Huge i18n bug that must be corrected

ResolvedModerator tobifjellner (Tor-Bjorn Fjellner)
(@tobifjellner)

1 year, 10 months ago

Hi,
I was just checking some pending translations (to Swedish) when I noted a case where your developers seems to have totally misunderstood how the plural handling functions (and can be used).
The source string pair contains: (singular) span; (plural) replace with

The function _n() should be used for sentences where there is a number and there is something that can be counted. For instance “We have deleted n files.
In this case, the _n() makes it possible to create a correct translation, despite the fact that various languages handle this kind of situations in very different ways. In some languages the rules are so complicated that, depending of the specific number, one out of five-six versions will be chosen in the target language. On the other hand, some languages don’t bother at all, they exactly the same words for 0, 1 or a million…

So you can’t use _n() as some kind of “handy matrix”.

Viewing 8 replies - 1 through 8 (of 8 total)

Plugin Author Rogier Lankhorst
(@rogierlankhorst)

1 year, 10 months ago
Hi @tobifjellner,

Thanks for your feedback.

What string(s) are you referring to? For example, looking at the example implementation here https://developer.wordpress.org/reference/functions/_n/:

printf( _n( '%s person', '%s people', $count, 'text-domain' ), number_format_i18n( $count ) );

Which is what I based this implementation on. But maybe I got it wrong?
printf(_n( '%s is located in Australia.', '%s are located in Australia.', cmplz_count_socialmedia(), 'complianz-gdpr' ), '[comma_socialmedia_on_site]')
- This reply was modified 1 year, 10 months ago by Rogier Lankhorst.
Plugin Author Rogier Lankhorst
(@rogierlankhorst)

1 year, 10 months ago
Or do you mean that the %s also has to be a number? Why would that be a problem?

Is there an alternative for _n which does not require the replacement to be a number?

Because these sentences all use something that can be counted, and therefore need a plural and a singular form. Looking for a solution I found this function, which gave the job nicely.

But always looking for improvements of course.
- This reply was modified 1 year, 10 months ago by Rogier Lankhorst.
Moderator tobifjellner (Tor-Bjorn Fjellner)
(@tobifjellner)

1 year, 10 months ago

Hi @rogierlankhorst

As you know, it’s not easy to post images here, but I’ll give you a link to the string that concerns me:
– https://translate.wordpress.org/projects/wp-plugins/complianz-gdpr/stable/sv/default/?filters%5Boriginal_id%5D=13065885

Here the “Singular” strings reads “span”, and the plural says “replace with”.

Unfortunately, they corresponding ref to the string is not very helpful: https://plugins.trac.wordpress.org/browser/complianz-gdpr/tags/6.1.6.1/assets/ace/keybinding-vim.js#L1

For instance Japanese doesn’t change words in sentences if an included number changes. So their translation of this string pair offers only one field, see: https://translate.wordpress.org/projects/wp-plugins/complianz-gdpr/stable/ja/default/?filters%5Boriginal_id%5D=13065885

Arabic is closer to the other end of the spectrum. Each string that uses _n() needs to be translated with 6 (six!) different strings, depending on which exact number is used as the argument.

So _n() should only ever be used with the number included in the string. There are some very popular packages out there (TGMPA, I’m looking at you) that use _n() just as a tool to split between one and “generic plural”, but that turns very wrong. In Russian, for instance, “generic plural, without use of numbers” is a different form than anything with numbers. Oh, and if you use numbers, then they use singular for 21, 31, 191 and many other numbers that end in “1”.

I’ve drafted an article about this that not yet has made its way into Polyglots handbook. I hope my moderator peers excuse me for linking to my own site: https://gsm.fjellner.com/hints-about-i18n-let-your-code-to-shine-all-over-the-world/

Plugin Author Rogier Lankhorst
(@rogierlankhorst)

1 year, 10 months ago

Ah. It looks like this string is part of the ace.js library. This is not code that was written by us, and I think it wasn’t even intended to be translated.

Not sure if I can fix that without removing the ace.js, but I’ll have a look.
Plugin Author Rogier Lankhorst
(@rogierlankhorst)

1 year, 10 months ago
Is there a way to exclude files from the WordPress translation parser? Otherwise I’m afraid it can’t be fixed, as it’s included in a third party .js file.

Anyway, this file isn’t used with React/Gutenberg so it won’t be an actual issue.
- This reply was modified 1 year, 10 months ago by Rogier Lankhorst.
Moderator tobifjellner (Tor-Bjorn Fjellner)
(@tobifjellner)

1 year, 10 months ago

Actually, there IS 🙂
See https://github.com/wp-cli/i18n-command/blob/b02ecdc9a57f9633740c254d19749118b7411f0f/src/MakePotCommand.php#L44 and you’ll figure out that if you make sure that the filename includes the term vendor then it won’t be scanned for strings.
Moderator tobifjellner (Tor-Bjorn Fjellner)
(@tobifjellner)

1 year, 10 months ago
ok. Sorry for attacking you for a chance coincidence between our string import scanners and an external library’s code. And I totally understand you had not idea what I was talking about 🙂

But for generic plural, you simply just cover it in your code:
pseudocode:
```
switch ($i) {
    case 0:
        echo "nothing here";
        break;
    case 1:
        echo "one comment";
        break;
    default:
        echo "wow. Lots of comments";
  endswitch;
```
Plugin Author Rogier Lankhorst
(@rogierlankhorst)

1 year, 10 months ago
Great! I’ll exclude the files then. Thanks!
- This reply was modified 1 year, 10 months ago by Rogier Lankhorst.

Viewing 8 replies - 1 through 8 (of 8 total)

The topic ‘Huge i18n bug that must be corrected’ is closed to new replies.