require 'stringio'
require 'time'
require 'rmail'
+require 'cgi'
-## fucking imap fucking sucks. what the FUCK kind of committee of
-## dunces designed this shit.
-
+## fucking imap fucking sucks. what the FUCK kind of committee of dunces
+## designed this shit.
+##
## imap talks about 'unique ids' for messages, to be used for
-## cross-session identification. great---just what sup needs! except
-## it turns out the uids can be invalidated every time the
-## 'uidvalidity' value changes on the server, and 'uidvalidity' can
-## change without restriction. it can change any time you log in. it
-## can change EVERY time you log in. of course the imap spec "strongly
-## recommends" that it never change, but there's nothing to stop
-## people from just setting it to the current timestamp, and in fact
-## that's exactly what the one imap server i have at my disposal
-## does. thus the so-called uids are absolutely useless and imap
-## provides no cross-session way of uniquely identifying a
-## message. but thanks for the "strong recommendation", guys!
-
+## cross-session identification. great---just what sup needs! except it
+## turns out the uids can be invalidated every time the 'uidvalidity'
+## value changes on the server, and 'uidvalidity' can change without
+## restriction. it can change any time you log in. it can change EVERY
+## time you log in. of course the imap spec "strongly recommends" that it
+## never change, but there's nothing to stop people from just setting it
+## to the current timestamp, and in fact that's exactly what the one imap
+## server i have at my disposal does. thus the so-called uids are
+## absolutely useless and imap provides no cross-session way of uniquely
+## identifying a message. but thanks for the "strong recommendation",
+## guys!
+##
## so right now i'm using the 'internal date' and the size of each
## message to uniquely identify it, and i scan over the entire mailbox
## each time i open it to map those things to message ids. that can be
-## slow for large mailboxes, and we'll just have to hope that there
-## are no collisions. ho ho! a perfectly reasonable solution!
-
+## slow for large mailboxes, and we'll just have to hope that there are
+## no collisions. ho ho! a perfectly reasonable solution!
+##
+## and here's another thing. check out RFC2060 2.2.2 paragraph 5:
+##
+## A client MUST be prepared to accept any server response at all
+## times. This includes server data that was not requested.
+##
+## yeah. that totally makes a lot of sense. and once again, the idiocy of
+## the spec actually happens in practice. you'll request flags for one
+## message, and get it interspersed with a random bunch of flags for some
+## other messages, including a different set of flags for the same
+## message! totally ok by the imap spec. totally retarded by any other
+## metric.
+##
## fuck you, imap committee. you managed to design something nearly as
## shitty as mbox but goddamn THIRTY YEARS LATER.
module Redwood
RECOVERABLE_ERRORS = [ Errno::EPIPE, Errno::ETIMEDOUT, OpenSSL::SSL::SSLError ]
attr_accessor :username, :password
+ yaml_properties :uri, :username, :password, :cur_offset, :usual,
+ :archived, :id, :labels
- def initialize uri, username, password, last_idate=nil, usual=true, archived=false, id=nil
+ def initialize uri, username, password, last_idate=nil, usual=true, archived=false, id=nil, labels=[]
raise ArgumentError, "username and password must be specified" unless username && password
raise ArgumentError, "not an imap uri" unless uri =~ %r!imaps?://!
@username = username
@password = password
@imap = nil
- @imap_ids = {}
+ @imap_state = {}
@ids = []
@last_scan = nil
- @labels = [:unread]
- @labels << mailbox.intern unless mailbox =~ /inbox/i
+ @labels = ((labels || []) - LabelManager::RESERVED_LABELS).uniq.freeze
+ @say_id = nil
@mutex = Mutex.new
end
+ def self.suggest_labels_for path
+ path =~ /([^\/]*inbox[^\/]*)/i ? [$1.downcase.intern] : []
+ end
+
def host; @parsed_uri.host; end
def port; @parsed_uri.port || (ssl? ? 993 : 143); end
def mailbox
x = @parsed_uri.path[1..-1]
- x.nil? || x.empty? ? 'INBOX' : x
+ (x.nil? || x.empty?) ? 'INBOX' : CGI.unescape(x)
end
def ssl?; @parsed_uri.scheme == 'imaps' end
- def check
- ids =
- @mutex.synchronize do
- unsynchronized_scan_mailbox
- @ids
- end
-
- start = ids.index(cur_offset || start_offset) or raise OutOfSyncSourceError, "Unknown message id #{cur_offset || start_offset}."
