When render the investment list component with the link "see all
investments", now we redirect to groups index page when a budget has
multiple headings.
In general, slow system tests requiring no interaction from the user are
good candidates to be moved to component tests because component tests
are much faster.
In this case, the system tests were also updating the database after
starting the browser, which might cause concurrency issues. We could
split the test and have one system test per phase, but IMHO there's no
need.
We're still having a couple of system tests for the happy path, in order
to make sure users actually see the list of investments.
While we use Pronto to detect offenses in the lines changed in our pull
request, sometimes our changes introduce offenses in other lines, and we
don't detect them.
In commit 0488b3735, we removed the only usage of the `heading` method
in a test, which caused a `RSpec/LetSetup` offense.
In commit 287c48873, we changed some lines from `fill_in` to
`fill_in_ckeditor`. Some of these lines were aligned with the following
ones, which after that change had extra spacing for no reason.
Finally, in commit 8d38ed58c we added a line before two lines which had
their equals signs aligned. Since, after adding this line, the block was
no longer aligned, there was no reason for the extra space in one of the
lines.
Using `flex` instead of a fixed width for the navigation, the elements
take all the available space when the search form isn't present. That
wasn't the case before and produced a strange effect on medium-sized
screens.
This way we also align the search to the right.
As mentioned in commit 5214d89c8, there are several issues with
submitting a form when a `<select>` tag changes. In particular, keyboard
users might accidentally fire the event while browsing the options, and
screen reader users will find a form with no obvious way to submit it.
In this case, there's an extra problem: in commit be8a0dbe8 we added a
second `<select>` field to this form, which also submitted on change.
Sometimes users changed one of the values and wanted to change the other
value as well before submitting the form. However, it wasn't possible,
because we would submit it before they had a chance to change the second
value.
So now we don't submit the form on change and add a submit button. This
is similar to what we do in the "Advanced filters" we use in several
places.
Note we don't cast negative votes when users remove their support. That
way we provide compatibility for institutions who have implemented real
negative votes (in case there are / will be any), and we also keep the
database meaningful: it's not that users downvoted something; they
simply removed their upvote.
Co-Authored-By: Javi Martín <javim@elretirao.net>
Co-Authored-By: Julian Nicolas Herrero <microweb10@gmail.com>
It was hard to notice what was going on when supporting one investment
which was at the bottom of the investment index page.
I wonder whether we should add the title of the investment to this text;
I'm not doing so because we don't do that anywhere else.
Using links combined with JavaScript to generate POST requests to the
browser has a few issues.
An obvious one is that it doesn't work for users without JavaScript
enabled (which lately I've noticed are more common than I thought, even
though I've been one of them for years). These users will reach a 404
page.
Since CONSUL isn't currently designed to work without JavaScript
enabled, let's ignore this one for now.
A not so obvious issue is screen reader users might expect the link to
take them somewhere instead of performing an action (in this case,
sending a form to the server).
There might be more issues I'm unaware of. Quoting DHH [1]:
"Turning ahrefs into POSTs is a bit of an anti-pattern, especially for
a11y reasons. Better to use button_to with a styling."
So we're using a button instead. This way we can also simplify the code
and make the button disabled for unidentified users, which automatically
makes it impossible to focus it using the keyboard.
A possible disadvantage of using `button_to` is it will create a form
tag which will be announced to screen readers as a form landmark. I've
tested it with my screen reader and everything worked fine for me, but
some screen reader users might interact with these forms in a different
way and find it confusing, particularly in the case where the button is
disabled.
With this change, we're only changing links for buttons in one place.
There are other places where we should do similar changes.
[1] See issue 33 in https://github.com/hotwired/turbo-rails/
We meant to check that the link to support a different investment didn't
have an alert dialog. However, we were testing by trying to support the
same investment twice, and here there was no link at all, so the
assertion "there's no link with a dialog" always passed even when the
application didn't behave properly.
Now we check there's a link there, and that link doesn't have an
associated alert dialog.
In regular system tests, we prefer testing against the text of the link
because it's easier to recognize which link we're talking about.
I was wondering whether we should remove the `title` attribute
completely and use just the `aria-label` attribute. Quoting The Paciello
Group blog [1]:
"If you want to hide content from mobile and tablet users as well as
assistive tech users and keyboard only users, use the title attribute."
However, there's a case where mouse users might find the title attribute
useful. When visiting an investment page, the connection between the
"Support" link and the investment is not as clear as it is in the
investments index page, so it might not be clear what you're supporting.
As mentioned, though, this information will only be relevant for mouse
users, who nowadays represent less than half of our users.
[1] https://www.tpgi.com/using-the-html-title-attribute-updated/
The key was declared twice and so the first one ("Support this project")
was overwritten.
