#354

backend/app/controllers/post_versions_controller.rb

@@ -23,17 +23,19 @@ class PostVersionsController < ApplicationController
               prev.post_id = post_versions.post_id
               AND prev.version_no = post_versions.version_no - 1
         SQL
-        .select('post_versions.*', 'prev.title AS prev_title', 'prev.url AS prev_url',
-                'prev.thumbnail_base AS prev_thumbnail_base', 'prev.tags AS prev_tags',
+        .select('post_versions.*',
+                'prev.title AS prev_title',
+                'prev.url AS prev_url',
+                'prev.thumbnail_base AS prev_thumbnail_base',
+                'prev.tags_json AS prev_tags_json',
                 'prev.video_ms AS prev_video_ms',
                 'prev.original_created_from AS prev_original_created_from',
                 'prev.original_created_before AS prev_original_created_before')
     q = q.where('post_versions.post_id = ?', post_id) if post_id
     if tag_name
       escaped = ActiveRecord::Base.sanitize_sql_like(tag_name.name)
-      q = q.where(("CONCAT(' ', post_versions.tags, ' ') LIKE :kw " +
-                   "OR CONCAT(' ', prev.tags, ' ') LIKE :kw"),
-                  kw: "% #{ escaped } %")
+      q = q.where(("JSON_CONTAINTS(post_version.tags_json, JSON_OBJECT('id', :tag_id)) " +
+                   "OR JSON_CONTAINS(prev.tags_json, JSON_OBJECT('id', :tag_id))"), tag_id:)
     end
 
     count = q.except(:select, :order, :limit, :offset).count
@@ -47,56 +49,47 @@ class PostVersionsController < ApplicationController
 
   private
 
+  def tag_literal tag
+    sections = tag.fetch('sections', []).map do |sec|
+      begin_ms = sec.fetch('begin_ms')
+      end_ms = sec['end_ms']
+
+      "[#{ Post.ms_to_time(begin_ms) }-#{ end_ms ? Post.ms_to_time(end_ms) : '' }]"
+    end
+
+    "#{ tag.fetch('name') }#{ sections.join }"
+  end
+
   def serialise_versions rows
     user_ids = rows.map(&:created_by_user_id).compact.uniq
     users_by_id = User.where(id: user_ids).pluck(:id, :name).to_h
 
     rows.map do |row|
-      cur_tags = split_tags(row.tags)
-      prev_tags = split_tags(row.attributes['prev_tags'])
+      cur_tags = split_tags(row.tags_json.map { tag_literal(_1) })
+      prev_tags = split_tags(row.attributes['prev_tags_json'].map { tag_literal(_1) })
 
-      {
-        post_id: row.post_id,
+      { post_id: row.post_id,
         version_no: row.version_no,
         event_type: row.event_type,
-        title: {
-          current: row.title,
-          prev: row.attributes['prev_title']
-        },
-        url: {
-          current: row.url,
-          prev: row.attributes['prev_url']
-        },
-        thumbnail: {
-          current: nil,
-          prev: nil
-        },
-        thumbnail_base: {
-          current: row.thumbnail_base,
-          prev: row.attributes['prev_thumbnail_base']
-        },
-        video_ms: {
-          current: row.video_ms,
-          prev: row.attributes['prev_video_ms']
-        },
+        title: { current: row.title, prev: row.attributes['prev_title'] },
+        url: { current: row.url, prev: row.attributes['prev_url'] },
+        thumbnail: { current: nil, prev: nil },
+        thumbnail_base: { current: row.thumbnail_base,
+                          prev: row.attributes['prev_thumbnail_base'] },
+        video_ms: { current: row.video_ms, prev: row.attributes['prev_video_ms'] },
         tags: build_version_tags(cur_tags, prev_tags),
         original_created_from: {
-          current: row.original_created_from&.iso8601,
-          prev: row.attributes['prev_original_created_from']&.iso8601
-        },
+            current: row.original_created_from&.iso8601,
+            prev: row.attributes['prev_original_created_from']&.iso8601 },
         original_created_before: {
-          current: row.original_created_before&.iso8601,
-          prev: row.attributes['prev_original_created_before']&.iso8601
-        },
+            current: row.original_created_before&.iso8601,
+            prev: row.attributes['prev_original_created_before']&.iso8601 },
         created_at: row.created_at.iso8601,
         created_by_user:
-          if row.created_by_user_id
-            {
-              id: row.created_by_user_id,
-              name: users_by_id[row.created_by_user_id]
-            }
-          end
-      }
+            if row.created_by_user_id
+              { id: row.created_by_user_id,
+                name: users_by_id[row.created_by_user_id] }
+            end }
     end
   end
 

backend/app/controllers/posts_controller.rb

@@ -699,7 +699,10 @@ class PostsController < ApplicationController
   end
 
   def editable_tag_names_from_version version
-    version.tags.to_s.split.reject { |name| name.downcase.start_with?('nico:') }.sort
+    version.tags_json
+        .reject { _1[:name].downcase.start_with?('nico:') }
+        .map { tag_literal(_1) }
+        .sort
   end
 
   def post_snapshot_from_record post

backend/app/models/post.rb

@@ -137,6 +137,21 @@ class Post < ApplicationRecord
     end
   end
 
+  def snapshot_tags_json
+    post_tags
+        .kept
+        .joins(tag: :tag_name)
+        .includes(:sections, tag: :tag_name)
+        .order('tag_names.name')
+        .map do |pt|
+      { id: pt.tag.id,
+        version_no: pt.tag.version_no,
+        name: pt.tag.name,
+        category: pt.tag.category,
+        sections: pt.sections.map { { begin_ms: _1.begin_ms, end_ms: _1.end_ms } } }
+    end
+  end
+
   def self.section_literal section
     end_ms =
         section.end_ms ? Post.ms_to_time(section.end_ms) : ''

backend/app/services/post_version_recorder.rb

@@ -25,6 +25,7 @@ class PostVersionRecorder < VersionRecorder
       thumbnail_base: @record.thumbnail_base,
       video_ms: @record.video_ms,
       tags: @record.snapshot_tag_names.join(' '),
+      tags_json: @record.snapshot_tags_json,
       parent_post_ids: @record.snapshot_parent_post_ids.join(' '),
       original_created_from: @record.original_created_from,
       original_created_before: @record.original_created_before }

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"