Closed Bug 1423138 Opened 5 years ago Closed 5 years ago

[cy] Switch plural rule from #1 (2 forms) to #18 (6 forms)

Categories

(Mozilla Localizations :: cy / Welsh, defect)

Product:
Mozilla Localizations
Component:
cy / Welsh
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

(Not tracked)

Status:
RESOLVED FIXED

People

(Reporter: flod, Assigned: rprys)

References

Details

Welsh has 6 plural forms according to CLDR
http://www.unicode.org/cldr/charts/dev/supplemental/language_plural_rules.html#cy

But Firefox and Firefox for Android are set to use the plural form 1, which means 2 plural forms (like English).

Translate House has 4 forms
http://docs.translatehouse.org/projects/localization-guide/en/latest/l10n/pluralforms.html

In comparison, Irish uses 5 forms
https://developer.mozilla.org/en-US/docs/Mozilla/Localization/Localization_and_Plurals

At this point I have no clue which form is correct, I only know for sure that the one used in Firefox is wrong.

Once change, about 50/60 strings will need to be fixed.
Flags: needinfo?(rprys)
It's complicated and depends what you're asking for...
Delyth is offering you a tutorial on the fringes of Y'all Hands next week
:-)
Flags: needinfo?(rprys)
(In reply to Rhoslyn Prys from comment #1)
> Delyth is offering you a tutorial on the fringes of Y'all Hands next week

That sounds like a great idea :)
I fixed the following errors.

Locale: cy (5)
- charts.totalSecondsNonBlocking (devtools/client/netmonitor.properties) has 3 plural forms (requested: 2)
- shortSeconds (toolkit/chrome/mozapps/downloads/downloads.properties) has 4 plural forms (requested: 2)
- shortMinutes (toolkit/chrome/mozapps/downloads/downloads.properties) has 4 plural forms (requested: 2)
- shortHours (toolkit/chrome/mozapps/downloads/downloads.properties) has 4 plural forms (requested: 2)
- shortDays (toolkit/chrome/mozapps/downloads/downloads.properties) has 4 plural forms (requested: 2)

We'll still need to figure out what's the best rule for Welsh.
Hi, the correct rule is the six-category one in the CLDR.

From a computational perspective, there are six categories but no single word distinguishes all six categories. That is why the CLDR minimal pairs use both "ci" (dog) and "cath" (cat) at the same time. (Disclaimer: I wrote those examples for Mark Davis many years ago).

The confusion arises because, from a traditional grammar perspective, the form changes that mark agreement with the numbers are not called "plural forms" but "mutations". So Welsh speakers would never say that Welsh "has six plural forms". But to generate translations that Welsh speakers regard as correct for any number, the six-category rule must be used.
Oh, the Translate House data is a very old version of the CLDR data which was incorrect for Welsh (e.g. you can see it in CLDR 1.8.1). So it should be regarded as a pre-correction copy, and not as an authority in itself.
Thanks David, this sounds reasonable to me.

@Rhoslyn
Any concern in moving forward with 6 forms? You'll need to verify about plural 50 strings

Pontoon is also set to use 4 forms right now, missing zero and many, and it doesn't look like something that can be automated.
Flags: needinfo?(rprys)
No, I accept David'a recommendation.
No problem with dealing with the changes - will they appear in Pontoon as fuzzy?
Flags: needinfo?(rprys)
(In reply to Rhoslyn Prys from comment #7)
> No problem with dealing with the changes - will they appear in Pontoon as
> fuzzy?

No, you'll need to go through them for Firefox, Firefox for Android, etc. It's not as painful as it sounds, since you can search all plural strings and show them in one screen.

I'm afraid it's going to be more complicated for .po based products and Pontoon, but we can deal with that later.

First of all I need to patch our pluralform system, since we don't have any rule similar to this. I'll ping you as soon as we're ready on the code side.
Depends on: 1428366
Blocks: 1429072
(In reply to Francesco Lodolo [:flod] from comment #9)
> @Rhos
> I've switched the plural rule in Gecko to 18.

Forgot the link: https://hg.mozilla.org/l10n-central/cy/rev/5933d78c2433bc4063b828a1880143bf46a9cf0d
Ok, I've updated the strings. Thanks.
(In reply to Rhoslyn Prys from comment #11)
> Ok, I've updated the strings. Thanks.

At first glance, I don't see anything updated in Firefox or Fennec in Pontoon.

