[JSC] Date parse should accept narrow-no-break-space

https://bugs.webkit.org/show_bug.cgi?id=247986
rdar://102427399

Reviewed by Alexey Shvayka.

After ICU 72, CLDR generates narrow-no-break-space for date time output. So, `new Date().toLocaleString('en-US')` starts generating
a string including narrow-no-break-space instead of simple spaces. However since code in the wild assumes `new Date(new Date().toLocaleString('en-US'))`
works, we need to maintain the ability to parse string including narrowNoBreakSpaces. Rough consensus among implementaters (V8, SpiderMonkey, and us)
is replacing narrowNoBreakSpaces with simple spaces before parsing.

* JSTests/complex.yaml:
* JSTests/complex/date-parse-with-narrow-no-break-space.js: Added.
(shouldBe):
* Source/JavaScriptCore/runtime/JSDateMath.cpp:
(JSC::DateCache::parseDate):
* Source/WTF/wtf/unicode/CharacterNames.h:

Canonical link: https://commits.webkit.org/256754@main

JSTests/complex.yaml

@@ -64,3 +64,6 @@
 
 - path: complex/intl-date-time-format-date-parse.js
   cmd: runComplexTest [], [], "TZ=America/New_York"
+
+- path: complex/date-parse-with-narrow-no-break-space.js
+  cmd: runComplexTest [], [], "TZ=America/Los_Angeles"

JSTests/complex/date-parse-with-narrow-no-break-space.js

@@ -0,0 +1,11 @@
+function shouldBe(actual, expected) {
+    if (actual !== expected)
+        throw new Error('bad value: ' + actual);
+}
+var space = ' ';
+var narrowNoBreakSpace = '\u202f';
+
+shouldBe(new Date(`11/16/2022, 10:33:58${space}AM`).getTime(), 1668623638000);
+shouldBe(new Date(`11/16/2022, 10:33:58${narrowNoBreakSpace}AM`).getTime(), 1668623638000);
+shouldBe(Date.parse(`11/16/2022, 10:33:58${space}AM`), 1668623638000);
+shouldBe(Date.parse(`11/16/2022, 10:33:58${narrowNoBreakSpace}AM`), 1668623638000);

Source/JavaScriptCore/runtime/JSDateMath.cpp

@@ -76,6 +76,7 @@
 #include "VM.h"
 #include <limits>
 #include <wtf/Language.h>
+#include <wtf/unicode/CharacterNames.h>
 #include <wtf/unicode/icu/ICUHelpers.h>
 
 #if U_ICU_VERSION_MAJOR_NUM >= 69 || (U_ICU_VERSION_MAJOR_NUM == 68 && USE(APPLE_INTERNAL_SDK))
@@ -317,7 +318,14 @@ double DateCache::parseDate(JSGlobalObject* globalObject, VM& vm, const String&
 
     if (date == m_cachedDateString)
         return m_cachedDateStringValue;
-    auto expectedString = date.tryGetUTF8();
+
+    // After ICU 72, CLDR generates narrowNoBreakSpace for date time format. Thus, `new Date().toLocaleString('en-US')` starts generating
+    // a string including narrowNoBreakSpaces instead of simple spaces. However since code in the wild assumes `new Date(new Date().toLocaleString('en-US'))`
+    // works, we need to maintain the ability to parse string including narrowNoBreakSpaces. Rough consensus among implementaters is replacing narrowNoBreakSpaces
+    // with simple spaces before parsing.
+    String updatedString = makeStringByReplacingAll(date, narrowNoBreakSpace, space);
+
+    auto expectedString = updatedString.tryGetUTF8();
     if (!expectedString) {
         if (expectedString.error() == UTF8ConversionError::OutOfMemory)
             throwOutOfMemoryError(globalObject, scope);

Source/WTF/wtf/unicode/CharacterNames.h

@@ -89,7 +89,6 @@ constexpr UChar lowDoublePrimeQuotationMark = 0x301F;
 constexpr UChar lowLine = 0x005F;
 constexpr UChar minusSign = 0x2212;
 constexpr UChar narrowNoBreakSpace = 0x202F;
-constexpr UChar narrowNonBreakingSpace = 0x202F;
 constexpr UChar newlineCharacter = 0x000A;
 constexpr UChar noBreakSpace = 0x00A0;
 constexpr UChar nullCharacter = 0x0;
@@ -206,7 +205,6 @@ using WTF::Unicode::leftToRightOverride;
 using WTF::Unicode::lowLine;
 using WTF::Unicode::minusSign;
 using WTF::Unicode::narrowNoBreakSpace;
-using WTF::Unicode::narrowNonBreakingSpace;
 using WTF::Unicode::newlineCharacter;
 using WTF::Unicode::noBreakSpace;
 using WTF::Unicode::nullCharacter;