To get the heading where a user voted, we were relying on the
`balloted_heading_id` field.
Our guess is this was done so the total number of users is the same as
the sum of users who voted on a heading. That is, if 2000 people voted
just on the "All city" heading, 1000 voted just on the "North district"
heading, and 500 people voted on both, instead of showing "3500 people
voted in total, 2500 voted in all city, 1500 voted in north district",
we show something like "3500 people voted in total, 2250 voted in all
city, and 1250 voted in north district".
However, this approach has some disadvantages.
The first disadvantage is, the stats aren't correct. In the case above,
2500 voted on the "All city heading", so the statistics for this heading
don't show reality.
The second one is we weren't considering the last heading where users
voted inside the budget being displayed, but the last heading where
users voted, period. That means that, if all the people above voted on a
later budget, the stats for the budget above would become "3500 people
voted in total, 0 voted in all city, and 0 voted in north district".
That also means we were including headings from previous budgets in the
statistics for more recent budgets when people hadn't voted on the
recent ones.
So we're removing the `balloted_heading_id` since its data is lost once
people vote on a new budget. And, in order to show the right stats and
simplify the code, we're no longer trying to add votes just to one
heading when users vote on several headings.
Co-Authored-By: Julian Nicolas Herrero <microweb10@gmail.com>
We're also moving the tests, but we're keeping one system test in order
to test the controller and the navigation to get to this page.
Note we're slightly changing the order of the methods in the component;
the order of the instance variables was `user_`, `vote_`, `vote_`,
`user_`, which was hard to follow.
We don't need groups to create headings since commit 20b1085dc, and
`let!` was used as a substitute for instance variables in commit
91c21b098, but it isn't necessary in this case.
Since the change on commit cbbe188d6 we added a Poll.current.any?
condition to show the officing link on admin menu to officers.
That condition doesn't have much sense since Poll results only can be
added after a poll has ended, and there may be only one active poll.
In most sections, we had two specs testing what happens after accessing
one of the privileged areas. We're grouping the expectations and so
we've only got one test per area, making these tests faster.
This way we reduce the number of system tests or, in some cases,
requests during system tests, making the tests faster.
We're still testing the interaction with the menu when users have the
right permissions.
The page was crashing when at least part of the content of the page had
been translated between the request showing the remote translations
button and the moment people pressed the button.
We were using the same code for the button in both the public and admin
headers, so we're removing the duplication.
Since the menu and the button must go together, and the contents of the
menu are different for different layouts, we're passing these contents
using a block.
Note the ID of the menu was `responsive-menu` in the public section but
`responsive_menu` in the admin section. Since we usually use underscores
for IDs and dashes for classes, we're keeping the one with the
underscore.
Since we allow many active budgets at the same time, the
controller should now check the budget given by params.
Before this change the controller was checking the latest
published budget, ignoring the request parameter `budget_id`.
We've noticed the following warning while testing the upgrade to
Ruby 3.0:
warning: File.exists? is deprecated; use File.exist? instead
We're adding a Rubocop rule so we don't call the deprecated method
in the future.
We were getting a warning with Ruby 2.7:
```
ruby/gems/2.7.0/gems/capybara-3.37.1/lib/capybara/session.rb:377:
warning: Using the last argument as keyword parameters is deprecated;
maybe ** should be added to the call
```
On Ruby 3.0, the test failed with `Unable to find fieldset
{:text=>"Draft phase"}` and we were also getting another warning:
```
Locator Hash:{:text=>"Draft phase"} for selector :fieldset must be an
instance of String or Symbol. This will raise an error in a future
version of Capybara
```
We were getting a warning with Ruby 2.7 due to the change in the way
keyword arguments are handled in Ruby 3.0.
```
ruby/gems/2.7.0/gems/rspec-support-3.11.0/lib/rspec/support/with_keywords_when_needed.rb:18:
warning: Passing the keyword argument as the last hash parameter is
deprecated
```
As hinted by the warning, this code crashes with Ruby 3.0:
```
ArgumentError:
unknown keyword: :budget_id
```
I'm not sure why this is the case, though, since we were already
explicitely passing a hash first before passing the keyword parameters.
I guess there are some cases in this whole keyword VS hash parameters
incompatibility that I haven't completely understood.
We weren't showing the year when a page was created/updated, and we
were displaying the created date instead of the updated one.
Co-Authored-By: Diego Calvo <diego.calvo@enreda.coop>
In the management section, `current_user` is the user impersonated by
the manager. We were deciding whether to show the admin menu depending
on the privileges of the current user, but this menu should be shown
according to the privileges of the manager who is impersonating the
user.
