1 ## the index structure for redwood. interacts with ferret.
9 Redwood::log "'chronic' library not found. run 'gem install chronic' to install."
16 class LockError < StandardError
21 def method_missing m; @h[m.to_s] end
28 def initialize dir=BASE_DIR
31 @sources_dirty = false
33 wsa = Ferret::Analysis::WhiteSpaceAnalyzer.new false
34 sa = Ferret::Analysis::StandardAnalyzer.new [], true
35 @analyzer = Ferret::Analysis::PerFieldAnalyzer.new wsa
37 @analyzer[:subject] = sa
38 @qparser ||= Ferret::QueryParser.new :default_field => :body, :analyzer => @analyzer, :or_default => false
39 @lock = Lockfile.new lockfile, :retries => 0, :max_age => nil
41 self.class.i_am_the_instance self
44 def lockfile; File.join @dir, "lock" end
47 Redwood::log "locking #{lockfile}..."
50 rescue Lockfile::MaxTriesLockError
51 raise LockError, @lock.lockinfo_on_disk
55 def start_lock_update_thread
56 @lock_update_thread = Redwood::reporting_thread do
64 def stop_lock_update_thread
65 @lock_update_thread.kill if @lock_update_thread
66 @lock_update_thread = nil
69 def fancy_lock_error_message_for e
70 secs = Time.now - e.mtime
74 "#{secs.to_i} seconds"
80 Error: the sup index is locked by another process! User '#{e.user}' on
81 host '#{e.host}' is running #{e.pname} with pid #{e.pid}. The process was alive
90 $stderr.puts fancy_lock_error_message_for(e)
93 You can wait for the process to finish, or, if it crashed and left a
94 stale lock file behind, you can manually delete #{@lock.path}.
101 if @lock && @lock.locked?
102 Redwood::log "unlocking #{lockfile}..."
113 Redwood::log "saving index and sources..."
114 FileUtils.mkdir_p @dir unless File.exists? @dir
119 def add_source source
120 raise "duplicate source!" if @sources.include? source
121 @sources_dirty = true
122 source.id ||= @sources.size
123 ##TODO: why was this necessary?
124 ##source.id += 1 while @sources.member? source.id
125 @sources[source.id] = source
128 def source_for uri; @sources.values.find { |s| s.is_source_for? uri }; end
129 def usual_sources; @sources.values.find_all { |s| s.usual? }; end
130 def sources; @sources.values; end
132 def load_index dir=File.join(@dir, "ferret")
134 Redwood::log "loading index..."
135 @index = Ferret::Index::Index.new(:path => dir, :analyzer => @analyzer)
136 Redwood::log "loaded index of #{@index.size} messages"
138 Redwood::log "creating index..."
139 field_infos = Ferret::Index::FieldInfos.new :store => :yes
140 field_infos.add_field :message_id
141 field_infos.add_field :source_id
142 field_infos.add_field :source_info
143 field_infos.add_field :date, :index => :untokenized
144 field_infos.add_field :body, :store => :no
145 field_infos.add_field :label
146 field_infos.add_field :subject
147 field_infos.add_field :from
148 field_infos.add_field :to
149 field_infos.add_field :refs
150 field_infos.add_field :snippet, :index => :no, :term_vector => :no
151 field_infos.create_index dir
152 @index = Ferret::Index::Index.new(:path => dir, :analyzer => @analyzer)
156 ## Syncs the message to the index: deleting if it's already there,
157 ## and adding either way. Index state will be determined by m.labels.
159 ## docid and entry can be specified if they're already known.
160 def sync_message m, docid=nil, entry=nil
161 docid, entry = load_entry_for_id m.id unless docid && entry
163 raise "no source info for message #{m.id}" unless m.source && m.source_info
164 raise "trying to delete non-corresponding entry #{docid} with index message-id #{@index[docid][:message_id].inspect} and parameter message id #{m.id.inspect}" if docid && @index[docid][:message_id] != m.id
167 if m.source.is_a? Integer
170 m.source.id or raise "unregistered source #{m.source} (id #{m.source.id.inspect})"
173 to = (m.to + m.cc + m.bcc).map { |x| x.email }.join(" ")
176 :source_id => source_id,
177 :source_info => m.source_info,
178 :date => m.date.to_indexable_s,
180 :snippet => m.snippet,
181 :label => m.labels.uniq.join(" "),
182 :from => m.from ? m.from.email : "",
183 :to => (m.to + m.cc + m.bcc).map { |x| x.email }.join(" "),
184 :subject => wrap_subj(Message.normalize_subj(m.subj)),
185 :refs => (m.refs + m.replytos).uniq.join(" "),
188 @index.delete docid if docid
189 @index.add_document d
191 docid, entry = load_entry_for_id m.id
192 ## this hasn't been triggered in a long time. TODO: decide whether it's still a problem.
