Commas should be internationalized

[huji]

There are many places where commas should be internationalized. A very typical example is this: You have to separate tags using "," (Latin comma). RTL languages like Arabic and Persian use a different character for comma ("،") so there should be away to internationalize it let Wordpress understand it correctly, per content language.

[DD32]

The theme controls the seperators, There is no reason why a given theme cannot internationalise itself, or the theme author internationalise the theme.
A user can also replace the characters in question in their choice of theme if they choose.

[huji]

Correct; however, at least the default theme should support this.

[nbachiyski]

Reopen if there are places, where the text is i18n-ed and a coma can't be translated.

[huji]

I have to reopen it. It was not about themes in the first place. It was about dashboard which is i18n-ed. When you right a new post, you can add tags, but you have to separate them with Latin commas; if the interface is in a non-Latin language, there should be a way to use non-Latin commas instead.

[janeforshort]

Punting due to feature freeze. Reconsider with next release.

[AhmadSherif]

The patch assume that there's a message called "," which contains the localized comma.

Two notes on that patch:

• The patch doesn't contain diff for wp-admin/js/post.js 'cause I don't know how you guys are doing that compression stuff.

• Some changes in post-dev.js might look ugly (or it's ugly), but I made it that way to ease any possible debugging.

[westi]

This looks like a worthwhile fix.

Couple of questions/observations:

• Why do we need the evals() in the javascript code?
• It looks like there are still some hardcoded commas in the javascript code as well.

[nacin]

We should be able to yank the code in question in westi's comment, this looks pretty good otherwise.

[Denis-de-Bernardy]

punting pending new patch

[uadroid]

Is there any progress on this? I'm using wordpress.com and I have to switch my keyboard to Latin when ever I want to separate tags.

It's 'ARABIC COMMA' (U+060C)

[nacin]

7897.diff​ approaches this holistically. It works for bulk edit, quick edit, and the tags meta box (including without JS). It also works for both an internationalized comma and the existing standard comma, intermingled, in case they are different, or if force of habit causes someone to continue to use the standard comma.

Standardization occurs in very specific places.

Rendering tag strings (PHP displays):

• get_terms_to_edit(), which joins tags with a comma and escapes the result, is not modified. This is for compatibility reasons with plugins, who may expect to do a PHP explode or JS split on this data.

• get_inline_data(), which uses get_terms_to_edit(), is not modified. It currently does a replacement on the result from get_terms_to_edit(), adding a space after each comma. Again, this is not modified for compatibility reasons.

• What is modified is the tags meta box code. In this case, the terms string is going directly into a textarea. So, we find all commas and replace it with the internationalized comma, along with a trailing space. (This did not occur in 3.3, resulting in "apple,pear,orange" with no JS.)

Saving tag strings (PHP saves):

• wp_set_post_terms() is where the explode-by-commas may take place. It will now first find all L10n commas and replace them with standard commas before continuing with the explode. This is done in part to allow both commas to work, and in part because we can't reliably trim() an html entity as well as we can trim a leading or trailing comma.

• bulk_edit_posts() is where the other explode-by-commas takes place on save. It will now do the same thing as wp_set_post_terms(), improving its cleaning regex in the process.

Working with tag strings (in JS):

• In post.dev.js, care is taken to properly clean/sanitize the new string. Since the comma is sometimes used in a character class (tagBox.clean), and an html entity could cause problems here, the L10n comma is first replaced with a standard comma, and then switched back when done.

• In inline-edit-post.dev.js, there is a find-replace for converting the standard comma to the L10n comma. This is because it is relying on a standard comma-separated term string from get_inline_data() (see above for more).

• In post.dev.js and inline-edit-post.dev.js, the only other changes are to specify the localized separator for tag suggestions, and properly split or join on the new localized comma.

Tag suggestions:

• In ajax-tag-search, L10n commas are first replaced with standard commas before exploding.

This was the code I was using to test:

add_filter( 'gettext_with_context', 'nacin_test_l10n_comma', 10, 3 );
function nacin_test_l10n_comma( $translated, $text, $context ) {
	if ( $text == ',' && $context == 'tag delimiter' )
		return '`';
	return $translated;
}

Let's soak this in trunk.

[nacin]

In [19853]:

Allow localized commas to be used as tag separators. see #7897.

[nacin]

It should be noted that there is an assumption made about localized commas — that, like the standard comma, they use word-comma-space-word.

I'll be looking into this with the translators to see if this needs to be modified. It will not take much, only a new translated string that is leveraged in a few places.

[nacin]

The existing locales that have translated ", " (with space) do include a trailing space after their character. Arabic, Persian, Burmese, Chinese (Taiwan), Chinese (China), Uighur, and Kurdish.

[nacin]

Pending bug reports, closing as fixed.