require 'thread'
require 'fileutils'
-#require 'ferret'
-require_gem 'ferret', ">= 0.10.13"
+require 'ferret'
module Redwood
-class IndexError < StandardError
- attr_reader :source
-
- def initialize source, s
- super s
- @source = source
- end
-end
-
class Index
include Singleton
- attr_reader :index # debugging only
-
+ attr_reader :index
def initialize dir=BASE_DIR
@dir = dir
@sources = {}
@sources_dirty = false
- @qparser ||= Ferret::QueryParser.new :default_field => :body, :analyzer => Ferret::Analysis::WhiteSpaceAnalyzer.new
+
+ wsa = Ferret::Analysis::WhiteSpaceAnalyzer.new false
+ sa = Ferret::Analysis::StandardAnalyzer.new Ferret::Analysis::FULL_ENGLISH_STOP_WORDS, true
+ @analyzer = Ferret::Analysis::PerFieldAnalyzer.new wsa
+ @analyzer[:body] = sa
+ @qparser ||= Ferret::QueryParser.new :default_field => :body, :analyzer => @analyzer
self.class.i_am_the_instance self
end
@sources[source.id] = source
end
- def source_for name; @sources.values.find { |s| s.is_source_for? name }; 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 load_index dir=File.join(@dir, "ferret")
- wsa = Ferret::Analysis::WhiteSpaceAnalyzer.new false
- sa = Ferret::Analysis::StandardAnalyzer.new
- analyzer = Ferret::Analysis::PerFieldAnalyzer.new wsa
- analyzer[:body] = sa
-
if File.exists? dir
- Redwood::log "loading index"
- @index = Ferret::Index::Index.new(:path => dir, :analyzer => analyzer)
+ 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"
+ Redwood::log "creating index..."
field_infos = Ferret::Index::FieldInfos.new :store => :yes
field_infos.add_field :message_id
field_infos.add_field :source_id
field_infos.add_field :refs
field_infos.add_field :snippet, :index => :no, :term_vector => :no
field_infos.create_index dir
- @index = Ferret::Index::Index.new(:path => dir, :analyzer => analyzer)
+ @index = Ferret::Index::Index.new(:path => dir, :analyzer => @analyzer)
end
end
- ## update the message by deleting and re-adding
- def update_message m, source=nil, source_info=nil
- docid, entry = load_entry_for_id m.id
- if entry
- source ||= entry[:source_id].to_i
- source_info ||= entry[:source_info].to_i
+ ## Update the message state on disk, by deleting and re-adding it.
+ ## The message must exist in the index. docid and entry are found
+ ## unless given.
+ ##
+ ## Overwrites the labels on disk with the new labels in 'm', so that
+ ## we can actually change message state.
+ def update_message m, docid=nil, entry=nil
+ unless docid && entry
+ docid, entry = load_entry_for_id m.id
+ raise ArgumentError, "cannot find #{m.id} in the index" unless entry
end
- ## this happens sometimes. i'm not sure why. ferret bug?
- raise "no entry and no source info for message #{m.id}" unless source && source_info
+ raise "no entry and no source info for message #{m.id}" unless m.source && m.source_info
raise "deleting non-corresponding entry #{docid}" unless @index[docid][:message_id] == m.id
+
@index.delete docid
add_message m
+ docid, entry = load_entry_for_id m.id
end
def save_index fn=File.join(@dir, "ferret")
- # don't have to do anything apparently
+ # don't have to do anything, apparently
end
def contains_id? id
def size; @index.size; end
## you should probably not call this on a block that doesn't break
- ## rather quickly because the results will probably be, as we say
- ## in scotland, frikkin' huuuge.
+ ## rather quickly because the results can be very large.
EACH_BY_DATE_NUM = 100
def each_id_by_date opts={}
return if @index.size == 0 # otherwise ferret barfs ###TODO: remove this once my ferret patch is accepted
end
def num_results_for opts={}
- query = build_query opts
- x = @index.search(query).total_hits
- Redwood::log "num_results_for: have #{x} for query #{query}"
- x
+ return 0 if @index.size == 0 # otherwise ferret barfs ###TODO: remove this once my ferret patch is accepted
+ q = build_query opts
+ index.search(q).total_hits
end
+ ## yield all messages in the thread containing 'm' by repeatedly
+ ## querying the index. uields pairs of message ids and
+ ## message-building lambdas, so that building an unwanted message
+ ## can be skipped in the block if desired.
+ ##
+ ## stops loading any thread if a message with a :killed flag is found.
+
SAME_SUBJECT_DATE_LIMIT = 7
def each_message_in_thread_for m, opts={}
messages = {}
searched = {}
num_queries = 0
- ## temporarily disabling subject searching because it's a
- ## significant slowdown.
