5 ## a Message is what's threaded.
7 ## it is also where the parsing for quotes and signatures is done, but
8 ## that should be moved out to a separate class at some point (because
9 ## i would like, for example, to be able to add in a ruby-talk
10 ## specific module that would detect and link to /ruby-talk:\d+/
11 ## sequences in the text of an email. (how sweet would that be?)
13 ## this class catches all source exceptions. if the underlying source
14 ## throws an error, it is caught and handled.
18 RE_PATTERN = /^((re|re[\[\(]\d[\]\)]):\s*)+/i
20 ## some utility methods
22 def normalize_subj s; s.gsub(RE_PATTERN, ""); end
23 def subj_is_reply? s; s =~ RE_PATTERN; end
24 def reify_subj s; subj_is_reply?(s) ? s : "Re: " + s; end
27 QUOTE_PATTERN = /^\s{0,4}[>|\}]/
28 BLOCK_QUOTE_PATTERN = /^-----\s*Original Message\s*----+$/
29 SIG_PATTERN = /(^-- ?$)|(^\s*----------+\s*$)|(^\s*_________+\s*$)|(^\s*--~--~-)|(^\s*--\+\+\*\*==)/
31 MAX_SIG_DISTANCE = 15 # lines from the end
33 DEFAULT_SENDER = "(missing sender)"
35 attr_reader :id, :date, :from, :subj, :refs, :replytos, :to, :source,
36 :cc, :bcc, :labels, :attachments, :list_address, :recipient_email, :replyto,
37 :source_info, :list_subscribe, :list_unsubscribe
39 bool_reader :dirty, :source_marked_read, :snippet_contains_encrypted_content
41 ## if you specify a :header, will use values from that. otherwise,
42 ## will try and load the header from the source.
44 @source = opts[:source] or raise ArgumentError, "source can't be nil"
45 @source_info = opts[:source_info] or raise ArgumentError, "source_info can't be nil"
46 @snippet = opts[:snippet]
47 @snippet_contains_encrypted_content = false
48 @have_snippet = !(opts[:snippet].nil? || opts[:snippet].empty?)
49 @labels = (opts[:labels] || []).to_set_of_symbols
55 ## we need to initialize this. see comments in parse_header as to
59 #parse_header(opts[:header] || @source.load_header(@source_info))
62 def parse_header header
63 ## forcibly decode these headers from and to the current encoding,
64 ## which serves to strip out characters that aren't displayable
65 ## (and which would otherwise be screwing up the display)
66 %w(from to subject cc bcc).each do |f|
67 header[f] = Iconv.easy_decode($encoding, $encoding, header[f]) if header[f]
70 @id = if header["message-id"]
71 mid = header["message-id"] =~ /<(.+?)>/ ? $1 : header["message-id"]
72 sanitize_message_id mid
74 id = "sup-faked-" + Digest::MD5.hexdigest(raw_header)
76 #Redwood::log "faking non-existent message-id for message from #{from}: #{id}"
80 @from = Person.from_address(if header["from"]
83 name = "Sup Auto-generated Fake Sender <sup@fake.sender.example.com>"
84 #Redwood::log "faking non-existent sender for message #@id: #{name}"
88 @date = case(date = header["date"])
94 rescue ArgumentError => e
95 #Redwood::log "faking mangled date header for #{@id} (orig #{header['date'].inspect} gave error: #{e.message})"
99 #Redwood::log "faking non-existent date header for #{@id}"
103 @subj = header.member?("subject") ? header["subject"].gsub(/\s+/, " ").gsub(/\s+$/, "") : DEFAULT_SUBJECT
104 @to = Person.from_address_list header["to"]
105 @cc = Person.from_address_list header["cc"]
106 @bcc = Person.from_address_list header["bcc"]
108 ## before loading our full header from the source, we can actually
109 ## have some extra refs set by the UI. (this happens when the user
110 ## joins threads manually). so we will merge the current refs values
112 refs = (header["references"] || "").scan(/<(.+?)>/).map { |x| sanitize_message_id x.first }
113 @refs = (@refs + refs).uniq
114 @replytos = (header["in-reply-to"] || "").scan(/<(.+?)>/).map { |x| sanitize_message_id x.first }
116 @replyto = Person.from_address header["reply-to"]
118 if header["list-post"]
119 @list_address = Person.from_address header["list-post"].gsub(/^<mailto:|>$/, "")
124 @recipient_email = header["envelope-to"] || header["x-original-to"] || header["delivered-to"]
125 @source_marked_read = header["status"] == "RO"
126 @list_subscribe = header["list-subscribe"]
127 @list_unsubscribe = header["list-unsubscribe"]
130 ## Expected index entry format:
131 ## :message_id, :subject => String
133 ## :refs, :replytos => Array of String
135 ## :to, :cc, :bcc => Array of Person
136 def load_from_index! entry
137 @id = entry[:message_id]
140 @subj = entry[:subject]
144 @refs = (@refs + entry[:refs]).uniq
145 @replytos = entry[:replytos]
149 @recipient_email = nil
150 @source_marked_read = false
151 @list_subscribe = nil
152 @list_unsubscribe = nil
161 @dirty = true if @refs.delete ref
164 def snippet; @snippet || (chunks && @snippet); end
165 def is_list_message?; !@list_address.nil?; end
166 def is_draft?; @source.is_a? DraftLoader; end
168 raise "not a draft" unless is_draft?
