When we use Gitlab::SQL::Pattern, this typically relates to a trigram
index. As the 'tri' indicates, we need at least three characters to be
able to use that index.
Labels don't have a trigram index, because we never allow you to search
for them globally: it's always in the context of a project or a
group. In that context, it's just as fast to search for a single
character (in general) because there is already a pretty specific index
being used.
This optimises searching for users when using queries consisting out of
one or two characters such as "ab". We optimise such cases by searching
for `LOWER(name)` and `LOWER(username)` instead of using `ILIKE`. Using
`LOWER` produces a _much_ better performing query.
For example, when searching for all users matching the term "a" we'd
produce the following plan:
Limit (cost=637.69..637.74 rows=20 width=805) (actual time=41.983..41.995 rows=20 loops=1)
Buffers: shared hit=8330
-> Sort (cost=637.69..638.61 rows=368 width=805) (actual time=41.982..41.990 rows=20 loops=1)
Sort Key: (CASE WHEN ((name)::text = 'a'::text) THEN 0 WHEN ((username)::text = 'a'::text) THEN 1 WHEN ((email)::text = 'a'::text) THEN 2 ELSE 3 END), name
Sort Method: top-N heapsort Memory: 35kB
Buffers: shared hit=8330
-> Bitmap Heap Scan on users (cost=75.47..627.89 rows=368 width=805) (actual time=9.452..41.305 rows=277 loops=1)
Recheck Cond: (((name)::text ~~* 'a'::text) OR ((username)::text ~~* 'a'::text) OR ((email)::text = 'a'::text))
Rows Removed by Index Recheck: 7601
Heap Blocks: exact=7636
Buffers: shared hit=8327
-> BitmapOr (cost=75.47..75.47 rows=368 width=0) (actual time=8.290..8.290 rows=0 loops=1)
Buffers: shared hit=691
-> Bitmap Index Scan on index_users_on_name_trigram (cost=0.00..38.85 rows=180 width=0) (actual time=4.369..4.369 rows=4071 loops=1)
Index Cond: ((name)::text ~~* 'a'::text)
Buffers: shared hit=360
-> Bitmap Index Scan on index_users_on_username_trigram (cost=0.00..34.41 rows=188 width=0) (actual time=3.896..3.896 rows=4140 loops=1)
Index Cond: ((username)::text ~~* 'a'::text)
Buffers: shared hit=328
-> Bitmap Index Scan on users_email_key (cost=0.00..1.94 rows=1 width=0) (actual time=0.022..0.022 rows=0 loops=1)
Index Cond: ((email)::text = 'a'::text)
Buffers: shared hit=3
Planning time: 3.912 ms
Execution time: 42.171 ms
With the changes in this commit we now produce the following plan
instead:
Limit (cost=13257.48..13257.53 rows=20 width=805) (actual time=1.567..1.579 rows=20 loops=1)
Buffers: shared hit=287
-> Sort (cost=13257.48..13280.93 rows=9379 width=805) (actual time=1.567..1.572 rows=20 loops=1)
Sort Key: (CASE WHEN ((name)::text = 'a'::text) THEN 0 WHEN ((username)::text = 'a'::text) THEN 1 WHEN ((email)::text = 'a'::text) THEN 2 ELSE 3 END), name
Sort Method: top-N heapsort Memory: 35kB
Buffers: shared hit=287
-> Bitmap Heap Scan on users (cost=135.66..13007.91 rows=9379 width=805) (actual time=0.194..1.107 rows=277 loops=1)
Recheck Cond: ((lower((name)::text) = 'a'::text) OR (lower((username)::text) = 'a'::text) OR ((email)::text = 'a'::text))
Heap Blocks: exact=277
Buffers: shared hit=287
-> BitmapOr (cost=135.66..135.66 rows=9379 width=0) (actual time=0.152..0.152 rows=0 loops=1)
Buffers: shared hit=10
-> Bitmap Index Scan on yorick_test_users (cost=0.00..124.75 rows=9377 width=0) (actual time=0.101..0.101 rows=277 loops=1)
Index Cond: (lower((name)::text) = 'a'::text)
Buffers: shared hit=4
-> Bitmap Index Scan on index_on_users_lower_username (cost=0.00..1.94 rows=1 width=0) (actual time=0.035..0.035 rows=1 loops=1)
Index Cond: (lower((username)::text) = 'a'::text)
Buffers: shared hit=3
-> Bitmap Index Scan on users_email_key (cost=0.00..1.94 rows=1 width=0) (actual time=0.014..0.014 rows=0 loops=1)
Index Cond: ((email)::text = 'a'::text)
Buffers: shared hit=3
Planning time: 0.303 ms
Execution time: 1.687 ms
Here we can see the new query is 25 times faster compared to the old
query.
In this particular case the use of UNION ALL leads to a better query
plan compared to using 1 big query that uses an OR statement to combine
different data sources.
See https://gitlab.com/gitlab-org/gitlab-ce/issues/38508 for more
information.
This commit introduces the usage of Common Table Expressions (CTEs) to
efficiently retrieve nested group hierarchies, without having to rely on
the "routes" table (which is an _incredibly_ inefficient way of getting
the data). This requires a patch to ActiveRecord (found in the added
initializer) to work properly as ActiveRecord doesn't support WITH
statements properly out of the box.
