diff options
Diffstat (limited to 'vendor/ruby-msg/lib')
| -rw-r--r-- | vendor/ruby-msg/lib/mapi.rb | 109 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mapi/convert.rb | 61 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mapi/convert/contact.rb | 142 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mapi/convert/note-mime.rb | 274 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mapi/convert/note-tmail.rb | 287 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mapi/msg.rb | 440 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mapi/property_set.rb | 269 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mapi/pst.rb | 1806 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mapi/rtf.rb | 169 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mapi/types.rb | 51 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/mime.rb | 165 | ||||
| -rw-r--r-- | vendor/ruby-msg/lib/orderedhash.rb | 218 | ||||
| -rwxr-xr-x | vendor/ruby-msg/lib/rtf.rb | 109 |
13 files changed, 4100 insertions, 0 deletions
diff --git a/vendor/ruby-msg/lib/mapi.rb b/vendor/ruby-msg/lib/mapi.rb new file mode 100644 index 000000000..b9d3413f7 --- /dev/null +++ b/vendor/ruby-msg/lib/mapi.rb @@ -0,0 +1,109 @@ +require 'mapi/types' +require 'mapi/property_set' + +module Mapi + VERSION = '1.4.0' + + # + # Mapi::Item is the base class used for all mapi objects, and is purely a + # property set container + # + class Item + attr_reader :properties + alias props properties + + # +properties+ should be a PropertySet instance. + def initialize properties + @properties = properties + end + end + + # a general attachment class. is subclassed by Msg and Pst attachment classes + class Attachment < Item + def filename + props.attach_long_filename || props.attach_filename + end + + def data + @embedded_msg || @embedded_ole || props.attach_data + end + + # with new stream work, its possible to not have the whole thing in memory at one time, + # just to save an attachment + # + # a = msg.attachments.first + # a.save open(File.basename(a.filename || 'attachment'), 'wb') + def save io + raise "can only save binary data blobs, not ole dirs" if @embedded_ole + data.each_read { |chunk| io << chunk } + end + + def inspect + "#<#{self.class.to_s[/\w+$/]}" + + (filename ? " filename=#{filename.inspect}" : '') + + (@embedded_ole ? " embedded_type=#{@embedded_ole.embedded_type.inspect}" : '') + ">" + end + end + + class Recipient < Item + # some kind of best effort guess for converting to standard mime style format. + # there are some rules for encoding non 7bit stuff in mail headers. should obey + # that here, as these strings could be unicode + # email_address will be an EX:/ address (X.400?), unless external recipient. the + # other two we try first. + # consider using entry id for this too. + def name + name = props.transmittable_display_name || props.display_name + # dequote + name[/^'(.*)'/, 1] or name rescue nil + end + + def email + props.smtp_address || props.org_email_addr || props.email_address + end + + RECIPIENT_TYPES = { 0 => :orig, 1 => :to, 2 => :cc, 3 => :bcc } + def type + RECIPIENT_TYPES[props.recipient_type] + end + + def to_s + if name = self.name and !name.empty? and email && name != email + %{"#{name}" <#{email}>} + else + email || name + end + end + + def inspect + "#<#{self.class.to_s[/\w+$/]}:#{self.to_s.inspect}>" + end + end + + # i refer to it as a message (as does mapi), although perhaps Item is better, as its a more general + # concept than a message, as used in Pst files. though maybe i'll switch to using + # Mapi::Object as the base class there. + # + # IMessage essentially, but there's also stuff like IMAPIFolder etc. so, for this to form + # basis for PST Item, it'd need to be more general. + class Message < Item + # these 2 collections should be provided by our subclasses + def attachments + raise NotImplementedError + end + + def recipients + raise NotImplementedError + end + + def inspect + str = %w[message_class from to subject].map do |key| + " #{key}=#{props.send(key).inspect}" + end.compact.join + str << " recipients=#{recipients.inspect}" + str << " attachments=#{attachments.inspect}" + "#<#{self.class.to_s[/\w+$/]}#{str}>" + end + end +end + diff --git a/vendor/ruby-msg/lib/mapi/convert.rb b/vendor/ruby-msg/lib/mapi/convert.rb new file mode 100644 index 000000000..4c7a0d298 --- /dev/null +++ b/vendor/ruby-msg/lib/mapi/convert.rb @@ -0,0 +1,61 @@ +# we have two different "backends" for note conversion. we're sticking with +# the current (home grown) mime one until the tmail version is suitably +# polished. +require 'mapi/convert/note-mime' +require 'mapi/convert/contact' + +module Mapi + class Message + CONVERSION_MAP = { + 'text/x-vcard' => [:to_vcard, 'vcf'], + 