193 raise "just added message #{m.id.inspect} but couldn't find it in a search" unless docid
197 def save_index fn=File.join(@dir, "ferret")
198 # don't have to do anything, apparently
202 @index.search(Ferret::Search::TermQuery.new(:message_id, id)).total_hits > 0
204 def contains? m; contains_id? m.id; end
205 def size; @index.size; end
207 ## you should probably not call this on a block that doesn't break
208 ## rather quickly because the results can be very large.
209 EACH_BY_DATE_NUM = 100
210 def each_id_by_date opts={}
211 return if @index.size == 0 # otherwise ferret barfs ###TODO: remove this once my ferret patch is accepted
212 query = build_query opts
215 results = @index.search(query, :sort => "date DESC", :limit => EACH_BY_DATE_NUM, :offset => offset)
216 Redwood::log "got #{results.total_hits} results for query (offset #{offset}) #{query.inspect}"
217 results.hits.each { |hit| yield @index[hit.doc][:message_id], lambda { build_message hit.doc } }
218 break if offset >= results.total_hits - EACH_BY_DATE_NUM
219 offset += EACH_BY_DATE_NUM
223 def num_results_for opts={}
224 return 0 if @index.size == 0 # otherwise ferret barfs ###TODO: remove this once my ferret patch is accepted
227 index.search(q, :limit => 1).total_hits
230 ## yield all messages in the thread containing 'm' by repeatedly
231 ## querying the index. yields pairs of message ids and
232 ## message-building lambdas, so that building an unwanted message
233 ## can be skipped in the block if desired.
235 ## only two options, :limit and :skip_killed. if :skip_killed is
236 ## true, stops loading any thread if a message with a :killed flag
238 SAME_SUBJECT_DATE_LIMIT = 7
239 def each_message_in_thread_for m, opts={}
240 #Redwood::log "Building thread for #{m.id}: #{m.subj}"
245 if $config[:thread_by_subject] # do subject queries
246 date_min = m.date - (SAME_SUBJECT_DATE_LIMIT * 12 * 3600)
247 date_max = m.date + (SAME_SUBJECT_DATE_LIMIT * 12 * 3600)
249 q = Ferret::Search::BooleanQuery.new true
250 sq = Ferret::Search::PhraseQuery.new(:subject)
251 wrap_subj(Message.normalize_subj(m.subj)).split(/\s+/).each do |t|
254 q.add_query sq, :must
255 q.add_query Ferret::Search::RangeQuery.new(:date, :>= => date_min.to_indexable_s, :<= => date_max.to_indexable_s), :must
257 q = build_query :qobj => q
259 pending = @index.search(q).hits.map { |hit| @index[hit.doc][:message_id] }
260 Redwood::log "found #{pending.size} results for subject query #{q}"
265 until pending.empty? || (opts[:limit] && messages.size >= opts[:limit])
267 next if searched.member? id
269 q = Ferret::Search::BooleanQuery.new true
270 q.add_query Ferret::Search::TermQuery.new(:message_id, id), :should
271 q.add_query Ferret::Search::TermQuery.new(:refs, id), :should
273 q = build_query :qobj => q
277 @index.search_each(q, :limit => :all) do |docid, score|
278 break if opts[:limit] && messages.size >= opts[:limit]
279 if @index[docid][:label].split(/\s+/).include?("killed") && opts[:skip_killed]
283 mid = @index[docid][:message_id]
284 unless messages.member?(mid)
285 #Redwood::log "got #{mid} as a child of #{id}"
286 messages[mid] ||= lambda { build_message docid }
287 refs = @index[docid][:refs].split(" ")
293 Redwood::log "thread for #{m.id} is killed, ignoring"
296 Redwood::log "ran #{num_queries} queries to build thread of #{messages.size + 1} messages for #{m.id}: #{m.subj}" if num_queries > 0
297 messages.each { |mid, builder| yield mid, builder }
302 ## builds a message object from a ferret result
303 def build_message docid
305 source = @sources[doc[:source_id].to_i]
306 #puts "building message #{doc[:message_id]} (#{source}##{doc[:source_info]})"
307 raise "invalid source #{doc[:source_id]}" unless source
310 "date" => Time.at(doc[:date].to_i),
311 "subject" => unwrap_subj(doc[:subject]),
312 "from" => doc[:from],
314 "message-id" => doc[:message_id],
315 "references" => doc[:refs].split(/\s+/).map { |x| "<#{x}>" }.join(" "),
318 Message.new :source => source, :source_info => doc[:source_info].to_i,
319 :labels => doc[:label].split(" ").map { |s| s.intern },
320 :snippet => doc[:snippet], :header => fake_header
323 def fresh_thread_id; @next_thread_id += 1; end
