Reset 'seen' flag in log watcher after check

[melanieclarke]

Small PR to make LogWatcher reusable for multiple messages.

Using this helper to develop tests for #8669, I found it's convenient to use the same watcher to check for different messages, but the flag needs to be reset between calls. I can do it manually, but that's failure prone. With this change, I can do something like:

log_watcher.message = 'test 1'
function_call()
log_watcher.assert_seen()

log_watcher.message = 'test 2'
function_call()
log_watcher.assert_seen()

Checklist for PR authors (skip items if you don't have permissions or they are not applicable)

• added entry in CHANGES.rst within the relevant release section
• updated or added relevant tests
• updated relevant documentation
• added relevant milestone
• added relevant label(s)
• ran regression tests, post a link to the Jenkins job below.
  How to run regression tests on a PR
• All comments are resolved
• Make sure the JIRA ticket is resolved properly

[braingram]

The change looks good to me. Would you update the documentation to describe that assert_seen resets the seen flag?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 60.36%. Comparing base (351157e) to head (4c50426).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #8686   +/-   ##
=======================================
  Coverage   60.36%   60.36%           
=======================================
  Files         372      372           
  Lines       38325    38325           
=======================================
  Hits        23136    23136           
  Misses      15189    15189           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[braingram]

Thanks for opening this PR.

Alternatively, do you think documenting that the LogWatcher is one time use would work? There are reasons to strive for a single assert per test and having a one-time use watcher would work for all of these. I'm ok with either way (documenting it as one-time-use or the changes in this PR) and as this PR revealed the class documentation could be improved either way.

[melanieclarke]

I added some documentation.

Personally, I've found that single-assert-per-test is unrealistic when working with science functions. There's often a lot of setup and preprocessing that has to happen before a function can run, and sometimes the functions themselves take a long time to run. It would make unit tests take a lot longer if we could only test a single condition for each test.

[braingram]

Thanks!

The expensive setup and run can be put in a fixture. Here's one example in the jwst regression tests.
https://github.com/spacetelescope/jwst/blob/master/jwst/regtest/test_fgs_image2.py