gitlab-org--gitlab-foss/spec/finders/issues_finder_spec.rb

782 lines
26 KiB
Ruby
Raw Normal View History

require 'spec_helper'
describe IssuesFinder do
include_context 'IssuesFinder context'
describe '#execute' do
include_context 'IssuesFinder#execute context'
context 'scope: all' do
let(:scope) { 'all' }
it 'returns all issues' do
expect(issues).to contain_exactly(issue1, issue2, issue3, issue4)
end
context 'assignee filtering' do
let(:issuables) { issues }
it_behaves_like 'assignee ID filter' do
let(:params) { { assignee_id: user.id } }
let(:expected_issuables) { [issue1, issue2] }
end
it_behaves_like 'assignee username filter' do
before do
project2.add_developer(user3)
issue3.assignees = [user2, user3]
end
set(:user3) { create(:user) }
let(:params) { { assignee_username: [user2.username, user3.username] } }
let(:expected_issuables) { [issue3] }
end
it_behaves_like 'no assignee filter' do
set(:user3) { create(:user) }
let(:expected_issuables) { [issue4] }
2018-10-25 22:47:14 -04:00
end
it_behaves_like 'any assignee filter' do
let(:expected_issuables) { [issue1, issue2, issue3] }
2018-10-25 22:47:14 -04:00
end
end
context 'filtering by group_id' do
let(:params) { { group_id: group.id } }
context 'when include_subgroup param not set' do
it 'returns all group issues' do
expect(issues).to contain_exactly(issue1)
end
end
context 'when include_subgroup param is true', :nested_groups do
before do
params[:include_subgroups] = true
end
it 'returns all group and subgroup issues' do
expect(issues).to contain_exactly(issue1, issue4)
end
end
end
context 'filtering by author ID' do
let(:params) { { author_id: user2.id } }
it 'returns issues created by that user' do
expect(issues).to contain_exactly(issue3)
end
end
context 'filtering by milestone' do
let(:params) { { milestone_title: milestone.title } }
it 'returns issues assigned to that milestone' do
expect(issues).to contain_exactly(issue1)
end
end
2017-07-07 11:08:49 -04:00
context 'filtering by group milestone' do
let!(:group) { create(:group, :public) }
let(:group_milestone) { create(:milestone, group: group) }
let!(:group_member) { create(:group_member, group: group, user: user) }
let(:params) { { milestone_title: group_milestone.title } }
before do
project2.update(namespace: group)
issue2.update(milestone: group_milestone)
issue3.update(milestone: group_milestone)
end
it 'returns issues assigned to that group milestone' do
expect(issues).to contain_exactly(issue2, issue3)
end
end
context 'filtering by no milestone' do
2018-10-26 08:37:52 -04:00
let(:params) { { milestone_title: 'None' } }
it 'returns issues with no milestone' do
expect(issues).to contain_exactly(issue2, issue3, issue4)
end
2018-10-26 08:37:52 -04:00
it 'returns issues with no milestone (deprecated)' do
params[:milestone_title] = Milestone::None.title
expect(issues).to contain_exactly(issue2, issue3, issue4)
end
end
context 'filtering by any milestone' do
2018-10-26 08:37:52 -04:00
let(:params) { { milestone_title: 'Any' } }
it 'returns issues with any assigned milestone' do
expect(issues).to contain_exactly(issue1)
end
2018-10-26 08:37:52 -04:00
it 'returns issues with any assigned milestone (deprecated)' do
params[:milestone_title] = Milestone::Any.title
expect(issues).to contain_exactly(issue1)
end
end
context 'filtering by upcoming milestone' do
let(:params) { { milestone_title: Milestone::Upcoming.name } }
let!(:group) { create(:group, :public) }
let!(:group_member) { create(:group_member, group: group, user: user) }
let(:project_no_upcoming_milestones) { create(:project, :public) }
let(:project_next_1_1) { create(:project, :public) }
let(:project_next_8_8) { create(:project, :public) }
let(:project_in_group) { create(:project, :public, namespace: group) }
let(:yesterday) { Date.today - 1.day }
let(:tomorrow) { Date.today + 1.day }
let(:two_days_from_now) { Date.today + 2.days }
let(:ten_days_from_now) { Date.today + 10.days }
let(:milestones) do
[
create(:milestone, :closed, project: project_no_upcoming_milestones),