I only see one landing in Mercurial for Thunderbird, and it's incorrect.
https://hg.mozilla.org/l10n-central/cy/rev/b30b368f8094

glodaFacetView.facets.mode.top.listAllLabel=Rhestru'r #1i gyd;Rhestru'r #1 i gyd;

There should be six forms, each one separated by a semicolon

stringID = form1;form2;form3;form4;form5;form6

The string above has 3 form, and the 3rd form is empty, since the string ends with a semicolon.
OK... Like this?
Mae #1 tab wedi cyrraedd o #2.;Mae #1 tab wedi cyrraedd o #2;Mae #1 tab wedi cyrraedd o #2;Mae #1 tab wedi cyrraedd o #2;Mae #1 tab wedi cyrraedd o #2;Mae #1 tab wedi cyrraedd o #2
(In reply to Rhoslyn Prys from comment #13)
> OK... Like this?
> Mae #1 tab wedi cyrraedd o #2.;Mae #1 tab wedi cyrraedd o #2;Mae #1 tab wedi
> cyrraedd o #2;Mae #1 tab wedi cyrraedd o #2;Mae #1 tab wedi cyrraedd o
> #2;Mae #1 tab wedi cyrraedd o #2

Exactly, remembering the the forms are used in this order:

- 0
- 1
- 2
- 3
- 6
- anything else
Does this follow the same pattern? The first part looks like - 1
#1 has been installed successfully.;#2 add-ons have been installed successfully.
(In reply to Rhoslyn Prys from comment #15)
> Does this follow the same pattern? The first part looks like - 1
> #1 has been installed successfully.;#2 add-ons have been installed
> successfully.

That's an old string that was poorly done, and it's been fixed in Firefox 59. 

Ignore the first form, translate everything as "#2 add-ons have been installed"
Does this follow the same pattern?
Delete Page;Delete Pages
(In reply to Rhoslyn Prys from comment #17)
> Does this follow the same pattern?
> Delete Page;Delete Pages

Yes. That's another poor string, they should always use the number explicitly.

Anyhow, you still need 6 forms in Welsh, e.g.

Dileu Tudalen;Dileu Tudalennau;Dileu Tudalennau;Dileu Tudalennau;Dileu Tudalennau;Dileu Tudalennau
e;e
s;s
m;m
d;d
?
Sadly, you need to repeat it 6 times

e;e;e;e;e;e

The idea is that the code doesn't know if the string is dumb (like this) or more complex: it's going to search for 6 forms for Welsh, falling back to the first form and log an error if it doesn't find the one it's searching for.

The plural is there to accommodate for locales that won't use a single character as unit of measurement.
I'm a bit confused by this type of string - See one result;See all #1 results. 

If:

- 0
- 1
- 2
- 3
- 6
- anything else

Then the first part appears to be - 1. Where is - 0? How do I deal with - 0?
In English:

1st form: 1
2nd form: 0, anything else

So it uses the 2nd form, "See all 0 results". The code should probably know not to show this specific string for 0, since it doesn't make a lot of sense.

In case of Welsh, you have a specific case for 0 that you still need to add explicitly. So, you will end up with something similar to:

> Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad

The 1st form is used for 0. You could also translate this as

> ;Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad

Notice the starting semicolon? It means that there's a 1st form, and it's empty. Or even:

> No results;Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad;Gweld y #1 canlyniad

Up to you what you prefer, I'd just go with the full 6 forms for sake of simplicity.

Side note: remember that the semicolon is just a separator, not proper punctuation. For example:

> Gweld y canlyniad;gweld y #1 canlyniad

The second sentence should start uppercase too.
Thanks, I'll go with the full 6 forms to avoid confusion. I'm easily confused. ;-)
Please do not change this string: the correct value is 18 (number of rule, not number of forms)
https://pontoon.mozilla.org/cy/firefox/toolkit/chrome/global/intl.properties/?search=pluralRule&string=81020
These are the remaining errors:
- addonsInstalledNeedsRestart (browser/chrome/browser/browser.properties) has 2 plural forms (requested: 6)
- addonsGenericInstalledNeedsRestart (browser/chrome/browser/browser.properties) has 5 plural forms (requested: 6)
- removeContainerMsg (browser/chrome/browser/preferences/preferences.properties) has 5 plural forms (requested: 6)
- tabs.closeWarningMultiple (browser/chrome/browser/tabbrowser.properties) has 7 plural forms (requested: 6)
- crashesTitle (toolkit/chrome/global/aboutSupport.properties) has 5 plural forms (requested: 6)
- webextPerms.hostDescription.tooManySites (mobile/android/chrome/browser.properties) has 5 plural forms (requested: 6)

I've already fixed these two (they had double ;):
- showAllSearchResults (toolkit/chrome/mozapps/extensions/extensions.properties) has 9 plural forms (requested: 6)
- webrtcIndicator.sharingWindowWithNTabs.menuitem (browser/chrome/browser/webrtcIndicator.properties) has 8 plural forms (requested: 6)
Thanks, I've checked the above and corrected them - these however appeared to be correct ???

