Work towards replacing recursive namespace queries with linear queries.
This commit tracks the path to each namespace, and future work will
replace recursive queries with queries that utilize this new column.

Move recursive namespace query to INNER JOIN

This change adds validation to check the nature
of parent for Group & Namespace records

After a successful enabling of the recursive_approach_for_all_projects
feature flag in production, the feature flag can be removed from the
code.

Drew Blessing authored 4 years ago and Drew Blessing committed 4 years ago

Migrate the column to namespace_settings in preparation for
cascading namespace settings support for this setting.

Removes the validates_associated call which calls `#valid?` on each
runner of the group which triggers an N+1 query

This improves the validation for RunnerNamespace so that we don't need
to fire multiple SQL queries

David Fernandez authored 4 years ago and Alexis Kalderimis committed 4 years ago

The backend will now consider only root namespaces which will
properly exclude subgroups.

to replace ci_syntax_templates experiment

alex pooley authored 4 years ago and Alexis Kalderimis committed 4 years ago

Refactor Namespace traversal queries in to their own module. Moved
related specs to a shared context.

Because we search for permissioned labels which all have the same
top-level group, we can preset root_ancestor for them before permission
check. This optimization saves us extra query to fetch root ancestor
(called from group policy).

Because it is no longer needed

With individual columns for actions

Create table group_package_settings with
boolean setting to allow or disallow Maven
package duplicates and a text setting to
store a regex of exceptions to the disallowed
duplicates.

Add graphql foundation to update and fetch
the new settings.

For tracking onboarding progress

This table will be used to record any actions associated with a
namespace when an action happened for the first time. Actions could be
the first MR that got pushed, the first issue that got created or the
first time a trial started.

Mario de la Ossa authored 4 years ago and Mayra Cabrera committed 4 years ago

Up to now we have not been taking uploads into consideration for
computing total storage usage for repositories. This changes that.

Add 2 feature flags:
* 1 for using deployments
* 1 for allowing to use deployments stored in disk

Also rename old feature flag to match the naming schema

This change introduces a fix to Groups API search auery parameter that
was returning invalid results.

Alan (Maciej) Paruszewski authored 4 years ago and Tiger Watson committed 4 years ago

This change adds new column to fuzzy-search in Groups API to filter on
full path.

The recursive query for personal namespaces is not performing efficient,
there we'll not use the recursive project lookup for personal
namespaces.

Add a recursive approach to load all projects for a namespace next to
the existing route path wildcard approach. The new approach is guarded
by a feature flag.

Include namespace's route and gitlab_subscription's hosted_plan since
those are required for every namespace and gitlab_subscription.

Change code and add validations according to code review
Regactor helper methods

When we change a namespace's path, we only need to move pages
directories on disk for its projects if it has any projects with
deployed pages sites.

Vladlena Shumilo authored 4 years ago and Peter Leitzen committed 4 years ago

Update the logic that assigns a default group
name for the new signup flow

mksionek authored 4 years ago and Imre (Admin) committed 4 years ago

Fix schema file

Fix schema file

Add new settings while creating new record

Fix specs

Fix structure file

Remove obsolete comment

Add changelog entry

Fix specs

Fix rubocop offences

Dmitriy Zaporozhets authored 4 years ago and E'zeki&el Kigbo committed 4 years ago

This will affect Group URL and username (in URL).

This will allow to keep one letter routes for service needs.
Also it should improve autocomplete quality. Also most of 1
and 2 character length usernames are usually taken by bots.

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

Rajendra Kadam authored 4 years ago and Peter Leitzen committed 4 years ago

author Rajendra Kadam <rajendrakadam249@gmail.com> 1580653953 +0530
committer Rajendra Kadam <rajendrakadam249@gmail.com> 1586670416 +0530

parent 6c39d8d1
author Rajendra Kadam <rajendrakadam249@gmail.com> 1580653953 +0530
committer Rajendra Kadam <rajendrakadam249@gmail.com> 1586670412 +0530

parent 6c39d8d1
author Rajendra Kadam <rajendrakadam249@gmail.com> 1580653953 +0530
committer Rajendra Kadam <rajendrakadam249@gmail.com> 1586670402 +0530

parent 6c39d8d1
author Rajendra Kadam <rajendrakadam249@gmail.com> 1580653953 +0530
committer Rajendra Kadam <rajendrakadam249@gmail.com> 1586670397 +0530

parent 6c39d8d1
author Rajendra Kadam <rajendrakadam249@gmail.com> 1580653953 +0530
committer Rajendra Kadam <rajendrakadam249@gmail.com> 1586670392 +0530

Add custom emoji schema and model spec

Add custom emoji factory for specs

Add migration file

Add custom emoji model spec

update emoji specs

Fix failures in model

Remove index creation using concurrent algorithm

Change db/schema.rb

Revert "Change db/schema.rb"

This reverts commit a5d492f738b839b165dd7e3fff5f02bedb9aabab.

In this commit we add the total snippets size to
project and namespace statistics.

Both columns have been ignored for at least 1 release

As MySQL is no longer supported we can omit
mentioning it in search method descriptions

Because we have foreign_key for parent_id now, we are sure
that parent_id can't be nil anymore so we can check parent_id
directly.

Only for free gitlab.com users for group billing we direct the owners
to the new paid signup flow

Use PlanLimits to control the maximum value allowed to be retreived
through offset pagination for API endpoints that support keyset
pagination.

Fabio Pitino authored 4 years ago and Bob Van Landuyt :neckbeard: committed 4 years ago

with a feature flag, ClearSharedRunnersMinutesWorker defines the
batches based on ID ranges and spawns a Ci::BatchResetMinutesWorker
per batch. This will parallelize the processing having each worker
taking less and avoiding running for long time.