create(:milestone, project: project_next_1_1, title: '1.1', due_date: two_days_from_now),
create(:milestone, project: project_next_1_1, title: '8.9', due_date: ten_days_from_now),
create(:milestone, project: project_next_8_8, title: '1.2', due_date: yesterday),
create(:milestone, project: project_next_8_8, title: '8.8', due_date: tomorrow),
create(:milestone, group: group, title: '9.9', due_date: tomorrow)
]
end
before do
milestones.each do |milestone|
create(:issue, project: milestone.project || project_in_group, milestone: milestone, author: user, assignees: [user])
end
end
it 'returns issues in the upcoming milestone for each project or group' do
expect(issues.map { |issue| issue.milestone.title }).to contain_exactly('1.1', '8.8', '9.9')
expect(issues.map { |issue| issue.milestone.due_date }).to contain_exactly(tomorrow, two_days_from_now, tomorrow)
end
end
context 'filtering by started milestone' do
let(:params) { { milestone_title: Milestone::Started.name } }
let(:project_no_started_milestones) { create(:project, :public) }
let(:project_started_1_and_2) { create(:project, :public) }
let(:project_started_8) { create(:project, :public) }
let(:yesterday) { Date.today - 1.day }
let(:tomorrow) { Date.today + 1.day }
let(:two_days_ago) { Date.today - 2.days }
let(:three_days_ago) { Date.today - 3.days }
let(:milestones) do
[
create(:milestone, project: project_no_started_milestones, start_date: tomorrow),
create(:milestone, project: project_started_1_and_2, title: '1.0', start_date: two_days_ago),
create(:milestone, project: project_started_1_and_2, title: '2.0', start_date: yesterday),
create(:milestone, project: project_started_1_and_2, title: '3.0', start_date: tomorrow),
create(:milestone, :closed, project: project_started_1_and_2, title: '4.0', start_date: three_days_ago),
create(:milestone, :closed, project: project_started_8, title: '6.0', start_date: three_days_ago),
create(:milestone, project: project_started_8, title: '7.0'),
create(:milestone, project: project_started_8, title: '8.0', start_date: yesterday),
create(:milestone, project: project_started_8, title: '9.0', start_date: tomorrow)
]
end
before do
milestones.each do |milestone|
create(:issue, project: milestone.project, milestone: milestone, author: user, assignees: [user])
end
end
it 'returns issues in the started milestones for each project' do
expect(issues.map { |issue| issue.milestone.title }).to contain_exactly('1.0', '2.0', '8.0')
expect(issues.map { |issue| issue.milestone.start_date }).to contain_exactly(two_days_ago, yesterday, yesterday)
end
end
context 'filtering by label' do
let(:params) { { label_name: label.title } }
it 'returns issues with that label' do
expect(issues).to contain_exactly(issue2)
end
end
context 'filtering by multiple labels' do
let(:params) { { label_name: [label.title, label2.title].join(',') } }
let(:label2) { create(:label, project: project2) }
before do
create(:label_link, label: label2, target: issue2)
end
it 'returns the unique issues with all those labels' do
expect(issues).to contain_exactly(issue2)
end
end
context 'filtering by a label that includes any or none in the title' do
let(:params) { { label_name: [label.title, label2.title].join(',') } }
let(:label) { create(:label, title: 'any foo', project: project2) }
let(:label2) { create(:label, title: 'bar none', project: project2) }
it 'returns the unique issues with all those labels' do
create(:label_link, label: label2, target: issue2)
expect(issues).to contain_exactly(issue2)
end
end
context 'filtering by no label' do
let(:params) { { label_name: described_class::FILTER_NONE } }
it 'returns issues with no labels' do
expect(issues).to contain_exactly(issue1, issue3, issue4)
end
end
context 'filtering by legacy No+Label' do
let(:params) { { label_name: Label::NONE } }
it 'returns issues with no labels' do
expect(issues).to contain_exactly(issue1, issue3, issue4)
end
end
context 'filtering by any label' do
let(:params) { { label_name: described_class::FILTER_ANY } }
it 'returns issues that have one or more label' do
2.times do