- end
+ def check; end # do nothing because anything we do will be too slow,
+ # and we'll catch the errors later.
## is this necessary? TODO: remove maybe
def == o; o.is_a?(IMAP) && o.uri == self.uri && o.username == self.username; end
end
def load_message id
- RMail::Parser.read raw_full_message(id)
+ RMail::Parser.read raw_message(id)
+ end
+
+ def each_raw_message_line id
+ StringIO.new(raw_message(id)).each { |l| yield l }
end
def raw_header id
unsynchronized_scan_mailbox
- header, flags = get_imap_fields id, 'RFC822.HEADER', 'FLAGS'
- header = header + "Status: RO\n" if flags.include? :Seen # fake an mbox-style read header # TODO: improve source-marked-as-read reporting system
+ header, flags = get_imap_fields id, 'RFC822.HEADER'
header.gsub(/\r\n/, "\n")
end
synchronized :raw_header
- def raw_full_message id
+ def raw_message id
unsynchronized_scan_mailbox
get_imap_fields(id, 'RFC822').first.gsub(/\r\n/, "\n")
end
- synchronized :raw_full_message
+ synchronized :raw_message
def connect
return if @imap
return if last_id == @ids.length
- Redwood::log "fetching IMAP headers #{(@ids.length + 1) .. last_id}"
- values = safely { @imap.fetch((@ids.length + 1) .. last_id, ['RFC822.SIZE', 'INTERNALDATE']) }
- values.each do |v|
+ range = (@ids.length + 1) .. last_id
+ Redwood::log "fetching IMAP headers #{range}"
+ fetch(range, ['RFC822.SIZE', 'INTERNALDATE', 'FLAGS']).each do |v|
id = make_id v
@ids << id
- @imap_ids[id] = v.seqno
+ @imap_state[id] = { :id => v.seqno, :flags => v.attr["FLAGS"] }
end
+ Redwood::log "done fetching IMAP headers"
end
synchronized :scan_mailbox
def each
+ return unless start_offset
+
ids =
@mutex.synchronize do
unsynchronized_scan_mailbox
start = ids.index(cur_offset || start_offset) or raise OutOfSyncSourceError, "Unknown message id #{cur_offset || start_offset}."
- start.upto(ids.length - 1) do |i|
+ start.upto(ids.length - 1) do |i|
id = ids[i]
- self.cur_offset = id
- yield id, @labels.clone
+ state = @mutex.synchronize { @imap_state[id] } or next
+ self.cur_offset = id
+ labels = { :Flagged => :starred,
+ :Deleted => :deleted
+ }.inject(@labels) do |cur, (imap, sup)|
+ cur + (state[:flags].include?(imap) ? [sup] : [])
+ end
+
+ labels += [:unread] unless state[:flags].include?(:Seen)
+
+ yield id, labels
end
end
private
+ def fetch ids, fields
+ results = safely { @imap.fetch ids, fields }
+ good_results =
+ if ids.respond_to? :member?
+ results.find_all { |r| ids.member?(r.seqno) && fields.all? { |f| r.attr.member?(f) } }
+ else
+ results.find_all { |r| ids == r.seqno && fields.all? { |f| r.attr.member?(f) } }
+ end
+
+ if good_results.empty?
+ raise FatalSourceError, "no IMAP response for #{ids} containing all fields #{fields.join(', ')} (got #{results.size} results)"
+ elsif good_results.size < results.size
+ Redwood::log "Your IMAP server sucks. It sent #{results.size} results for a request for #{good_results.size} messages. What are you using, Binc?"
