Copilot f0956a622f

Refactor regex subset logic into seq_subset with depth-bounded recursion and optimized concat traversal (#9777) ...

`seq_rewriter::is_subset` was too localized and missed key subset
implications for regex concatenations. This change extracts subset
reasoning into a dedicated component and adds heuristic
closure/monotonicity rules, then tunes the recursion strategy based on
profiling feedback.

- **Architecture: isolate subset reasoning**
  - Introduce `seq_subset` in `src/ast/rewriter` (`seq_subset.h/.cpp`).
- Add `seq_subset` as an attribute on `seq_rewriter` and route
`seq_rewriter::is_subset` through it.
- Keep `seq_rewriter` focused on rewrite orchestration while subset
logic evolves independently.

- **Subset rules: broaden inferable cases**
- Add derive-style subset decomposition across `union`, `intersection`,
`complement`, `concat`, and bounded `loop`.
  - Add E3-style closure rules:
    - `R ⊆ R*`
    - `R1* ⊆ R2*  ⇐  R1 ⊆ R2`
    - `R1+ ⊆ R2+  ⇐  R1 ⊆ R2`
  - Add missing cheap cases:
    - `ε ⊆ R` when `R` is nullable
    - `R ⊆ R+`
    - `R+ ⊆ R*`
    - Range containment: `[c1–c2] ⊆ [c3–c4]` when `c3 ≤ c1 ∧ c2 ≤ c4`
    - `to_re(s) ⊆ range` for single-character string constants
    - Difference monotonicity: `a1 \ a2 ⊆ b` when `a1 ⊆ b`
- Star absorption checks for concat/star combinations (`R·R* ⊆ R*`,
`R*·R ⊆ R*`)
- Preserve nullable-based `. +` handling and top/bottom regular-language
shortcuts.

- **Concatenation reasoning and traversal tuning**
- Remove `flatten_concat` and assume right-associative concatenation
traversal.
- Keep containment shortcuts for both `R ⊆ Σ*·R'` and `R ⊆ R'·Σ*` when
`R ⊆ R'`.
  - Make concat/concat handling tail-recursive on second arguments.

- **Depth-bounded recursion (profiling follow-up)**
- Replace visited-pair hash-table recursion state with an explicit depth
parameter in `is_subset_rec`.
  - Add `m_max_depth = 3` and return `false` when the bound is reached.
- Increment depth on recursive calls, except for the tail-recursive
concat-second-argument step.

- **Build integration**
  - Register `seq_subset.cpp` in `src/ast/rewriter/CMakeLists.txt`.

```cpp
// seq_rewriter.cpp
bool seq_rewriter::is_subset(expr* r1, expr* r2) const {
    return m_subset.is_subset(r1, r2);
}
```

---------

Signed-off-by: Nikolaj Bjorner <nbjorner@microsoft.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: Nikolaj Bjorner <nbjorner@microsoft.com>

2026-06-09 13:42:28 -07:00

14 MiB

Raw History

View raw