create(:label_link, label: create(:label, project: project2), target: issue3)
end
expect(issues).to contain_exactly(issue2, issue3)
end
end
2016-06-27 16:22:19 -04:00
context 'filtering by issue term' do
let(:params) { { search: 'git' } }
it 'returns issues with title and description match for search term' do
expect(issues).to contain_exactly(issue1, issue2)
end
end
context 'filtering by issue term in title' do
let(:params) { { search: 'git', in: 'title' } }
it 'returns issues with title match for search term' do
expect(issues).to contain_exactly(issue1)
end
end
2017-02-17 13:28:32 -05:00
context 'filtering by issues iids' do
let(:params) { { iids: issue3.iid } }
2016-06-27 16:22:19 -04:00
2017-02-17 13:28:32 -05:00
it 'returns issues with iids match' do
2016-06-27 16:22:19 -04:00
expect(issues).to contain_exactly(issue3)
end
end
context 'filtering by state' do
context 'with opened' do
let(:params) { { state: 'opened' } }
it 'returns only opened issues' do
expect(issues).to contain_exactly(issue1, issue2, issue3, issue4)
end
end
context 'with closed' do
let(:params) { { state: 'closed' } }
it 'returns only closed issues' do
expect(issues).to contain_exactly(closed_issue)
end
end
context 'with all' do
let(:params) { { state: 'all' } }
it 'returns all issues' do
expect(issues).to contain_exactly(issue1, issue2, issue3, closed_issue, issue4)
end
end
context 'with invalid state' do
let(:params) { { state: 'invalid_state' } }
it 'returns all issues' do
expect(issues).to contain_exactly(issue1, issue2, issue3, closed_issue, issue4)
end
end
end
context 'filtering by created_at' do
context 'through created_after' do
let(:params) { { created_after: issue3.created_at } }
it 'returns issues created on or after the given date' do
expect(issues).to contain_exactly(issue3)
end
end
context 'through created_before' do
let(:params) { { created_before: issue1.created_at } }
it 'returns issues created on or before the given date' do
expect(issues).to contain_exactly(issue1)
end
end
context 'through created_after and created_before' do
let(:params) { { created_after: issue2.created_at, created_before: issue3.created_at } }
it 'returns issues created between the given dates' do
expect(issues).to contain_exactly(issue2, issue3)
end
end
end
context 'filtering by updated_at' do
context 'through updated_after' do
let(:params) { { updated_after: issue3.updated_at } }
it 'returns issues updated on or after the given date' do
expect(issues).to contain_exactly(issue3)
end
end
context 'through updated_before' do
let(:params) { { updated_before: issue1.updated_at } }
it 'returns issues updated on or before the given date' do
expect(issues).to contain_exactly(issue1)
end
end
context 'through updated_after and updated_before' do
let(:params) { { updated_after: issue2.updated_at, updated_before: issue3.updated_at } }
it 'returns issues updated between the given dates' do
expect(issues).to contain_exactly(issue2, issue3)
end
end
end
context 'filtering by closed_at' do
let!(:closed_issue1) { create(:issue, project: project1, state: :closed, closed_at: 1.week.ago) }
let!(:closed_issue2) { create(:issue, project: project2, state: :closed, closed_at: 1.week.from_now) }
let!(:closed_issue3) { create(:issue, project: project2, state: :closed, closed_at: 2.weeks.from_now) }
context 'through closed_after' do
let(:params) { { state: :closed, closed_after: closed_issue3.closed_at } }
it 'returns issues closed on or after the given date' do
expect(issues).to contain_exactly(closed_issue3)
end
end
context 'through closed_before' do
let(:params) { { state: :closed, closed_before: closed_issue1.closed_at } }
it 'returns issues closed on or before the given date' do
expect(issues).to contain_exactly(closed_issue1)
end
end
context 'through closed_after and closed_before' do
let(:params) { { state: :closed, closed_after: closed_issue2.closed_at, closed_before: closed_issue3.closed_at } }
it 'returns issues closed between the given dates' do
expect(issues).to contain_exactly(closed_issue2, closed_issue3)
end
end
