1 ## the index structure for redwood. interacts with ferret.
6 #require_gem 'ferret', ">= 0.10.13"
10 class IndexError < StandardError
13 def initialize source, s
22 attr_reader :index # debugging only
24 def initialize dir=BASE_DIR
27 @sources_dirty = false
29 wsa = Ferret::Analysis::WhiteSpaceAnalyzer.new false
30 sa = Ferret::Analysis::StandardAnalyzer.new Ferret::Analysis::FULL_ENGLISH_STOP_WORDS, true
31 @analyzer = Ferret::Analysis::PerFieldAnalyzer.new wsa
33 @qparser ||= Ferret::QueryParser.new :default_field => :body, :analyzer => @analyzer
35 self.class.i_am_the_instance self
44 FileUtils.mkdir_p @dir unless File.exists? @dir
50 raise "duplicate source!" if @sources.include? source
52 source.id ||= @sources.size
53 ##TODO: why was this necessary?
54 ##source.id += 1 while @sources.member? source.id
55 @sources[source.id] = source
58 def source_for name; @sources.values.find { |s| s.is_source_for? name }; end
59 def usual_sources; @sources.values.find_all { |s| s.usual? }; end
61 def load_index dir=File.join(@dir, "ferret")
63 Redwood::log "loading index..."
64 @index = Ferret::Index::Index.new(:path => dir, :analyzer => @analyzer)
65 Redwood::log "loaded index of #{@index.size} messages"
67 Redwood::log "creating index..."
68 field_infos = Ferret::Index::FieldInfos.new :store => :yes
69 field_infos.add_field :message_id
70 field_infos.add_field :source_id
71 field_infos.add_field :source_info
72 field_infos.add_field :date, :index => :untokenized
73 field_infos.add_field :body, :store => :no
74 field_infos.add_field :label
75 field_infos.add_field :subject
76 field_infos.add_field :from
77 field_infos.add_field :to
78 field_infos.add_field :refs
79 field_infos.add_field :snippet, :index => :no, :term_vector => :no
80 field_infos.create_index dir
81 @index = Ferret::Index::Index.new(:path => dir, :analyzer => @analyzer)
85 ## update the message by deleting and re-adding
86 def update_message m, source=nil, source_info=nil
87 docid, entry = load_entry_for_id m.id
89 source ||= entry[:source_id].to_i
90 source_info ||= entry[:source_info].to_i
93 ## this happens sometimes. i'm not sure why. ferret bug?
94 raise "no entry and no source info for message #{m.id}: source #{source.inspect}, info #{source_info.inspect}, entry #{entry.inspect}, query #{Ferret::Search::TermQuery.new(:message_id, m.id)}, results #{@index.search(Ferret::Search::TermQuery.new(:message_id, m.id)).inspect}" unless source && source_info
96 raise "deleting non-corresponding entry #{docid}" unless @index[docid][:message_id] == m.id
101 def save_index fn=File.join(@dir, "ferret")
102 # don't have to do anything apparently
106 @index.search(Ferret::Search::TermQuery.new(:message_id, id)).total_hits > 0
108 def contains? m; contains_id? m.id; end
109 def size; @index.size; end
111 ## you should probably not call this on a block that doesn't break
112 ## rather quickly because the results will probably be, as we say
113 ## in scotland, frikkin' huuuge.
114 EACH_BY_DATE_NUM = 100
115 def each_id_by_date opts={}
116 return if @index.size == 0 # otherwise ferret barfs ###TODO: remove this once my ferret patch is accepted
117 query = build_query opts
120 results = @index.search(query, :sort => "date DESC", :limit => EACH_BY_DATE_NUM, :offset => offset)
121 Redwood::log "got #{results.total_hits} results for query (offset #{offset}) #{query.inspect}"
122 results.hits.each { |hit| yield @index[hit.doc][:message_id], lambda { build_message hit.doc } }
123 break if offset >= results.total_hits - EACH_BY_DATE_NUM
124 offset += EACH_BY_DATE_NUM
128 def num_results_for opts={}
129 return 0 if @index.size == 0 # otherwise ferret barfs ###TODO: remove this once my ferret patch is accepted
131 index.search(q).total_hits
134 ## yield all messages in the thread containing 'm' by repeatedly
135 ## querying the index. yields pairs of message ids and
136 ## message-building lambdas, so that building an unwanted message
137 ## can be skipped in the block if desired.
138 SAME_SUBJECT_DATE_LIMIT = 7
139 def each_message_in_thread_for m, opts={}
144 ## temporarily disabling subject searching because it's a
145 ## significant slowdown.