- addonsGenericInstalledNeedsRestart (browser/chrome/browser/browser.properties) has 5 plural forms (requested: 6)
- removeContainerMsg (browser/chrome/browser/preferences/preferences.properties) has 5 plural forms (requested: 6)
- tabs.closeWarningMultiple (browser/chrome/browser/tabbrowser.properties) has 7 plural forms (requested: 6)
(In reply to Rhoslyn Prys from comment #26)
> Thanks, I've checked the above and corrected them - these however appeared
> to be correct ???
> - addonsGenericInstalledNeedsRestart
> (browser/chrome/browser/browser.properties) has 5 plural forms (requested: 6)

There was still a period instead of semicolon between form 2 and 3. I've replaced it.

> - removeContainerMsg
> (browser/chrome/browser/preferences/preferences.properties) has 5 plural
> forms (requested: 6)

No separator between form 1 and 2 ('… Cynhwysydd hwn?Os fyddwch yn …'), added it.

> - tabs.closeWarningMultiple (browser/chrome/browser/tabbrowser.properties)
> has 7 plural forms (requested: 6)

This one looks correct now. Before, since it started with ";", it had an empty form 1.
(In reply to Francesco Lodolo [:flod] from comment #25)
> These are the remaining errors:
> - addonsInstalledNeedsRestart (browser/chrome/browser/browser.properties)
> has 2 plural forms (requested: 6)

This one still has 2 forms. While it's unused in 59, it would be good to fix for 58, translating only the second part (as per comment 16)
https://pontoon.mozilla.org/cy/firefox/browser/chrome/browser/browser.properties/?search=addonsInstalledNeedsRestart&string=74463

Side question: other projects based on other formats (Gettext, Fluent). Right now they're using 4 forms, we need to move them to 6. Would it be safe to copy one of existing values into the 2 missing forms? Should those strings be marked as suggested/fuzzy to be double checked?

For example: "x minutes ago"
https://pontoon.mozilla.org/cy/firefox-screenshots/server.ftl/?search=ago&string=168505
Ok I've corrected the issues you outlined, I think they all now have 6 forms. 

In the meanwhile I've received a useful e-mail from David outlining some further changes in relation to mutations and making the initial form more natural and less a variation of forms 2-6. I'll be giving those some attention over the next few days, if that's OK.

Side question - yes, they need to change to 6 and having the changes marked as suggested/fuzzy would enable me to identify and check them.
(In reply to Rhoslyn Prys from comment #29)
> Ok I've corrected the issues you outlined, I think they all now have 6
> forms. 

Thanks. I can confirm that there are no more errors for Firefox and Firefox for Android.

The dashboard will start showing wrong number plurals as warnings in the near future, so they will be easier to spot in all products, also for new strings.

> In the meanwhile I've received a useful e-mail from David outlining some
> further changes in relation to mutations and making the initial form more
> natural and less a variation of forms 2-6. I'll be giving those some
> attention over the next few days, if that's OK.

Yes, that's absolutely OK.

> Side question - yes, they need to change to 6 and having the changes marked
> as suggested/fuzzy would enable me to identify and check them.

OK. Let's close this one and wait for bug 1423679 to fix that for other projects.
Assignee: nobody → rprys
Blocks: 1423679
Status: NEW → RESOLVED
Closed: 5 years ago
Resolution: --- → FIXED
For now please ignore warnings reported in the dashboard for plural rules, everything is correct
https://l10n.mozilla.org/teams/cy

We need to release a new version of compare-locales that knows about this bug. It should happen in a few days.
(In reply to Francesco Lodolo [:flod] from comment #32)
> For now please ignore warnings reported in the dashboard for plural rules,
> everything is correct
> https://l10n.mozilla.org/teams/cy

OK, new version of compare-locales is up, and errors reported from now on are actual errors ;-)
https://l10n.mozilla.org/dashboard/compare?run=898422#issue1

String: https://pontoon.mozilla.org/cy/firefox/toolkit/chrome/global/findbar.properties/?search=FoundMatchesCountLimit&string=80344
findbar.properties
↓
expecting 6 plurals, found 5 at line 22, column 24 for FoundMatchesCountLimit

Ok, I fixed that one about an hour ago, it should be correct now - I've just counted 6.
I've also included suggestions by David on the plural forms.
Confirming that now it's fixed.
No longer blocks: 1423679
Depends on: 1423679
You need to log in before you can comment on or make changes to this bug.