Class: SearchIndexer

Inherits:
Object
  • Object
show all
Defined in:
app/services/search_indexer.rb

Defined Under Namespace

Classes: HtmlScrubber

Constant Summary collapse

MIN_POST_BLURB_INDEX_VERSION = 
4
POST_INDEX_VERSION = 
5
TOPIC_INDEX_VERSION = 
4
CATEGORY_INDEX_VERSION = 
3
USER_INDEX_VERSION = 
3
TAG_INDEX_VERSION = 
3
REINDEX_VERSION =

version to apply when issuing a background reindex

0
TS_VECTOR_PARSE_REGEX = 
/('([^']*|'')*'\:)(([0-9]+[A-D]?,?)+)/

Class Method Summary collapse

Class Method Details

.disable ⇒ Object 



16
17
18
 
# File 'app/services/search_indexer.rb', line 16

def self.disable
  @disabled = true
end

.enable ⇒ Object 



20
21
22
 
# File 'app/services/search_indexer.rb', line 20

def self.enable
  @disabled = false
end

.index(obj, force: false) ⇒ Object 



253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
 
# File 'app/services/search_indexer.rb', line 253

def self.index(obj, force: false)
  return if @disabled

  category_name = nil
  tag_names = nil
  topic = nil

  if Topic === obj
    topic = obj
  elsif Post === obj
    topic = obj.topic
  end

  category_name = topic.category&.name if topic

  if topic
    tags = topic.tags.select(:id, :name).to_a

    if tags.present?
      tag_names =
        (tags.map(&:name) + Tag.where(target_tag_id: tags.map(&:id)).pluck(:name)).join(" ")
    end
  end

  if Post === obj && obj.raw.present? &&
       (force || obj.saved_change_to_cooked? || obj.saved_change_to_topic_id?)
    if topic
      SearchIndexer.update_posts_index(
        post_id: obj.id,
        topic_title: topic.title,
        category_name: category_name,
        topic_tags: tag_names,
        cooked: obj.cooked,
        private_message: topic.private_message?,
      )

      SearchIndexer.update_topics_index(topic.id, topic.title, obj.cooked) if obj.is_first_post?
    end
  end

  if User === obj && (obj.saved_change_to_username? || obj.saved_change_to_name? || force)
    SearchIndexer.update_users_index(
      obj.id,
      obj.username_lower || "",
      obj.name ? obj.name.downcase : "",
      obj.user_custom_fields.searchable.map(&:value).join(" "),
    )
  end

  if Topic === obj && (obj.saved_change_to_title? || force)
    if obj.posts
      if post = obj.posts.find_by(post_number: 1)
        SearchIndexer.update_posts_index(
          post_id: post.id,
          topic_title: obj.title,
          category_name: category_name,
          topic_tags: tag_names,
          cooked: post.cooked,
          private_message: obj.private_message?,
        )

        SearchIndexer.update_topics_index(obj.id, obj.title, post.cooked)
      end
    end
  end

  if Category === obj && (obj.saved_change_to_name? || force)
    SearchIndexer.queue_category_posts_reindex(obj.id)
    SearchIndexer.update_categories_index(obj.id, obj.name)
  end

  if Tag === obj && (obj.saved_change_to_name? || force)
    SearchIndexer.update_tags_index(obj.id, obj.name)
  end
end

.queue_category_posts_reindex(category_id) ⇒ Object 



217
218
219
220
221
222
223
224
225
226
227
228
229
 
# File 'app/services/search_indexer.rb', line 217

def self.queue_category_posts_reindex(category_id)
  return if @disabled

  DB.exec(<<~SQL, category_id: category_id, version: REINDEX_VERSION)
    UPDATE post_search_data
    SET version = :version
    FROM posts
    INNER JOIN topics ON posts.topic_id = topics.id
    INNER JOIN categories ON topics.category_id = categories.id
    WHERE post_search_data.post_id = posts.id
    AND categories.id = :category_id
  SQL
end

