Add missing translated strings

[frjo]

(This contribution was sponsored by DigitalHub.sh)

Fixes #4753

I've split this PR over several commits which are mostly independent of each other. I'm happy to open separate PRs if you think that'd be easier to review.

Explanations of the separate commits

Mark missing strings for translation

After a few unsuccessful attempts at scripting the identification of untranslated strings, I ended up reviewing all Python files manually.

The command I used to identify the relevant files was:

git ls-files -z '**.py' ':!**/migrations/*.py' ':!**/urls.py' ':!hypha/settings/*.py' ':!docs/*' ':!**/test*.py' | find -files0-from - -not -empty

(That's all Python files, but excluding stuff like migrations, settings, ... (and empty files))

This change should have no impact for a site running the English version of Hypha.

Add missing strings for translations (templates)

Similar to the previous commit, I ended up reviewing all templates manually (git ls-files '*.html'), adding {% trans %}, {% blocktrans %}, or _(...) where needed.

Again, this change should have no impact.

Add trimmed option to blocktranslate

This removes irrelevant whitespace in source strings, making it a bit nicer.

I used git grep blocktrans | grep -vw trimmed | grep -v endblocktrans to identify uses of {% blocktrans %} without the trimmed argument, excluding lines that also included {% endblocktrans %} on the same line.

This is technically a breaking change, but it only introduces whitespace changes to HTML templates, which is irrelevant.

Use {% blocktranslate trimmed %} to break down long lines

This one just makes templates a little bit more readable as it breaks very long lines into more manageable pieces.

Again, it can introduce whitespace changes to the HTML, but should not have any effect to the source strings.

Added verbose_name and verbose_name_plural to all models

Some models already has a translated verbose_name or verbose_name_plural, but it was inconsistent (and quite rare). This change adds these attributes to all models.

I used git ls-files '**/models/*.py' and git ls-files '**/models.py' to find all models, then went through the files manually to add the missing verbose_name and verbose_name_plural.

I decided to omit the plural form for model classes inheriting from BaseSiteSettings since it didn't seem to make sense.

I also opted to use lower case (like _("foo bar") for model FooBar) to match Django's own default behavior, but I used Foo Bar in a few instances where the model already had a verbose_name that used titlecase.

Next steps

I ran out of time to complete all my objectives and I've still got some translation-related tasks I wanted to get to. I'm happy to open issues for these if you think they're relevant:

• Javascript files include non-translated user-facing strings #4752 Fix missing translations in javascript files: I found about 10 instances but fixing them would require either using the Django javascript translation functions, or moving the strings to the HTML (data attributes for example).
• Model and form fields should have a translated verbose_name/label #4753 Add explicit verbose_name to all declared model fields (and maybe form fields as well). This can probably be scripted somewhat, ideally with a linter that would catch the introducing of future untranslated strings.
• Review cases where gettext is used at the module level instead of gettext_lazy as this could indicate some possible bugs (visible in a multi-language setup where the user is not using the default language).
• Fix translations that build up a sentence word by word, as this doesn't really work in practice. Something like {% trans "Application" %} {{ application_id }} {% trans "updated on" %} {{ date }} {% trans "by" %} {{ author.name }} should be rewritten to have the whole sentence in a single blocktrans, or when not possible small words like by should be given a context because they're likely to be translated differently for different sentences.

[frjo]

I created a new PR on a Hypha repo branch so I could rebase it. Thanks @bmispelon for all the work on this one!

[frjo]

@wes-otf Please have a look through the changes. The PR touches many files and creates many migrations (verbose_name on fields) but should not change anything in how Hypha works. There is one minor bugfix I noticed, correcting to PermissionDenied.

When you also looked at this and do not find anything that stands out you can merge it in.

This makes many more strings translatable in Hypha! Thanks to @bmispelon and DigitalHub.sh.

After this is in I will update all the po files so translations can be updated/created.

With the resent rapid improvements in LLM:s I think they could do a decent job of creating ruff translation of Hypha. I think it will be easier for a translator to correct an LLM translation that start from scratch and translate nearly 5000 strings.

[wes-otf]

@frjo I did a rough check of it when it first dropped and things seemed pretty good but I'll heavily test it again today and make sure we're in the clear - definitely a fantastic improvement though!