Add three tests for the setting of tracks' medium and medium_total on
the discogs plugin. test_parse_medium_numbers_single_medium is meant
to fail due to #587.
Modify coalesce_tracks() in order to handle the case where subtracks
are defined inside an index track (as `sub_tracks` attribute),
reorganizing the if logic to avoid duplicated code.
Add a try..catch block enclosing the call to clean_tracklist, as a
measure for avoiding side effects (and reverting to just parsing the
raw_list if there are any problems).
Revise the regular expression used for extracting information from a
discogs position, making the subtrack_index only be matched if one of
the other groups is matched as well. Split the definition into several
lines in order to try to document it a bit more clearly.
Add unit tests for position parsing and non standard positions.
Fix the underscore on tox dependency.
Add a first skeleton for providing unit tests for the discogs plugin,
heavily inspired by test_mb, which adds a couple of method for creating
very simple and not comprehensive Bags that mimic discogs_client.Release
and discogs_client.Track.
Add a test for checking the problem of setting media for the Tracks.
Add a comparison at DateQuery.match in order to avoid comparing if the
field is not in the item.
Revise the existing types_plugin unit tests for explicitely checking the
behaviour when the values are unset.
Rather than using an unordered set for storing pairs, append to a list
and build an OrderedDict from it to filter duplicated strings while
keeping order.
This was breaking the `info` command in the click branch. I assume
that was because `run_command` was aliased to `run_with_output`.
It was also confusing in its own right, so I consider it a win
either way.
Fallout from #2188. Following @jrobeson's suggestion, I just turned the check
in LibModel.__format__ into an assert. This turned up a few badly-behaved
clients, which are also fixed.
There was a problem with the authentication header in the latest
versions. The header creation function changed to fix that. Username and
passwort authentication should work again.
The `host` config variable takes now a full hostname. For example
`http://localhost` instead of just `localhost`. This makes it easier to
use https hosts.
It worked in previous versions to put `noqa` on the non-import line that
preceded the imports. As of flake8 3.0.0, we apparently need to ignore the
import-after-non-import error on every `import` line.
On my machine, the "leak" of other bash_completion scripts here recently
started doing a *lot* of work for other packages. Most damningly, it started
compiling a bunch of stuff with rustc! Now we only load beets' own completion
stuff.