.queue_post_reindex(topic_id) ⇒ Object 



241
242
243
244
245
246
247
248
249
250
251
 
# File 'app/services/search_indexer.rb', line 241

def self.queue_post_reindex(topic_id)
  return if @disabled

  DB.exec(<<~SQL, topic_id: topic_id, version: REINDEX_VERSION)
    UPDATE post_search_data
    SET version = :version
    FROM posts
    WHERE post_search_data.post_id = posts.id
    AND posts.topic_id = :topic_id
  SQL
end

.queue_users_reindex(user_ids) ⇒ Object 



231
232
233
234
235
236
237
238
239
 
# File 'app/services/search_indexer.rb', line 231

def self.queue_users_reindex(user_ids)
  return if @disabled

  DB.exec(<<~SQL, user_ids: user_ids, version: REINDEX_VERSION)
    UPDATE user_search_data
    SET version = :version
    WHERE user_search_data.user_id IN (:user_ids)
  SQL
end

.update_categories_index(category_id, name) ⇒ Object 



209
210
211
 
# File 'app/services/search_indexer.rb', line 209

def self.update_categories_index(category_id, name)
  update_index(table: "category", id: category_id, a_weight: name)
end

.update_index(table:, id:, a_weight: nil, b_weight: nil, c_weight: nil, d_weight: nil) ⇒ Object 



24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
 
# File 'app/services/search_indexer.rb', line 24

def self.update_index(table:, id:, a_weight: nil, b_weight: nil, c_weight: nil, d_weight: nil)
  raw_data = { a: a_weight, b: b_weight, c: c_weight, d: d_weight }

  # The version used in excerpts
  search_data = raw_data.transform_values { |data| Search.prepare_data(data || "", :index) }

  # The version used to build the index
  indexed_data =
    search_data.transform_values do |data|
      data.gsub(/\S+/) { |word| word[0...SiteSetting.search_max_indexed_word_length] }
    end

  table_name = "#{table}_search_data"
  foreign_key = "#{table}_id"

  # for user login and name use "simple" lowercase stemmer
  stemmer = table == "user" ? "simple" : Search.ts_config

  ranked_index = <<~SQL
    setweight(to_tsvector('#{stemmer}', #{Search.wrap_unaccent("coalesce(:a,''))")}, 'A') ||
    setweight(to_tsvector('#{stemmer}', #{Search.wrap_unaccent("coalesce(:b,''))")}, 'B') ||
    setweight(to_tsvector('#{stemmer}', #{Search.wrap_unaccent("coalesce(:c,''))")}, 'C') ||
    setweight(to_tsvector('#{stemmer}', #{Search.wrap_unaccent("coalesce(:d,''))")}, 'D')
  SQL

  tsvector = DB.query_single("SELECT #{ranked_index}", indexed_data)[0]
  additional_lexemes = []

  # we also want to index parts of a domain name
  # that way stemmed single word searches will match
  additional_words = []

  tsvector
    .scan(/'(([a-zA-Z0-9]+\.)+[a-zA-Z0-9]+)'\:([\w+,]+)/)
    .reduce(additional_lexemes) do |array, (lexeme, _, positions)|
      count = 0

      if lexeme !~ /\A(\d+\.)?(\d+\.)*(\*|\d+)\z/
        loop do
          count += 1
          break if count >= 10 # Safeguard here to prevent infinite loop when a term has many dots
          term, _, remaining = lexeme.partition(".")
          break if remaining.blank?

