AI: address review on PR #461

backend/app/controllers/tags_controller.rb

@@ -218,16 +218,63 @@ class TagsController < ApplicationController
     locale = resolve_locale!
 
     q = params[:q].to_s.strip.sub(/\Anot:/i, '')
-    literal = ActiveRecord::Base.sanitize_sql_like(q)
+    prefix = "#{ ActiveRecord::Base.sanitize_sql_like(q) }%"
-    nico = bool?(:nico, default: true)
-    present = bool?(:present, default: true)
 
-    rows = autocomplete_rows(locale:, pattern: "#{ literal }%", nico:, present:)
+    with_nico = bool?(:nico, default: true)
-    if rows.length < 20 && q.present?
+    present_only = bool?(:present, default: true)
-      suffix_rows = autocomplete_rows(locale:, pattern: "%#{ literal }", nico:, present:)
+
-      rows = (rows + suffix_rows).uniq { [_1['category'], _1['id']] }.first(20)
+    alias_rows =
+        TagName
+        .where(language_code: locale.language_code, primary_flg: false)
+        .where('name LIKE ?', prefix)
+        .pluck(:tag_id, :name)
+
+    matched_alias_by_tag_id = { }
+    tag_ids = []
+    alias_rows.each do |tag_id, alias_name|
+      tag_ids << tag_id
+      matched_alias_by_tag_id[tag_id] ||= alias_name
     end
 
+    base =
+        Tag
+        .joins(:tag_names)
+        .preload(:tag_names, :materials, :deerjikists, :tag_descriptions,
+                 parents: [:tag_names, :materials, :deerjikists, :tag_descriptions])
+        .where(tag_names: { language_code: locale.language_code, primary_flg: true },
+               deprecated_at: nil)
+    base = base.where('tags.post_count > 0') if present_only
+
+    canonical_hit = base.where('tag_names.name LIKE ?', prefix)
+
+    internal_tags = canonical_hit.or(base.where(id: tag_ids.uniq))
+
+    internal_rows =
+        internal_tags
+        .order(Arel.sql('tags.post_count DESC, tag_names.name'))
+        .limit(20)
+        .map { |tag|
+          TagRepr.base(tag, language_code: locale.language_code)
+                 .merge(matched_alias: matched_alias_by_tag_id[tag.id])
+        }
+
+    return render json: internal_rows unless with_nico
+
+    external_base = ExternalTag.all
+    external_base = external_base.where('post_count > 0') if present_only
+
+    external_rows =
+        external_base
+        .where("CONCAT(platform, ':', name) LIKE ? OR name LIKE ?", prefix, prefix)
+        .order(post_count: :desc, name: :asc)
+        .limit(20)
+        .map { external_tag_json(_1) }
+
+    rows =
+    (internal_rows + external_rows)
+        .sort_by { |row| [-row['post_count'], row['name'] || row[:name]] }
+        .first(20)
+
     render json: rows
   end
 
@@ -498,59 +545,6 @@ class TagsController < ApplicationController
 
   private
 
-  def autocomplete_rows locale:, pattern:, nico:, present:
-    alias_rows =
-        TagName
-        .where(language_code: locale.language_code, primary_flg: false)
-        .where('name LIKE ?', pattern)
-        .pluck(:tag_id, :name)
-
-    matched_alias_by_tag_id = { }
-    tag_ids = []
-    alias_rows.each do |tag_id, alias_name|
-      tag_ids << tag_id
-      matched_alias_by_tag_id[tag_id] ||= alias_name
-    end
-
-    base =
-        Tag
-        .joins(:tag_names)
-        .preload(:tag_names, :materials, :deerjikists, :tag_descriptions,
-                 parents: [:tag_names, :materials, :deerjikists, :tag_descriptions])
-        .where(tag_names: { language_code: locale.language_code, primary_flg: true },
-               deprecated_at: nil)
-    base = base.where('tags.post_count > 0') if present
-
-    canonical_hit = base.where('tag_names.name LIKE ?', pattern)
-
-    internal_tags = canonical_hit.or(base.where(id: tag_ids.uniq))
-
-    internal_rows =
-        internal_tags
-        .order(Arel.sql('tags.post_count DESC, tag_names.name'))
-        .limit(20)
-        .map { |tag|
-          TagRepr.base(tag, language_code: locale.language_code)
-                 .merge(matched_alias: matched_alias_by_tag_id[tag.id])
-        }
-
-    return internal_rows unless nico
-
-    external_base = ExternalTag.all
-    external_base = external_base.where('post_count > 0') if present
-
-    external_rows =
-        external_base
-        .where("CONCAT(platform, ':', name) LIKE ? OR name LIKE ?", pattern, pattern)
-        .order(post_count: :desc, name: :asc)
-        .limit(20)
-        .map { external_tag_json(_1) }
-
-    (internal_rows + external_rows)
-        .sort_by { |row| [-row['post_count'], row['name'] || row[:name]] }
-        .first(20)
-  end
-
   def material_filter_param default:
     value = params[:material_filter].to_s.presence
     return default unless ['present', 'missing', 'any'].include?(value)

