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 = Set.new(opts[:labels] || [])
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"] =~ /<(.+?)>/ ? $1 : header["message-id"]
72 sanitize_message_id mid
74 id = "sup-faked-" + Digest::MD5.hexdigest(raw_header)
76 #debug "faking non-existent message-id for message from #{from}: #{id}"
80 @from = Person.from_address(if header["from"]
83 name = "Sup Auto-generated Fake Sender <sup@fake.sender.example.com>"
84 #debug "faking non-existent sender for message #@id: #{name}"
88 @date = case(date = header["date"])
94 rescue ArgumentError => e
95 #debug "faking mangled date header for #{@id} (orig #{header['date'].inspect} gave error: #{e.message})"
99 #debug "faking non-existent date header for #{@id}"
103 @subj = header.member?("subject") ? header["subject"].gsub(/\s+/, " ").gsub(/\s+$/, "") : 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.update_message_state self
192 def has_label? t; @labels.member? t; end
194 return if @labels.member? l
199 return unless @labels.member? l
209 raise ArgumentError, "not a set" unless l.is_a?(Set)
210 return if @labels == l
220 ## this is called when the message body needs to actually be loaded.
221 def load_from_source!
223 if @source.respond_to?(:has_errors?) && @source.has_errors?
224 [Chunk::Text.new(error_message(@source.error.message).split("\n"))]
227 ## we need to re-read the header because it contains information
228 ## that we don't store in the index. actually i think it's just
229 ## the mailing list address (if any), so this is kinda overkill.
230 ## i could just store that in the index, but i think there might
231 ## be other things like that in the future, and i'd rather not
233 ## actually, it's also the differentiation between to/cc/bcc,
234 ## so i will keep this.
235 parse_header @source.load_header(@source_info)
236 message_to_chunks @source.load_message(@source_info)
237 rescue SourceError, SocketError => e
238 warn "problem getting messages from #{@source}: #{e.message}"
239 ## we need force_to_top here otherwise this window will cover
240 ## up the error message one
242 Redwood::report_broken_sources :force_to_top => true
243 [Chunk::Text.new(error_message(e.message).split("\n"))]
248 def error_message msg
252 ***********************************************************************
253 An error occurred while loading this message. It is possible that
254 the source has changed, or (in the case of remote sources) is down.
255 You can check the log for errors, though hopefully an error window
256 should have popped up at some point.
258 The message location was:
259 #@source##@source_info
260 ***********************************************************************
262 The error message was:
267 ## wrap any source methods that might throw sourceerrors
268 def with_source_errors_handled
271 rescue SourceError => e
272 warn "problem getting messages from #{@source}: #{e.message}"
274 Redwood::report_broken_sources :force_to_top => true
275 error_message e.message
280 with_source_errors_handled { @source.raw_header @source_info }
284 with_source_errors_handled { @source.raw_message @source_info }
287 ## much faster than raw_message
288 def each_raw_message_line &b
289 with_source_errors_handled { @source.each_raw_message_line(@source_info, &b) }
292 ## returns all the content from a message that will be indexed
293 def indexable_content
296 from && from.indexable_content,
297 to.map { |p| p.indexable_content },
298 cc.map { |p| p.indexable_content },
299 bcc.map { |p| p.indexable_content },
300 indexable_chunks.map { |c| c.lines },
302 ].flatten.compact.join " "
306 indexable_chunks.map { |c| c.lines }.flatten.compact.join " "
310 chunks.select { |c| c.is_a? Chunk::Text }
313 def indexable_subject
314 Message.normalize_subj(subj)
317 def quotable_body_lines
318 chunks.find_all { |c| c.quotable? }.map { |c| c.lines }.flatten
321 def quotable_header_lines
322 ["From: #{@from.full_address}"] +
323 (@to.empty? ? [] : ["To: " + @to.map { |p| p.full_address }.join(", ")]) +
324 (@cc.empty? ? [] : ["Cc: " + @cc.map { |p| p.full_address }.join(", ")]) +
325 (@bcc.empty? ? [] : ["Bcc: " + @bcc.map { |p| p.full_address }.join(", ")]) +
326 ["Date: #{@date.rfc822}",
330 def self.build_from_source source, source_info
331 m = Message.new :source => source, :source_info => source_info
338 ## here's where we handle decoding mime attachments. unfortunately
339 ## but unsurprisingly, the world of mime attachments is a bit of a
340 ## mess. as an empiricist, i'm basing the following behavior on
341 ## observed mail rather than on interpretations of rfcs, so probably
342 ## this will have to be tweaked.
