Twenty Twelve and Twenty Fifteen: maintain font subset translations

[sabernhardt]

Changesets [55277] and [55268] removed font subset translations from the font URL functions because the subsets do not affect the new URLs. Google Fonts had supported all character sets regardless of the URL query string, and the new stylesheets also serve any character sets (without a query string).

However, developers may have a use for the subset information in a custom stylesheet or in a preload resource link.

Creates new functions to output the string from no-subset translations that previously were included in Twenty Twelve and Twenty Fifteen.

Trac 57807

[sabernhardt]

I first tried ​returning an array with Latin characters, including any extended subsets, but simpler may be better. Returning the string (translation) lets developers decide exactly what to do with the output. Someone who wants an array could build one from the string.

I do not think creating similar functions in the other themes is very valuable. If I'm wrong about that, we could open another ticket to add them in a future release.

This is not a bug ticket because it would create new functions. Maybe it could be a task for 6.2, though.

[sabernhardt]

I was worried about possibly losing the translations, but checking known language codes might be better anyway.

Looking at no-subset ​translations again, the collection only seems reliable for a few—yet not all—Cyrillic languages. Most translations are either the default no-subset or else invalid translations of the word instead of the setting (including Greek and Vietnamese).

Valid, non-default translations:

• Twenty Twelve has cyrillic for eight languages: Belarusian, Bengali, Korean(?), Kyrgyz, Russian, Serbian, Ukrainian and Uzbek. Belarusian, Bengali and Korean turned off the Open Sans font anyway.
• Twenty Fifteen has devanagari for Bengali and cyrillic for four languages: Belarusian, Bulgarian, Russian and Serbian. Bengali and Belarusian disable all the packaged theme fonts; Russian and Serbian only turn off Inconsolata.

Notes:

• Devanagari was not an option for Twenty Twelve's Open Sans, but these proposed functions are not only for sites with the default font selection(s).
• Irish and Lithuanian specify latin-ext, which technically was invalid in the original stylesheet functions, though it might have some use with these.

[sabernhardt]

The dev note for #55985 could recommend using something like Twenty Twenty's language code collection for any of the six themes edited on that ticket instead of having a separate method for these two.

Side note: Twenty Twenty should be updated to include at least Persian (Afghanistan) and Uzbek.

[sabernhardt]

Example of a language code array (for themes without these proposed functions):

<?php
$script_families = array(
	// Cyrillic.
	'bel'   => 'cyrillic',
	'bg-BG' => 'cyrillic',
	'kk'    => 'cyrillic',
	'mk-MK' => 'cyrillic',
	'mn'    => 'cyrillic',
	'ru-RU' => 'cyrillic',
	'sah'   => 'cyrillic',
	'sr-RS' => 'cyrillic',
	'tt-RU' => 'cyrillic',
	'uk'    => 'cyrillic',
	'uz-UZ' => 'cyrillic',

	// Devanagari.
	'bn-BD' => 'devanagari',
	'hi-IN' => 'devanagari',
	'mr'    => 'devanagari',
	'ne-NP' => 'devanagari',

	// Greek.
	'el'    => 'greek',

	// Hebrew.
	'he-IL' => 'hebrew',

	// Vietnamese.
	'vi'    => 'vietnamese',
);
$locale = get_bloginfo( 'language' );
if ( array_key_exists( $locale, $script_families ) ) {
	// Do something with the special script family name.
	echo $script_families[$locale];
}

[sabernhardt]

The no-subset translations are not included in the PO files anymore for new sites. It probably is not worth restoring them now.