end
2017-08-30 03:48:55 -04:00
context 'filtering by reaction name' do
2018-10-26 23:18:31 -04:00
context 'user searches by no reaction' do
let(:params) { { my_reaction_emoji: 'None' } }
it 'returns issues that the user did not react to' do
expect(issues).to contain_exactly(issue2, issue4)
end
end
context 'user searches by any reaction' do
let(:params) { { my_reaction_emoji: 'Any' } }
it 'returns issues that the user reacted to' do
expect(issues).to contain_exactly(issue1, issue3)
end
end
2017-08-30 03:48:55 -04:00
context 'user searches by "thumbsup" reaction' do
let(:params) { { my_reaction_emoji: 'thumbsup' } }
it 'returns issues that the user thumbsup to' do
expect(issues).to contain_exactly(issue1)
end
end
context 'user2 searches by "thumbsup" reaction' do
let(:search_user) { user2 }
let(:params) { { my_reaction_emoji: 'thumbsup' } }
it 'returns issues that the user2 thumbsup to' do
expect(issues).to contain_exactly(issue2)
end
end
context 'user searches by "thumbsdown" reaction' do
let(:params) { { my_reaction_emoji: 'thumbsdown' } }
it 'returns issues that the user thumbsdown to' do
expect(issues).to contain_exactly(issue3)
end
end
end
context 'filtering by confidential' do
set(:confidential_issue) { create(:issue, project: project1, confidential: true) }
context 'no filtering' do
it 'returns all issues' do
expect(issues).to contain_exactly(issue1, issue2, issue3, issue4, confidential_issue)
end
end
context 'user filters confidential issues' do
let(:params) { { confidential: true } }
it 'returns only confdential issues' do
expect(issues).to contain_exactly(confidential_issue)
end
end
context 'user filters only public issues' do
let(:params) { { confidential: false } }
it 'returns only confdential issues' do
expect(issues).to contain_exactly(issue1, issue2, issue3, issue4)
end
end
end
context 'when the user is unauthorized' do
let(:search_user) { nil }
it 'returns no results' do
expect(issues).to be_empty
end
end
context 'when the user can see some, but not all, issues' do
let(:search_user) { user2 }
it 'returns only issues they can see' do
expect(issues).to contain_exactly(issue2, issue3)
end
end
it 'finds issues user can access due to group' do
group = create(:group)
project = create(:project, group: group)
issue = create(:issue, project: project)
group.add_user(user, :owner)
expect(issues).to include(issue)
end
end
context 'personal scope' do
let(:scope) { 'assigned_to_me' }
it 'returns issue assigned to the user' do
expect(issues).to contain_exactly(issue1, issue2)
end
context 'filtering by project' do
let(:params) { { project_id: project1.id } }
it 'returns issues assigned to the user in that project' do
expect(issues).to contain_exactly(issue1)
end
end
end
context 'when project restricts issues' do
let(:scope) { nil }
it "doesn't return team-only issues to non team members" do
project = create(:project, :public, :issues_private)
issue = create(:issue, project: project)
expect(issues).not_to include(issue)
end
it "doesn't return issues if feature disabled" do
[project1, project2, project3].each do |project|
project.project_feature.update!(issues_access_level: ProjectFeature::DISABLED)
end
expect(issues.count).to eq 0
end
end
context 'external authorization' do
it_behaves_like 'a finder with external authorization service' do
let!(:subject) { create(:issue, project: project) }
let(:project_params) { { project_id: project.id } }
end
end
end
describe '#row_count', :request_store do
it 'returns the number of rows for the default state' do
finder = described_class.new(user)
expect(finder.row_count).to eq(5)
end
it 'returns the number of rows for a given state' do
finder = described_class.new(user, state: 'closed')
expect(finder.row_count).to be_zero
end
end
describe '#with_confidentiality_access_check' do
let(:guest) { create(:user) }
set(:authorized_user) { create(:user) }
set(:project) { create(:project, namespace: authorized_user.namespace) }
set(:public_issue) { create(:issue, project: project) }
set(:confidential_issue) { create(:issue, project: project, confidential: true) }
