When we were visiting a page showing the content of a record which uses
globalize and our locale was the default one and there was no
translation for the default locale, the application was crashing in some
places because there are no fallbacks for the default locale.
For example, when visiting a legislation process, the line with
`CGI.escape(title)` was crashing because `title` was `nil` for the
default locale.
We've decided to solve this issue by using any available translations as
globalize fallbacks instead of showing a 404 error or a translation
missing error because these solutions would (we thinkg) either require
modifying many places in the application or making the translatable
logic even more complex.
Initially we tried to add this solution to an initializer, but it must
be called after initializing the application so I18n.fallbacks[locale]
gets the value defined in config.i18n.fallbacks.
Also note the line:
fallbacks[locale] = I18n.fallbacks[locale] + I18n.available_locales
Doesn't mention `I18n.default_locale` because the method
`I18n.fallbacks[locale]` automatically adds the default locale.
Having translations makes this validation too complex and not worth the
effort, since now we can't just scope in a single column but we need to
scope in the translations locale and the question ID.
This way we can show/hide that div when displaying translations, and we
can remove the duplication applying the same logic to the label, the
input, the error and the CKEditor.
This way we also solve the problem of the textarea of the CKEditor
taking space when we switch locales, as well as CKEditor itself taking
space even when not displayed.
In order to be consistent with similar specs, the I18nContent#begins_with_key
spec was improved by explicitly specifying an I18n key for one (1) factory
that relied on its default value
We needed to bring back support for CKEditor in our translatable form,
which we had temporarily remove.
And now we support CKEditor in our translatable specs, and so we can
remove the duplicated specs for poll question answers.
We need to replace ".title=" by ".title_#{locale}=" in one place because
for some reason globalize builds a new translation record when using the
latter but it doesn't build one when using the former.
Updating it required reorganizing the form so translatable fields are
together.
We also needed to add a `hint` option to the form label and input
methods so the hint wouldn't show up for every language.
Finally, the markdown editor needed to use the same globalize attributes
as inputs, labels and hints, which adds a bit of duplication.
Note the title field was hidden since commit 01b9aa8, even though it was
required and translatable. I've removed the required validation rule,
since it doesn't seem to make much sense and made the translatable
tests harder to write.
Also note the method `I18n.localize`, which is used to set the
milestone's title, uses `I18n.locale` even if it's inside a
`Globalize.with_locale` block, and so the same format is generated for
every locale.
The same way we did for banners.
We needed to add new translation keys so the labels are displayed in the
correct language. I've kept the original `title` and `body` attributes
so they can be used in other places.
While backporting, we also added the original translations because they
hadn't been backported yet.
After removing a translation while editing another one with invalid data
and sending the form, we were displaying the removed translation to the
user.
We now remove that translation from the form, but we don't remove it
from the database until the form has been sent without errors.
After adding a new translation with invalid data and sending the form,
we were disabling the new translation when displaying the form again to
the user, which was confusing.
This change forces us to use nested attributes for translations, instead
of using the more convenient `:"title_#{locale}"` methods.
On the other hand, we can use Rails' native `_destroy` attribute to
remove existing translations, so we don't have to use our custom
`delete_translations`, which was a bit buggy since it didn't consider
failed updates.
It was accidentally deleted in commit 914bfa6.
Note the following spec passes on my machine if we add a `sleep 0.1`
call in the `:wait_readable` part of ruby's `Net::Protocol#rbuf_fill`.
Otherwise, it hangs forever after clicking the `.fileupload-file` div,
which closes its window. It might be solved when upgrading rails,
capybara, selenium or chromedriver.
scenario "Allows images in CKEditor", :js do
visit edit_admin_site_customization_page_path(custom_page)
within(".ckeditor") do
within_frame(0) { expect(page).not_to have_css("img") }
expect(page).to have_css(".cke_toolbar .cke_button__image_icon")
find(".cke_toolbar .cke_button__image_icon").click
end
within_window(window_opened_by { click_link "Browse Server" }) do
attach_file :file,
Rails.root.join('spec/fixtures/files/clippy.jpg'),
visible: false
find(".fileupload-file").click
end
click_link "OK"
within(".ckeditor") do
within_frame(0) { expect(page).to have_css("img") }
end
end
The test was failing sometimes because of the sequence:
within('#side_menu') do
click_link "Collaborative Legislation"
end
click_link "All"
expect(page).to have_content 'An example legislation process'
click_link 'An example legislation process'
Right after clicking the "Collaborative Legislation" link, the link 'An
example legislation process' is already available. So sometimes Capybara
might click the links "All" and 'An example legislation process' at more
or less the same time, causing the second link not to be correctly
clicked.
Making sure the "All" link doesn't exist anymore we guarantee Capybara
will wait for the previous AJAX request to finish before clicking the
next link.
Note the test to "Create Valid legislation question" is almost identical
but it doesn't fail because it doesn't use Capybara's JavaScript driver.
Skipping a spec in the middle of it, particularly after doing some
requests, caused Capybara to keep using the same driver in the following
spec.
Since the current spec uses the JavaScript driver, the next test would
also use the JavaScript driver, causing apparently random failures if
that test was supposed to use the Rack driver.
This spec was causing a side effect on another spec[2], making it fail 😌
I think it was because no translation had been called yet, in the failing spec, and so the the i18n backend translations had not been initialized, and was always returning empty translations for any locale. This might have been due to tampering with translations in the this newly introduced spec.
By forcing translations to load after this new spec, the other spec passes again
[2] https://github.com/AyuntamientoMadrid/consul/blob/master/spec/features/localization_spec.rb#L20
There where two issues with the current implementation:
- There was a possible duplication between looking up the language name in key "locale" and in key "i18n.language.name"
- The "default" option was not being picked up, as the fallback always returned the default locale's translation, "English"
With this implementation there is only a single place to put the language name: i18n.language.name. I think this place is easier to find and understand for Crowdin translators than a "locale" key hidden in general.yml
If the translation is not found we display the language key, instead of English, which makes more sense to me too 😌
Solution based on recent comments[1] on a related I18n issue
[1] https://github.com/svenfuchs/i18n/issues/365#issuecomment-419263847
We work with many languages using Crowdin[1]
Sometimes translators forget to fill in all the necessary plural forms of a translation (zero, one, other) and in those cases we were seing the exception InvalidPluralizationData being raised
There are a number of approches to fix this... from being more strict when approving translations, to automatically extrapolating what those plural forms should be
For now, we've gone for a simple approach to display the actual count(0,1,2,3,4, etc) instead of the whole translation
So, if the plural form of "1 comment" is missing, just a "1" will be displayed and no exceptions raised
Note: The first two specs, test what is really Rails' functionalities. However as we are monkey patching the pluralize method, I thought it was appropriate to doble check it
[1]https://crowdin.com/project/consul
