Module: Mongoid::FullTextSearch::ClassMethods

Defined in:
lib/mongoid_fulltext.rb

Instance Method Summary collapse

Instance Method Details

#all_ngrams(str, config, bound_number_returned = true) ⇒ Object 



188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
 
# File 'lib/mongoid_fulltext.rb', line 188

def all_ngrams(str, config, bound_number_returned = true)
  return {} if str.nil?

  if config[:remove_accents]
    if defined?(UnicodeUtils)
      str = UnicodeUtils.nfkd(str)
    elsif defined?(DiacriticsFu)
      str = DiacriticsFu::escape(str)
    end
  end

  # Remove any characters that aren't in the alphabet and aren't word separators
  filtered_str = str.mb_chars.downcase.to_s.split('').find_all{ |ch| config[:alphabet][ch] or config[:word_separators][ch] }.join('')
  
  # Figure out how many ngrams to extract from the string. If we can't afford to extract all ngrams,
  # step over the string in evenly spaced strides to extract ngrams. For example, to extract 3 3-letter
  # ngrams from 'abcdefghijk', we'd want to extract 'abc', 'efg', and 'ijk'.
  if bound_number_returned
    step_size = [((filtered_str.length - config[:ngram_width]).to_f / config[:max_ngrams_to_search]).ceil, 1].max
  else
    step_size = 1
  end
  
  # Create an array of records of the form {:ngram => x, :score => y} for all ngrams that occur in the
  # input string using the step size that we just computed. Let score(x,y) be the score of string x
  # compared with string y - assigning scores to ngrams with the square root-based scoring function
  # below and multiplying scores of matching ngrams together yields a score function that has the
  # property that score(x,y) > score(x,z) for any string z containing y and score(x,y) > score(x,z)
  # for any string z contained in y.
  ngram_array = (0..filtered_str.length - config[:ngram_width]).step(step_size).map do |i|
    if i == 0 or (config[:apply_prefix_scoring_to_all_words] and \
                  config[:word_separators].has_key?(filtered_str[i-1].chr))
      score = Math.sqrt(1 + 1.0/filtered_str.length)
    else
      score = Math.sqrt(2.0/filtered_str.length)
    end
    {:ngram => filtered_str[i..i+config[:ngram_width]-1], :score => score}
  end

  # If an ngram appears multiple times in the query string, keep the max score
  ngram_array = ngram_array.group_by{ |h| h[:ngram] }.map{ |key, values| {:ngram => key, :score => values.map{ |v| v[:score] }.max} }
  
  if config[:index_short_prefixes] or config[:index_full_words]
    split_regex_def = config[:word_separators].keys.map{ |k| Regexp.escape(k) }.join
    split_regex = Regexp.compile("[#{split_regex_def}]")
    all_words = filtered_str.split(split_regex)
  end

  # Add 'short prefix' records to the array: prefixes of the string that are length (ngram_width - 1)
  if config[:index_short_prefixes]
    prefixes_seen = {}
    all_words.each do |word|
      next if word.length < config[:ngram_width]-1
      prefix = word[0...config[:ngram_width]-1]
      if prefixes_seen[prefix].nil? and (config[:stop_words][word].nil? or word == filtered_str)
        ngram_array << {:ngram => prefix, :score => 1 + 1.0/filtered_str.length}
        prefixes_seen[prefix] = true
      end
    end
  end

  # Add records to the array of ngrams for each full word in the string that isn't a stop word
  if config[:index_full_words]
    full_words_seen = {}
    all_words.each do |word|
      if word.length > 1 and full_words_seen[word].nil? and (config[:stop_words][word].nil? or word == filtered_str)
        ngram_array << {:ngram => word, :score => 1 + 1.0/filtered_str.length}
        full_words_seen[word] = true
      end
    end
  end

  # If an ngram appears as any combination of full word, short prefix, and ngram, keep the sum of the two scores
  Hash[ngram_array.group_by{ |h| h[:ngram] }.map{ |key, values| [key, values.map{ |v| v[:score] }.sum] }]
end

#create_fulltext_indexes ⇒ Object 



63
64
65
66
67
68
 
# File 'lib/mongoid_fulltext.rb', line 63

def create_fulltext_indexes
  return unless self.mongoid_fulltext_config
  self.mongoid_fulltext_config.each_pair do |index_name, fulltext_config|
    fulltext_search_ensure_indexes(index_name, fulltext_config)
  end
end

#fulltext_search(query_string, options = {}) ⇒ Object 



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
165
166
167
168
169
170
171
172
173
174
 
# File 'lib/mongoid_fulltext.rb', line 107

def fulltext_search(query_string, options={})
  max_results = options.has_key?(:max_results) ? options.delete(:max_results) : 10
  return_scores = options.has_key?(:return_scores) ? options.delete(:return_scores) : false
  if self.mongoid_fulltext_config.count > 1 and !options.has_key?(:index)
    error_message = '%s is indexed by multiple full-text indexes. You must specify one by passing an :index_name parameter'
    raise UnspecifiedIndexError, error_message % self.name, caller
  end
  index_name = options.has_key?(:index) ? options.delete(:index) : self.mongoid_fulltext_config.keys.first

  # Options hash should only contain filters after this point

  ngrams = all_ngrams(query_string, self.mongoid_fulltext_config[index_name])
  return [] if ngrams.empty?