169 @source.fn_for_offset @source_info
172 ## sanitize message ids by removing spaces and non-ascii characters.
173 ## also, truncate to 255 characters. all these steps are necessary
174 ## to make ferret happy. of course, we probably fuck up a couple
175 ## valid message ids as well. as long as we're consistent, this
176 ## should be fine, though.
178 ## also, mostly the message ids that are changed by this belong to
181 ## an alternative would be to SHA1 or MD5 all message ids on a regular basis.
183 def sanitize_message_id mid; mid.gsub(/(\s|[^\000-\177])+/, "")[0..254] end
187 index.sync_message self
192 def has_label? t; @labels.member? t; end
194 return if @labels.member? t
195 @labels = (@labels + [t]).to_set_of_symbols
199 return unless @labels.member? t
209 @labels = l.to_set_of_symbols
218 ## this is called when the message body needs to actually be loaded.
219 def load_from_source!
221 if @source.respond_to?(:has_errors?) && @source.has_errors?
222 [Chunk::Text.new(error_message(@source.error.message).split("\n"))]
225 ## we need to re-read the header because it contains information
226 ## that we don't store in the index. actually i think it's just
227 ## the mailing list address (if any), so this is kinda overkill.
228 ## i could just store that in the index, but i think there might
229 ## be other things like that in the future, and i'd rather not
231 ## actually, it's also the differentiation between to/cc/bcc,
232 ## so i will keep this.
233 parse_header @source.load_header(@source_info)
234 message_to_chunks @source.load_message(@source_info)
235 rescue SourceError, SocketError => e
236 Redwood::log "problem getting messages from #{@source}: #{e.message}"
237 ## we need force_to_top here otherwise this window will cover
238 ## up the error message one
240 Redwood::report_broken_sources :force_to_top => true
241 [Chunk::Text.new(error_message(e.message).split("\n"))]
246 def error_message msg
250 ***********************************************************************
251 An error occurred while loading this message. It is possible that
252 the source has changed, or (in the case of remote sources) is down.
253 You can check the log for errors, though hopefully an error window
254 should have popped up at some point.
