Previously this controller would in multiple places load tons (read:
around 65000) project and/or group IDs into memory. These changes in
combination with the previous commits significantly cut down loading
times of user profile pages and the Atom feeds of users.
These methods no longer include public groups/projects (that don't
belong to the actual user) as this is handled by the various finder
classes now. This also removes the need for passing extra arguments.
Note that memoizing was removed _explicitly_. For whatever reason doing
so messes up the users controller to a point where it claims a certain
user does _not_ have access to certain groups/projects when it does have
access. Existing code shouldn't be affected as these methods are only
called in ways that they'd run queries anyway (e.g. a combination of
"any?" and "each" which would run 2 queries regardless of memoizing).
This class now uses a UNION (when needed) instead of plucking tens of
thousands of project IDs into memory. The tests have also been
re-written to ensure all different use cases are tested properly
(assuming I didn't forget any cases).
The finder has also been broken up into 3 different finder classes:
* ContributedProjectsFinder: class for getting the projects a user
contributed to.
* PersonalProjectsFinder: class for getting the personal projects of a
user.
* ProjectsFinder: class for getting generic projects visible to a given
user.
Previously a lot of the logic of these finders was handled directly in
the users controller.
In the previous setup the GroupsFinder class had two distinct tasks:
1. Finding the projects user A could see
2. Finding the projects of user A that user B could see
Task two was actually handled outside of the GroupsFinder (in the
UsersController) by restricting the returned list of groups to those the
viewed user was a member of. Moving all this logic into a single finder
proved to be far too complex and confusing, hence there are now two
finders:
* GroupsFinder: for finding groups a user can see
* JoinedGroupsFinder: for finding groups that user A is a member of,
restricted to either public groups or groups user B can also see.
This new setup no longer loads any IDs into memory using "pluck",
instead using SQL UNIONs to merge the various datasets together. This
results in greatly improved query performance as well as a reduction of
memory usage.
The old setup was in particular problematic when requesting the
authorized projects _including_ public/internal projects as this would
result in roughly 65000 project IDs being loaded into memory. These IDs
would in turn be passed to other queries.
By moving the default sort order into a separate scope (and calling this
from the default scope) we can more easily re-apply a default order
without having to specify the exact column/ordering all over the place.
This allows retrieving of the list of authorized projects using a single
query, without having to load any IDs into Ruby. This in turn also means
we can remove the method User#authorized_projects_id.
This class can be used to join multiple AcitveRecord::Relation objects
together using a SQL UNION statement. ActiveRecord < 5.0 sadly doesn't
support UNION and existing Gems out there don't handle prepared
statements (e.g. they never incremented the variable bindings).
Instead of using MAX(events.updated_at) we can simply sort the events in
descending order by the "id" column and grab the first row. In other
words, instead of this:
SELECT max(events.updated_at) AS max_id
FROM events
LEFT OUTER JOIN projects ON projects.id = events.project_id
LEFT OUTER JOIN namespaces ON namespaces.id = projects.namespace_id
WHERE events.author_id IS NOT NULL
AND events.project_id IN (13083);
we can use this:
SELECT events.updated_at AS max_id
FROM events
LEFT OUTER JOIN projects ON projects.id = events.project_id
LEFT OUTER JOIN namespaces ON namespaces.id = projects.namespace_id
WHERE events.author_id IS NOT NULL
AND events.project_id IN (13083)
ORDER BY events.id DESC
LIMIT 1;
This has the benefit that on PostgreSQL a backwards index scan can be
used, which due to the "LIMIT 1" will at most process only a single row.
This in turn greatly speeds up the process of grabbing the latest update
time. This can be confirmed by looking at the query plans. The first
query produces the following plan:
Aggregate (cost=43779.84..43779.85 rows=1 width=12) (actual time=2142.462..2142.462 rows=1 loops=1)
-> Index Scan using index_events_on_project_id on events (cost=0.43..43704.69 rows=30060 width=12) (actual time=0.033..2138.086 rows=32769 loops=1)
Index Cond: (project_id = 13083)
Filter: (author_id IS NOT NULL)
Planning time: 1.248 ms
Execution time: 2142.548 ms
The second query in turn produces the following plan:
Limit (cost=0.43..41.65 rows=1 width=16) (actual time=1.394..1.394 rows=1 loops=1)
-> Index Scan Backward using events_pkey on events (cost=0.43..1238907.96 rows=30060 width=16) (actual time=1.394..1.394 rows=1 loops=1)
Filter: ((author_id IS NOT NULL) AND (project_id = 13083))
Rows Removed by Filter: 2104
Planning time: 0.166 ms
Execution time: 1.408 ms
According to the above plans the 2nd query is around 1500 times faster.
However, re-running the first query produces timings of around 80 ms,
making the 2nd query "only" around 55 times faster.
Yorick Peterse
Douwe Maan
Drew Blessing
Dmitriy Zaporozhets
Marin Jankovski
Douwe Maan
Arseny Zarechnev
Zeger-Jan van de Weg