344 ## the general behavior i want is: ignore content-disposition, at
345 ## least in so far as it suggests something being inline vs being an
346 ## attachment. (because really, that should be the recipient's
347 ## decision to make.) if a mime part is text/plain, OR if the user
348 ## decoding hook converts it, then decode it and display it
349 ## inline. for these decoded attachments, if it has associated
350 ## filename, then make it collapsable and individually saveable;
351 ## otherwise, treat it as regular body text.
353 ## everything else is just an attachment and is not displayed
356 ## so, in contrast to mutt, the user is not exposed to the workings
357 ## of the gruesome slaughterhouse and sausage factory that is a
358 ## mime-encoded message, but need only see the delicious end
361 def multipart_signed_to_chunks m
363 warn "multipart/signed with #{m.body.size} parts (expecting 2)"
367 payload, signature = m.body
368 if signature.multipart?
369 warn "multipart/signed with payload multipart #{payload.multipart?} and signature multipart #{signature.multipart?}"
373 ## this probably will never happen
374 if payload.header.content_type == "application/pgp-signature"
375 warn "multipart/signed with payload content type #{payload.header.content_type}"
379 if signature.header.content_type != "application/pgp-signature"
380 ## unknown signature type; just ignore.
381 #warn "multipart/signed with signature content type #{signature.header.content_type}"
385 [CryptoManager.verify(payload, signature), message_to_chunks(payload)].flatten.compact
388 def multipart_encrypted_to_chunks m
390 warn "multipart/encrypted with #{m.body.size} parts (expecting 2)"
394 control, payload = m.body
395 if control.multipart?
396 warn "multipart/encrypted with control multipart #{control.multipart?} and payload multipart #{payload.multipart?}"
400 if payload.header.content_type != "application/octet-stream"
401 warn "multipart/encrypted with payload content type #{payload.header.content_type}"
405 if control.header.content_type != "application/pgp-encrypted"
406 warn "multipart/encrypted with control content type #{signature.header.content_type}"
410 decryptedm, sig, notice = CryptoManager.decrypt payload
411 children = message_to_chunks(decryptedm, true) if decryptedm
412 [notice, sig, children].flatten.compact
415 ## takes a RMail::Message, breaks it into Chunk:: classes.
416 def message_to_chunks m, encrypted=false, sibling_types=[]
419 case m.header.content_type
420 when "multipart/signed"
421 multipart_signed_to_chunks m
422 when "multipart/encrypted"
423 multipart_encrypted_to_chunks m
427 sibling_types = m.body.map { |p| p.header.content_type }
428 chunks = m.body.map { |p| message_to_chunks p, encrypted, sibling_types }.flatten.compact
432 elsif m.header.content_type == "message/rfc822"
433 payload = RMail::Parser.read(m.body)
434 from = payload.header.from.first ? payload.header.from.first.format : ""
435 to = payload.header.to.map { |p| p.format }.join(", ")
436 cc = payload.header.cc.map { |p| p.format }.join(", ")
437 subj = payload.header.subject
438 subj = subj ? Message.normalize_subj(payload.header.subject.gsub(/\s+/, " ").gsub(/\s+$/, "")) : subj
439 if Rfc2047.is_encoded? subj
440 subj = Rfc2047.decode_to $encoding, subj
442 msgdate = payload.header.date
443 from_person = from ? Person.from_address(from) : nil
444 to_people = to ? Person.from_address_list(to) : nil
445 cc_people = cc ? Person.from_address_list(cc) : nil
446 [Chunk::EnclosedMessage.new(from_person, to_people, cc_people, msgdate, subj)] + message_to_chunks(payload, encrypted)
449 ## first, paw through the headers looking for a filename
450 if m.header["Content-Disposition"] && m.header["Content-Disposition"] =~ /filename="?(.*?[^\\])("|;|$)/
452 elsif m.header["Content-Type"] && m.header["Content-Type"] =~ /name="?(.*?[^\\])("|;|$)/
455 ## haven't found one, but it's a non-text message. fake
458 ## TODO: make this less lame.