256 The message location was:
257 #@source##@source_info
258 ***********************************************************************
260 The error message was:
265 ## wrap any source methods that might throw sourceerrors
266 def with_source_errors_handled
269 rescue SourceError => e
270 Redwood::log "problem getting messages from #{@source}: #{e.message}"
272 Redwood::report_broken_sources :force_to_top => true
273 error_message e.message
278 with_source_errors_handled { @source.raw_header @source_info }
282 with_source_errors_handled { @source.raw_message @source_info }
285 ## much faster than raw_message
286 def each_raw_message_line &b
287 with_source_errors_handled { @source.each_raw_message_line(@source_info, &b) }
290 ## returns all the content from a message that will be indexed
291 def indexable_content
294 from && from.indexable_content,
295 to.map { |p| p.indexable_content },
296 cc.map { |p| p.indexable_content },
297 bcc.map { |p| p.indexable_content },
298 indexable_chunks.map { |c| c.lines },
300 ].flatten.compact.join " "
304 indexable_chunks.map { |c| c.lines }.flatten.compact.join " "
308 chunks.select { |c| c.is_a? Chunk::Text }
311 def indexable_subject
312 Message.normalize_subj(subj)
315 def quotable_body_lines
316 chunks.find_all { |c| c.quotable? }.map { |c| c.lines }.flatten
319 def quotable_header_lines
320 ["From: #{@from.full_address}"] +
321 (@to.empty? ? [] : ["To: " + @to.map { |p| p.full_address }.join(", ")]) +
322 (@cc.empty? ? [] : ["Cc: " + @cc.map { |p| p.full_address }.join(", ")]) +
323 (@bcc.empty? ? [] : ["Bcc: " + @bcc.map { |p| p.full_address }.join(", ")]) +
324 ["Date: #{@date.rfc822}",
328 def self.build_from_source source, source_info
329 m = Message.new :source => source, :source_info => source_info
336 ## here's where we handle decoding mime attachments. unfortunately
337 ## but unsurprisingly, the world of mime attachments is a bit of a
338 ## mess. as an empiricist, i'm basing the following behavior on
339 ## observed mail rather than on interpretations of rfcs, so probably
340 ## this will have to be tweaked.
342 ## the general behavior i want is: ignore content-disposition, at
343 ## least in so far as it suggests something being inline vs being an
344 ## attachment. (because really, that should be the recipient's
345 ## decision to make.) if a mime part is text/plain, OR if the user
346 ## decoding hook converts it, then decode it and display it
347 ## inline. for these decoded attachments, if it has associated
348 ## filename, then make it collapsable and individually saveable;
349 ## otherwise, treat it as regular body text.
351 ## everything else is just an attachment and is not displayed
354 ## so, in contrast to mutt, the user is not exposed to the workings
355 ## of the gruesome slaughterhouse and sausage factory that is a
356 ## mime-encoded message, but need only see the delicious end
359 def multipart_signed_to_chunks m
361 Redwood::log "warning: multipart/signed with #{m.body.size} parts (expecting 2)"
365 payload, signature = m.body
366 if signature.multipart?
367 Redwood::log "warning: multipart/signed with payload multipart #{payload.multipart?} and signature multipart #{signature.multipart?}"
371 ## this probably will never happen
372 if payload.header.content_type == "application/pgp-signature"
373 Redwood::log "warning: multipart/signed with payload content type #{payload.header.content_type}"
377 if signature.header.content_type != "application/pgp-signature"
378 ## unknown signature type; just ignore.
379 #Redwood::log "warning: multipart/signed with signature content type #{signature.header.content_type}"
383 [CryptoManager.verify(payload, signature), message_to_chunks(payload)].flatten.compact
386 def multipart_encrypted_to_chunks m
388 Redwood::log "warning: multipart/encrypted with #{m.body.size} parts (expecting 2)"
392 control, payload = m.body
393 if control.multipart?
394 Redwood::log "warning: multipart/encrypted with control multipart #{control.multipart?} and payload multipart #{payload.multipart?}"
398 if payload.header.content_type != "application/octet-stream"
399 Redwood::log "warning: multipart/encrypted with payload content type #{payload.header.content_type}"
403 if control.header.content_type != "application/pgp-encrypted"
404 Redwood::log "warning: multipart/encrypted with control content type #{signature.header.content_type}"
408 decryptedm, sig, notice = CryptoManager.decrypt payload
409 children = message_to_chunks(decryptedm, true) if decryptedm
410 [notice, sig, children].flatten.compact
413 ## takes a RMail::Message, breaks it into Chunk:: classes.
