In some countries, postal codes are defined with a dash in the middle,
so we're using a colon to define ranges instead. We could also use two
dots, like in Ruby ranges, but IMHO this would cause typos since people
would enter codes separated with three dots or just one dot.
We weren't using the `be_valid` matcher because we had errors in the
census data.
Removing the `before_validation` callback and using a method to get the
census data instead allows us to stub the `census_data` method in the
tests, and so we can use the `be_valid` matcher instead of calling
`valid?` manually and then checking for errors.
Using a field with `type="tel"` causes most mobile phone browsers to
show a numeric keyboard which makes it easy to enter phone numbers.
We aren't using a number field because some browsers show "spinners" to
increment/decrement numbers (which doesn't make much sense in a phone
number) and because they don't allow characters like spaces. Phone
numbers can have characters like spaces, dashes, parenthesis, asterisks,
the plus sign, ...
Since phone number formats vary depending on the region, and sometimes
even within the same region several different formats can be used, for
now we aren't offering any kind of format validation. We could offer a
format validation that rejects just letters (although there are cases
where letters are actually allowed). However, this causes issues at both
the server side and the client side.
We could use a client-side validation adding a `pattern` attribute to
the field, but that would be something unlike anything else we use in
the application, and we would need to to write some JavaScript in order
to use a proper error message.
On the other hand, if we add a server-side validation, we might find out
existing users are invalid, and it would be impossible to update them in
the many places where we update users assuming they were valid in the
first place. We'd have to add a rake task to update existing user
records and make sure they contain a valid phone number or create a
setting so this validation only applies to new CONSUL installations.
Another option would be to add a per-form validation, so the phone
number is only validated in pages where it is introduced.
All the mentioned scenarios offer certain complexities. So, for now,
we're keeping things simple.
Co-Authored-By: decabeza <alberto@decabeza.es>
Now that we also have the "go back to CONSUL" link, the layouts are so
similar that it isn't worth it maintaining both of them separately.
With this change, people using small screens also get the "menu" button
in the management section, just like they do everywhere else.
We're adding the `namespace != "management"` condition so the menu still
shows up in the officing namespace.
We were making some typos during development in the name of the keys and
tests were still passing.
We're also removing some texts that were never used.
Some developers work on CONSUL installations where Spanish and/or
English aren't part of the available locales. In those cases, the
`dev_seed` task was crashing because we were using attributes like
`name_en` and `name_es`.
So we're using attributes for random locales instead.
We're using a proc so we don't have code like this all over the place:
random_locales.map do |locale|
I18n.with_locale(locale) do
phase.name = I18n.t("budgets.phase.#{phase.kind}")
phase.save!
end
end
This would make the code harder to read and would execute a `save!` once
per locale, which would make the task much slower.
We could also avoid the procs writing something like:
def random_locales_attributes(**attribute_names_with_values)
random_locales.each_with_object({}) do |locale, attributes|
I18n.with_locale(locale) do
attribute_names_with_values.each do |attribute_name, (i18n_key, i18n_args)|
value = I18n.t(i18n_key, (i18n_args || {}).merge(language: I18n.t("i18n.language.name")))
attributes["#{attribute_name}_#{locale.to_s.underscore}"] = value
end
end
end
end
And calling the method with with:
random_locales_attributes(name: ["seeds.budgets.name", year: Date.current.year - 1])
However, this code would also be different that what we usually do, we'd
have to apply some magic to pass the `language:` parameter, and the
strings wouldn't be recognized by i18n-tasks, so we aren't sure we're
really gaining anything.
These locales are officially maintained by CONSUL developers, so we're
using them when available.
This way we'll be able to use `random_locales` in places where we're
manually using English and Spanish, giving support to other locales
while maintaining compatibility with the current version.
Originally there was a link pointing to the FAQ page but it was removed
in commit e14b7b67fb because by default the FAQ page in CONSUL only
contains a placeholder text.
We aren't sure where this link should point:
* FAQ page, only if the FAQ page is published
* Help page, only when the help feature is enabled
* CONSUL technical documentation page
So, for now, we're choosing the easiest solution which is removing the
text completely.
We already support Errbit and Airbrake as error monitoring services.
Since some people might not want to setup Errbit and might prefer
Rollbar over Airbrake, we're referencing it in the custom gemfile.
We haven't updated the gem for years and don't know whether it
still works with our current Ruby and Rails versions.
Besides, dependabot keeps opening pull requests to update it. In theory
we could just ignore the dependabot pull requests for this dependency,
but unfortunately right now we can't add a dependabot config file
because it would open pull requests on forks as well.
Finally, there are other companies offering similar services for Rails
applications, and it's up to each CONSUL installation to decide which
one is better for them. We might add a self-hosted performance
monitoring tool in the future.
Since other CONSUL installations might be using Newrelic, and in general
we recommend adding an application monitoring tool, we're suggesting it
in the custom gemfile. In the name of neutrality, we're also adding
Sentry. We might add other services in the future.
The `commentable_url` method wasn't updated when we added legislation
proposals.
Back when we first created this method, we couldn't pass budget
investments or topics directly to `polymorphic_url` because they are
nested resources. That isn't the case since commit ff93f5a59, so now we
can simplify this method.
We're keeping the `commentable_url` method for now in order to keep
compatibility with custom changes that might use it, although this
method isn't consistent with the `commentable_path` method (which
receives a comment, and not a commentable), and so we might have to
revisit this code in the future.
With the `update_without_changes`, previous translations are kept
even when the source string changed utterly. Using this option caused
some problems when updating source language strings, as the
translators do not get any notification, and the old translation keeps
approved, so translators do not have an easy way to find those strings
in Crowdin.
With the `update_as_unapproved` option, translators and Crowdin
managers can find those translations that need to be updated by
searching among not approved translations.
Quoting Crowdin's docs [1]:
> update_without_changes - preserve translations and validations of
changed strings
> update_as_unapproved - preserve translations of changed strings
and remove validations of those translations if they exist
[1] https://support.crowdin.com/configuration-file/#changed-strings-update
This test was failing sometimes. One possible cause (although it might
not be the only one) is we were querying the database with
`Campaing.last` after starting the process running the browser with a
`visit`. In the past doing so has resulted in database inconsistencies
while running the tests.
Since after running the test more than 1500 times we weren't able to
reproduce the failure, it's possible that this change doesn't fix the
issue which caused the test to fail, but in the worst case scenario we
reduce the number of possible reasons why it fails.