We're grouping all keys related to the investment list together in order
to reduce the chances of this issue happening again (or, at least, in
this part of the code).
Note one of the tests dealing with random results is a bit flaky; since
it's a permutation selecting 7 objects out of 12, it will fail about
once every 4 million times. We think this is acceptable.
Co-Authored-By: Julian Nicolas Herrero <microweb10@gmail.com>
In the form of creating a new investment was hiding the name of the
group if it had only one heading, but could be confusing to users if
there are, for example, five different groups of one heading.
The solution:
- If the budget has one group and one heading, the heading selector is
hidden.
- If the budget has one group and more than one heading, the group name
is hidden.
- If the budget has more than one group, the group name appears
regardless of the number of headings.
Note we're using an extra `<span>` element but we could use a CSS grid
layout instead. We're not using it because browser compatibility is only
94.56% at the time of writing.
We only need to define one `in_browser`, which is the one opening the
session as an administrator.
This change is done to simplify the code, although there's a small
chance it might also make the test stop failing in our CI. Sometimes in
our CI the first `visit` in the `in_browser(:admin)` block fails for
unknown reasons, rendering a blank page.
We want to make sure the request is finished after clicking a button and
before visiting a different page, so we need to check the page has
changed.
Usually this shouldn't be a problem because most of our forms are sent
with regular HTTP requests instead of AJAX ones, so the `visit` method
wouldn't be called before the request is finished.
However, we're experiencing problems with certain version of
Chromedriver, and in general it's a good practice because we might send
forms using AJAX/Turbolinks in the future.
It's strange to create records without assigning them to a variable and
then query the database to fetch the very same records. Assigning them
to a variable makes the tests easier to understand.
Besides, this way we avoid querying the database after the browser has
started.
This way we avoid modifying the database in the middle of a system test
(after we've started the browser), which can lead to database
inconsistencies.
In the case of the reclassification specs we're simply removing part of
the test because that part is already tested by other specs.
Even after the previous changes, this test is still failing sometimes
(although now it fails for a different reason). We're doing this change
in order to discard it as the reason why the test is failing.
There seems to be an issue with capybara or chromedriver which results
in `fill_in` sometimes appending to an input rather than overwriting
[1], causing some tests to fail under certain circumstances.
Clearing fields before filling them in solves the issue.
Note we're now getting warnings on all tests using the rack driver. I
haven't found a way to avoid the `clear: :backspace` option in
non-JavaScript tests, so to avoid the annoying warnings we should reduce
the number of tests using the rack driver even more.
[1] See issue 2419 in https://github.com/teamcapybara/capybara/issues
JavaScript is used by about 98% of web users, so by testing without it
enabled, we're only testing that the application works for a very
reduced number of users.
We proceeded this way in the past because CONSUL started using Rails 4.2
and truncating the database between JavaScript tests with database
cleaner, which made these tests terribly slow.
When we upgraded to Rails 5.1 and introduced system tests, we started
using database transactions in JavaScript tests, making these tests much
faster. So now we can use JavaScript tests everywhere without critically
slowing down our test suite.
Content like lowercase letters with `text-transform: uppercase` or
spaces after elements with `display: block` or "You're on page:" are not
seen that way by users with a browser supporting CSS.
So we're testing what most users actually experience.
These fields are hidden for users with a browser supporting CSS and is
only there to fool bots, so we're testing the case of an attack by bots
using browsers with no CSS support.
The method `formatted_heading_price` depends on the current locale. When
we make a request to `visit budgets_path locale: :es`, the request
modifies `I18n.locale` as well.
However, if we use JavaScript tests, the process running the test is
different than the process handling the request, and so the change in
`I18n.locale` does not affect the test.
Checking against the actual value we expect makes the test work with and
without JavaScript.
Using `have_current_path`, Capybara waits until the condition is true,
while using `include` the expectation is evaluated immediately and so
tests might fail when using a driver supporting JavaScript.
Besides, using `have_current_path` the error message is more readable
when the test fails.
System tests are used to test the application from the user's point of
view. To test for specific exceptions, particularly regarding
authorization permissions, controller tests fit better.
Another option would be to test the page displayed shows a certain text,
like "Internal server error". I'm choosing controller tests because
they're faster and we're basically testing the same scenario many times
and we've already got a test checking what happens when users access a
page raising an exception.
Note we're absolutely positioning the link instead of the icon; that way
keyboard users will be able to focus on the icon. Until now, users would
focus on an empty link.
For the same reason, we couldn't use `click_link` with Capybara, since
it would fail to click an empty link. Now we can.
Co-authored-by: Javi Martín <javim@elretirao.net>