context 'when no project filter is given' do
let(:params) { {} }
context 'for an anonymous user' do
subject { described_class.new(nil, params).with_confidentiality_access_check }
it 'returns only public issues' do
expect(subject).to include(public_issue)
expect(subject).not_to include(confidential_issue)
end
end
context 'for a user without project membership' do
subject { described_class.new(user, params).with_confidentiality_access_check }
it 'returns only public issues' do
expect(subject).to include(public_issue)
expect(subject).not_to include(confidential_issue)
end
end
context 'for a guest user' do
subject { described_class.new(guest, params).with_confidentiality_access_check }
before do
project.add_guest(guest)
end
it 'returns only public issues' do
expect(subject).to include(public_issue)
expect(subject).not_to include(confidential_issue)
end
end
context 'for a project member with access to view confidential issues' do
subject { described_class.new(authorized_user, params).with_confidentiality_access_check }
it 'returns all issues' do
expect(subject).to include(public_issue, confidential_issue)
end
end
context 'for an admin' do
let(:admin_user) { create(:user, :admin) }
subject { described_class.new(admin_user, params).with_confidentiality_access_check }
it 'returns all issues' do
expect(subject).to include(public_issue, confidential_issue)
end
end
end
context 'when searching within a specific project' do
let(:params) { { project_id: project.id } }
context 'for an anonymous user' do
subject { described_class.new(nil, params).with_confidentiality_access_check }
it 'returns only public issues' do
expect(subject).to include(public_issue)
expect(subject).not_to include(confidential_issue)
end
it 'does not filter by confidentiality' do
expect(Issue).not_to receive(:where).with(a_string_matching('confidential'), anything)
subject
end
end
context 'for a user without project membership' do
subject { described_class.new(user, params).with_confidentiality_access_check }
it 'returns only public issues' do
expect(subject).to include(public_issue)
expect(subject).not_to include(confidential_issue)
end
it 'filters by confidentiality' do
expect(Issue).to receive(:where).with(a_string_matching('confidential'), anything)
subject
end
end
context 'for a guest user' do
subject { described_class.new(guest, params).with_confidentiality_access_check }
before do
project.add_guest(guest)
end
it 'returns only public issues' do
expect(subject).to include(public_issue)
expect(subject).not_to include(confidential_issue)
end
it 'filters by confidentiality' do
expect(Issue).to receive(:where).with(a_string_matching('confidential'), anything)
subject
end
end
context 'for a project member with access to view confidential issues' do
subject { described_class.new(authorized_user, params).with_confidentiality_access_check }
it 'returns all issues' do
expect(subject).to include(public_issue, confidential_issue)
end
it 'does not filter by confidentiality' do
expect(Issue).not_to receive(:where).with(a_string_matching('confidential'), anything)
subject
end
end
context 'for an admin' do
let(:admin_user) { create(:user, :admin) }
subject { described_class.new(admin_user, params).with_confidentiality_access_check }
it 'returns all issues' do
expect(subject).to include(public_issue, confidential_issue)
end
it 'does not filter by confidentiality' do
expect(Issue).not_to receive(:where).with(a_string_matching('confidential'), anything)
subject
end
end
end
end
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
describe '#use_cte_for_search?' do
let(:finder) { described_class.new(nil, params) }
before do
allow(Gitlab::Database).to receive(:postgresql?).and_return(true)
stub_feature_flags(attempt_group_search_optimizations: true)
end
context 'when there is no search param' do
let(:params) { { attempt_group_search_optimizations: true } }
it 'returns false' do
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
expect(finder.use_cte_for_search?).to be_falsey
end
end
