helper.pyg: document 'nseq' option
- smt_setup.h/.cpp: add setup_nseq(), wire into setup_QF_S() and
  setup_seq_str()
- smt/CMakeLists.txt: add new sources and smt_seq dependency

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-03-03 21:50:21 +00:00

66 lines
2 KiB
C++

/*++
Copyright (c) 2026 Microsoft Corporation
Module Name:
nseq_regex.h
Abstract:
Regex membership handling using Brzozowski derivatives.
Processes str_mem constraints after character consumption.
Author:
Nikolaj Bjorner (nbjorner) 2026-03-01
--*/
#pragma once
#include "ast/euf/euf_sgraph.h"
#include "smt/seq/seq_nielsen.h"
namespace smt {
class nseq_regex {
euf::sgraph& m_sg;
public:
nseq_regex(euf::sgraph& sg) : m_sg(sg) {}
// check if a regex snode represents the empty language
bool is_empty_regex(euf::snode* re) const {
return re && re->is_fail();
}
// compute derivative of regex re with respect to char elem and
// return a new str_mem for the resulting constraint
seq::str_mem derive(seq::str_mem const& mem, euf::snode* elem) {
euf::snode* deriv = m_sg.brzozowski_deriv(mem.m_regex, elem);
euf::snode* new_str = m_sg.drop_first(mem.m_str);
return seq::str_mem(new_str, deriv, mem.m_history, mem.m_id, mem.m_dep);
}
// process a regex membership constraint after one character has been consumed
// returns false if the resulting regex is empty (conflict)
bool process_str_mem(seq::str_mem const& mem,
vector<seq::str_mem>& out_mems) {
if (!mem.m_str || !mem.m_regex)
return true;
// if regex does not accept the empty string and the string side is empty, conflict
if (mem.m_str->is_empty()) {
return mem.m_regex->is_nullable();
}
// compute minterms for the regex
euf::snode_vector minterms;
m_sg.compute_minterms(mem.m_regex, minterms);
for (euf::snode* ch : minterms) {
seq::str_mem new_mem = derive(mem, ch);
if (!is_empty_regex(new_mem.m_regex))
out_mems.push_back(new_mem);
}
return true;
}
};
}