Bug 76714 – implement Date.toLocaleString() using ICU

• Component: JavaScriptCore
• Status: RESOLVED FIXED

[Eli Fidler]

Some platforms don't localize strftime well, so use ICU to implement Date.toLocaleString()

[Eli Fidler]

Created attachment 123341 [details]
Patch

[Darin Adler]

Comment on attachment 123341 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=123341&action=review

> Source/JavaScriptCore/runtime/DatePrototype.cpp:210
> +    UDateFormat* df = udat_open(timeStyle, dateStyle, 0, 0, -1, 0, 0, &status);

Since the point of these functions is to format for the current locale, it seems critical to pass in a suitable locale string. See, for example, what is done for text break locales in the currentTextBreakLocaleID function.

I’m surprised that is OK to pass 0 for the locale string.

[Eli Fidler]

0 is the "default locale". We rely on ICU for localization in many areas, so apps will have the default locale set appropriately to the desired locale.

I guess that's not true for OSX or IOS, but there's a different path here for that.

I could put the strftime path into HAVE(STRFTIME) or make a USE(ICU_DEFAULT_LOCALE) if you prefer.

[Darin Adler]

(In reply to comment #3)
> 0 is the "default locale"

Is that a general ICU feature, or one specific to RIM platforms? I ask because I don’t see it mentioned in ICU documentation.

> We rely on ICU for localization in many areas

Sure, OS X and iOS do too. My question is about how the ICU API works.

> I could put the strftime path into HAVE(STRFTIME) or make a USE(ICU_DEFAULT_LOCALE) if you prefer.

Since this code passes 0 for the locale it should be used only on platforms where that will do the right thing. To get the #if right we need to know how where that “0 means default locale” feature exists.

[Alexey Proskuryakov]

This may not be a regression from strftime, but the behavior would not be correct when locale changes at runtime, especially with WebKit2.

In Mac Safari, output of (new Date).toLocaleString() changes immediately after a user changes the format in system preferences.

[Eli Fidler]

NULL for default locale is an ICU-standard feature: "Using the ICU C functions, NULL can be passed for a locale parameter to specify the default locale."

from: http://userguide.icu-project.org/locale#TOC-Default-Locales


For Alexey's point, Date.toLocaleString() will always use the current default locale. If that is changed using uloc_setDefault(), the new value should be immediately take effect.

[WebKit Review Bot]

Comment on attachment 123341 [details]
Patch

Rejecting attachment 123341 [details] from commit-queue.

New failing tests:
media/audio-garbage-collect.html
Full output: http://queues.webkit.org/results/11342538

[Eli Fidler]

flaky test looks unrelated, so let's try cq+ again

[WebKit Review Bot]

Comment on attachment 123341 [details]
Patch

Clearing flags on attachment: 123341

Committed r105939: <http://trac.webkit.org/changeset/105939>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Mark Rowe (bdash)]

This broke all of the Mac builds on build.webkit.org as unicode/udat.h is not available.

[Mark Rowe (bdash)]

I checked in a fix in r105943 to try and fix the Mac builds.

[Alexey Proskuryakov]

This has reportedly caused a regression, bug 80262.