'message/rfc822' => [:to_mime, 'eml'], + 'text/plain' => [:to_post, 'txt'] + # ... + } + + # get the mime type of the message. + def mime_type + case props.message_class #.downcase <- have a feeling i saw other cased versions + when 'IPM.Contact' + # apparently "text/directory; profile=vcard" is what you're supposed to use + 'text/x-vcard' + when 'IPM.Note' + 'message/rfc822' + when 'IPM.Post' + 'text/plain' + when 'IPM.StickyNote' + 'text/plain' # hmmm.... + else + Mapi::Log.warn 'unknown message_class - %p' % props.message_class + nil + end + end + + def convert + type = mime_type + unless pair = CONVERSION_MAP[type] + raise 'unable to convert message with mime type - %p' % type + end + send pair.first + end + + # should probably be moved to mapi/convert/post + class Post + # not really sure what the pertinent properties are. we just do nothing for now... + def initialize message + @message = message + end + + def to_s + # should maybe handle other types, like html body. need a better format for post + # probably anyway, cause a lot of meta data is getting chucked. + @message.props.body + end + end + + def to_post + Post.new self + end + end +end + diff --git a/vendor/ruby-msg/lib/mapi/convert/contact.rb b/vendor/ruby-msg/lib/mapi/convert/contact.rb new file mode 100644 index 000000000..838ae6498 --- /dev/null +++ b/vendor/ruby-msg/lib/mapi/convert/contact.rb @@ -0,0 +1,142 @@ +require 'rubygems' +require 'vpim/vcard' + +# patch Vpim. TODO - fix upstream, or verify old behaviour was ok +def Vpim.encode_text v + # think the regexp was wrong + v.to_str.gsub(/(.)/m) do + case $1 + when "\n" + "\\n" + when "\\", ",", ";" + "\\#{$1}" + else + $1 + end + end +end + +module Mapi + class Message + class VcardConverter + include Vpim + + # a very incomplete mapping, but its a start... + # can't find where to set a lot of stuff, like zipcode, jobtitle etc + VCARD_MAP = { + # these are all standard mapi properties + :name => [ + { + :given => :given_name, + :family => :surname, + :fullname => :subject + } + ], + # outlook seems to eschew the mapi properties this time, + # like postal_address, street_address, home_address_city + # so we use the named properties + :addr => [ + { + :location => 'work', + :street => :business_address_street, + :locality => proc do |props| + [props.business_address_city, props.business_address_state].compact * ', ' + end + } + ], + + # right type? maybe date + :birthday => :birthday, + :nickname => :nickname + + # photo available? + # FIXME finish, emails, telephones etc + } + + attr_reader :msg + def initialize msg + @msg = msg + end + + def field name, *args + DirectoryInfo::Field.create name, Vpim.encode_text_list(args) + end + + def get_property key + if String === key + return key + elsif key.respond_to? :call + value = key.call msg.props + else + value = msg.props[key] + end + if String === value and value.empty? + nil + else + value + end + end + + def get_properties hash + constants = {} + others = {} + hash.each do |to, from| + if String === from + constants[to] = from + else + value = get_property from + others[to] = value if value + end + end + return nil if others.empty? + others.merge constants + end + + def convert + Vpim::Vcard::Maker.make2 do |m| + # handle name + [:name, :addr].each do |type| + VCARD_MAP[type].each do |hash| + next unless props = get_properties(hash) + m.send "add_#{type}" do |n| + props.each { |key, value| n.send "#{key}=", value } + end + end + end + + (VCARD_MAP.keys - [:name, :addr]).each do |key| + value = get_property VCARD_MAP[key] + m.send "#{key}=", value if value + end + + # the rest of the stuff is custom + + url = get_property(:webpage) || get_property(:business_home_page) + m.add_field field('URL', url) if url + m.add_field field('X-EVOLUTION-FILE-AS', get_property(:file_under)) if get_property(:file_under) + + addr = get_property(:email_email_address) || get_property(:email_original_display_name) + if addr + m.add_email addr do |e| + e.format ='x400' unless msg.props.email_addr_type == 'SMTP' + end + end + + if org = get_property(:company_name) + m.add_field field('ORG', get_property(:company_name)) + end + + # TODO: imaddress + end + end + end + + def to_vcard + #p props.raw.reject { |key, value| key.guid.inspect !