backend/app/services/post_tag_search.rb

@@ -3,8 +3,8 @@ class PostTagSearch
   def self.literals(locale, raw)
     tokens = raw.split.reject { _1.downcase == 'type:or' }
     names = tokens.map { _1.sub(/\Anot:/i, '') }
-    registered = TagName.where(language_code: locale.language_code, name: names)
+    registered = matching_names(TagName.where(language_code: locale.language_code),
-        .pluck(:name, :tag_id).to_h
+        names, 'tag_names.name', :tag_id).to_h
     primary = TagName.where(language_code: locale.language_code,
         tag_id: registered.values, primary_flg: true).pluck(:tag_id, :name).to_h
     tokens.zip(names).map do |token, name|
@@ -19,12 +19,25 @@ class PostTagSearch
 
   def self.references_for(locale, literals)
     names = literals.map { _1[:name] }.uniq
-    internal = TagName.where(language_code: locale.language_code, name: names)
+    internal = matching_names(TagName.where(language_code: locale.language_code),
-        .pluck(:name, :tag_id).group_by(&:first)
+        names, 'tag_names.name', :tag_id).group_by(&:first)
-    external = names.empty? ? {} : ExternalTag.where(
+    external = matching_names(ExternalTag.all, names,
-        "CONCAT(platform, ':', name) IN (?)", names)
+        "CONCAT(external_tags.platform, ':', external_tags.name)", :id).group_by(&:first)
-        .pluck(Arel.sql("CONCAT(platform, ':', name)"), :id).group_by(&:first)
     names.to_h { |name| [name, [internal.fetch(name, []).map(&:last).uniq,
                               external.fetch(name, []).map(&:last).uniq]] }
   end
+
+  def self.matching_names(scope, names, expression, identifier)
+    return [] if names.empty?
+
+    queries = names.uniq.map do |name|
+      scope.klass.sanitize_sql_array(['SELECT ? AS name', name])
+    end
+    # Return the input spelling so Ruby does not recheck database equivalence.
+    scope.joins(<<~SQL).pluck(Arel.sql('search_names.name'), identifier)
+      INNER JOIN (#{ queries.join(' UNION ALL ') }) AS search_names
+        ON #{ expression } = search_names.name COLLATE utf8mb4_0900_ai_ci
+    SQL
+  end
+  private_class_method :matching_names
 end

backend/spec/requests/posts_spec.rb

@@ -206,6 +206,25 @@ RSpec.describe 'Posts API', type: :request do
     end
 
     context 'when q is provided' do
+      ['SPEC_TAG', 'SpEc_TaG', 'spec_tág', 'MANKO', 'MaNkO', 'mánko'].each do |name|
+        it "matches #{ name } using database collation" do
+          get '/posts', params: { locale: 'ja', tags: name }
+
+          expect(response).to have_http_status(:ok)
+          expect(json.fetch('count')).to eq(2)
+          expect(json.fetch('posts').map { _1.fetch('id') })
+              .to contain_exactly(post_record.id, hit_post.id)
+        end
+
+        it "excludes #{ name } using database collation" do
+          get '/posts', params: { locale: 'ja', tags: "not:#{ name }" }
+
+          expect(response).to have_http_status(:ok)
+          expect(json.fetch('count')).to eq(1)
+          expect(json.fetch('posts').map { _1.fetch('id') }).to eq([miss_post.id])
+        end
+      end
+
       it 'filters posts by q (hit case)' do
         get '/posts', params: { tags: 'spec_tag' }
 
@@ -266,6 +285,26 @@ RSpec.describe 'Posts API', type: :request do
         PostExternalTag.create!(post: miss_post, external_tag:)
       end
 