- ##
- ## TODO: make this configurable, i guess
- if false
+ ## todo: make subject querying configurable
+ if true # do subject queries
date_min = m.date - (SAME_SUBJECT_DATE_LIMIT * 12 * 3600)
date_max = m.date + (SAME_SUBJECT_DATE_LIMIT * 12 * 3600)
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
q.add_query Ferret::Search::TermQuery.new(:message_id, id), :should
q.add_query Ferret::Search::TermQuery.new(:refs, id), :should
+ q = build_query :qobj => q, :load_killed => true
+
num_queries += 1
@index.search_each(q, :limit => :all) do |docid, score|
break if opts[:limit] && messages.size >= opts[:limit]
+ break if @index[docid][:label].split(/\s+/).include? "killed" unless opts[:load_killed]
mid = @index[docid][:message_id]
unless messages.member? mid
messages[mid] ||= lambda { build_message docid }
end
end
end
- Redwood::log "ran #{num_queries} queries to build thread of #{messages.size} messages for #{m.id}"
+ Redwood::log "ran #{num_queries} queries to build thread of #{messages.size} messages for #{m.id}" if num_queries > 0
messages.each { |mid, builder| yield mid, builder }
end
#puts "building message #{doc[:message_id]} (#{source}##{doc[:source_info]})"
raise "invalid source #{doc[:source_id]}" unless source
- m =
- if source.broken?
- nil
- else
- begin
- Message.new :source => source, :source_info => doc[:source_info].to_i,
- :labels => doc[:label].split(" ").map { |s| s.intern },
- :snippet => doc[:snippet]
- rescue MessageFormatError => e
- raise IndexError.new(source, "error building message #{doc[:message_id]} at #{source}/#{doc[:source_info]}: #{e.message}")
- rescue SourceError => e
- nil
- end
- end
+ fake_header = {
+ "date" => Time.at(doc[:date].to_i),
+ "subject" => unwrap_subj(doc[:subject]),
+ "from" => doc[:from],
+ "to" => doc[:to],
+ "message-id" => doc[:message_id],
+ "references" => doc[:refs],
+ }
- unless m
- fake_header = {
- "date" => Time.at(doc[:date].to_i),
- "subject" => unwrap_subj(doc[:subject]),
- "from" => doc[:from],
- "to" => doc[:to],
- "message-id" => doc[:message_id],
- "references" => doc[:refs],
- }
-
- m = Message.new :labels => doc[:label].split(" ").map { |s| s.intern },
- :snippet => doc[:snippet], :header => fake_header,
- :body => <<EOS
-#{doc[:snippet]}...
-
-An error occurred while loading this message. It is possible that the source
-has changed, or (in the case of remote sources) is down.
-
-The error message was:
- #{source.broken_msg}
-EOS
- end
- m
+ Message.new :source => source, :source_info => doc[:source_info].to_i,
+ :labels => doc[:label].split(" ").map { |s| s.intern },
+ :snippet => doc[:snippet], :header => fake_header
end
def fresh_thread_id; @next_thread_id += 1; end
def wrap_subj subj; "__START_SUBJECT__ #{subj} __END_SUBJECT__"; end
def unwrap_subj subj; subj =~ /__START_SUBJECT__ (.*?) __END_SUBJECT__/ && $1; end
+ ## Adds a message to the index. The message cannot already exist in
+ ## the index.
def add_message m
- return false if contains? m
+ raise ArgumentError, "index already contains #{m.id}" if contains? m
source_id =
if m.source.is_a? Integer
m.source
else
- m.source.id or raise "unregistered source #{m.source}"
+ 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(" ")
@index.add_document d
- ## TODO: figure out why this is sometimes triggered
- #docid, entry = load_entry_for_id m.id
- #raise "just added message #{m.id} but couldn't find it in a search" unless docid
+ docid, entry = load_entry_for_id m.id
+ ## this hasn't been triggered in a long time. TODO: decide whether it's still a problem.
+ raise "just added message #{m.id} but couldn't find it in a search" unless docid
true
end
def parse_user_query_string str; @qparser.parse str; end
def build_query opts
-
query = Ferret::Search::BooleanQuery.new
query.add_query opts[:qobj], :must if opts[:qobj]
labels = ([opts[:label]] + (opts[:labels] || [])).compact
end
query.add_query Ferret::Search::TermQuery.new("label", "spam"), :must_not unless opts[:load_spam] || labels.include?(:spam)
+ query.add_query Ferret::Search::TermQuery.new("label", "deleted"), :must_not unless opts[:load_deleted] || labels.include?(:deleted)
query.add_query Ferret::Search::TermQuery.new("label", "killed"), :must_not unless opts[:load_killed] || labels.include?(:killed)
query
end
projects / sup / blobdiff