[POC] Explore reading regex shapes with PHPRegex

.typos.toml

@@ -1,8 +1,6 @@
 [files]
 extend-exclude = [
     ".git/",
-    # generated by `make regex-parser`, contains lexer regexes
-    "src/Type/Regex/Parser/RegexParser.php",
 ]
 ignore-hidden = false
 

CLAUDE.md

@@ -187,14 +187,13 @@ Collects structural information about a class during AST traversal: properties,
 
 ### Regex parser (`src/Type/Regex/`)
 
-`RegexGroupParser` parses regex patterns to infer `preg_match()` array shapes (via `RegexArrayShapeMatcher`) and `IgnoredRegexValidator` uses it to validate `ignoreErrors` patterns. The parser itself is generated by [phplrt](https://github.com/phplrt/phplrt):
+`RegexGroupParser` reads regex patterns to infer `preg_match()` array shapes (via `RegexArrayShapeMatcher`) and `IgnoredRegexValidator` uses the same reader to validate `ignoreErrors` patterns. Patterns are parsed by [PHPRegex](https://github.com/php-regex/php-regex), from the `php-regex/regex-parser-php74` package (its PHP 7.4 compatible build, so it works after the PHAR is downgraded):
 
-- **Grammar**: `build/regex-parser/RegexGrammar.pp3` (phplrt `.pp3` format, derived from Hoa's regex grammar). The reducers in the grammar build a tree of `PHPStan\Type\Regex\Parser\RegexNode` objects (`#capturing`, `#alternation`, `#quantification`, `token`, ...) that `RegexGroupParser` walks.
-- **Generated parser**: `src/Type/Regex/Parser/RegexParser.php`. Never edit it by hand. Change the grammar and regenerate, then commit both files together.
-- **Regenerating**: `make regex-parser`. It installs the phplrt compiler into `build/regex-parser/vendor` (separate `build/regex-parser/composer.json`, needs PHP 8.4+) and runs `build/regex-parser/generate.php`. The script targets PHP 8.1 syntax and post-processes the output so that the PHAR's PHP 7.4 downgrader can handle it (first-class callables become `\Closure::fromCallable()`) and adds `@throws` to `parse()`.
-- **Runtime**: the generated parser only needs the phplrt runtime, which comes from the `ondrejmirtes/phplrt-runtime` fork (a PHP 7.4 compatible downgrade of the `phplrt/*` runtime packages that replaces them). The compiler is a build-time dependency only and is never shipped.
-- Parse failures throw `Phplrt\Contracts\Parser\Exception\ParserExceptionInterface` or `RegexSyntaxException`; callers treat both as "pattern not supported" and infer no shape.
-- After changing the grammar or `RegexGroupParser`, run the regex tests: `tests/PHPStan/Analyser/nsrt/preg_*.php`, `tests/PHPStan/Type/Php/`, `tests/PHPStan/Type/Regex/` and `tests/PHPStan/Rules/Regexp/`.
+- **Reader**: `PHPRegexReader` parses a pattern for the analysed PHP version and the PCRE2 version it bundles (`pcreVersion` parameter, or `config.platform.lib-pcre` in composer.json, see `PcreVersion`). A pattern the analysed engine refuses, or that the library cannot read, gets no shape.
+- **Shapes**: the type of each group comes from the library's capture shape analysis (`CaptureShapeAnalyzer`: participation, values, length, digit and truthiness facts, per-branch arms). `RegexGroupParser` only walks the tree to place each group: its parent group and the alternation branch it belongs to.
+- Library exceptions implement `PHPRegex\Parser\Exception\ExceptionInterface`; callers treat them as "pattern not supported" and infer no shape.
+- `RegexArrayShapeMatcherEngineTest` matches every constant pattern of the `preg_*` fixtures with the engine and checks that the inferred type accepts what the engine writes.
+- After changing `RegexGroupParser` or `RegexArrayShapeMatcher`, run the regex tests: `tests/PHPStan/Analyser/nsrt/preg_*.php`, `tests/PHPStan/Type/Php/`, `tests/PHPStan/Type/Regex/` and `tests/PHPStan/Rules/Regexp/`.
 
 ## Extension points
 
@@ -350,4 +349,4 @@ When adding or editing PHPDoc comments in this codebase, follow these guidelines
 - `nette/neon` - Configuration file format
 - `react/child-process`, `react/async` - Parallel analysis
 - `symfony/console` - CLI interface
-- `ondrejmirtes/phplrt-runtime` - PHP 7.4 compatible fork of the phplrt runtime, runs the regex parser generated from `build/regex-parser/RegexGrammar.pp3` (`make regex-parser`)
+- `php-regex/regex-parser-php74` - PHP 7.4 compatible build of the PHPRegex parser, reads regex patterns for `preg_*()` shapes and `ignoreErrors`

Makefile

@@ -177,14 +177,6 @@ phpstan-generate-baseline-php:
 phpstan-pro:
 	php -d memory_limit=450M bin/phpstan --pro
 
-# Regenerates src/Type/Regex/Parser/RegexParser.php from
-# build/regex-parser/RegexGrammar.pp3 with the phplrt compiler. The compiler
-# needs PHP 8.4+, the generated parser only needs the phplrt runtime.
-.PHONY: regex-parser
-regex-parser:
-	composer install --working-dir build/regex-parser
-	php build/regex-parser/generate.php
-
 name-collision:
 	php vendor/bin/detect-collisions --configuration build/collision-detector.json
 

build/phpstan.neon

@@ -39,9 +39,6 @@ parameters:
 		- ../tests/PHPStan/Reflection/MixedTypeTest.php
 		- ../tests/e2e/magic-setter/*
 		- ../tests/PHPStan/Rules/Properties/UninitializedPropertyRuleTest.php
-		- ../tests/PHPStan/Command/IgnoredRegexValidatorTest.php
-		- ../src/Command/IgnoredRegexValidator.php
-		- ../src/Type/Regex/Parser/RegexParser.php
 	exceptions:
 		uncheckedExceptionClasses:
 			- 'PHPStan\ShouldNotHappenException'