# SLeeLa Regex Native Foundation **Regex API:** 1.0.0 draft **Native layer:** C11 ## Implemented now The native foundation provides Level 1 literal `contains`, `starts`, `ends`, and `exact`, plus Level 2 ASCII `digit`, `word`, `space`, and `any` with `+`, `*`, `{n}`, and `{n,m}`. The result includes match status and byte offsets. Levels 3 and 4 currently return an explicit unsupported status until their structural, capture, resource-limit, and Unicode semantics are finalized. ## Unicode boundary The first native Level 2 predicates are intentionally ASCII-oriented. They do not claim full Unicode character-property semantics. Level 4 reserves explicit Unicode properties and advanced expression behavior for the completed contract. ## Safety The foundation uses a small forward scanner for its executable subset. It has no network, file, process, or code-execution side effects. ## Smoke test `impl/tests/core/regex_smoke.c` exercises Level 1 and Level 2 and verifies that Level 3 is reported as unsupported rather than silently misinterpreted. ## Corrections The standalone `regex/` subproject (which builds under `-Werror`) was repaired: a `Diagnostic*=nullptr` default argument that tokenized as `*=`, an over-escaped `'\\0'` char constant, misleading-indentation warnings, a Java `Map.of` call exceeding its 10-pair overload, a non-executable test script, a locale-dependent `sort` / missing `cmp`, a buggy `awk` pattern, and genuinely wrong test expectations (an anchored `^...$` pattern used for a substring search, plus a wrong match span and capture count). `make -C regex` now builds and tests clean. See the 2026-10-03 entry in [`../REVISIONS.md`](../REVISIONS.md).