Rebase master branch so that this PR can
be updated with the latest changes.
Conflicts has been solved and some specs
updated to fit the new changes. dev_seeds
has been also adapted to the new format.
Why:
Heading filter where not being correctly displayed
How:
Increasing scenario to cover all possible combinations, and fixing the
heading_filters method of the Valuation Budget Investment Controller to
correctly:
* Find how many investments the valuator can access
* Count investments for each heading
By using a random seed value smaller than 1, we solve the previous
situation[1] in a simpler way
This test is now obsolete.
It’s hard to write a tests to verify that even with a big seed in
params, we will covert it to a float smaller than 1.
We should refactor these `set_random_seed` methods into a nice model or
controller concern and test it thoroughly
[1]
https://github.com/AyuntamientoMadrid/consul/commit/ba3bf11526fc6ce9c66f
647c414946c61ff945fe
This is a defensive test, just in case we decide to go back to using
`setseed` instead of the `modulus`[1] approach to display investments
in random order
The reason for this test is that `setseed` only ~works in the next
`select` statement. And as when loading a user’s votes for investments
we do a second `select` it does not work as expected 😌
To solve this… we could call `set_random_seed` before loading a user’s
votes for an investment[2]
[1] https://github.com/consul/consul/pull/2131
[2]
https://github.com/AyuntamientoMadrid/consul/blob/master/app/controllers
/budgets/investments_controller.rb#L37
We are trying out a modulus function to return investments in random
order https://github.com/consul/consul/pull/2131
However we ran into the gotcha of having a seed value too big for the
modulus function to work as expected
If the seed is bigger than the investment id, the records are returned
ordered by id
By dividing the seed by a big number, this problem seems to get fixed
Why:
ValuatorGroup name should be unique and present to be able to identify
correctly each of them.
How:
- Adding a presence & uniqueness validation at the model
- Adding a sequenced value for name attribute at its factory
- Adding missing model spec that covers validations
When there are no budgets we were seeing an exception in the budgets’
index
There are two parts to take into account here:
1) Making sure there is a current_budget present, otherwise we display
the “no budgets” message
2) The map helper is called from the controller, so we need to make
sure current_budget is present there too
Note: We could have added a bunch of `try` statements in the budgets’s
index, instead of using a conditional, however there are quite a few
`current_budget` calls so it seems more appropriate to use a conditional
Date.new(...) does not take into account the current timezone, while other
parts of the application do. By default always parsing any date with the
default timezone and converting the resulting Time to Date would prevent
this kind of issues
DateTime.parse(...).in_time_zone gives an unexpected result, as the
DateTime.parse(...) will create a DateTime with +0000 time zone and the
`in_time_zone` will modify the DateTime to adjust to the default zone.
Maybe its better explained with an example, using 'Lima' as timezone:
DateTime.parse("2015-01-01")
> Thu, 01 Jan 2015 00:00:00 +0000
DateTime.parse("2015-01-01").in_time_zone
> Wed, 31 Dec 2014 19:00:00 -05 -05:00
And that's not the desired date but the previous day!
Not sure how this error creeped in 😕 probably a new gem version or
other conflicting code
The problem was we were getting an `unpermitted param email` when
updating a user’s email address
This stackoverflow solution seems to work nicely 😌https://stackoverflow.com/questions/17384289/unpermitted-parameters-addi
ng-new-fields-to-devise-in-rails-4-0#answer-19036427
The tests that will check the featute
is working well added. There are four test:
1. Checks that the emails are been send to
the user. The test looks for the link in there
and takes the token. Visits the page and
changes the password.
2 and 3. Both change the password by hand. One
uses a password written by the manager, whilst
the other uses the 'Generate random password'
option. Both tests check that the user can
sign in with the new passwords.
4. Checks that the password can be printed
when it is saved.
The example tests if a certain selector is hidden after adding
one image but the assertion expected said selector to be visible,
which made the scenario to fail at random
The now-deprecated `.trigger('click')` API simulated a click against
the DOM rather a click on the UI, which made our tests fragile and
wouldn't simulate actual user interaction
