* galaxy: improve logging and performance reporting
* galaxy: add a full galaxy state test
* galaxy: optimize user self score computation
* galaxy: add 'generate_galaxy_test_data' command for development at scale
* galaxy: big refactor
Main changes:
- Multiple Galaxy objects can now exist at the same time in DB. This allows for ruling a new galaxy while still
displaying the old one.
- The criteria to quickly know whether a user is a possible citizen is now a simple query on picture count. This
avoids a very complicated query to database, that could often result in huge working memory load. With this change,
it should be possible to run the galaxy even on a vanilla Postgres that didn't receive fine tuning for the Sith's
galaxy.
* galaxy: template: make the galaxy graph work and be usable with a lot of stars
- Display focused star and its connections clearly
- Display star label faintly by default for other stars to avoid overloading the graph
- Hide non-focused lanes
- Avoid clicks on non-highlighted, too far stars
- Make the canva adapt its width to initial screen size, doesn't work dynamically
* galaxy: better docstrings
* galaxy: use bulk_create whenever possible
This is a big performance gain, especially for the tests.
Examples:
----
`./manage.py test galaxy.tests.GalaxyTest.test_full_galaxy_state`
Measurements averaged over 3 run on *my machine*™:
Before: 2min15s
After: 1m41s
----
`./manage.py generate_galaxy_test_data --user-pack-count 1`
Before: 48s
After: 25s
----
`./manage.py rule_galaxy` (for 600 citizen, corresponding to 1 user-pack)
Before: 14m4s
After: 12m34s
* core: populate: use a less ambiguous 'timezone.now()'
When running the tests around midnight, the day is changing, leading to some values being offset to the next day
depending on the timezone, and making some tests to fail. This ensure to use a less ambiguous `now` when populating
the database.
* write more extensive documentation
- add documentation to previously documented classes and functions and refactor some of the documented one, in accordance to the PEP257 and ReStructuredText standards ;
- add some type hints ;
- use a NamedTuple for the `Galaxy.compute_users_score` method instead of a raw tuple. Also change a little bit the logic in the function which call the latter ;
- add some additional parameter checks on a few functions ;
- change a little bit the logic of the log level setting for the galaxy related commands.
* galaxy: tests: split Model and View for more efficient data usage
---------
Co-authored-by: maréchal <thgirod@hotmail.com>
* Better usage of cache for group retrieval
* Cache clearing on object deletion or update
* replace signals by save and delete override
* add is_anonymous check in is_owned_by
Add in many is_owned_by(self, user) methods that user is not anonymous. Since many of those functions do db queries, this should reduce a little bit the load of the db.
* Stricter usage of User.is_in_group
Constrain the parameters that can be passed to the function to make sure only a str or an int can be used. Also force to explicitly specify if the group id or the group name is used.
* write test and correct bugs
* remove forgotten populate commands
* Correct test
klmp200
imperosol
thomas girod
Skia
Julien Constant
lsacienne
tleb
