This was a helper for situations when Python 2 and 3 APIs returned bytes
and unicode, respectively. In these situation, we should nowadays know
which of the two we receive, so there's no need to wrap & hide the
`bytes.decode()` anymore (when it is still required).
Detailed justification:
beets/ui/__init__.py:
- command line options are always parsed to str
beets/ui/commands.py:
- confuse's config.dump always returns str
- open(...) defaults to text mode, read()ing str
beetsplug/keyfinder.py:
- ...
beetsplug/web/__init__.py:
- internally, paths are always bytestrings
- additionally, I took the liberty to slighlty re-arrange the code: it
makes sense to split off the basename first, since we're only
interested in the unicode conversion of that part.
test/helper.py:
- capture_stdout() gives a StringIO, which yields str
test/test_ui.py:
- self.io, from _common.TestCase, ultimately contains a
_common.DummyOut, which appears to be dealing with str (cf.
DummyOut.get)
The previous test worked (on my machine, and on Github CI and AppVeyor),
but it is not obvious whether the order is really guaranteed (given that
the full beets database stack and sqlite are involved). Thus, to prevent
this from exploding at some point, only verify the number of deletions
for now.
The test `CommonOptionsParserCliTest.test_version` was passing with nose
but failing with pytest (see output below). The reason for the failure
seemed to be that the `test` plugin was loaded when it wasn't expected
to be loaded, changing the output of the `version` command. I'm not sur
exactly why that was happening, but since that test already inherited
from `TestHelper`, just invoking the plugin load/unload helper was
enough to fix it. I also removed the line setting the `self.lib`
variable since that's already done in the helper.
---
self = <test.test_ui.CommonOptionsParserCliTest testMethod=test_version>
def test_version(self):
l = self.run_with_output(u'version')
self.assertIn(u'Python version', l)
> self.assertIn(u'no plugins loaded', l)
E AssertionError: 'no plugins loaded' not found in 'beets version 1.4.8\nPython version 3.7.3rc1\nplugins: test\n'
test/test_ui.py:1292: AssertionError
The move functions in library.py and manipule_files in importer.py where
changed to use a single parameter for the file operation instead of
multiple boolean flags.
A typo in the documentation of the Album.move and Item.move functions
confusing True and False when describing the store parameter was fixed
as well.
The old tests were wrong but the incorrectness was hidden by the incorrect
parameter passing fixed in the previous commit. Now we actually test that the
item's path did not change.
Related to #1966. Previously, we used a `syspath` call inside MediaFile, which
probably wasn't right: the constructor should behave like `open` in that we
need to use pass an OS path.
This was a vestige from when we used to need the unittest2 library for pre-2.7
compatibility. Now that we require Python 2.7, we aren't using that library
and this indirection wasn't doing any good.
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.
This rolls back the places in #2094 where `six.text_type(s)` was used where
`s` might be a bytestring. We almost never want that conversion, because it
uses the default encoding (almost always ASCII). The need for it indicates a
problem somewhere else, where bytes were produced when we were expecting
unicode strings.
PEP 414 didn't add `ur"..."` literals to Python 3. So for hybrid 2/3
compatibility, these had to be replaced with ordinary Unicode string literals.
This was only painful for one string; the others were only raw strings by
convention. (All were regexes.)
This reverts commit 9c41c39913.
That commit used byte strings for the `if __name__ == '__main__'` pattern,
which was necessary when we were doing unicode_literals. But it is wrong on
Python 3, and now that we're liberated from unicode_literals, we need to go
back to native strings for this comparison.
I found some infelicities while investigating #1826. This test was hard-coding
a specific path for bash-completion, which was the wrong one for my system, so
it was being skipped every time. This searches the list of paths used by the
actual command.
