Changelog¶

v2.5.2 (2017-10-06)¶

Happy birthday to me.
Correct incorrect information in the NoAffinityError docstring, which stated that the minimum shared rated anime threshold for calculating affinity was ten, when it’s actually eleven.

Add the conventions section to extras_require, so conventions dependencies can easily be installed if needed.
Add the travis section to extras_require, so Travis can easily install any additional dependencies it needs.
Rename the doc and test sections in extras_require to docs and tests respectively.

Create the Affinity namedtuple, and return that instead of a normal tuple when calculating affinity.
Rename the “their_list” var in MALAffinity.calculate_affinity to “user_list”. It’s an improvement, but still a bad var name.

Tidy up the MALAffinity __repr__, and dynamically specify the class name, in case the class needs to be renamed (which it probably won’t).
Use double backticks instead of a single backtick in the calculate_affinity function’s docstring.
Add a one-off comparison function (+ docs).
Move the “shared rated anime threshold” check to the calculate_affinity method, as a comparison can be done no matter how many shared rated anime two users share.
Add the error/warning codes to ignore next to the # noqa comments.
Rewrite the contributing page in the docs, and create CONTRIBUTING.rst to link to it.

Use Xinil and Luna‘s usernames to demonstrate how MALAffinity can be used, in the docstring examples.
Use different anime in the comparison method’s key:value example (in the docstring), so a wider range of ratings can be shown.

Rewrite calcs.pearson to make it 5x faster.
Do the “standard deviation is zero” checking in calcs.pearson, so the standard deviation doesn’t actually have to be calculated.
Don’t make MALRateLimitExceededError a descendant of MALAffinityException.

Move test dependencies to extras_require in setup.py, and have tests_require mirror that.
Add documentation dependencies to extras_require in setup.py.
Make all docstrings PEP 257 compliant.
Add more classifiers to setup.py.

This release exists solely to test that the above is working. No code changes have been made.

Fix tests up to use the mock module, and be less hacky.
Hook tests up to setup.py, so python setup.py test can be called, for the test suite to run.
Add version constraints in for dependencies.

Create the comparison method, and have calculate_affinity use that to retrieve both sets of scores.
Rewrite the docstrings in the MALAffinity class, to be more useful and more compliant with le Sphinx syntax.
Add docs to the project, and have them hosted on readthedocs.io.
Add (badly written) tests, and hook them up to Travis and Coveralls.
Get rid of Markdown altogether, and rewrite the README, as most of the info is now in the docs.

Fix a typo in the calcs.pearson docstring, which incorrectly said the :rtype was a bool.
Use the find_all BS4 method instead of findAll.
Fix a typo in the “Shared rated anime count is less than required” exception message, which incorrectly stated that the minimum required was ten, when it’s actually eleven.
Add a docstring for .__about__.
Add a docstring for MALAffinity._retrieve_scores.
Remove the useless kwargs from MALAffinity._retrieve_scores.
Make the statistics pypi package a requirement for all Python versions.
Add a __repr__ to the MALAffinity class.
Move the _retrieve_scores method in the MALAffinity class to its own file (endpoints.py).
Create a const.py file for constants.
Add a # NOQA comment to the .__about__ imports in __init__, to suppress the F401 flake8 warnings.

Move the MALAffinity class to its own separate file (malaffinity.malaffinity) and import that into the malaffinity namespace via __init__.
Move exceptions to its own separate file (malaffinity.exceptions).
Modify description of the package slightly (“Calculate affinity between two MyAnimeList users” => “Calculate affinity between MyAnimeList users”).
Add exception message for when the standard deviation of one of the two users’ scores is zero, and affinity can’t be calculated.
Create the base MALAffinityException class and derive all malaffinity exceptions from that.
Add docstrings for malaffinity.calcs.
Modify docstrings to remove typos and unnecessary information, and reword some sections.
Reword exception messages to be more useful.
Have the init method return self, to allow for chaining.
Make all code PEP8-compliant (ignoring F401 for meta reasons).

Remove scipy (and numpy) as a dependency. Pearson’s correlation code is now in malaffinity.calcs and stdev checking is handled by the statistics module.
Use lxml for XML parsing, instead of the default html.parser.
Add return types for components inside the return tuple into the docstring.

Change ‘base user has been set’ testing to also check if self._base_scores has been set as well.
Use zip to create the scores1 and scores2 arrays that calculations are done with.
Check if the standard deviation of scores1 or scores2 is zero, and thrown an error if so.
Use scipy.asscalar as opposed to .item() for numpy.float64 => float conversion.

Don’t count rated anime on a user’s PTW. MAL didn’t count this, so our affinity values were a bit off when a user did this.