324 def wrap_subj subj; "__START_SUBJECT__ #{subj} __END_SUBJECT__"; end
325 def unwrap_subj subj; subj =~ /__START_SUBJECT__ (.*?) __END_SUBJECT__/ && $1; end
327 def drop_entry docno; @index.delete docno; end
329 def load_entry_for_id mid
330 results = @index.search(Ferret::Search::TermQuery.new(:message_id, mid))
331 return if results.total_hits == 0
332 docid = results.hits[0].doc
333 [docid, @index[docid]]
336 def load_contacts emails, h={}
337 q = Ferret::Search::BooleanQuery.new true
339 qq = Ferret::Search::BooleanQuery.new true
340 qq.add_query Ferret::Search::TermQuery.new(:from, e), :should
341 qq.add_query Ferret::Search::TermQuery.new(:to, e), :should
344 q.add_query Ferret::Search::TermQuery.new(:label, "spam"), :must_not
346 Redwood::log "contact search: #{q}"
349 @index.search_each(q, :sort => "date DESC", :limit => :all) do |docid, score|
350 break if contacts.size >= num
351 #Redwood::log "got message #{docid} to: #{@index[docid][:to].inspect} and from: #{@index[docid][:from].inspect}"
352 f = @index[docid][:from]
353 t = @index[docid][:to]
355 if AccountManager.is_account_email? f
356 t.split(" ").each { |e| contacts[PersonManager.person_for(e)] = true }
358 contacts[PersonManager.person_for(f)] = true
362 contacts.keys.compact
365 def load_sources fn=Redwood::SOURCE_FN
366 source_array = (Redwood::load_yaml_obj(fn) || []).map { |o| Recoverable.new o }
367 @sources = Hash[*(source_array).map { |s| [s.id, s] }.flatten]
368 @sources_dirty = false
371 def has_any_from_source_with_label? source, label
372 q = Ferret::Search::BooleanQuery.new
373 q.add_query Ferret::Search::TermQuery.new("source_id", source.id.to_s), :must
374 q.add_query Ferret::Search::TermQuery.new("label", label.to_s), :must
375 index.search(q, :limit => 1).total_hits > 0
380 ## do any specialized parsing
381 ## returns nil and flashes error message if parsing failed
382 def parse_user_query_string str
383 result = str.gsub(/\b(to|from):(\S+)\b/) do
385 if(p = ContactManager.contact_for(name))
393 chronic_failure = false
394 result = result.gsub(/\b(before|on|in|after):(\((.+?)\)\B|(\S+)\b)/) do
395 break if chronic_failure
396 field, datestr = $1, ($3 || $4)
397 realdate = Chronic.parse(datestr, :guess => false, :context => :none)
401 Redwood::log "chronic: translated #{field}:#{datestr} to #{realdate.end}"
402 "date:(>= #{sprintf "%012d", realdate.end.to_i})"
404 Redwood::log "chronic: translated #{field}:#{datestr} to #{realdate.begin}"
405 "date:(<= #{sprintf "%012d", realdate.begin.to_i})"
407 Redwood::log "chronic: translated #{field}:#{datestr} to #{realdate}"
408 "date:(<= #{sprintf "%012d", realdate.end.to_i}) date:(>= #{sprintf "%012d", realdate.begin.to_i})"
411 BufferManager.flash "Don't understand date #{datestr.inspect}!"
412 chronic_failure = true
415 result = nil if chronic_failure
418 Redwood::log "translated #{str.inspect} to #{result}" unless result == str
419 @qparser.parse result if result
423 query = Ferret::Search::BooleanQuery.new
424 query.add_query opts[:qobj], :must if opts[:qobj]
425 labels = ([opts[:label]] + (opts[:labels] || [])).compact
426 labels.each { |t| query.add_query Ferret::Search::TermQuery.new("label", t.to_s), :must }
427 if opts[:participants]
428 q2 = Ferret::Search::BooleanQuery.new
429 opts[:participants].each do |p|
430 q2.add_query Ferret::Search::TermQuery.new("from", p.email), :should
431 q2.add_query Ferret::Search::TermQuery.new("to", p.email), :should
433 query.add_query q2, :must
436 query.add_query Ferret::Search::TermQuery.new("label", "spam"), :must_not unless opts[:load_spam] || labels.include?(:spam)
437 query.add_query Ferret::Search::TermQuery.new("label", "deleted"), :must_not unless opts[:load_deleted] || labels.include?(:deleted)
438 query.add_query Ferret::Search::TermQuery.new("label", "killed"), :must_not if opts[:skip_killed]
442 def save_sources fn=Redwood::SOURCE_FN
443 if @sources_dirty || @sources.any? { |id, s| s.dirty? }
447 FileUtils.mv fn, bakfn, :force => true unless File.exists?(bakfn) && File.size(fn) == 0
449 Redwood::save_yaml_obj @sources.values.sort_by { |s| s.id.to_i }, fn, true
452 @sources_dirty = false