  # For each ngram, construct the query we'll use to pull index documents and
  # get a count of the number of index documents containing that n-gram
  ordering = {'score' => -1}
  limit = self.mongoid_fulltext_config[index_name][:max_candidate_set_size]
  coll = collection.database[index_name]
  cursors = ngrams.map do |ngram|
    query = {'ngram' => ngram[0]}
    query.update(map_query_filters options)
    count = coll.find(query).count
    {:ngram => ngram, :count => count, :query => query}
  end.sort!{ |record1, record2| record1[:count] <=> record2[:count] }

  # Using the queries we just constructed and the n-gram frequency counts we
  # just computed, pull in about *:max_candidate_set_size* candidates by
  # considering the n-grams in order of increasing frequency. When we've
  # spent all *:max_candidate_set_size* candidates, pull the top-scoring
  # *max_results* candidates for each remaining n-gram.
  results_so_far = 0
  candidates_list = cursors.map do |doc|
    next if doc[:count] == 0
    query_result = coll.find(doc[:query])
    if results_so_far >= limit
      query_result = query_result.sort(ordering).limit(max_results)
    elsif doc[:count] > limit - results_so_far
      query_result = query_result.sort(ordering).limit(limit - results_so_far)
    end
    results_so_far += doc[:count]
    ngram_score = ngrams[doc[:ngram][0]]
    Hash[query_result.map do |candidate|
           [candidate['document_id'],
            {:clazz => candidate['class'], :score => candidate['score'] * ngram_score}]
         end]
  end.compact
  
  # Finally, score all candidates by matching them up with other candidates that are
  # associated with the same document. This is similar to how you might process a
  # boolean AND query, except that with an AND query, you'd stop after considering
  # the first candidate list and matching its candidates up with candidates from other
  # lists, whereas here we want the search to be a little fuzzier so we'll run through
  # all candidate lists, removing candidates as we match them up.
  all_scores = []
  while !candidates_list.empty?
    candidates = candidates_list.pop
    scores = candidates.map do |candidate_id, data|
      {:id => candidate_id,
       :clazz => data[:clazz],
       :score => data[:score] + candidates_list.map{ |others| (others.delete(candidate_id) || {:score => 0})[:score] }.sum
       }
    end
    all_scores.concat(scores)
  end
  all_scores.sort!{ |document1, document2| -document1[:score] <=> -document2[:score] }
  instantiate_mapreduce_results(all_scores[0..max_results-1], { :return_scores => return_scores })
end

#fulltext_search_ensure_indexes(index_name, config) ⇒ Object 



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
 
# File 'lib/mongoid_fulltext.rb', line 70

def fulltext_search_ensure_indexes(index_name, config)
  db = collection.database
  coll = db[index_name]

  # The order of filters matters when the same index is used from two or more collections.
  filter_indexes = (config[:filters] || []).map do |key,value|
    ["filter_values.#{key}", 1]
  end.sort_by { |filter_index| filter_index[0] }
  
  index_definition = [['ngram', 1], ['score', -1]].concat(filter_indexes)

  # Since the definition of the index could have changed, we'll clean up by
  # removing any indexes that aren't on the exact.
  correct_keys = index_definition.map{ |field_def| field_def[0] }
  all_filter_keys = filter_indexes.map{ |field_def| field_def[0] }
  coll.indexes.each do |idef|
    keys = idef['key'].keys
    next if !keys.member?('ngram')
    all_filter_keys |= keys.find_all{ |key| key.starts_with?('filter_values.') }
    if keys & correct_keys != correct_keys
      Mongoid.logger.info "Dropping #{idef['name']} [#{keys & correct_keys} <=> #{correct_keys}]" if Mongoid.logger
      coll.indexes.drop(idef['key'])
    end
  end

  if all_filter_keys.length > filter_indexes.length
    filter_indexes = all_filter_keys.map {|key| [key, 1] }.sort_by { |filter_index| filter_index[0] }
    index_definition = [['ngram', 1], ['score', -1]].concat(filter_indexes)
  end
  
