#354

Reviewed-on: #413
Co-authored-by: miteruzo <miteruzo@naver.com>
Co-committed-by: miteruzo <miteruzo@naver.com>

AGENTS.md

@@ -574,8 +574,32 @@ and layout reuse, follow `frontend/AGENTS.md`.
   wording and placement before implementing it.
 - Do not invent replacement copy when removing unrequested wording.
 - Do not create, modify, or run tests unless the user explicitly asks for
-  test work. When the user asks for tests, keep working and rerun them until
-  they pass or the remaining failure is clearly blocked.
+  test work. When the user asks for tests, keep working within the permitted
+  test-file scope and rerun them until they pass or the remaining failure is
+  clearly blocked.
+- Test-only work includes adding, updating, deleting, reorganising, or fixing
+  SyntaxError in tests. During test-only work, do not modify production code.
+- During test-only work, do not change production constants, behaviour, API
+  contracts, validation, routes, authentication, permissions, UI, copy,
+  dependencies, limits, thresholds, defaults, migrations, schema, or
+  environment settings to satisfy tests.
+- Do not make production code match failing tests, mock assumptions, fixtures,
+  snapshots, old expectations, or stale setup. This includes changing
+  production constants merely because a test expects a different value.
+- If test work reveals a production bug, spec mismatch, or missing behaviour,
+  stop without modifying production code and report: the failing test or
+  discovered issue, the related production file, the actual behaviour, the
+  expected behaviour, and why a production change appears necessary.
+- Modify production code for test failures only when the user explicitly asks
+  for that production change. Do not expand a test task into a production task
+  on your own authority.
+- If the user explicitly asks for both production implementation and test
+  updates, implement production code to the confirmed specification first,
+  then add or update tests to verify that specification. Never roll production
+  behaviour back to satisfy old tests.
+- If it is unclear whether the test or the production implementation is stale,
+  or a test cannot be corrected without changing production code, ask the user
+  instead of guessing.
 
