Include joins into names suggestions

For database sourced product intelligence metrics
we should include joined relations into names
suggestons to make them more descriptive

lib/gitlab/usage/metrics/names_suggestions/generator.rb

@@ -48,48 +48,144 @@ module Gitlab
 
             def name_suggestion(relation:, column: nil, prefix: nil, distinct: nil)
               parts = [prefix]
-
-              if column
-                parts << parse_target(column)
+              arel_column = arelize_column(relation, column)
+
+              # nil as column indicates that the counting would use fallback value of primary key.
+              # Because counting primary key from relation is the conceptual equal to counting all
+              # records from given relation, in order to keep name suggestion more condensed
+              # primary key column is skipped.
+              # eg: SELECT COUNT(id) FROM issues would translate as count_issues and not
+              # as count_id_from_issues since it does not add more information to the name suggestion
+              if arel_column != Arel::Table.new(relation.table_name)[relation.primary_key]
+                parts << arel_column.name
                 parts << 'from'
               end
 
-              source = parse_source(relation)
-              constraints = parse_constraints(relation: relation, column: column, distinct: distinct)
-
-              if constraints.include?(source)
+              arel = arel_query(relation: relation, column: arel_column, distinct: distinct)
+              constraints = parse_constraints(relation: relation, arel: arel)
+
+              # In some cases due to performance reasons metrics are instrumented with joined relations
+              # where relation listed in FROM statement is not the one that includes counted attribute
+              # in such situations to make name suggestion more intuitive source should be inferred based
+              # on the relation that provide counted attribute
+              # EG: SELECT COUNT(deployments.environment_id) FROM clusters
+              #       JOIN deployments ON deployments.cluster_id = cluster.id
+              # should be translated into:
+              #   count_environment_id_from_deployments_with_clusters
+              # instead of
+              #   count_environment_id_from_clusters_with_deployments
+              actual_source = parse_source(relation, arel_column)
+
+              if constraints.include?(actual_source)
                 parts << "<adjective describing: '#{constraints}'>"
               end
 
-              parts << source
+              parts << actual_source
+              parts += process_joined_relations(actual_source, arel, relation)
               parts.compact.join('_')
             end
 
-            def parse_constraints(relation:, column: nil, distinct: nil)
+            def parse_constraints(relation:, arel:)
               connection = relation.connection
               ::Gitlab::Usage::Metrics::NamesSuggestions::RelationParsers::Constraints
                 .new(connection)
-                .accept(arel(relation: relation, column: column, distinct: distinct), collector(connection))
+                .accept(arel, collector(connection))
                 .value
             end
 
-            def parse_target(column)
-              if column.is_a?(Arel::Attribute)
-                "#{column.relation.name}.#{column.name}"
-              else
+            # TODO: joins with `USING` keyword
+            def process_joined_relations(actual_source, arel, relation)
+              joins = parse_joins(connection: relation.connection, arel: arel)
+              return [] unless joins.any?
+
+              sources = [relation.table_name, *joins.map { |join| join[:source] }]
+              joins = extract_joins_targets(joins, sources)
+
+              relations = if actual_source != relation.table_name
+                            build_relations_tree(joins + [{ source: relation.table_name }], actual_source)
+                          else
+                            # in case where counter attribute comes from joined relations, the relations
+                            # diagram has to be built bottom up, thus source and target are reverted
+                            build_relations_tree(joins + [{ source: relation.table_name }], actual_source, source_key: :target, target_key: :source)
+                          end
+
+              collect_join_parts(relations[actual_source])
+            end
+
+            def parse_joins(connection:, arel:)
+              ::Gitlab::Usage::Metrics::NamesSuggestions::RelationParsers::Joins
+                .new(connection)
+                .accept(arel)
+            end
+
+            def extract_joins_targets(joins, sources)
+              joins.map do |join|
+                source_regex = /(#{join[:source]})\.(\w+_)*id/i
+
+                tables_except_src = (sources - [join[:source]]).join('|')
+                target_regex = /(?<target>#{tables_except_src})\.(\w+_)*id/i
+
+                join_cond_regex = /(#{source_regex}\s+=\s+#{target_regex})|(#{target_regex}\s+=\s+#{source_regex})/i
+                matched = join_cond_regex.match(join[:constraints])
+
+                join[:target] = matched[:target] if matched
+                join
+              end
+            end
+
+            def build_relations_tree(joins, parent, source_key: :source, target_key: :target)
+              return [] if joins.blank?
+
+              tree = {}
+              tree[parent] = []
+
+              joins.each do |join|
+                if join[source_key] == parent
+                  tree[parent] << build_relations_tree(joins - [join], join[target_key], source_key: source_key, target_key: target_key)
+                end
+              end
+              tree
+            end
+
+            def collect_join_parts(joined_relations, parts = [], conjunctions = %w[with having including].cycle)
+              conjunction = conjunctions.next
+              joined_relations.each do |subtree|
+                subtree.each do |parent, children|
+                  parts << "<#{conjunction}>"
+                  parts << parent
+                  collect_join_parts(children, parts, conjunctions)
+                end
+              end
+              parts
+            end
+
+            def arelize_column(relation, column)
+              case column
+              when Arel::Attribute
                 column
+              when NilClass
+                Arel::Table.new(relation.table_name)[relation.primary_key]
+              when String
+                if column.include?('.')
+                  table, col = column.split('.')
+                  Arel::Table.new(table)[col]
+                else
+                  Arel::Table.new(relation.table_name)[column]
+                end
+              when Symbol
+                arelize_column(relation, column.to_s)
               end
             end
 
-            def parse_source(relation)
-              relation.table_name
+            def parse_source(relation, column)
+              column.relation.name || relation.table_name
             end
 
             def collector(connection)
               Arel::Collectors::SubstituteBinds.new(connection, Arel::Collectors::SQLString.new)
             end
 