414 def message_to_chunks m, encrypted=false, sibling_types=[]
417 case m.header.content_type
418 when "multipart/signed"
419 multipart_signed_to_chunks m
420 when "multipart/encrypted"
421 multipart_encrypted_to_chunks m
425 sibling_types = m.body.map { |p| p.header.content_type }
426 chunks = m.body.map { |p| message_to_chunks p, encrypted, sibling_types }.flatten.compact
430 elsif m.header.content_type == "message/rfc822"
431 payload = RMail::Parser.read(m.body)
432 from = payload.header.from.first
433 from_person = from ? Person.from_address(from.format) : nil
434 [Chunk::EnclosedMessage.new(from_person, payload.to_s)] +
435 message_to_chunks(payload, encrypted)
438 ## first, paw through the headers looking for a filename
439 if m.header["Content-Disposition"] && m.header["Content-Disposition"] =~ /filename="?(.*?[^\\])("|;|$)/
441 elsif m.header["Content-Type"] && m.header["Content-Type"] =~ /name="?(.*?[^\\])("|;|$)/
444 ## haven't found one, but it's a non-text message. fake
447 ## TODO: make this less lame.
448 elsif m.header["Content-Type"] && m.header["Content-Type"] !~ /^text\/plain/
450 case m.header["Content-Type"]
451 when /text\/html/ then "html"
452 when /image\/(.*)/ then $1
455 ["sup-attachment-#{Time.now.to_i}-#{rand 10000}", extension].join(".")
458 ## if there's a filename, we'll treat it as an attachment.
460 # add this to the attachments list if its not a generated html
461 # attachment (should we allow images with generated names?).
462 # Lowercase the filename because searches are easier that way
463 @attachments.push filename.downcase unless filename =~ /^sup-attachment-/
464 add_label :attachment unless filename =~ /^sup-attachment-/
465 content_type = m.header.content_type || "application/unknown" # sometimes RubyMail gives us nil
466 [Chunk::Attachment.new(content_type, filename, m, sibling_types)]
468 ## otherwise, it's body text
470 ## if there's no charset, use the current encoding as the charset.
471 ## this ensures that the body is normalized to avoid non-displayable
473 body = Iconv.easy_decode($encoding, m.charset || $encoding, m.decode) if m.body
474 text_to_chunks((body || "").normalize_whitespace.split("\n"), encrypted)
479 ## parse the lines of text into chunk objects. the heuristics here
480 ## need tweaking in some nice manner. TODO: move these heuristics
481 ## into the classes themselves.
482 def text_to_chunks lines, encrypted
483 state = :text # one of :text, :quote, or :sig
487 lines.each_with_index do |line, i|
488 nextline = lines[(i + 1) ... lines.length].find { |l| l !~ /^\s*$/ } # skip blank lines
494 ## the following /:$/ followed by /\w/ is an attempt to detect the
495 ## start of a quote. this is split into two regexen because the
496 ## original regex /\w.*:$/ had very poor behavior on long lines
497 ## like ":a:a:a:a:a" that occurred in certain emails.
498 if line =~ QUOTE_PATTERN || (line =~ /:$/ && line =~ /\w/ && nextline =~ QUOTE_PATTERN)
500 elsif line =~ SIG_PATTERN && (lines.length - i) < MAX_SIG_DISTANCE
502 elsif line =~ BLOCK_QUOTE_PATTERN
503 newstate = :block_quote
507 chunks << Chunk::Text.new(chunk_lines) unless chunk_lines.empty?
517 if line =~ QUOTE_PATTERN || (line =~ /^\s*$/ && nextline =~ QUOTE_PATTERN)
519 elsif line =~ SIG_PATTERN && (lines.length - i) < MAX_SIG_DISTANCE
526 if chunk_lines.empty?
529 chunks << Chunk::Quote.new(chunk_lines)
535 when :block_quote, :sig
539 if !@have_snippet && state == :text && (@snippet.nil? || @snippet.length < SNIPPET_LEN) && line !~ /[=\*#_-]{3,}/ && line !~ /^\s*$/
541 @snippet += " " unless @snippet.empty?
542 @snippet += line.gsub(/^\s+/, "").gsub(/[\r\n]/, "").gsub(/\s+/, " ")
543 @snippet = @snippet[0 ... SNIPPET_LEN].chomp
544 @dirty = true unless encrypted && $config[:discard_snippets_from_encrypted_messages]
545 @snippet_contains_encrypted_content = true if encrypted
551 when :quote, :block_quote
552 chunks << Chunk::Quote.new(chunk_lines) unless chunk_lines.empty?
554 chunks << Chunk::Text.new(chunk_lines) unless chunk_lines.empty?
556 chunks << Chunk::Signature.new(chunk_lines) unless chunk_lines.empty?