Unfortunately MySQL provides no efficient way of getting nested groups.
For example, the old routes setup could easily take 5-10 seconds
depending on the amount of "routes" in a database. Providing vastly
different logic for both MySQL and PostgreSQL will negatively impact the
development process. Because of this the various nested groups related
methods return empty relations when used in combination with MySQL.
For project authorizations the logic is split up into two classes:
* Gitlab::ProjectAuthorizations::WithNestedGroups
* Gitlab::ProjectAuthorizations::WithoutNestedGroups
Both classes get the fresh project authorizations (= as they should be
in the "project_authorizations" table), including nested groups if
PostgreSQL is used. The logic of these two classes is quite different
apart from their public interface. This complicates development a bit,
but unfortunately there is no way around this.
This commit also introduces Gitlab::GroupHierarchy. This class can be
used to get the ancestors and descendants of a base relation, or both by
using a UNION. This in turn is used by methods such as:
* Namespace#ancestors
* Namespace#descendants
* User#all_expanded_groups
Again this class relies on CTEs and thus only works on PostgreSQL. The
Namespace methods will return an empty relation when MySQL is used,
while User#all_expanded_groups will return only the groups a user is a
direct member of.
Performance wise the impact is quite large. For example, on GitLab.com
Namespace#descendants used to take around 580 ms to retrieve data for a
particular user. Using CTEs we are able to reduce this down to roughly 1
millisecond, returning the exact same data.
== On The Fly Refreshing
Refreshing of authorizations on the fly (= when
users.authorized_projects_populated was not set) is removed with this
commit. This simplifies the code, and ensures any queries used for
authorizations are not mutated because they are executed in a Rails
scope (e.g. Project.visible_to_user).
This commit includes a migration to schedule refreshing authorizations
for all users, ensuring all of them have their authorizations in place.
Said migration schedules users in batches of 5000, with 5 minutes
between every batch to smear the load around a bit.
== Spec Changes
This commit also introduces some changes to various specs. For example,
some specs for ProjectTeam assumed that creating a personal project
would _not_ lead to the owner having access, which is incorrect. Because
we also no longer refresh authorizations on the fly for new users some
code had to be added to the "empty_project" factory. This chunk of code
ensures that the owner's permissions are refreshed after creating the
project, something that is normally done in Projects::CreateService.
Fix missing Note access checks in by moving Note#search to updated NoteFinder
Split from !2024 to partially solve https://gitlab.com/gitlab-org/gitlab-ce/issues/23867
## Which fixes are in this MR?
⚠️ - Potentially untested
💣 - No test coverage
🚥 - Test coverage of some sort exists (a test failed when error raised)
🚦 - Test coverage of return value (a test failed when nil used)
✅ - Permissions check tested
### Note lookup without access check
- [x] ✅ app/finders/notes_finder.rb:13 :download_code check
- [x] ✅ app/finders/notes_finder.rb:19 `SnippetsFinder`
- [x] ✅ app/models/note.rb:121 [`Issue#visible_to_user`]
- [x] ✅ lib/gitlab/project_search_results.rb:113
- This is the only use of `app/models/note.rb:121` above, but importantly has no access checks at all. This means it leaks MR comments and snippets when those features are `team-only` in addition to the issue comments which would be fixed by `app/models/note.rb:121`.
- It is only called from SearchController where `can?(current_user, :download_code, @project)` is checked, so commit comments are not leaked.
### Previous discussions
- [x] https://dev.gitlab.org/gitlab/gitlabhq/merge_requests/2024/diffs#b915c5267a63628b0bafd23d37792ae73ceae272_13_13 `: download_code` check on commit
- [x] https://dev.gitlab.org/gitlab/gitlabhq/merge_requests/2024/diffs#b915c5267a63628b0bafd23d37792ae73ceae272_19_19 `SnippetsFinder` should be used
- `SnippetsFinder` should check if the snippets feature is enabled -> https://gitlab.com/gitlab-org/gitlab-ce/issues/25223
### Acceptance criteria met?
- [x] Tests added for new code
- [x] TODO comments removed
- [x] Squashed and removed skipped tests
- [x] Changelog entry
- [ ] State Gitlab versions affected and issue severity in description
- [ ] Create technical debt issue for NotesFinder.
- Either split into `NotesFinder::ForTarget` and `NotesFinder::Search` or consider object per notable type such as `NotesFinder::OnIssue`. For the first option could create `NotesFinder::Base` which is either inherited from or which can be included in the other two.
- Avoid case statement anti-pattern in this finder with use of `NotesFinder::OnCommit` etc. Consider something on the finder for this? `Model.finder(user, project)`
- Move `inc_author` to the controller, and implement `related_notes` to replace `non_diff_notes`/`mr_and_commit_notes`
See merge request !2035
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).
GitLab Bot
Mario de la Ossa
Stan Hu
Sean McGivern
Andrew Newdigate
Douglas Barbosa Alexandre
Jacopo
gfyoung
Yorick Peterse
Douwe Maan
Bob Van Landuyt
Hiroyuki Sato
Lin Jen-Shin
Grzegorz Bizon
Douwe Maan