-            def arel(relation:, column: nil, distinct: nil)
+            def arel_query(relation:, column: nil, distinct: nil)
               column ||= relation.primary_key
 
               if column.is_a?(Arel::Attribute)

spec/lib/gitlab/usage/metrics/names_suggestions/generator_spec.rb

@@ -12,7 +12,7 @@ RSpec.describe Gitlab::Usage::Metrics::NamesSuggestions::Generator do
   describe '#generate' do
     shared_examples 'name suggestion' do
       it 'return correct name' do
-        expect(described_class.generate(key_path)).to eq name_suggestion
+        expect(described_class.generate(key_path)).to match name_suggestion
       end
     end
 
@@ -20,7 +20,7 @@ RSpec.describe Gitlab::Usage::Metrics::NamesSuggestions::Generator do
       it_behaves_like 'name suggestion' do
         # corresponding metric is collected with count(Board)
         let(:key_path) { 'counts.boards' }
-        let(:name_suggestion) { 'count_boards' }
+        let(:name_suggestion) { /count_boards/ }
       end
     end
 
@@ -28,7 +28,32 @@ RSpec.describe Gitlab::Usage::Metrics::NamesSuggestions::Generator do
       it_behaves_like 'name suggestion' do
         # corresponding metric is collected with distinct_count(ZoomMeeting, :issue_id)
         let(:key_path) { 'counts.issues_using_zoom_quick_actions' }
-        let(:name_suggestion) { 'count_distinct_issue_id_from_zoom_meetings' }
+        let(:name_suggestion) { /count_distinct_issue_id_from_zoom_meetings/ }
+      end
+    end
+
+    context 'joined relations' do
+      context 'counted attribute comes from  joined relation' do
+        it_behaves_like 'name suggestion' do
+          # corresponding metric is collected with:
+          # distinct_count(
+          #   ::Clusters::Applications::Ingress.modsecurity_enabled.logging
+          #     .joins(cluster: :deployments)
+          #     .merge(::Clusters::Cluster.enabled)
+          #     .merge(Deployment.success),
+          #   ::Deployment.arel_table[:environment_id]
+          # )
+          let(:key_path) { 'counts.ingress_modsecurity_logging' }
+          let(:name_suggestion) { /count_distinct_environment_id_from_<adjective describing\: '\(clusters_applications_ingress\.modsecurity_enabled = TRUE AND clusters_applications_ingress\.modsecurity_mode = \d+ AND clusters.enabled = TRUE AND deployments.status = \d+\)'>_deployments_<with>_clusters_<having>_clusters_applications_ingress/ }
+        end
+      end
+
+      context 'counted attribute comes from source relation' do
+        it_behaves_like 'name suggestion' do
+          # corresponding metric is collected with count(Issue.with_alert_management_alerts.not_authored_by(::User.alert_bot), start: issue_minimum_id, finish: issue_maximum_id)
+          let(:key_path) { 'counts.issues_created_manually_from_alerts' }
+          let(:name_suggestion) { /count_<adjective describing\: '\(issues\.author_id != \d+\)'>_issues_<with>_alert_management_alerts/ }
+        end
       end
     end
 
@@ -36,7 +61,7 @@ RSpec.describe Gitlab::Usage::Metrics::NamesSuggestions::Generator do
       it_behaves_like 'name suggestion' do
         # corresponding metric is collected with sum(JiraImportState.finished, :imported_issues_count)
         let(:key_path) { 'counts.jira_imports_total_imported_issues_count' }
-        let(:name_suggestion) { "sum_imported_issues_count_from_<adjective describing: '(jira_imports.status = 4)'>_jira_imports" }
+        let(:name_suggestion) { /sum_imported_issues_count_from_<adjective describing\: '\(jira_imports\.status = \d+\)'>_jira_imports/ }
       end
     end
 
@@ -44,7 +69,7 @@ RSpec.describe Gitlab::Usage::Metrics::NamesSuggestions::Generator do
       it_behaves_like 'name suggestion' do
         # corresponding metric is collected with add(data[:personal_snippets], data[:project_snippets])
         let(:key_path) { 'counts.snippets' }
-        let(:name_suggestion) { "add_count_<adjective describing: '(snippets.type = 'PersonalSnippet')'>_snippets_and_count_<adjective describing: '(snippets.type = 'ProjectSnippet')'>_snippets" }
+        let(:name_suggestion) { /add_count_<adjective describing\: '\(snippets\.type = 'PersonalSnippet'\)'>_snippets_and_count_<adjective describing\: '\(snippets\.type = 'ProjectSnippet'\)'>_snippets/ }
       end
     end
 
@@ -52,7 +77,7 @@ RSpec.describe Gitlab::Usage::Metrics::NamesSuggestions::Generator do
       it_behaves_like 'name suggestion' do
         # corresponding metric is collected with redis_usage_data { unique_visit_service.unique_visits_for(targets: :analytics) }
         let(:key_path) { 'analytics_unique_visits.analytics_unique_visits_for_any_target' }
-        let(:name_suggestion) { '<please fill metric name>' }
+        let(:name_suggestion) { /<please fill metric name>/ }
       end
     end
 
@@ -60,7 +85,7 @@ RSpec.describe Gitlab::Usage::Metrics::NamesSuggestions::Generator do
       it_behaves_like 'name suggestion' do
         # corresponding metric is collected with alt_usage_data(fallback: nil) { operating_system }
         let(:key_path) { 'settings.operating_system' }
-        let(:name_suggestion) { '<please fill metric name>' }
+        let(:name_suggestion) { /<please fill metric name>/ }
       end
     end
   end