context 'when the database is not Postgres' do
let(:params) { { search: 'foo', attempt_group_search_optimizations: true } }
before do
allow(Gitlab::Database).to receive(:postgresql?).and_return(false)
end
it 'returns false' do
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
expect(finder.use_cte_for_search?).to be_falsey
end
end
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
context 'when the force_cte param is falsey' do
let(:params) { { search: 'foo' } }
it 'returns false' do
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
expect(finder.use_cte_for_search?).to be_falsey
end
end
context 'when the attempt_group_search_optimizations flag is disabled' do
let(:params) { { search: 'foo', attempt_group_search_optimizations: true } }
before do
stub_feature_flags(attempt_group_search_optimizations: false)
end
it 'returns false' do
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
expect(finder.use_cte_for_search?).to be_falsey
end
end
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
context 'when attempt_group_search_optimizations is unset and attempt_project_search_optimizations is set' do
let(:params) { { search: 'foo', attempt_project_search_optimizations: true } }
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
context 'and the corresponding feature flag is disabled' do
before do
stub_feature_flags(attempt_project_search_optimizations: false)
end
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
it 'returns false' do
expect(finder.use_cte_for_search?).to be_falsey
end
end
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
context 'and the corresponding feature flag is enabled' do
before do
stub_feature_flags(attempt_project_search_optimizations: true)
end
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
it 'returns true' do
expect(finder.use_cte_for_search?).to be_truthy
end
end
end
context 'when all conditions are met' do
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
let(:params) { { search: 'foo', attempt_group_search_optimizations: true } }
it 'returns true' do
Extend CTE search optimisation to projects When we use the `search` param on an `IssuableFinder`, we can run into issues. We have trigram indexes to support these searches. On GitLab.com, we often see Postgres's optimiser prioritise the (global) trigram indexes over the index on `project_id`. For group and project searches, we know that it will be quicker to filter by `project_id` first, as it returns fewer rows in most cases. For group issues search, we ran into this issue previously, and went through the following iterations: 1. Use a CTE on the project IDs as an optimisation fence. This prevents the planner from disregarding the index on `project_id`. Unfortunately it breaks some types of sorting, like priority and popularity, as they sort on a joined table. 2. Use a subquery for listing issues, and a CTE for counts. The subquery - in the case of group lists - didn't help as much as the CTE, but was faster than not including it. We can safely use a CTE for counts as they don't have sorting. Now, however, we're seeing the same issue in a project context. The subquery doesn't help at all there (it would only return one row, after all). In an attempt to keep total code complexity under control, this commit removes the subquery optimisation and applies the CTE optimisation only for sorts we know that are safe. This means that for more complicated sorts (like priority and popularity), the search will continue to be very slow. If this is a high-priority issue, we can consider introducing further optimisations, but this finder is already very complicated and additional complexity has a cost. The group CTE optimisation is controlled by the same feature flag as before, `attempt_group_search_optimizations`, which is enabled by default. The new project CTE optimisation is controlled by a new feature flag, `attempt_project_search_optimizations`, which is disabled by default.
2019-04-03 05:46:13 -04:00
expect(finder.use_cte_for_search?).to be_truthy
end
end
end
end