~ /00062004-0000-0000-c000-000000000046/ }. + # map { |key, value| [key.to_sym, value] }.reject { |a, b| b.respond_to? :read } + #y props.to_h.reject { |a, b| b.respond_to? :read } + VcardConverter.new(self).convert + end + end +end + diff --git a/vendor/ruby-msg/lib/mapi/convert/note-mime.rb b/vendor/ruby-msg/lib/mapi/convert/note-mime.rb new file mode 100644 index 000000000..deb035f2c --- /dev/null +++ b/vendor/ruby-msg/lib/mapi/convert/note-mime.rb @@ -0,0 +1,274 @@ +require 'base64' +require 'mime' +require 'time' + +# there is still some Msg specific stuff in here. + +module Mapi + class Message + def mime + return @mime if @mime + # if these headers exist at all, they can be helpful. we may however get a + # application/ms-tnef mime root, which means there will be little other than + # headers. we may get nothing. + # and other times, when received from external, we get the full cigar, boundaries + # etc and all. + # sometimes its multipart, with no boundaries. that throws an error. so we'll be more + # forgiving here + @mime = SimpleMime.new props.transport_message_headers.to_s, true + populate_headers + @mime + end + + def headers + mime.headers + end + + # copy data from msg properties storage to standard mime. headers + # i've now seen it where the existing headers had heaps on stuff, and the msg#props had + # practically nothing. think it was because it was a tnef - msg conversion done by exchange. + def populate_headers + # construct a From value + # should this kind of thing only be done when headers don't exist already? maybe not. if its + # sent, then modified and saved, the headers could be wrong? + # hmmm. i just had an example where a mail is sent, from an internal user, but it has transport + # headers, i think because one recipient was external. the only place the senders email address + # exists is in the transport headers. so its maybe not good to overwrite from. + # recipients however usually have smtp address available. + # maybe we'll do it for all addresses that are smtp? (is that equivalent to + # sender_email_address !~ /^\// + name, email = props.sender_name, props.sender_email_address + if props.sender_addrtype == 'SMTP' + headers['From'] = if name and email and name != email + [%{"#{name}" <#{email}>}] + else + [email || name] + end + elsif !headers.has_key?('From') + # some messages were never sent, so that sender stuff isn't filled out. need to find another + # way to get something + # what about marking whether we thing the email was sent or not? or draft? + # for partition into an eventual Inbox, Sent, Draft mbox set? + # i've now seen cases where this stuff is missing, but exists in transport message headers, + # so maybe i should inhibit this in that case. + if email + # disabling this warning for now + #Log.warn "* no smtp sender email address available (only X.400). creating fake one" + # this is crap. though i've specially picked the logic so that it generates the correct + # email addresses in my case (for my organisation). + # this user stuff will give valid email i think, based on alias. + user = name ? name.sub(/(.*), (.*)/, "\\2.\\1") : email[/\w+$/].downcase + domain = (email[%r{^/O=([^/]+)}i, 1].downcase + '.com' rescue email) + headers['From'] = [name ? %{"#{name}" <#{user}@#{domain}>} : "<#{user}@#{domain}>" ] + elsif name + # we only have a name? thats screwed up. + # disabling this warning for now + #Log.warn "* no smtp sender email address available (only name). creating fake one" + headers['From'] = [%{"#{name}"}] + else + # disabling this warning for now + #Log.warn "* no sender email address available at all. FIXME" + end + # else we leave the transport message header version + end + + # for all of this stuff, i'm assigning in utf8 strings. + # thats ok i suppose, maybe i can say its the job of the mime class to handle that. + # but a lot of the headers are overloaded in different ways. plain string, many strings + # other stuff. what happens to a person who has a " in their name etc etc. encoded words + # i suppose. but that then happens before assignment. and can't be automatically undone + # until the header is decomposed into recipients. + recips_by_type = recipients.group_by { |r| r.type } + # i want to the the types in a specific order. + [:to, :cc, :bcc].each do |type| + # don't know why i bother, but if we can, we try to sort recipients by the numerical part + # of the ole name, or just leave it if we can't + recips = recips_by_type[type] + recips = (recips.sort_by { |r| r.obj.name[/\d{8}$/].hex } rescue recips) + # switched to using , for separation, not ;. see issue #4 + # recips.empty? is strange. i wouldn't have thought it possible, but it was right? + headers[type.to_s.sub(/^(.)/) { $1.upcase }] = [recips.join(', ')] if recips and !recips.empty? + end + headers['Subject'] = [props.subject] if props.subject + + # fill in a date value. by default, we won't mess with existing value hear + if !headers.has_key?