In response to: [^1].
The stacktrace conveys the same and more information. We don't do
anything different for each exception class, so there's no value
in having three of them over one exception.
I did think about DRYing-up the duplicate exception behaviour into
the base class one. This isn't ideal because the base class would
be making assumptions about how inheriting classes make requests,
which might change with future providers. Although it might be nice
to have more info in the top-level message, we'll still get it in
the stacktrace e.g.
ValueError: Expected 'code' to be '0'
During handling of the above exception, another exception occurred:
app.clients.sms.SmsClientResponseException: SMS client error (Invalid response JSON)
requests.exceptions.ReadTimeout
During handling of the above exception, another exception occurred:
app.clients.sms.SmsClientResponseException: SMS client error (Request failed)
[^1]: https://github.com/alphagov/notifications-api/pull/3493#discussion_r837363717
This avoids duplicating it as we add a new provider and means we
can test it all in one place (although it wasn't tested before).
I'm not sure why the previous code did "super(..)__init__" in a
non-init function - it's a bit late! - so I've just replaced it
with a call to the new "init_app" function in the parent class.
This reduces the code to copy when we add a new provider. I don't
think we need to log the URL or status code each time:
- The URL is always the same.
- A "200" status code is implicit in "success".
- Other status codes will be reported as exceptions.
Removing these specific elements means "record_outcome" is generic
and can be de-duplicated in the base class.
Previously we used a combination of "provider.name" and "get_name()"
which was confusing. Using a non-property function also gave me the
impression that the name was more dynamic than it actually is.
