[Form] Fix ICU 72+ whitespace handling in DateTimeToLocalizedStringTransformer

roukmoute · roukmoute · commit b9f24b7b6334 · 2026-01-18T21:43:36.000+01:00
ICU 72+ uses Unicode whitespace characters (such as U+202F narrow no-break space) in formatted date strings instead of regular ASCII spaces. This caused parsing failures when users typed dates with regular spaces. This fix: - Normalizes transform() output to use regular ASCII spaces for consistent display - Makes reverseTransform() accept both regular spaces and ICU 72+ whitespace by trying both formats when parsing fails Fix symfony#63113 (review)
diff --git a/src/Symfony/Component/Form/Extension/Core/DataTransformer/DateTimeToLocalizedStringTransformer.php b/src/Symfony/Component/Form/Extension/Core/DataTransformer/DateTimeToLocalizedStringTransformer.php
@@ -25,6 +25,15 @@
  */
 class DateTimeToLocalizedStringTransformer extends BaseDateTimeTransformer
 {
+    /**
+     * Unicode whitespace characters used by ICU in formatted date strings.
+     *
+     * @see https://unicode-org.atlassian.net/browse/CLDR-14032
+     */
+    private const NO_BREAK_SPACE = "\u{00A0}";
+    private const NARROW_NO_BREAK_SPACE = "\u{202F}"; // Used by ICU 72+ before AM/PM
+    private const THIN_SPACE = "\u{2009}";
+
     private int $dateFormat;
     private int $timeFormat;
 
@@ -85,7 +94,7 @@ public function transform(mixed $dateTime): string
             throw new TransformationFailedException(intl_get_error_message());
         }
 
-        return $value;
+        return self::normalizeWhitespace($value);
     }
 
     public function reverseTransform(mixed $value): ?\DateTime
@@ -103,11 +112,7 @@ public function reverseTransform(mixed $value): ?\DateTime
         $dateOnly = $this->isPatternDateOnly();
         $dateFormatter = $this->getIntlDateFormatter($dateOnly);
 
-        try {
-            $timestamp = @$dateFormatter->parse($value);
-        } catch (\IntlException $e) {
-            throw new TransformationFailedException($e->getMessage(), $e->getCode(), $e);
-        }
+        $timestamp = $this->parse($dateFormatter, $value);
 
         if (0 != intl_get_error_code()) {
             throw new TransformationFailedException(intl_get_error_message(), intl_get_error_code());
@@ -177,4 +182,47 @@ protected function isPatternDateOnly(): bool
         // check for the absence of time-related placeholders
         return 0 === preg_match('#[ahHkKmsSAzZOvVxX]#', $pattern);
     }
+
+    /**
+     * Normalizes various Unicode whitespace characters to regular ASCII spaces.
+     *
+     * ICU 72+ uses special Unicode whitespace characters (such as narrow no-break space U+202F)
+     * in formatted date strings. This method ensures consistent handling regardless of ICU version
+     * by normalizing these characters to regular ASCII spaces (U+0020).
+     */
+    private static function normalizeWhitespace(string $string): string
+    {
+        return str_replace([self::NO_BREAK_SPACE, self::NARROW_NO_BREAK_SPACE, self::THIN_SPACE], ' ', $string);
+    }
+
+    /**
+     * Parses a localized date string, handling ICU version differences in whitespace.
+     *
+     * ICU 72+ uses special Unicode whitespace characters (such as narrow no-break space U+202F)
+     * that users typically don't type. This method first tries parsing the input as-is, then
+     * tries with whitespace normalization to ensure compatibility across ICU versions.
+     *
+     * @throws TransformationFailedException When the input cannot be parsed
+     */
+    private function parse(\IntlDateFormatter $dateFormatter, string $value): int|float|false
+    {
+        try {
+            $timestamp = @$dateFormatter->parse($value);
+        } catch (\IntlException $e) {
+            throw new TransformationFailedException($e->getMessage(), $e->getCode(), $e);
+        }
+
+        // If parsing failed and the value contains regular spaces, try with ICU 72+ whitespace
+        if ((false === $timestamp || 0 !== intl_get_error_code()) && str_contains($value, ' ')) {
+            $icuValue = str_replace(' ', self::NARROW_NO_BREAK_SPACE, $value);
+
+            try {
+                $timestamp = @$dateFormatter->parse($icuValue);
+            } catch (\IntlException) {
+                // Ignore, we'll use the original error below
+            }
+        }
+
+        return $timestamp;
+    }
 }
diff --git a/src/Symfony/Component/Form/Tests/Extension/Core/DataTransformer/DateTimeToLocalizedStringTransformerTest.php b/src/Symfony/Component/Form/Tests/Extension/Core/DataTransformer/DateTimeToLocalizedStringTransformerTest.php
@@ -19,7 +19,6 @@
 use Symfony\Component\Form\Extension\Core\DataTransformer\BaseDateTimeTransformer;
 use Symfony\Component\Form\Extension\Core\DataTransformer\DateTimeToLocalizedStringTransformer;
 use Symfony\Component\Form\Tests\Extension\Core\DataTransformer\Traits\DateTimeEqualsTrait;
-use Symfony\Component\Intl\Intl;
 use Symfony\Component\Intl\Util\IntlTestHelper;
 
 class DateTimeToLocalizedStringTransformerTest extends BaseDateTimeTransformerTestCase
@@ -131,7 +130,7 @@ public function testTransformToDifferentLocale()
 
         $transformer = new DateTimeToLocalizedStringTransformer('UTC', 'UTC');
 
-        $this->assertMatchesRegularExpression('/^Feb 3, 2010, 4:05\s+AM$/u', $transformer->transform($this->dateTime));
+        $this->assertSame('Feb 3, 2010, 4:05 AM', $transformer->transform($this->dateTime));
     }
 
     public function testTransformEmpty()
@@ -235,15 +234,11 @@ public function testReverseTransformFullTime()
 
     public function testReverseTransformFromDifferentLocale()
     {
-        if (version_compare(Intl::getIcuVersion(), '71.1', '>')) {
-            $this->markTestSkipped('ICU version 71.1 or lower is required.');
-        }
-
         \Locale::setDefault('en_US');
 
         $transformer = new DateTimeToLocalizedStringTransformer('UTC', 'UTC');
 
-        $this->assertDateTimeEquals($this->dateTimeWithoutSeconds, $transformer->reverseTransform('Feb 3, 2010, 04:05 AM'));
+        $this->assertDateTimeEquals($this->dateTimeWithoutSeconds, $transformer->reverseTransform('Feb 3, 2010, 4:05 AM'));
     }
 
     public function testReverseTransformWithDifferentTimezones()
