lib/sup/index.rb in sup-0.4 vs lib/sup/index.rb in sup-0.5

- old
+ new

@@ -123,27 +123,31 @@ source.id ||= (max || 0) + 1 ##source.id += 1 while @sources.member? source.id @sources[source.id] = source end - def source_for uri; @sources.values.find { |s| s.is_source_for? uri }; end - def usual_sources; @sources.values.find_all { |s| s.usual? }; end - def sources; @sources.values; end + def sources + ## favour the inbox by listing non-archived sources first + @sources.values.sort_by { |s| s.id }.partition { |s| !s.archived? }.flatten + end + def source_for uri; sources.find { |s| s.is_source_for? uri }; end + def usual_sources; sources.find_all { |s| s.usual? }; end + def load_index dir=File.join(@dir, "ferret") if File.exists? dir Redwood::log "loading index..." @index = Ferret::Index::Index.new(:path => dir, :analyzer => @analyzer) Redwood::log "loaded index of #{@index.size} messages" else Redwood::log "creating index..." field_infos = Ferret::Index::FieldInfos.new :store => :yes - field_infos.add_field :message_id + field_infos.add_field :message_id, :index => :untokenized field_infos.add_field :source_id field_infos.add_field :source_info field_infos.add_field :date, :index => :untokenized - field_infos.add_field :body, :store => :no + field_infos.add_field :body field_infos.add_field :label field_infos.add_field :subject field_infos.add_field :from field_infos.add_field :to field_infos.add_field :refs @@ -155,11 +159,11 @@ ## Syncs the message to the index: deleting if it's already there, ## and adding either way. Index state will be determined by m.labels. ## ## docid and entry can be specified if they're already known. - def sync_message m, docid=nil, entry=nil + def sync_message m, docid=nil, entry=nil, opts={} docid, entry = load_entry_for_id m.id unless docid && entry raise "no source info for message #{m.id}" unless m.source && m.source_info 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 @@ -168,30 +172,63 @@ m.source else m.source.id or raise "unregistered source #{m.source} (id #{m.source.id.inspect})" end - to = (m.to + m.cc + m.bcc).map { |x| x.email }.join(" ") snippet = if m.snippet_contains_encrypted_content? && $config[:discard_snippets_from_encrypted_messages] "" else m.snippet end + ## write the new document to the index. if the entry already exists in the + ## index, reuse it (which avoids having to reload the entry from the source, + ## which can be quite expensive for e.g. large threads of IMAP actions.) + ## + ## exception: if the index entry belongs to an earlier version of the + ## message, use everything from the new message instead, but union the + ## flags. this allows messages sent to mailing lists to have their header + ## updated and to have flags set properly. + ## + ## minor hack: messages in sources with lower ids have priority over + ## messages in sources with higher ids. so messages in the inbox will + ## override everyone, and messages in the sent box will be overridden + ## by everyone else. + ## + ## written in this manner to support previous versions of the index which + ## did not keep around the entry body. upgrading is thus seamless. + entry ||= {} + labels = m.labels.uniq # override because this is the new state, unless... + + ## if we are a later version of a message, ignore what's in the index, + ## but merge in the labels. + if entry[:source_id] && entry[:source_info] && entry[:label] && + ((entry[:source_id].to_i > source_id) || (entry[:source_info].to_i < m.source_info)) + labels = (entry[:label].split(/\s+/).map { |l| l.intern } + m.labels).uniq + #Redwood::log "found updated version of message #{m.id}: #{m.subj}" + #Redwood::log "previous version was at #{entry[:source_id].inspect}:#{entry[:source_info].inspect}, this version at #{source_id.inspect}:#{m.source_info.inspect}" + #Redwood::log "merged labels are #{labels.inspect} (index #{entry[:label].inspect}, message #{m.labels.inspect})" + entry = {} + end + + ## if force_overwite is true, ignore what's in the index. this is used + ## primarily by sup-sync to force index updates. + entry = {} if opts[:force_overwrite] + d = { :message_id => m.id, :source_id => source_id, :source_info => m.source_info, - :date => m.date.to_indexable_s, - :body => m.content, - :snippet => snippet, - :label => m.labels.uniq.join(" "), - :from => m.from ? m.from.email : "", - :to => (m.to + m.cc + m.bcc).map { |x| x.email }.join(" "), - :subject => wrap_subj(m.subj), - :refs => (m.refs + m.replytos).uniq.join(" "), + :date => (entry[:date] || m.date.to_indexable_s), + :body => (entry[:body] || m.indexable_content), + :snippet => snippet, # always override + :label => labels.uniq.join(" "), + :from => (entry[:from] || (m.from ? m.from.indexable_content : "")), + :to => (entry[:to] || (m.to + m.cc + m.bcc).map { |x| x.indexable_content }.join(" ")), + :subject => (entry[:subject] || wrap_subj(Message.normalize_subj(m.subj))), + :refs => (entry[:refs] || (m.refs + m.replytos).uniq.join(" ")), } @index.delete docid if docid @index.add_document d @@ -248,10 +285,11 @@ #Redwood::log "Building thread for #{m.id}: #{m.subj}" messages = {} searched = {} num_queries = 0 + pending = [m.id] if $config[:thread_by_subject] # do subject queries date_min = m.date - (SAME_SUBJECT_DATE_LIMIT * 12 * 3600) date_max = m.date + (SAME_SUBJECT_DATE_LIMIT * 12 * 3600) q = Ferret::Search::BooleanQuery.new true @@ -262,14 +300,17 @@ q.add_query sq, :must q.add_query Ferret::Search::RangeQuery.new(:date, :>= => date_min.to_indexable_s, :<= => date_max.to_indexable_s), :must q = build_query :qobj => q - pending = @index.search(q).hits.map { |hit| @index[hit.doc][:message_id] } - Redwood::log "found #{pending.size} results for subject query #{q}" - else - pending = [m.id] + p1 = @index.search(q).hits.map { |hit| @index[hit.doc][:message_id] } + Redwood::log "found #{p1.size} results for subject query #{q}" + + p2 = @index.search(q.to_s, :limit => :all).hits.map { |hit| @index[hit.doc][:message_id] } + Redwood::log "found #{p2.size} results in string form" + + pending = (pending + p1 + p2).uniq end until pending.empty? || (opts[:limit] && messages.size >= opts[:limit]) q = Ferret::Search::BooleanQuery.new true # this disappeared in newer ferrets... wtf. @@ -394,22 +435,11 @@ ## do any specialized parsing ## returns nil and flashes error message if parsing failed def parse_user_query_string s extraopts = {} - ## this is a little hacky, but it works, at least until ferret changes - ## its api. we parse the user query string with ferret twice: the first - ## time we just turn the resulting object back into a string, which has - ## the next effect of transforming the original string into a nice - ## normalized form with + and - instead of AND, OR, etc. then we do some - ## string substitutions which depend on this normalized form, re-parse - ## the string with Ferret, and return the resulting query object. - - norms = @qparser.parse(s).to_s - Redwood::log "normalized #{s.inspect} to #{norms.inspect}" unless s == norms - - subs = norms.gsub(/\b(to|from):(\S+)\b/) do + subs = s.gsub(/\b(to|from):(\S+)\b/) do field, name = $1, $2 if(p = ContactManager.contact_for(name)) [field, p.email] elsif name == "me" [field, "(" + AccountManager.user_emails.join("||") + ")"] @@ -475,11 +505,10 @@ end end subs = nil if chronic_failure end - Redwood::log "translated #{norms.inspect} to #{subs.inspect}" unless subs == norms if subs [@qparser.parse(subs), extraopts] else nil end @@ -510,10 +539,10 @@ bakfn = fn + ".bak" if File.exists? fn File.chmod 0600, fn FileUtils.mv fn, bakfn, :force => true unless File.exists?(bakfn) && File.size(fn) == 0 end - Redwood::save_yaml_obj @sources.values.sort_by { |s| s.id.to_i }, fn, true + Redwood::save_yaml_obj sources.sort_by { |s| s.id.to_i }, fn, true File.chmod 0600, fn end @sources_dirty = false end end