('Date') + # we want to get a received date, as i understand it. + # use this preference order, or pull the most recent? + keys = %w[message_delivery_time client_submit_time last_modification_time creation_time] + time = keys.each { |key| break time if time = props.send(key) } + time = nil unless Date === time + + # now convert and store + # this is a little funky. not sure about time zone stuff either? + # actually seems ok. maybe its always UTC and interpreted anyway. or can be timezoneless. + # i have no timezone info anyway. + # in gmail, i see stuff like 15 Jan 2007 00:48:19 -0000, and it displays as 11:48. + # can also add .localtime here if desired. but that feels wrong. + headers['Date'] = [Time.iso8601(time.to_s).rfc2822] if time + end + + # some very simplistic mapping between internet message headers and the + # mapi properties + # any of these could be causing duplicates due to case issues. the hack in #to_mime + # just stops re-duplication at that point. need to move some smarts into the mime + # code to handle it. + mapi_header_map = [ + [:internet_message_id, 'Message-ID'], + [:in_reply_to_id, 'In-Reply-To'], + # don't set these values if they're equal to the defaults anyway + [:importance, 'Importance', proc { |val| val.to_s == '1' ? nil : val }], + [:priority, 'Priority', proc { |val| val.to_s == '1' ? nil : val }], + [:sensitivity, 'Sensitivity', proc { |val| val.to_s == '0' ? nil : val }], + # yeah? + [:conversation_topic, 'Thread-Topic'], + # not sure of the distinction here + # :originator_delivery_report_requested ?? + [:read_receipt_requested, 'Disposition-Notification-To', proc { |val| from }] + ] + mapi_header_map.each do |mapi, mime, *f| + next unless q = val = props.send(mapi) or headers.has_key?(mime) + next if f[0] and !(val = f[0].call(val)) + headers[mime] = [val.to_s] + end + end + + # redundant? + def type + props.message_class[/IPM\.(.*)/, 1].downcase rescue nil + end + + # shortcuts to some things from the headers + %w[From To Cc Bcc Subject].each do |key| + define_method(key.downcase) { headers[key].join(' ') if headers.has_key?(key) } + end + + def body_to_mime + # to create the body + # should have some options about serializing rtf. and possibly options to check the rtf + # for rtf2html conversion, stripping those html tags or other similar stuff. maybe want to + # ignore it in the cases where it is generated from incoming html. but keep it if it was the + # source for html and plaintext. + if props.body_rtf or props.body_html + # should plain come first? + mime = SimpleMime.new "Content-Type: multipart/alternative\r\n\r\n" + # its actually possible for plain body to be empty, but the others not. + # if i can get an html version, then maybe a callout to lynx can be made... + mime.parts << SimpleMime.new("Content-Type: text/plain\r\n\r\n" + props.body) if props.body + # this may be automatically unwrapped from the rtf if the rtf includes the html + mime.parts << SimpleMime.new("Content-Type: text/html\r\n\r\n" + props.body_html) if props.body_html + # temporarily disabled the rtf. its just showing up as an attachment anyway. + #mime.parts << SimpleMime.new("Content-Type: text/rtf\r\n\r\n" + props.body_rtf) if props.body_rtf + # its thus currently possible to get no body at all if the only body is rtf. that is not + # really acceptable FIXME + mime + else + # check no header case. content type? etc?. not sure if my SimpleMime class will accept + Log.debug "taking that other path" + # body can be nil, hence the to_s + SimpleMime.new "Content-Type: text/plain\r\n\r\n" + props.body.to_s + end + end + + def to_mime + # intended to be used for IPM.note, which is the email type. can use it for others if desired, + # YMMV + Log.warn "to_mime used on a #{props.message_class}" unless props.message_class == 'IPM.Note' + # we always have a body + mime = body = body_to_mime + + # If we have attachments, we take the current mime root (body), and make it the first child + # of a new tree that will contain body and attachments. + unless attachments.empty? + mime = SimpleMime.new "Content-Type: multipart/mixed\r\n\r\n" + mime.parts << body + # i don't know any better way to do this. need multipart/related for inline images + # referenced by cid: urls to work, but don't want to use it otherwise... + related = false + attachments.each do |attach| + part = attach.to_mime + related = true if part.headers.has_key?('Content-ID') or part.headers.has_key?('Content-Location') + mime.parts << part + end + mime.headers['Content-Type'] = ['multipart/related'] if related + end + + # at this point, mime is either + # - a single text/plain, consisting of the body ('taking that other path' above. rare) + # - a multipart/alternative, consiting of a few bodies (plain and html body. common) + # - a multipart/mixed, consisting of 1 of the above 2 types of bodies, and attachments. + # we add this standard preamble if its multipart + # FIXME preamble.replace, and body.replace both suck. + # preamble= is doable. body= wasn't being done because body will get rewritten from parts + # if multipart, and is only there readonly. can do that, or do a reparse... + # The way i do this means that only the first preamble will say it, not preambles of nested + # multipart chunks. + mime.preamble.replace "This is a multi-part message in MIME format.