The visual convention for redaction is heavy black crossing-out.
Our text colour isn’t always black exactly though (usually it’s very
dark grey, sometimes just dark grey). `currentColor` is a magical CSS
value that let’s us set the background colour to whatever the text
colour is. So both the text and redaction look like they are part of the
same thing.
Couple of subtle visual things here:
- opacity to make the colour better match the text colour – a filled box
naturally looks darker than thin text, so this knocks it back a bit
- an inset shadow to take a few pixels of the bottom edge – the visual
weight of text is biased upwards because text has ore capital letters
and ascenders than descenders – this make it look aligned visually
If a template has the `redact_personalisation` flag set, then this
commit removes the personalisation from the notification before
rehydrating the template.
We’re doing this because we have a need to not show things like one time
passwords or two factor codes when we show the content of messages.
By passing through empty personalisation, and the `redact_missing` flag,
the `Template` instance will make use of the work done in:
- [x] https://github.com/alphagov/notifications-utils/pull/171
This will let us break up this method a bit more, rather than make the
dictionary comprehension even more involved and nested.
Means we need to `list()` it, because generator expressions cast to
boolean are `True`, even if they’re empty – Python doesn’t evaluate
them. ie:
```python
bool(list())
>>> False
```
```python
bool((item for item in list()))
>>> True
```
```python
bool(list(item for item in list()))
>>> False
```
they're currently expecting a RecipientsCSV object - but we won't
always have that available if we're handling a single notification.
So make the partials take generic variables, and then use a jinja with
block to pass in the correct values from either the check csv page or
the check notification page.
Additionally set the notification check page to show errors nicely -
hide the send button if there were problems, and replace the header
The following errors may happen:
* Number outside of service if service in trial mode
* Message too long for sms
* Service over daily limit
We need to handle these. They only return on send, rather than in a
separate validation step (for now).
where we were previously setting the placeholder when going through
the send self a test - however, should be setting recipient. Also,
only do this on step-0 of the one-off route, not the send-test route,
since step-0 of send-test is the first normal placeholder. Phew!
note: in the case of letters, we still want to create a CSV file. This
only modifies the code flow when it's an email or template 😩
renamed `send_test_values` to `placeholders` because a) that's what
they are and b) this isn't just for sending a test message any more
rather than creating a job, after entering the placeholders, you now
send a single notification. This means we don't clog up s3 by creating
lots of one line CSV files.
it's confusing reassigining one template (json from api) to another
type (utils object) on one line.
Also removed an unnecessary bounds check (since if placeholders is
empty the IndexError will throw on the next line anyway and it'll
be handled the same), and moved get_back_link out to its own function
Works similarly to the delete template flow, because it’s a destructive,
one-way action.
Not on the edit template page, because it’s not something you want to be
considering every time you’re editing a template. And we saw that people
couldn’t find the delete button when it was on this page.
Adds a bit more CSS for the `dangerous` banner type, because the content
here is quite complicated. Breaking it into a list helps, but the
spacing didn’t look right, so needed some tweaking.
Can ship independently of the code that shows the redaction, but needs
the API first.
For some reason:
- notifications sent from CSV files have the recipient as part of the
personalisation
- notifications sent via the API don’t have the recipient as part of the
personalisation
I’d only tested it locally with CSV-sent files so didn’t spot this. The
`conftest.py` fixtures we set up like the API already, but we didn’t
have an explicit test.
This commit adds a method to append the recipient to the
personalisation, so we can populate the template with it.
There are a bunch of services that just have one text message template
and one email template.
In this case the template navigation is essentially the same as the
list of templates – a user can just differentiate by looking at the two
templates. Adding the nav would just be noise in this case.
There are lots of services that only send emails, or only send text
messages. For these services, being able to filter the list of templates
but type is pointless – it won’t cut the list down at all.
This commit adds some logic to only show the navigation if the service
has some variety of template types.
When users are trying to find a template there’s a fair chance that they
know whether or not it’s an email/text message/(letter) that they’re
looking for.
Making them scroll past a whole bunch of templates of a different type
means it will take them longer to find the template they are looking
for.
We already have search on the templates page, but this is only good for
where they can remember the name of the template. This will be
sometimes but not always.
This commit adds some navigation to filter down the list of templates to
only show one type at a time. By default it will show all templates. It
adapts the pattern we use for filtering notifications by
sending/failed/delivered, but without the counts of how many things are
in each bucket (I don’t think there’s any value in knowing you have X
text message templates; on this page you only really care
about the one template you’re looking for).
_Note: required re-arranging the functions in `templates.py`. The route
for `/template/:uuid` needs to come before the route for
`template/:string` otherwise Flask tries to interpret a template’s ID
as its type.
Bug was happening because:
```python
bool(list())
>>> False
```
```python
bool((item for item in list()))
>>> True
```
i.e. generator expressions cast to boolean are `True`, even if they’re
empty – Python doesn’t evaluate them.
This was causing the functional tests to fail because it was taking too
long for any table rows to appear on the page.