We're doing a similar (very subtle) change in the login items. We were
rendering the `login_items` partial passing `current_user: user`.
However, inside this method, we were using `user_signed_in`, which
ignored the `current_user` we were passing. The result was always the
same expect in tests where we manually sign in users, but we're changing
it anyway in order to reduce confusion.
This way it's easier to refactor it.
Note we're using `with_request_url` in the tests because the component
renders the locale switcher, which needs a URL in order to work. This
doesn't affect whether we're in the management section or not.
It was added in rubocop-performance 1.13.0. We were already applying it
in most places.
We aren't adding it for performance reasons but in order to make the
code more consistent.
Note we could use `acts_as_paranoid` with the `without_default_scope`
option, but we aren't doing so because it isn't possible to consider
deleted records in uniqueness validations with the paranoia gem [1].
I've added tests for these cases so we don't accidentally add
`acts_as_paranoid` in the future.
Also note we're extracting a `RowComponent` because, when
enabling/disabling a tenant, we're also enabling/disabling the link
pointing to its URL, and so we need to update the URL column after the
AJAX call.
[1] See issues 285 and 319 in https://github.com/rubysherpas/paranoia/
In general, we don't use links inside admin tables because we don't know
where the link will point to, and use "view" actions/links instead.
However, in this case, we're showing a URL, so it's perfectly obvious
where the link will point to. And so it makes sense to use the URL as a
link instead of using a "view" action/link.
Some institutions using CONSUL have expressed interest in this feature
since some of their tenants might already have their own domains.
We've considered many options for the user interface to select whether
we're using a subdomain or a domain, like having two separate fields,
using a check box, ... In the end we've chosen radio buttons because
they make it easier to follow a logical sequence: first you decide
whether you're introducing a domain or subdomain, and then you enter it.
We've also considered hiding this option and assuming "if it's got a
dot, it's a domain". However, this wouldn't work with nested subdomains
and it wouldn't work with domains which are simply machine names.
Note that a group of radio buttons (or check boxes) is difficult to
style when the text of the label might expand over more than one line
(as is the case here on small screens); in this case, most solutions
result in the second line of the label appearing immediately under the
radio button, instead of being aligned with the first line of the label.
That's why I've added a container for the input+label combination.
Creating a schema takes about 3-4 seconds on my machine, so omitting
the callbacks makes tests much faster.
To do so, we're using the `insert!` method added in Rails 6.0, which
inserts a record without executing callbacks or validations. To make the
tests look consistent, we're adding a FactoryBot strategy which uses
`insert!` instead of `create!`.
Note this strategy is useless in most cases because it doesn't work when
models have translatable attributes or associations. However, IMHO it's
worth it even if we only use it for tenants.
We could also use `Tenant.insert!` instead, but then we would have to
add all the mandatory attributes, and in this case the code is clearer
if we only add the attributes we need for the test.
Just like we respond with "not found" for any other record. This
improves the user experience because with the "Not found" error message
people realize the URL is wrong instead of thinking that they broke the
application.
While creating the "venus" tenant for every test makes the code cleaner,
it also makes the tests much slower, so we aren't doing so in tests
where we don't use this tenant.
We were displaying an icon showing that certain actions can't be
performed. However, people who can't see the icons were hearing that
they _can_ perform certain actions while the opposite is true.
We've considered other options to solve this problem. One was to split
the list in two: actions that can be performed and actions that can't be
performed. It was tricky because in some cases we're listing that
actions that can be performed now and in other cases we're displaying
the actions that people will be able to perform once they verify their
account.
Another option was to include the word "Cannot" as a prefix instead of
"Additional verification needed". We haven't done so because, while in
English we say "cannot do this thing", in other languages they say
"this thing cannot do".
So we've gone with a solution where people hearing what's on the screen
know what's going on and we don't have to make big changes in the code.
As mentioned in commit 925f04e3f, icon classes make screen readers
announce strange symbols and aren't properly displayed for people who
have changed their preferred font family.
Coveralls stopped working back in July when we reached build number 3790
because back when we used Travis we created builds from numbers 3791 to
35700.
After trying to change the numbers in several ways, all of them
resulting in a "No build matching CI build number" message, we're trying
the Github Action provided by Coveralls instead. In order to make it
work, we need to add the `simplecov-lcov` gem and generate the report at
`coverage/lcov.info`. Note that, for some reason, `simplecov-lcov`
doesn't seem to depend on `simplecov` and we need `simplecov` 0.18 or
later, so we're manually adding this dependency as well.
We were assigning variants in a controller, in the context of a request.
However, when sending emails, there is no request and no controller is
involved, so we also need to set the variant in the ApplicationMailer
class.