459 elsif m.header["Content-Type"] && m.header["Content-Type"] !~ /^text\/plain/
461 case m.header["Content-Type"]
462 when /text\/html/ then "html"
463 when /image\/(.*)/ then $1
466 ["sup-attachment-#{Time.now.to_i}-#{rand 10000}", extension].join(".")
469 ## if there's a filename, we'll treat it as an attachment.
471 # add this to the attachments list if its not a generated html
472 # attachment (should we allow images with generated names?).
473 # Lowercase the filename because searches are easier that way
474 @attachments.push filename.downcase unless filename =~ /^sup-attachment-/
475 add_label :attachment unless filename =~ /^sup-attachment-/
476 content_type = m.header.content_type || "application/unknown" # sometimes RubyMail gives us nil
477 [Chunk::Attachment.new(content_type, filename, m, sibling_types)]
479 ## otherwise, it's body text
481 ## if there's no charset, use the current encoding as the charset.
482 ## this ensures that the body is normalized to avoid non-displayable
484 body = Iconv.easy_decode($encoding, m.charset || $encoding, m.decode) if m.body
485 text_to_chunks((body || "").normalize_whitespace.split("\n"), encrypted)
490 ## parse the lines of text into chunk objects. the heuristics here
491 ## need tweaking in some nice manner. TODO: move these heuristics
492 ## into the classes themselves.
493 def text_to_chunks lines, encrypted
494 state = :text # one of :text, :quote, or :sig
498 lines.each_with_index do |line, i|
499 nextline = lines[(i + 1) ... lines.length].find { |l| l !~ /^\s*$/ } # skip blank lines
505 ## the following /:$/ followed by /\w/ is an attempt to detect the
506 ## start of a quote. this is split into two regexen because the
507 ## original regex /\w.*:$/ had very poor behavior on long lines
508 ## like ":a:a:a:a:a" that occurred in certain emails.
509 if line =~ QUOTE_PATTERN || (line =~ /:$/ && line =~ /\w/ && nextline =~ QUOTE_PATTERN)
511 elsif line =~ SIG_PATTERN && (lines.length - i) < MAX_SIG_DISTANCE
513 elsif line =~ BLOCK_QUOTE_PATTERN
514 newstate = :block_quote
518 chunks << Chunk::Text.new(chunk_lines) unless chunk_lines.empty?
528 if line =~ QUOTE_PATTERN || (line =~ /^\s*$/ && nextline =~ QUOTE_PATTERN)
530 elsif line =~ SIG_PATTERN && (lines.length - i) < MAX_SIG_DISTANCE
537 if chunk_lines.empty?
540 chunks << Chunk::Quote.new(chunk_lines)
546 when :block_quote, :sig
550 if !@have_snippet && state == :text && (@snippet.nil? || @snippet.length < SNIPPET_LEN) && line !~ /[=\*#_-]{3,}/ && line !~ /^\s*$/
552 @snippet += " " unless @snippet.empty?
553 @snippet += line.gsub(/^\s+/, "").gsub(/[\r\n]/, "").gsub(/\s+/, " ")
554 @snippet = @snippet[0 ... SNIPPET_LEN].chomp
555 @dirty = true unless encrypted && $config[:discard_snippets_from_encrypted_messages]
556 @snippet_contains_encrypted_content = true if encrypted
562 when :quote, :block_quote
563 chunks << Chunk::Quote.new(chunk_lines) unless chunk_lines.empty?
565 chunks << Chunk::Text.new(chunk_lines) unless chunk_lines.empty?
567 chunks << Chunk::Signature.new(chunk_lines) unless chunk_lines.empty?