Fix parsing for marker verbs, fixes #8948

Author: Seldaek

resources/RegexGrammar.pp

@@ -78,6 +78,11 @@
 %token  co:_comment              \)                 -> default
 %token  co:comment               .*?(?=(?<!\\)\))
 
+// Marker verbs
+%token  marker_                  \(\*:              -> mark
+%token  mark:name                [^)]+
+%token  mark:_marker             \)                 -> default
+
 // Capturing group.
 %token  named_capturing_         \(\?P?<            -> nc
 %token  nc:_named_capturing      >                  -> default
@@ -186,7 +191,8 @@
   | literal()
 
 #capturing:
-    ::comment_:: <comment>? ::_comment:: #comment
+    ::marker_:: <name> ::_marker:: #mark
+  | ::comment_:: <comment>? ::_comment:: #comment
   | (
         ::named_capturing_:: <capturing_name> ::_named_capturing:: #namedcapturing
       | ::non_capturing_:: #noncapturing

src/Type/Php/RegexArrayShapeMatcher.php

@@ -125,7 +125,7 @@ private function matchRegex(string $regex, ?int $flags, TrinaryLogic $wasMatched
 			// regex could not be parsed by Hoa/Regex
 			return null;
 		}
-		[$groupList, $groupCombinations] = $parseResult;
+		[$groupList, $groupCombinations, $hasMark] = $parseResult;
 
 		$trailingOptionals = 0;
 		foreach (array_reverse($groupList) as $captureGroup) {
@@ -152,6 +152,7 @@ private function matchRegex(string $regex, ?int $flags, TrinaryLogic $wasMatched
 				$wasMatched,
 				$trailingOptionals,
 				$flags ?? 0,
+				$hasMark,
 			);
 
 			if (!$this->containsUnmatchedAsNull($flags ?? 0)) {
@@ -189,6 +190,7 @@ private function matchRegex(string $regex, ?int $flags, TrinaryLogic $wasMatched
 					$wasMatched,
 					$trailingOptionals,
 					$flags ?? 0,
+					$hasMark,
 				);
 
 				$combiTypes[] = $combiType;
@@ -211,6 +213,7 @@ private function matchRegex(string $regex, ?int $flags, TrinaryLogic $wasMatched
 			$wasMatched,
 			$trailingOptionals,
 			$flags ?? 0,
+			$hasMark,
 		);
 	}
 
@@ -272,6 +275,7 @@ private function buildArrayType(
 		TrinaryLogic $wasMatched,
 		int $trailingOptionals,
 		int $flags,
+		bool $hasMark,
 	): Type
 	{
 		$builder = ConstantArrayTypeBuilder::createEmpty();
@@ -325,6 +329,14 @@ private function buildArrayType(
 			$i++;
 		}
 
+		if ($hasMark) {
+			$builder->setOffsetValueType(
+				$this->getKeyType('MARK'),
+				new IntersectionType([new StringType(), new AccessoryNonEmptyStringType()]),
+				true,
+			);
+		}
+
 		return $builder->getArray();
 	}
 
@@ -372,7 +384,7 @@ private function getValueType(Type $baseType, int $flags): Type
 	}
 
 	/**
-	 * @return array{array<int, RegexCapturingGroup>, array<int, array<int, int[]>>}|null
+	 * @return array{array<int, RegexCapturingGroup>, array<int, array<int, int[]>>, bool}|null
 	 */
 	private function parseGroups(string $regex): ?array
 	{
@@ -398,6 +410,7 @@ private function parseGroups(string $regex): ?array
 		$groupCombinations = [];
 		$alternationId = -1;
 		$captureGroupId = 100;
+		$hasMark = false;
 		$this->walkRegexAst(
 			$ast,
 			false,
@@ -408,9 +421,10 @@ private function parseGroups(string $regex): ?array
 			$captureGroupId,
 			$capturingGroups,
 			$groupCombinations,
+			$hasMark,
 		);
 
-		return [$capturingGroups, $groupCombinations];
+		return [$capturingGroups, $groupCombinations, $hasMark];
 	}
 
 	/**
@@ -427,6 +441,7 @@ private function walkRegexAst(
 		int &$captureGroupId,
 		array &$capturingGroups,
 		array &$groupCombinations,
+		bool &$hasMark,
 	): void
 	{
 		$group = null;
@@ -483,6 +498,11 @@ private function walkRegexAst(
 			$inAlternation = true;
 		}
 
+		if ($ast->getId() === '#mark') {
+			$hasMark = true;
+			return;
+		}
+
 		if ($group instanceof RegexCapturingGroup) {
 			$capturingGroups[$group->getId()] = $group;
 
@@ -506,6 +526,7 @@ private function walkRegexAst(
 				$captureGroupId,
 				$capturingGroups,
 				$groupCombinations,
+				$hasMark,
 			);
 
 			if ($ast->getId() !== '#alternation') {

tests/PHPStan/Analyser/nsrt/preg_match_shapes.php

@@ -521,6 +521,9 @@ function bug11323(string $s): void {
 	if (preg_match("{([\r\n]+)(\n)([\n])}", $s, $matches)) {
 		assertType('array{string, non-empty-string, non-empty-string, non-empty-string}', $matches);
 	}
+	if (preg_match('/foo(*:first)|bar(*:second)([x])/', $s, $matches)) {
+		assertType('array{0: string, 1?: non-empty-string, MARK?: non-empty-string}', $matches);
+	}
 }
 
 function (string $s): void {