This PR is currently a bit of a mess. It depends on both PR#170 and PR#174, and has some other related fixes in it too. So it will need rebasing and separating into individual PRs.

Also there are still some missing pieces: mainly I want to write proper migration tests, after the kerfuffle around #129. And the POST /api/v1.0/waivers/+filtered endpoint could use more tests.

But I wanted to put this up now so that people can start giving feedback.

Why not use subject_item instead of subject_identifier? Looks more consistent with subject_type filter name and the nested key name ("subject/item") which is listed by the GET request.

Oh, sorry, just noticed that it's for v2 and it lists subject_identifier. Never mind then.

Temporary variable for results['data'] would help in this section a bit.

Mmm, it is actually strange. But this test is failing for me. Basically, the "valid_results_list" called inside the "+by-subjects-and-testcases" API raises the:
ValueError('Must be a list of dictionaries with "subject" and "testcase"')

...because v (=> "case 1") is unicode type and not str as expected. But I don't see anything here in your changes that changed that thing.
Am I the only one with this error?

That's my mistake. It works with py.test-3, doesn't work with py.test. Didn't check it for Python 2. PR: #178

It looks ok to me

Good idea

Okay, now that PR#170 and PR#174 are merged, this is rebased.

Some of the commits in this PR are still not strictly related to the subject_type but they form a neat series and they would just conflict if I separated them out, so I think it's simpler to leave them here. I definitely recommend reviewing the commits individually.

Also, let's leave this unmerged until I finish off the corresponding Greenwave change, so that we can be sure this new API is what we definitely want.

Series ending in commit 395772f is tweaked as per Lukas's suggestion for result['data'], and to tidy up the handling of 'comment' required argument.

:+1:, @dcallagh - looking good.

+1

Note, this will need a rebase now.

Pushed a new series ending in commit a90846d. Rebased it on top of latest master, and fixed pylint's complaint about redefining filter.

Added an extra commit d225e14 to map 'brew-build' to 'koji_build', as we do in Greenwave.

Added one more commit, to accept unrecognised subject types in /api/v1.0/waivers/+by-subjects-and-testcases, for better backwards compatibility with current Greenwave. We need that because our internal greenwave-dev instance is currently making decisions about certain subjects that don't fit any of our recognised subject types.

Also rebased again due to conflicts with PR#201.