... continuing the test suite migration to `pytest` (#5361), this PR
replaces `unittest` method assertions with `assert` statement.
See the table below for a list of the replaced methods and their counts.
This should help clarify the size of the pull request 😆.
To keep things simple for the review, I have replaced each method in its
own commit. This
way, things should be easy to follow if you step through the commits one
by one. 🙂
method | count
--- | ---
**`assertEqual`** | 1747
**`assertIn`** | 200
**`assertTrue`** | 149
**`assertIsNone`** | 87
**`assertFalse`** | 85
**`assertRaises`** | 69
**`assertNotIn`** | 64
**`assertNotEqual`** | 39
**`assertIsNotNone`** | 35
**`assertIsInstance`** | 35
**`assertLessEqual`** | 23
**`assertGreater`** | 15
**`assertLess`** | 14
**`assertGreaterEqual`** | 14
**`assertAlmostEqual`** | 9
**`assertRaisesRegex`** | 4
**`assertCountEqual`** | 2
**`assertListEqual`** | 1
❗ Note that this *only* replaces methods provided by
`unittest` by default.
**Custom** assertion method replacements (like `assertExists` will be
addressed in a separate PR).
I surprisingly found that many queries for fields that exist in the
database are 'slow'. Querying items with `month:0` used a slow query
while `title:0` used a fast one.
I found the mistake in the logic: the field existence in the db check
was done *after* prepending the table name to the field. This meant that
the logic wrongly checked e.g. `items.field` instead of `field`.
This commit fixes the above and renames the variables for more clarity.
I surprisingly found that many queries for fields that exist in the
database are 'slow'. Querying items with 'month:0' used a slow query
while 'title:0' used a fast one.
I found the mistake in the logic: the field existence in the db check
was done *after* prepending the table name to the field. This meant that
the logic wrongly checked e.g. 'items.field' instead of 'field'.
This commit fixes the above and renames the variables for more clarity.
This PR deduplicates shared tests setup that was used across test files.
It attempts to centralise the setup into a single source of truth
implementation in `beets.test.helper`.
This works towards the eventual migration to `pytest` (#5361) making it
easier to replace the tests setup with `pytest` fixtures.
It builds upon the temporary files cleanup in #5345.
## Details
### Test setup
* Mostly duplicate test setup implementations in
`beets.test._common.TestCase` and `beets.test.helper.TestHelper` were
merged into a single implementation as
`beets.test.helper.BeetsTestCase`.
* Replaced `unittest.TestCase` and `beets.test.helper.TestHelper`
combination with `beets.test.helper.ImportTestCase`
* Refactored `test/test_plugins.py` removing duplicate import files
setup.
* Centralised setting up the database on disk.
* Centralised plugin test setup into `beets.test.helpers.PluginMixin`.
### Removals
* Around 1/3 of the removed lines correspond to the removal of `def
suite()` functions that were previously used for tests collection.
* Removed redundant `setUp` and `tearDown` methods from many test files
given that the functions that they performed had been centralised in
`beets.test.helper`.
* Removed redundant library initialisations
### Importing
* Centralised import directory creation (only gets created on the first
access).
* Deduplicated `_setup_import_session` implementations in
`TerminalImportSessionSetup` and `ImportHelper`.
* Merged `ImportHelper._create_import_dir` and
`TestHelper.create_importer` implementations into
`ImportHelper.setup_importer`.
* Merged various takes on import files prep into
`ImportHelper.prepare_albums_for_import` and
`ImportHelper.prepare_album_for_import`.
* Seeing that many tests used it, defined `AsIsImporterMixin` which
provides a method `run_asis_importer` to setup _asis_ (non-autotag)
importer and run it
A constant `preload_plugin` is used to disable loading the plugin in the
`setUp` initialisation, allowing the plugin to be loaded manually by the
tests.
Also added a cleanup instruction to remove listeners from configured
plugins, and removed this logic from several tests.
A couple of `ConfigTest` would previously fail locally since they
somehow depended on the local environment to work fine. This commmit
decouples them from any environment by setting up patching of the
environment variables properly.