          additional_words << [term, positions]

          array << "'#{remaining}':#{positions}"
          lexeme = remaining
        end
      end

      array
    end

  extra_domain_word_terms =
    if additional_words.length > 0
      DB
        .query_single(
          "SELECT to_tsvector(?, ?)",
          stemmer,
          additional_words.map { |term, _| term }.join(" "),
        )
        .first
        .scan(TS_VECTOR_PARSE_REGEX)
        .map do |term, _, indexes|
          new_indexes =
            indexes
              .split(",")
              .map do |index|
                existing_positions = additional_words[index.to_i - 1]
                if existing_positions
                  existing_positions[1]
                else
                  index
                end
              end
              .join(",")
          "#{term}#{new_indexes}"
        end
        .join(" ")
    end

  tsvector = "#{tsvector} #{additional_lexemes.join(" ")} #{extra_domain_word_terms}"

  if (max_dupes = SiteSetting.max_duplicate_search_index_terms) > 0
    reduced = []
    tsvector
      .scan(TS_VECTOR_PARSE_REGEX)
      .each do |term, _, indexes|
        family_counts = Hash.new(0)
        new_index_array = []

        indexes
          .split(",")
          .each do |index|
            family = nil
            family = index[-1] if index[-1].match?(/[A-D]/)
            # title dupes can completely dominate the index
            # so we limit them to 1
            if (family_counts[family] += 1) <= (family == "A" ? 1 : max_dupes)
              new_index_array << index
            end
          end
        reduced << "#{term.strip}#{new_index_array.join(",")}"
      end
    tsvector = reduced.join(" ")
  end

  indexed_data =
    if table.to_s == "post"
      clean_post_raw_data!(search_data[:d])
    else
      search_data.values.select { |d| d.length > 0 }.join(" ")
    end

  params = {
    "raw_data" => indexed_data,
    "#{foreign_key}" => id,
    "locale" => SiteSetting.default_locale,
    "version" => const_get("#{table.upcase}_INDEX_VERSION"),
    "search_data" => tsvector,
  }

  yield params if block_given?
  table_name.camelize.constantize.upsert(params)
rescue => e
  if Rails.env.test?
    raise
  else
    # TODO is there any way we can safely avoid this?
    # best way is probably pushing search indexer into a dedicated process so it no longer happens on save
    # instead in the post processor
    Discourse.warn_exception(
      e,
      message: "Unexpected error while indexing #{table} for search",
      env: {
        id: id,
      },
    )
  end
end

.update_posts_index(post_id:, topic_title:, category_name:, topic_tags:, cooked:, private_message:) ⇒ Object 



177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
 
# File 'app/services/search_indexer.rb', line 177

def self.update_posts_index(
  post_id:,
  topic_title:,
  category_name:,
  topic_tags:,
  cooked:,
  private_message:
)
  update_index(
    table: "post",
    id: post_id,
    a_weight: topic_title,
    b_weight: category_name,
    c_weight: topic_tags,
    # The tsvector resulted from parsing a string can be double the size of
    # the original string. Since there is no way to estimate the length of
    # the expected tsvector, we limit the input to ~50% of the maximum
    # length of a tsvector (1_048_576 bytes).
    d_weight: HtmlScrubber.scrub(cooked)[0..600_000],
  ) { |params| params["private_message"] = private_message }
end

.update_tags_index(tag_id, name) ⇒ Object 



213
214
215
 
# File 'app/services/search_indexer.rb', line 213

def self.update_tags_index(tag_id, name)
  update_index(table: "tag", id: tag_id, a_weight: name.downcase)
end

.update_topics_index(topic_id, title, cooked) ⇒ Object 



166
167
168
169
170
171
172
173
174
175
 
# File 'app/services/search_indexer.rb', line 166

def self.update_topics_index(topic_id, title, cooked)
  # a bit inconsistent that we use title as A and body as B when in
  # the post index body is D
  update_index(
    table: "topic",
    id: topic_id,
    a_weight: title,
    b_weight: HtmlScrubber.scrub(cooked)[0...Topic::MAX_SIMILAR_BODY_LENGTH],
  )
end

.update_users_index(user_id, username, name, custom_fields) ⇒ Object 



199
200
201
202
203
204
205
206
207
 
# File 'app/services/search_indexer.rb', line 199

def self.update_users_index(user_id, username, name, custom_fields)
  update_index(
    table: "user",
    id: user_id,
    a_weight: username,
    b_weight: name,
    c_weight: custom_fields,
  )
end