lib/sup/index.rb in sup-0.7 vs lib/sup/index.rb in sup-0.8
- old
+ new
@@ -175,35 +175,35 @@
@index = Ferret::Index::Index.new(:path => dir, :analyzer => @analyzer)
end
end
end
- ## Syncs the message to the index: deleting if it's already there,
- ## and adding either way. Index state will be determined by m.labels.
+ ## Syncs the message to the index, replacing any previous version. adding
+ ## either way. Index state will be determined by the message's #labels
+ ## accessor.
##
- ## docid and entry can be specified if they're already known.
- def sync_message m, docid=nil, entry=nil, opts={}
- docid, entry = load_entry_for_id m.id unless docid && entry
+ ## if need_load is false, docid and entry are assumed to be set to the
+ ## result of load_entry_for_id (which can be nil).
+ def sync_message m, need_load=true, docid=nil, entry=nil, opts={}
+ docid, entry = load_entry_for_id m.id if need_load
raise "no source info for message #{m.id}" unless m.source && m.source_info
@index_mutex.synchronize do
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
end
- source_id =
- if m.source.is_a? Integer
- m.source
- else
- m.source.id or raise "unregistered source #{m.source} (id #{m.source.id.inspect})"
- end
+ source_id = if m.source.is_a? Integer
+ m.source
+ else
+ m.source.id or raise "unregistered source #{m.source} (id #{m.source.id.inspect})"
+ end
- snippet =
- if m.snippet_contains_encrypted_content? && $config[:discard_snippets_from_encrypted_messages]
- ""
- else
- m.snippet
- end
+ 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.)
##
@@ -224,11 +224,11 @@
## 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
+ labels = (entry[:label].symbolistize + 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
@@ -244,25 +244,33 @@
:date => (entry[:date] || m.date.to_indexable_s),
:body => (entry[:body] || m.indexable_content),
:snippet => snippet, # always override
:label => labels.uniq.join(" "),
:attachments => (entry[:attachments] || m.attachments.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(" ")),
+
+ ## always override :from and :to.
+ ## older versions of Sup would often store the wrong thing in the index
+ ## (because they were canonicalizing email addresses, resulting in the
+ ## wrong name associated with each.) the correct address is read from
+ ## the original header when these messages are opened in thread-view-mode,
+ ## so this allows people to forcibly update the address in the index by
+ ## marking those threads for saving.
+ :from => (m.from ? m.from.indexable_content : ""),
+ :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_mutex.synchronize do
+ @index_mutex.synchronize do
@index.delete docid if docid
@index.add_document d
end
- 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.inspect} but couldn't find it in a search" unless docid
- true
+ ## this hasn't been triggered in a long time.
+ ## docid, entry = load_entry_for_id m.id
+ ## raise "just added message #{m.id.inspect} but couldn't find it in a search" unless docid
end
def save_index fn=File.join(@dir, "ferret")
# don't have to do anything, apparently
end
@@ -322,11 +330,11 @@
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
sq = Ferret::Search::PhraseQuery.new(:subject)
- wrap_subj(Message.normalize_subj(m.subj)).split(/\s+/).each do |t|
+ wrap_subj(Message.normalize_subj(m.subj)).split.each do |t|
sq.add_term t
end
q.add_query sq, :must
q.add_query Ferret::Search::RangeQuery.new(:date, :>= => date_min.to_indexable_s, :<= => date_max.to_indexable_s), :must
@@ -367,22 +375,22 @@
end
mid = @index[docid][:message_id]
unless messages.member?(mid)
#Redwood::log "got #{mid} as a child of #{id}"
messages[mid] ||= lambda { build_message docid }
- refs = @index[docid][:refs].split(" ")
+ refs = @index[docid][:refs].split
pending += refs.select { |id| !searched[id] }
end
end
end
end
if killed
Redwood::log "thread for #{m.id} is killed, ignoring"
false
else
- Redwood::log "ran #{num_queries} queries to build thread of #{messages.size + 1} messages for #{m.id}: #{m.subj}" if num_queries > 0
