Because:
- GOV.UK elements is now published with a package.json that only install the
SASS files (https://github.com/alphagov/govuk_elements/pull/156)
- We can drop Git submodules, so one less dependency management tool
This commit also changes the `gulpfile.js` and `main.scss` files to use the
assets from `node_modules` rather than the Git submodules.
> it is about consistency and updates, if that endpoint changes in the future
> we don't have to update hundreds of tests for a specific string. The actual
> url should be ambiguous we are testing a view endpoint.
> it is about consistency and updates, if that endpoint changes in the future
> we don't have to update hundreds of tests for a specific string. The actual
> url should be ambiguous we are testing a view endpoint.
This is a link not a button because:
- it’s less prominent—delete is an infrequent action
- it’s a two-step process, and only the second part changes any data (so it has
a button)
If the templates page contains text messages and emails then there’s two ways it
could be structured:
- into two sections, all text messages first, then all emails
- emails and text messages interleaved, sorted by date
I think the second one is better. Imagine a situation where you mostly do emails
but have a few text messages. You’d have to scroll past the text messages to get
to your emails. Every time.
I reckon that the most commonly accessed templates will be the most recent ones.
Because a user can have multiple services, they need a way to navigate between
them. Normally they can use the ▶ Switcher to do this, except when:
- they first sign in
- they are on a page which isn’t associated with a service (eg user profile) in
which case we can’t use the switcher because it won’t know what the ‘current’
service is
So this commit adds a new page with a (fake) list of services.
If Node tries to parse the Gulpfile as the wrong version of ECMAScript it will
choke.
This can be stopped by explicitly telling it what version of ECMAScript to use,
as suggested here: https://markgoodyear.com/2015/06/using-es6-with-gulp/Fixes#74
This commit replaces the previous `StringField` used for collecting mobile
phone numbers with the `UKMobileNumber` field.
This means changing a few of the preexisting tests to have more realistic mobile
numbers so that they still pass.
This commit:
- improves the tests to check for specific error messages, rather than just
pass/fail
- makes the error messages more human, and more suggestive of what the user
needs to do to fix the error
This field does two things:
- validates the format of the phone number
- outputs a consistent representation of the phone number
Because of this I think it’s better represented as a new field type, rather
than individual validators.
I also think that it’s better to do this without regular expression(s), because
it makes returning the specific error easier.
This commit also adds basic pass/fail test for a series of valid/invalid
phone numbers.
This commit adds an extra page or field for confirming your current password
when making important changes
Name | Email address | Mobile number | Password
---------------------|-------------------|-------------------|------------
No password required | As second page | As second page | On same page as new password
This commit adds a page or series of pages for changing your:
Name | Email address | Mobile number | Password
------------------|-------------------|-------------------|------------
Enter new value | Enter new value | Enter new value | Enter new value
| Enter 2fa code | Enter 2fa code |
Return to profile | Return to profile | Return to profile | Return to profile
(each row is a page)
Gulp was failing silently on Travis. I tracked this down to the task that
builds a custom, slimmed-down version of jQuery from source.
To fix this I’ve removed the task and replaced it with just `src`ing the
minified version of jQuery from `node_modules`.
Cons:
- A few more kb of JS
Pros:
- Less random, afternoon-chewing complexity
