RemoveFalseReturnTypeExtension: preserved |false for UTF-8 validation patterns like //u

Author: dg

src/Php/RemoveFalseReturnTypeExtension.php

@@ -19,7 +19,7 @@
 use PHPStan\Type\ExpressionTypeResolverExtension;
 use PHPStan\Type\Type;
 use PHPStan\Type\TypeCombinator;
-use function array_merge, explode, str_contains, str_starts_with, strtolower;
+use function array_merge, explode, str_contains, str_starts_with, strlen, strrpos, strtolower, substr;
 
 
 /**
@@ -90,9 +90,10 @@ private function resolveFuncCall(FuncCall $expr, Scope $scope): ?Type
 		}
 
 		// preg_* functions return false only for invalid patterns, so skip narrowing for non-constant patterns
+		// Also preserve |false for UTF-8 validation patterns like //u where false means invalid UTF-8
 		if (str_starts_with($functionName, 'preg_')) {
 			$args = $expr->getArgs();
-			if ($args === [] || $scope->getType($args[0]->value)->getConstantStrings() === []) {
+			if ($args === [] || self::isUtf8ValidationPattern($scope->getType($args[0]->value))) {
 				return null;
 			}
 		}
@@ -259,6 +260,32 @@ private function isMethodSupportedByType(Type $callerType, string $methodName):
 	}
 
 
+	/**
+	 * Returns true for non-constant patterns or UTF-8 validation patterns (empty body + u modifier).
+	 */
+	private static function isUtf8ValidationPattern(Type $patternType): bool
+	{
+		$constants = $patternType->getConstantStrings();
+		if ($constants === []) {
+			return true; // non-constant → preserve |false
+		}
+
+		foreach ($constants as $constant) {
+			$pattern = $constant->getValue();
+			if (
+				strlen($pattern) >= 2
+				&& ($lastPos = strrpos($pattern, $pattern[0], 1)) !== false
+				&& $lastPos === 1 // empty body → delimiter at pos 0 and 1
+				&& str_contains(substr($pattern, 2), 'u')
+			) {
+				return true;
+			}
+		}
+
+		return false;
+	}
+
+
 	private static function removeFalse(Type $type): Type
 	{
 		return TypeCombinator::remove($type, new ConstantBooleanType(false));

tests/data/narrow-return-type.php

@@ -1,4 +1,6 @@
-<?php declare(strict_types=1);
+<?php
+
+declare(strict_types=1);
 
 use function PHPStan\Testing\assertType;
 
@@ -36,6 +38,15 @@ function testRegexConstant(string $s): void
 }
 
 
+// Regex (UTF-8 validation pattern //u — |false preserved)
+function testRegexUtf8Validation(string $s): void
+{
+	assertType('0|1|false', preg_match('//u', $s));
+	assertType('list<string>|false', preg_split('//u', $s));
+	assertType('array|false', preg_grep('//u', [$s]));
+}
+
+
 // Regex (non-constant pattern — |false preserved)
 function testRegexDynamic(string $pattern, string $s): void
 {