+ Redwood::log "ran #{num_queries} queries to build thread of #{messages.size} messages for #{m.id}: #{m.subj}" if num_queries > 0
messages.each { |mid, builder| yield mid, builder }
true
end
end
@@ -398,18 +406,20 @@
fake_header = {
"date" => Time.at(doc[:date].to_i),
"subject" => unwrap_subj(doc[:subject]),
"from" => doc[:from],
- "to" => doc[:to].split(/\s+/).join(", "), # reformat
+ "to" => doc[:to].split.join(", "), # reformat
"message-id" => doc[:message_id],
- "references" => doc[:refs].split(/\s+/).map { |x| "<#{x}>" }.join(" "),
+ "references" => doc[:refs].split.map { |x| "<#{x}>" }.join(" "),
}
- 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
+ m = Message.new :source => source, :source_info => doc[:source_info].to_i,
+ :labels => doc[:label].symbolistize,
+ :snippet => doc[:snippet]
+ m.parse_header fake_header
+ m
end
end
def fresh_thread_id; @next_thread_id += 1; end
def wrap_subj subj; "__START_SUBJECT__ #{subj} __END_SUBJECT__"; end
@@ -447,13 +457,13 @@
#Redwood::log "got message #{docid} to: #{@index[docid][:to].inspect} and from: #{@index[docid][:from].inspect}"
f = @index[docid][:from]
t = @index[docid][:to]
if AccountManager.is_account_email? f
- t.split(" ").each { |e| contacts[PersonManager.person_for(e)] = true }
+ t.split(" ").each { |e| contacts[Person.from_address(e)] = true }
else
- contacts[PersonManager.person_for(f)] = true
+ contacts[Person.from_address(f)] = true
end
end
end
contacts.keys.compact
@@ -472,14 +482,31 @@
q.add_query Ferret::Search::TermQuery.new("source_id", source.id.to_s), :must
q.add_query Ferret::Search::TermQuery.new("label", label.to_s), :must
@index_mutex.synchronize { @index.search(q, :limit => 1).total_hits > 0 }
end
+ ## takes a user query string and returns the list of docids for messages
+ ## that match the query.
+ ##
+ ## messages can then be loaded from the index with #build_message.
+ ##
+ ## raises a ParseError if the parsing failed.
+ def run_query query
+ qobj, opts = Redwood::Index.parse_user_query_string query
+ query = Redwood::Index.build_query opts.merge(:qobj => qobj)
+ results = @index.search query, :limit => (opts[:limit] || :all)
+ results.hits.map { |hit| hit.doc }
+ end
+
protected
- ## do any specialized parsing
- ## returns nil and flashes error message if parsing failed
+ class ParseError < StandardError; end
+
+ ## parse a query string from the user. returns a query object and a set of
+ ## extra flags; both of these are meant to be passed to #build_query.
+ ##
+ ## raises a ParseError if something went wrong.
def parse_user_query_string s
extraopts = {}
subs = s.gsub(/\b(to|from):(\S+)\b/) do
field, name = $1, $2
@@ -537,15 +564,13 @@
"attachments:(*.#{name.downcase})"
end
end
if $have_chronic
- chronic_failure = false
subs = subs.gsub(/\b(before|on|in|during|after):(\((.+?)\)\B|(\S+)\b)/) do
- break if chronic_failure
field, datestr = $1, ($3 || $4)
- realdate = Chronic.parse(datestr, :guess => false, :context => :none)
+ realdate = Chronic.parse datestr, :guess => false, :context => :past
if realdate
case field
when "after"
Redwood::log "chronic: translated #{field}:#{datestr} to #{realdate.end}"
"date:(>= #{sprintf "%012d", realdate.end.to_i})"
@@ -555,32 +580,29 @@
else
Redwood::log "chronic: translated #{field}:#{datestr} to #{realdate}"
"date:(<= #{sprintf "%012d", realdate.end.to_i}) date:(>= #{sprintf "%012d", realdate.begin.to_i})"
end
else
- BufferManager.flash "Can't understand date #{datestr.inspect}!"
- chronic_failure = true
+ raise ParseError, "can't understand date #{datestr.inspect}"
end
end
- subs = nil if chronic_failure
end
## limit:42 restrict the search to 42 results
subs = subs.gsub(/\blimit:(\S+)\b/) do
lim = $1
if lim =~ /^\d+$/
extraopts[:limit] = lim.to_i
''
else
- BufferManager.flash "Can't understand limit #{lim.inspect}!"
- subs = nil
+ raise ParseError, "non-numeric limit #{lim.inspect}"
end
end
- if subs
+ begin
[@qparser.parse(subs), extraopts]
- else
- nil
+ rescue Ferret::QueryParser::QueryParseException => e
+ raise ParseError, e.message
end
end
def build_query opts
query = Ferret::Search::BooleanQuery.new