Class: Search::Simple::Searcher

Inherits:
Object
  • Object
show all
Defined in:
lib/search/simple/searcher.rb

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(dict, document_vectors, cache_file) ⇒ Searcher

Returns a new instance of Searcher.



61
62
63
64
65
# File 'lib/search/simple/searcher.rb', line 61

def initialize(dict, document_vectors, cache_file)
  @dict = dict
  @document_vectors = document_vectors
  @cache_file = cache_file
end

Class Method Details

.create_indices(entries, dict, document_vectors) ⇒ Object

Create a new dictionary and document vectors from a blog archive



168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
# File 'lib/search/simple/searcher.rb', line 168

def Searcher.create_indices(entries, dict, document_vectors)
  word_count = doc_count = 0

  entries.each do |entry|
    doc_count += 1
    vector = Vector.new
    extract_words_for_searcher(entry.content.downcase) do |word|
      word_index = dict.add_word(word)
      if word_index
        vector.add_word_index(word_index) 
        word_count += 1
      end
    end
    document_vectors[entry.identifier] = vector
  end
  
  $stderr.puts "#{dict.size} unique words out of #{word_count} " +
    "in #{doc_count} documents"
  
end

.extract_words_for_searcher(text) ⇒ Object



159
160
161
162
163
# File 'lib/search/simple/searcher.rb', line 159

def Searcher.extract_words_for_searcher(text)
  text.scan(/[-+]?\w[\-\w]{2,}/) do |word|
    yield word
  end
end

.load(contents, cache_file) ⇒ Object

Serialization support. At some point we’ll need to do incremental indexing. For now, however, the following seems to work fairly effectively on 1000 entry blogs, so I’ll defer the change until later.



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
# File 'lib/search/simple/searcher.rb', line 124

def Searcher.load(contents, cache_file)
  dict = document_vectors = nil
  modified = false
  loaded   = false
  begin
    File.open(cache_file, "r") do |f| 
      if f.mtime > contents.latest_mtime
        dict = Marshal.load(f)
        document_vectors = Marshal.load(f)
        loaded = true
      end
    end
  rescue
  ;
  end
    
  unless loaded
    dict = Dictionary.new
    document_vectors = {}
    create_indices(contents, dict, document_vectors)
    modified = true
  end
  
  s = Searcher.new(dict, document_vectors, cache_file)
  s.dump if modified
  s
end

Instance Method Details

#dumpObject



152
153
154
155
156
157
# File 'lib/search/simple/searcher.rb', line 152

def dump
  File.open(@cache_file, "w") do |fileInstance|
    Marshal.dump(@dict, fileInstance)
    Marshal.dump(@document_vectors, fileInstance)
  end
end

#find_words(words) ⇒ Object

Return SearchResults based on trying to find the array of words in our document vectors

A word beginning ‘+’ must appear in the target documents A word beginning ‘-’ must not appear other words are scored. The documents with the highest scores are returned first



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
# File 'lib/search/simple/searcher.rb', line 75

def find_words(words)
  search_results = SearchResults.new
    
  general = Vector.new
  must_match = Vector.new
  must_not_match = Vector.new
  
  Searcher.extract_words_for_searcher(words.join(' ')) do |word|
    case word[0]
    when ?+
      word = word[1,99]
      vector = must_match
    when ?-
   word = word[1,99]
      vector = must_not_match
    else
   vector = general
    end
    
    index = @dict.find(word.downcase)
    if index
      vector.add_word_index(index)
    else
   search_results.add_warning "'#{word}' does not occur in the documents"
    end
  end
    
    #    if (general.num_bits + must_match.num_bits).zero? 
    #      search_results.add_warning "No valid search terms given"
    #    else
    res = []
    @document_vectors.each do |entry, dvec|
      score = dvec.score_against(must_match, must_not_match, general)
      res << [ entry, score ] if score > 0
    end
    
    res.sort {|a,b| b[1] <=> a[1] }.each {|name, score|
      search_results.add_result(name, score)
    }
    
    search_results.add_warning "No matches" unless search_results.contains_matches
    #    end
  search_results
end