 ## Backend rules
 

backend/db/migrate/20260727123600_add_tags_json_to_post_versions.rb

@@ -0,0 +1,279 @@
+class AddTagsJsonToPostVersions < ActiveRecord::Migration[8.0]
+  RESOLUTION_GRACE = 1.second
+  SECTION_LITERAL_PATTERN = /\[[^\[\]\s]*-[^\[\]\s]*\]\z/
+
+  class MigrationPostVersion < ActiveRecord::Base
+    self.table_name = 'post_versions'
+  end
+
+  class MigrationTag < ActiveRecord::Base
+    self.table_name = 'tags'
+  end
+
+  class MigrationTagName < ActiveRecord::Base
+    self.table_name = 'tag_names'
+  end
+
+  class MigrationTagVersion < ActiveRecord::Base
+    self.table_name = 'tag_versions'
+  end
+
+  class MigrationNicoTagVersion < ActiveRecord::Base
+    self.table_name = 'nico_tag_versions'
+  end
+
+  def up
+    add_column :post_versions, :tags_json, :json, after: :tags
+    MigrationPostVersion.reset_column_information
+
+    backfill_missing_initial_tag_versions!
+    intervals_by_name = build_intervals_by_name
+
+    say_with_time 'Backfilling post_versions.tags_json' do
+      MigrationPostVersion
+          .where(tags_json: nil)
+          .find_each(batch_size: 500) do |version|
+        version.update_columns(
+            tags_json: build_tags_json(version, intervals_by_name))
+      end
+    end
+
+    assert_backfill_complete!
+    change_column_null :post_versions, :tags_json, false
+  end
+
+  def down
+    remove_column :post_versions, :tags_json
+  end
+
+  private
+
+  def backfill_missing_initial_tag_versions!
+    say_with_time 'Backfilling missing initial tag versions' do
+      tag_rows = missing_initial_version_rows(MigrationTagVersion, nico: false)
+      nico_rows = missing_initial_version_rows(MigrationNicoTagVersion, nico: true)
+
+      MigrationTagVersion.insert_all!(tag_rows) if tag_rows.any?
+      MigrationNicoTagVersion.insert_all!(nico_rows) if nico_rows.any?
+
+      tag_rows.length + nico_rows.length
+    end
+  end
+
+  def missing_initial_version_rows version_class, nico:
+    first_versions =
+        version_class
+        .order(:tag_id, :version_no)
+        .to_a
+        .group_by(&:tag_id)
+        .transform_values(&:first)
+    rows = []
+
+    MigrationTag.find_each do |tag|
+      next if (tag.category == 'nico') != nico
+
+      first_version = first_versions[tag.id]
+      next unless first_version
+      next if valid_initial_version?(first_version)
+
+      assert_inferable_initial_version!(tag, first_version)
+      rows << initial_version_row(tag, first_version, nico:)
+    end
+
+    rows
+  end
+
+  def valid_initial_version? version
+    version.version_no == 1 && version.event_type == 'create'
+  end
+
+  def assert_inferable_initial_version! tag, version
+    inferable =
+        version.version_no == 2 &&
+        version.event_type == 'discard' &&
+        tag.created_at < version.created_at
+    return if inferable
+
+    details = [
+        "tag_id=#{ tag.id }",
+        "version_no=#{ version.version_no }",
+        "event_type=#{ version.event_type.inspect }",
+        "tag_created_at=#{ tag.created_at.iso8601(6) }",
+        "version_created_at=#{ version.created_at.iso8601(6) }"]
+
+    raise "Cannot infer initial tag version: #{ details.join(', ') }"
+  end
+
+  def initial_version_row tag, discard_version, nico:
+    row = {
+        tag_id: tag.id,
+        version_no: 1,
+        event_type: 'create',
+        name: discard_version.name,
+        created_at: tag.created_at,
+        created_by_user_id: nil }
+
+    if nico
+      return row.merge(linked_tags: discard_version.linked_tags)
+    end
+
+    row.merge(
+        category: discard_version.category,
+        aliases: discard_version.aliases,
+        parent_tag_ids: discard_version.parent_tag_ids,
+        deprecated_at: discard_version.deprecated_at)
+  end
+
+  def build_intervals_by_name
+    intervals_by_name = Hash.new { |hash, name| hash[name] = [] }
+    versions_by_kind = {
+        tag: versions_by_tag_id(MigrationTagVersion),
+        nico: versions_by_tag_id(MigrationNicoTagVersion) }
+    current_names = current_names_by_tag_id
+
+    MigrationTag.find_each do |tag|
+      nico = tag.category == 'nico'
+      kind = nico ? :nico : :tag
+      versions = versions_by_kind.fetch(kind).fetch(tag.id, [])
+
+      intervals_for(
+          tag,
+          versions,
+          current_name: current_names.fetch(tag.id),
+          nico:).each do |interval|
+        name = interval.delete(:name)
+        intervals_by_name[name] << interval
+      end
+    end
+
+    intervals_by_name
+  end
+
+  def versions_by_tag_id version_class
+    version_class
+        .order(:tag_id, :version_no)
+        .to_a
+        .group_by(&:tag_id)
+  end
+
+  def current_names_by_tag_id
+    MigrationTagName
+        .joins('INNER JOIN tags ON tags.tag_name_id = tag_names.id')
+        .pluck('tags.id', 'tag_names.name')
+        .to_h
+  end
+
+  def intervals_for tag, versions, current_name:, nico:
+    if versions.empty?
+      return [{
+          name: current_name,
+          tag_id: tag.id,
+          version_no: tag.version_no,
+          category: nico ? 'nico' : tag.category,
+          from: tag.created_at,
+          to: tag.discarded_at }]
+    end
+
+    versions.each_with_index.filter_map do |version, index|
+      next if version.event_type == 'discard'
+
+      {
+          name: version.name,
+          tag_id: tag.id,
+          version_no: version.version_no,
+          category: nico ? 'nico' : version.category,
+          from: version.created_at,
+          to: versions[index + 1]&.created_at || tag.discarded_at }
+    end
+  end
+
+  def build_tags_json version, intervals_by_name
+    entries = version.tags.to_s.split.map do |literal|
+      name = tag_name_from_literal(literal)
+      interval = resolve_tag!(intervals_by_name.fetch(name, []), name:, version:)
+
+      { 'id' => interval.fetch(:tag_id),
+        'version_no' => interval.fetch(:version_no),
+        'name' => name,
+        'category' => interval.fetch(:category),
+        'sections' => [] }
+    end
+
+    assert_unique_tag_ids!(version, entries)
+
+    entries.sort_by { |entry| entry.fetch('id') }
+  end
+
+  def tag_name_from_literal literal
+    name = literal.dup
+    name.sub!(SECTION_LITERAL_PATTERN, '') while name.match?(
+        SECTION_LITERAL_PATTERN)
+
+    if name.empty? || name.include?('[') || name.include?(']')
+      raise "Invalid legacy tag literal: #{ literal.inspect }"
+    end
+
+    name
+  end
+
+  def resolve_tag! intervals, name:, version:
+    time = version.created_at
+    candidates = intervals.select do |interval|
+      interval.fetch(:from) <= time &&
+        (interval[:to].nil? || time < interval.fetch(:to))
+    end
+
+    candidates = future_candidates(intervals, time) if candidates.empty?
+
+    return candidates.first if candidates.one?
+
+    candidate_versions = candidates.map do |candidate|
+      [candidate.fetch(:tag_id), candidate.fetch(:version_no)]
+    end
+    details = [
+        "post_version_id=#{ version.id }",
+        "post_id=#{ version.post_id }",
+        "name=#{ name.inspect }",
+        "created_at=#{ time.iso8601(6) }",
+        "candidates=#{ candidate_versions.inspect }"].join(', ')
+
+    raise "Could not resolve tag snapshot: #{ details }"
+  end
+
+  def future_candidates intervals, time
+    candidates = intervals.select do |interval|
+      interval.fetch(:from) > time &&
+        interval.fetch(:from) <= time + RESOLUTION_GRACE
+    end
+    return [] if candidates.empty?
+
+    nearest_from = candidates.map { |interval| interval.fetch(:from) }.min
+
+    candidates.select do |interval|
+      interval.fetch(:from) == nearest_from
+    end
+  end
+
+  def assert_unique_tag_ids! version, entries
+    duplicate_tag_ids =
+        entries
+        .map { |entry| entry.fetch('id') }
+        .tally
+        .select { |_tag_id, count| count > 1 }
+        .keys
+    return if duplicate_tag_ids.empty?
+
+    details = [
+        "post_version_id=#{ version.id }",
+        "duplicate_tag_ids=#{ duplicate_tag_ids.inspect }"].join(', ')
+
+    raise "Duplicate tag IDs: #{ details }"
+  end
+
+  def assert_backfill_complete!
+    missing_count = MigrationPostVersion.where(tags_json: nil).count
+    return if missing_count.zero?
+
+    raise "#{ missing_count } post versions were not backfilled"
+  end
+end

