[wp-trac] [WordPress Trac] #39210: switch_to_locale() unloads all plugin and theme translations
WordPress Trac
noreply at wordpress.org
Thu Oct 20 12:13:17 UTC 2022
#39210: switch_to_locale() unloads all plugin and theme translations
-------------------------------------------------+-------------------------
Reporter: gchtr | Owner: ocean90
Type: defect (bug) | Status: reopened
Priority: normal | Milestone: 6.1
Component: I18N | Version: 4.7
Severity: normal | Resolution:
Keywords: has-unit-tests has-patch needs-dev- | Focuses:
note |
-------------------------------------------------+-------------------------
Comment (by flixos90):
Replying to [comment:67 ocean90]:
> Replying to [comment:66 flixos90]:
> > * When using the en_US locale, `load_textdomain()` is (unnecessarily)
called a ton of times (107 times in WP 6.1-RC2 vs 3 times in WP 6.0.3).
>
> Assuming you have no other plugins active, could you explain what calls
these are? And why there are unnecessary?
I used a vanilla WordPress site with no plugins and Twenty Twenty-One,
with just the initial Hello World post. Only the default and
twentytwentyone text domains are used in this context, but the commit here
results in load_textdomain() to be called 107 times in one page load, even
though all the site is en_US (no locale switching happening at all). So
this is clearly a bug that needs to be addressed.
> > * The regression is only present on en_US sites. For sites that use a
locale that typically has translation files present, performance remains
basically as before.
>
> That's actually a good point since the changes here will significantly
improve the translation handling
[https://wordpress.org/about/stats/#locales for the majority of our user
base]. ''If'' there's some tradeoff this should also be taken into
consideration.
It makes sense that this is a trade off. I would like to see some data
though under which conditions the commit improves translation handling and
how (much).
> I think for the moment we should identify what the problem actually is
and if it's not something that can be fixed before the release and/or in a
follow-up release.
As mentioned, I'm open to fixing this prior to release if feasible. But
the data I shared proves the regression coming from the commit. Without
having opposing data that would quantify any performance ''benefit'' for
non-en_US sites, this either needs to be fixed or reverted prior to
release.
--
Ticket URL: <https://core.trac.wordpress.org/ticket/39210#comment:70>
WordPress Trac <https://core.trac.wordpress.org/>
WordPress publishing platform
More information about the wp-trac
mailing list