147 ## TODO: make this configurable, i guess
149 date_min = m.date - (SAME_SUBJECT_DATE_LIMIT * 12 * 3600)
150 date_max = m.date + (SAME_SUBJECT_DATE_LIMIT * 12 * 3600)
152 q = Ferret::Search::BooleanQuery.new true
153 sq = Ferret::Search::PhraseQuery.new(:subject)
154 wrap_subj(Message.normalize_subj(m.subj)).split(/\s+/).each do |t|
157 q.add_query sq, :must
158 q.add_query Ferret::Search::RangeQuery.new(:date, :>= => date_min.to_indexable_s, :<= => date_max.to_indexable_s), :must
160 pending = @index.search(q).hits.map { |hit| @index[hit.doc][:message_id] }
161 Redwood::log "found #{pending.size} results for subject query #{q}"
166 until pending.empty? || (opts[:limit] && messages.size >= opts[:limit])
168 next if searched.member? id
170 q = Ferret::Search::BooleanQuery.new true
171 q.add_query Ferret::Search::TermQuery.new(:message_id, id), :should
172 q.add_query Ferret::Search::TermQuery.new(:refs, id), :should
175 @index.search_each(q, :limit => :all) do |docid, score|
176 break if opts[:limit] && messages.size >= opts[:limit]
177 mid = @index[docid][:message_id]
178 unless messages.member? mid
179 messages[mid] ||= lambda { build_message docid }
180 refs = @index[docid][:refs].split(" ")
185 Redwood::log "ran #{num_queries} queries to build thread of #{messages.size} messages for #{m.id}"
186 messages.each { |mid, builder| yield mid, builder }
189 ## builds a message object from a ferret result
190 def build_message docid
192 source = @sources[doc[:source_id].to_i]
193 #puts "building message #{doc[:message_id]} (#{source}##{doc[:source_info]})"
194 raise "invalid source #{doc[:source_id]}" unless source
197 "date" => Time.at(doc[:date].to_i),
198 "subject" => unwrap_subj(doc[:subject]),
199 "from" => doc[:from],
201 "message-id" => doc[:message_id],
202 "references" => doc[:refs],
205 Message.new :source => source, :source_info => doc[:source_info].to_i,
206 :labels => doc[:label].split(" ").map { |s| s.intern },
207 :snippet => doc[:snippet], :header => fake_header
210 def fresh_thread_id; @next_thread_id += 1; end
211 def wrap_subj subj; "__START_SUBJECT__ #{subj} __END_SUBJECT__"; end
212 def unwrap_subj subj; subj =~ /__START_SUBJECT__ (.*?) __END_SUBJECT__/ && $1; end
215 return false if contains? m
218 if m.source.is_a? Integer
221 m.source.id or raise "unregistered source #{m.source} (id #{m.source.id.inspect})"
224 to = (m.to + m.cc + m.bcc).map { |x| x.email }.join(" ")
227 :source_id => source_id,
228 :source_info => m.source_info,
229 :date => m.date.to_indexable_s,
231 :snippet => m.snippet,
232 :label => m.labels.join(" "),
233 :from => m.from ? m.from.email : "",
234 :to => (m.to + m.cc + m.bcc).map { |x| x.email }.join(" "),
235 :subject => wrap_subj(Message.normalize_subj(m.subj)),
236 :refs => (m.refs + m.replytos).uniq.join(" "),
239 @index.add_document d
241 ## TODO: figure out why this is sometimes triggered
242 #docid, entry = load_entry_for_id m.id
243 #raise "just added message #{m.id} but couldn't find it in a search" unless docid
247 def drop_entry docno; @index.delete docno; end
249 def load_entry_for_id mid
250 results = @index.search(Ferret::Search::TermQuery.new(:message_id, mid))
251 return if results.total_hits == 0
252 docid = results.hits[0].doc
253 [docid, @index[docid]]
256 def load_contacts emails, h={}
257 q = Ferret::Search::BooleanQuery.new true
259 qq = Ferret::Search::BooleanQuery.new true
260 qq.add_query Ferret::Search::TermQuery.new(:from, e), :should
261 qq.add_query Ferret::Search::TermQuery.new(:to, e), :should
264 q.add_query Ferret::Search::TermQuery.new(:label, "spam"), :must_not
266 Redwood::log "contact search: #{q}"
269 @index.search_each(q, :sort => "date DESC", :limit => :all) do |docid, score|
270 break if contacts.size >= num
271 #Redwood::log "got message with to: #{@index[docid][:to].inspect} and from: #{@index[docid][:from].inspect}"
272 f = @index[docid][:from]
273 t = @index[docid][:to]
275 if AccountManager.is_account_email? f
276 t.split(" ").each { |e| #Redwood::log "adding #{e} because there's a message to him from account email #{f}";
277 contacts[Person.for(e)] = true }
279 #Redwood::log "adding from #{f} because there's a message from him to #{t}"
280 contacts[Person.for(f)] = true
284 contacts.keys.compact
289 def parse_user_query_string str; @qparser.parse str; end
291 query = Ferret::Search::BooleanQuery.new
292 query.add_query opts[:qobj], :must if opts[:qobj]
293 labels = ([opts[:label]] + (opts[:labels] || [])).compact
294 labels.each { |t| query.add_query Ferret::Search::TermQuery.new("label", t.to_s), :must }
295 if opts[:participants]
296 q2 = Ferret::Search::BooleanQuery.new
297 opts[:participants].each do |p|
298 q2.add_query Ferret::Search::TermQuery.new("from", p.email), :should
299 q2.add_query Ferret::Search::TermQuery.new("to", p.email), :should
301 query.add_query q2, :must
304 query.add_query Ferret::Search::TermQuery.new("label", "spam"), :must_not unless opts[:load_spam] || labels.include?(:spam)
305 query.add_query Ferret::Search::TermQuery.new("label", "killed"), :must_not unless opts[:load_killed] || labels.include?(:killed)
309 def load_sources fn=Redwood::SOURCE_FN
310 @sources = Hash[*(Redwood::load_yaml_obj(fn) || []).map { |s| [s.id, s] }.flatten]
311 @sources_dirty = false
314 def save_sources fn=Redwood::SOURCE_FN
315 if @sources_dirty || @sources.any? { |id, s| s.dirty? }
319 FileUtils.mv fn, bakfn, :force => true unless File.exists?(bakfn) && File.size(bakfn) > File.size(fn)
321 Redwood::save_yaml_obj @sources.values, fn
324 @sources_dirty = false