backend/db/schema.rb

@@ -10,7 +10,7 @@
 #
 # It's strongly recommended that you check this file into your version control system.
 
-ActiveRecord::Schema[8.0].define(version: 2026_07_13_000000) do
+ActiveRecord::Schema[8.0].define(version: 2026_07_27_123600) do
   create_table "active_storage_attachments", charset: "utf8mb4", collation: "utf8mb4_0900_ai_ci", force: :cascade do |t|
     t.string "name", null: false
     t.string "record_type", null: false
@@ -349,6 +349,7 @@ ActiveRecord::Schema[8.0].define(version: 2026_07_13_000000) do
     t.string "url", limit: 768, null: false
     t.string "thumbnail_base", limit: 2000
     t.text "tags", null: false
+    t.json "tags_json", null: false
     t.text "parent_post_ids", null: false
     t.datetime "original_created_from"
     t.datetime "original_created_before"

frontend/src/lib/postNewQueryState.test.ts

@@ -29,15 +29,15 @@ describe ('post new review URL state', () => {
 	.toEqual (['one', 'two'])
   })
 
-  it ('allows at most a 4095-byte request target', () => {
+  it ('allows at most a 6 143 byte request target', () => {
     const baseUrl = 'https://example.com/'
     const baseLength = postNewReviewPathByteLength ([baseUrl])
-    const allowed = `${ baseUrl }${ 'a'.repeat (4_095 - baseLength) }`
+    const allowed = `${ baseUrl }${ 'a'.repeat (6_143 - baseLength) }`
     const denied = `${ allowed }a`
 
-    expect (postNewReviewPathByteLength ([allowed])).toBe (4_095)
+    expect (postNewReviewPathByteLength ([allowed])).toBe (6_143)
     expect (isPostNewReviewPathWithinLimit ([allowed])).toBe (true)
-    expect (postNewReviewPathByteLength ([denied])).toBe (4_096)
+    expect (postNewReviewPathByteLength ([denied])).toBe (6_144)
     expect (isPostNewReviewPathWithinLimit ([denied])).toBe (false)
   })
 })

frontend/src/lib/postNewQueryState.ts

@@ -1,5 +1,5 @@
 const POST_NEW_REVIEW_PATH_PREFIX = '/posts/new?urls='
-const MAX_POST_NEW_REVIEW_TARGET_BYTES = 4_096
+const MAX_POST_NEW_REVIEW_TARGET_BYTES = 6_144
 
 const textEncoder = new TextEncoder ()
 

frontend/src/pages/posts/PostImportSourcePage.test.tsx

@@ -60,7 +60,7 @@ describe ('PostImportSourcePage', () => {
     const input = screen.getByRole ('textbox', { name: '' })
 
     fireEvent.change (input, {
-      target: { value: `https://example.com/${ 'a'.repeat (4_100) }` } })
+      target: { value: `https://example.com/${ 'a'.repeat (6_200) }` } })
 
     expect (screen.getByRole ('button', { name: '次へ' })).toBeDisabled ()
   })