  Mongoid.logger.info "Ensuring fts_index on #{coll.name}: #{index_definition}" if Mongoid.logger
  coll.indexes.create(Hash[index_definition], { :name => 'fts_index' })

  Mongoid.logger.info "Ensuring document_id index on #{coll.name}" if Mongoid.logger
  coll.indexes.create('document_id' => 1) # to make removes fast
end

#fulltext_search_in(*args) ⇒ Object 



17
18
19
20
21
22
23
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
 
# File 'lib/mongoid_fulltext.rb', line 17

def fulltext_search_in(*args)
  self.mongoid_fulltext_config = {} if self.mongoid_fulltext_config.nil?
  options = args.last.is_a?(Hash) ? args.pop : {}
  if options.has_key?(:index_name)
    index_name = options[:index_name]
  else
    index_name = 'mongoid_fulltext.index_%s_%s' % [self.name.downcase, self.mongoid_fulltext_config.count]
  end

  config = {
    :alphabet => 'abcdefghijklmnopqrstuvwxyz0123456789 ',
    :word_separators => "-_ \n\t",
    :ngram_width => 3,
    :max_ngrams_to_search => 6,
    :apply_prefix_scoring_to_all_words => true,
    :index_full_words => true,
    :index_short_prefixes => false,
    :max_candidate_set_size => 1000,
    :remove_accents => true,
    :reindex_immediately => true,
    :stop_words => Hash[['i', 'a', 's', 't', 'me', 'my', 'we', 'he', 'it', 'am', 'is', 'be', 'do', 'an', 'if',
                         'or', 'as', 'of', 'at', 'by', 'to', 'up', 'in', 'on', 'no', 'so', 'our', 'you', 'him',
                         'his', 'she', 'her', 'its', 'who', 'are', 'was', 'has', 'had', 'did', 'the', 'and',
                         'but', 'for', 'out', 'off', 'why', 'how', 'all', 'any', 'few', 'nor', 'not', 'own',
                         'too', 'can', 'don', 'now', 'ours', 'your', 'hers', 'they', 'them', 'what', 'whom',
                         'this', 'that', 'were', 'been', 'have', 'does', 'with', 'into', 'from', 'down', 'over',
                         'then', 'once', 'here', 'when', 'both', 'each', 'more', 'most', 'some', 'such', 'only',
                         'same', 'than', 'very', 'will', 'just', 'yours', 'their', 'which', 'these', 'those',
                         'being', 'doing', 'until', 'while', 'about', 'after', 'above', 'below', 'under',
                         'again', 'there', 'where', 'other', 'myself', 'itself', 'theirs', 'having', 'during',
                         'before', 'should', 'himself', 'herself', 'because', 'against', 'between', 'through',
                         'further', 'yourself', 'ourselves', 'yourselves', 'themselves'].map{ |x| [x,true] }]
  }
  
  config.update(options)

  args = [:to_s] if args.empty?
  config[:ngram_fields] = args
  config[:alphabet] = Hash[config[:alphabet].split('').map{ |ch| [ch,ch] }]
  config[:word_separators] = Hash[config[:word_separators].split('').map{ |ch| [ch,ch] }]
  self.mongoid_fulltext_config[index_name] = config
  
  before_save(:update_ngram_index) if config[:reindex_immediately]
  before_destroy :remove_from_ngram_index
end

#instantiate_mapreduce_result(result) ⇒ Object 



176
177
178
 
# File 'lib/mongoid_fulltext.rb', line 176

def instantiate_mapreduce_result(result)
  result[:clazz].constantize.find(result[:id])
end

#instantiate_mapreduce_results(results, options) ⇒ Object 



180
181
182
183
184
185
186
 
# File 'lib/mongoid_fulltext.rb', line 180

def instantiate_mapreduce_results(results, options)
  if (options[:return_scores])
    results.map { |result| [ instantiate_mapreduce_result(result), result[:score] ] }.find_all { |result| ! result[0].nil? }
  else
    results.map { |result| instantiate_mapreduce_result(result) }.compact
  end
end

#remove_from_ngram_index ⇒ Object 



264
265
266
267
268
269
 
# File 'lib/mongoid_fulltext.rb', line 264

def remove_from_ngram_index
  self.mongoid_fulltext_config.each_pair do |index_name, fulltext_config|
    coll = collection.database[index_name]
    coll.find({'class' => self.name}).remove_all
  end
end

#update_ngram_index ⇒ Object 



271
272
273
274
275
 
# File 'lib/mongoid_fulltext.rb', line 271

def update_ngram_index
  self.all.each do |model|
    model.update_ngram_index
  end
end