+ end
+
+ good_results
+ end
+
def unsafe_connect
say "Connecting to IMAP server #{host}:#{port}..."
- ## apparently imap.rb does a lot of threaded stuff internally and
- ## if an exception occurs, it will catch it and re-raise it on the
- ## calling thread. but i can't seem to catch that exception, so
- ## i've resorted to initializing it in its own thread. surely
- ## there's a better way.
+ ## apparently imap.rb does a lot of threaded stuff internally and if
+ ## an exception occurs, it will catch it and re-raise it on the
+ ## calling thread. but i can't seem to catch that exception, so i've
+ ## resorted to initializing it in its own thread. surely there's a
+ ## better way.
exception = nil
::Thread.new do
begin
@imap = Net::IMAP.new host, port, ssl?
say "Logging in..."
- ## although RFC1730 claims that "If an AUTHENTICATE command
- ## fails with a NO response, the client may try another", in
- ## practice it seems like they can also send a BAD response.
+ ## although RFC1730 claims that "If an AUTHENTICATE command fails
+ ## with a NO response, the client may try another", in practice
+ ## it seems like they can also send a BAD response.
begin
+ raise Net::IMAP::NoResponseError unless @imap.capability().member? "AUTH=CRAM-MD5"
@imap.authenticate 'CRAM-MD5', @username, @password
rescue Net::IMAP::BadResponseError, Net::IMAP::NoResponseError => e
Redwood::log "CRAM-MD5 authentication failed: #{e.class}. Trying LOGIN auth..."
begin
+ raise Net::IMAP::NoResponseError unless @imap.capability().member? "AUTH=LOGIN"
@imap.authenticate 'LOGIN', @username, @password
rescue Net::IMAP::BadResponseError, Net::IMAP::NoResponseError => e
Redwood::log "LOGIN authentication failed: #{e.class}. Trying plain-text LOGIN..."
end
def get_imap_fields id, *fields
- imap_id = @imap_ids[id] or raise OutOfSyncSourceError, "Unknown message id #{id}"
-
- retried = false
- results = safely { @imap.fetch imap_id, (fields + ['RFC822.SIZE', 'INTERNALDATE']).uniq }.first
- got_id = make_id results
- raise OutOfSyncSourceError, "IMAP message mismatch: requested #{id}, got #{got_id}." unless got_id == id
-
- fields.map { |f| results.attr[f] or raise FatalSourceError, "empty response from IMAP server: #{f}" }
+ raise OutOfSyncSourceError, "Unknown message id #{id}" unless @imap_state[id]
+
+ imap_id = @imap_state[id][:id]
+ result = fetch(imap_id, (fields + ['RFC822.SIZE', 'INTERNALDATE']).uniq).first
+ got_id = make_id result
+
+ ## I've turned off the following sanity check because Microsoft
+ ## Exchange fails it. Exchange actually reports two different
+ ## INTERNALDATEs for the exact same message when queried at different
+ ## points in time.
+ ##
+ ## RFC2060 defines the semantics of INTERNALDATE for messages that
+ ## arrive via SMTP for via various IMAP commands, but states that
+ ## "All other cases are implementation defined.". Great, thanks guys,
+ ## yet another useless field.
+ ##
+ ## Of course no OTHER imap server I've encountered returns DIFFERENT
+ ## values for the SAME message. But it's Microsoft; what do you
+ ## expect? If their programmers were any good they'd be working at
+ ## Google.
+
+ # raise OutOfSyncSourceError, "IMAP message mismatch: requested #{id}, got #{got_id}." unless got_id == id
+
+ fields.map { |f| result.attr[f] or raise FatalSourceError, "empty response from IMAP server: #{f}" }
end
## execute a block, connected if unconnected, re-connected up to 3
end
-Redwood::register_yaml(IMAP, %w(uri username password cur_offset usual archived id))
-
end