Re-evaluate logic for deleting rows from `ratings` table

In #737, a WikiProject selection with multiple projects winds up with an article list that contains dozens of deleted articles.

This seems due to the fact that articles which have been deleted from English Wikipedia are never deleted from the ratings db table. The algorithm goes like this:

Find all the articles in the "... by quality" category for the project
Compare to all of the articles in the ratings table for that project
For any articles that are in the db (ratings table) but not the category:
1. Check if their quality/importance is already set to NotAClass. If so, skip
2. Check if they have been moved in 3 different ways.
3. If so, set the move data for that log
4. Regardless, set their quality or importance rating (or both) to NotAClass.

There is additional separate logic for deleting articles with this WHERE clause:

        WHERE r_project=%(r_project)s AND
              (r_quality IS NULL OR r_quality=%(not_a_class)s) AND
              (r_importance IS NULL OR r_importance=%(not_a_class)s)

So the bug is that articles in different namespaces like Category pages:

---------+---------+
| r_project | r_namespace | r_article  | r_quality      | r_quality_timestamp  | r_importance | r_importance_timestamp | r_score |
+-----------+-------------+------------+----------------+----------------------+--------------+------------------------+---------+
| Theatre   |          14 | 1725_plays | Category-Class | 2011-07-07T04:40:07Z | NA-Class     | 2011-07-07T04:40:07Z   |       0 |
| Years     |          14 | 1725_plays | Category-Class | 2016-12-31T09:22:04Z | NA-Class     | 2016-12-31T09:22:04Z   |       0 |
+-----------+-------------+------------+----------------+----------------------+--------------+------------------------+---------+

End up never being deleted!

My guess is that we could change the WHERE clause to include an OR r_namespace > 0 clause.

openzim / wp1

Re-evaluate logic for deleting rows from `ratings` table #738