+      ['NICO:SEARCH_EXTERNAL', 'NiCo:SeArCh_ExTeRnAl', 'nico:séarch_external'].each do |name|
+        it "matches #{ name } using database collation" do
+          get '/posts', params: { locale: 'ja', tags: name }
+
+          expect(response).to have_http_status(:ok)
+          expect(json.fetch('count')).to eq(2)
+          expect(json.fetch('posts').map { _1.fetch('id') })
+              .to contain_exactly(miss_post.id, both_post.id)
+        end
+
+        it "excludes #{ name } using database collation" do
+          get '/posts', params: { locale: 'ja', tags: "not:#{ name }" }
+
+          expect(response).to have_http_status(:ok)
+          expect(json.fetch('count')).to eq(2)
+          expect(json.fetch('posts').map { _1.fetch('id') })
+              .to contain_exactly(post_record.id, hit_post.id)
+        end
+      end
+
       it 'keeps internal name searches independent of colliding external ids' do
         get '/posts', params: { tags: primary_tag_name_for(tag, 'ja').name }
 

backend/spec/requests/tags_spec.rb

@@ -482,7 +482,6 @@ RSpec.describe 'Tags API', type: :request do
         create(:tag, category: :general, primary_name: name, post_count: 1)
         create(:external_tag, name:, post_count: 1)
       end
-      create(:external_tag, name: 'suffix_combined', post_count: 100)
 
       get '/tags/autocomplete', params: { q: 'combined' }
 
@@ -491,39 +490,6 @@ RSpec.describe 'Tags API', type: :request do
       expect(json.map { |row| row.fetch('name') }).to eq(expected.first(20))
     end
 
-    it 'appends primary, alias, and external suffix matches without duplicate resources' do
-      prefix = create(:tag, primary_name: 'tail', post_count: 1)
-      create(:tag_name, :alias, name: 'tail_alias', tag: prefix)
-      create(:tag_name, :alias, name: 'alias_tail', tag: prefix)
-      create(:tag, primary_name: 'primary_tail', post_count: 4)
-      alias_target = create(:tag, primary_name: 'alias_target', post_count: 3)
-      create(:tag_name, :alias, name: 'suffix_tail', tag: alias_target)
-      external = create(:external_tag, id: prefix.id, name: 'external_tail', post_count: 5)
-      create(:tag, primary_name: 'middle_tail_only', post_count: 100)
-      create(:external_tag, name: 'middle_tail_only', post_count: 100)
-
-      get '/tags/autocomplete', params: { q: 'tail' }
-
-      expect(response).to have_http_status(:ok)
-      expect(json.map { |row| row.fetch('name') }).to eq(
-          ['tail', 'nico:external_tail', 'primary_tail', 'alias_target'])
-      expect(json.first.fetch('matched_alias')).to eq('tail_alias')
-      expect(json.last.fetch('matched_alias')).to eq('suffix_tail')
-      expect(json[1]).to include('id' => external.id, 'category' => 'nico')
-    end
-
-    it 'fills only the remaining places with suffix matches' do
-      names = 19.times.map { |number| "capacity_#{ number.to_s.rjust(2, '0') }" }
-      names.each { create(:tag, primary_name: _1, post_count: 1) }
-      create(:tag, primary_name: 'first_capacity', post_count: 3)
-      create(:external_tag, name: 'second_capacity', post_count: 2)
-
-      get '/tags/autocomplete', params: { q: 'capacity' }
-
-      expect(response).to have_http_status(:ok)
-      expect(json.map { |row| row.fetch('name') }).to eq(names + ['first_capacity'])
-    end
-
     it 'excludes external tags but preserves internal alias matches when nico is false' do
       internal = create(:tag, category: :general, primary_name: 'switch_internal', post_count: 1)
       alias_target = create(:tag, category: :general, primary_name: 'alias_target', post_count: 1)
@@ -552,20 +518,12 @@ RSpec.describe 'Tags API', type: :request do
         create(:external_tag, name: literal, post_count: 1)
         create(:tag, category: :general, primary_name: 'literalXmatch', post_count: 2)
         create(:external_tag, name: 'literalXmatch', post_count: 2)
-        suffix = "suffix_literal#{ wildcard }"
-        create(:tag, primary_name: suffix, post_count: 1)
-        suffix_alias = create(:tag, primary_name: 'suffix_alias_target', post_count: 1)
-        create(:tag_name, :alias, name: "alias_#{ suffix }", tag: suffix_alias)
-        create(:external_tag, name: suffix, post_count: 1)
-        create(:tag, primary_name: 'suffix_literalX', post_count: 2)
-        create(:external_tag, name: 'suffix_literalX', post_count: 2)
 
         get '/tags/autocomplete', params: { q: "literal#{ wildcard }" }
 
         expect(json.map { |row| row.fetch('name') })
             .to contain_exactly(
-                literal, primary_tag_name_for(alias_target, 'ja').name, "nico:#{ literal }",
+                literal, primary_tag_name_for(alias_target, 'ja').name, "nico:#{ literal }")
-                suffix, 'suffix_alias_target', "nico:#{ suffix }")
       end
     end