\r\n" if mime.multipart? + + # now that we have a root, we can mix in all our headers + headers.each do |key, vals| + # don't overwrite the content-type, encoding style stuff + next if mime.headers.has_key? key + # some new temporary hacks + next if key =~ /content-type/i and vals[0] =~ /base64/ + next if mime.headers.keys.map(&:downcase).include? key.downcase + mime.headers[key] += vals + end + # just a stupid hack to make the content-type header last, when using OrderedHash + mime.headers['Content-Type'] = mime.headers.delete 'Content-Type' + + mime + end + end + + class Attachment + def to_mime + # TODO: smarter mime typing. + mimetype = props.attach_mime_tag || 'application/octet-stream' + mime = SimpleMime.new "Content-Type: #{mimetype}\r\n\r\n" + mime.headers['Content-Disposition'] = [%{attachment; filename="#{filename}"}] + mime.headers['Content-Transfer-Encoding'] = ['base64'] + mime.headers['Content-Location'] = [props.attach_content_location] if props.attach_content_location + mime.headers['Content-ID'] = [props.attach_content_id] if props.attach_content_id + # data.to_s for now. data was nil for some reason. + # perhaps it was a data object not correctly handled? + # hmmm, have to use read here. that assumes that the data isa stream. + # but if the attachment data is a string, then it won't work. possible? + data_str = if @embedded_msg + mime.headers['Content-Type'] = 'message/rfc822' + # lets try making it not base64 for now + mime.headers.delete 'Content-Transfer-Encoding' + # not filename. rather name, or something else right? + # maybe it should be inline?? i forget attach_method / access meaning + mime.headers['Content-Disposition'] = [%{attachment; filename="#{@embedded_msg.subject}"}] + @embedded_msg.to_mime.to_s + elsif @embedded_ole + # kind of hacky + io = StringIO.new + Ole::Storage.new io do |ole| + ole.root.type = :dir + Ole::Storage::Dirent.copy @embedded_ole, ole.root + end + io.string + else + # FIXME: shouldn't be required + data.read.to_s rescue '' + end + mime.body.replace @embedded_msg ? data_str : Base64.encode64(data_str).gsub(/\n/, "\r\n") + mime + end + end + + class Msg < Message + def populate_headers + super + if !headers.has_key?('Date') + # can employ other methods for getting a time. heres one in a similar vein to msgconvert.pl, + # ie taking the time from an ole object + time = @root.ole.dirents.map { |dirent| dirent.modify_time || dirent.create_time }.compact.sort.last + headers['Date'] = [Time.iso8601(time.to_s).rfc2822] if time + end + end + end +end + diff --git a/vendor/ruby-msg/lib/mapi/convert/note-tmail.rb b/vendor/ruby-msg/lib/mapi/convert/note-tmail.rb new file mode 100644 index 000000000..9ccc9e0b3 --- /dev/null +++ b/vendor/ruby-msg/lib/mapi/convert/note-tmail.rb @@ -0,0 +1,287 @@ +require 'rubygems' +require 'tmail' + +# these will be removed later +require 'time' +require 'mime' + +# there is some Msg specific stuff in here. + +class TMail::Mail + def quoted_body= str + body_port.wopen { |f| f.write str } + str + end +end + +module Mapi + class Message + def mime + return @mime if @mime + # if these headers exist at all, they can be helpful. we may however get a + # application/ms-tnef mime root, which means there will be little other than + # headers. we may get nothing. + # and other times, when received from external, we get the full cigar, boundaries + # etc and all. + # sometimes its multipart, with no boundaries. that throws an error. so we'll be more + # forgiving here + @mime = Mime.new props.transport_message_headers.to_s, true + populate_headers + @mime + end + + def headers + mime.headers + end + + # copy data from msg properties storage to standard mime. headers + # i've now seen it where the existing headers had heaps on stuff, and the msg#props had + # practically nothing. think it was because it was a tnef - msg conversion done by exchange. + def populate_headers + # construct a From value + # should this kind of thing only be done when headers don't exist already? maybe not. if its + # sent, then modified and saved, the headers could be wrong? + # hmmm. i just had an example where a mail is sent, from an internal user, but it has transport + # headers, i think because one recipient was external. the only place the senders email address + # exists is in the transport headers. so its maybe not good to overwrite from. + # recipients however usually have smtp address available. + # maybe we'll do it for all addresses that are smtp? (is that equivalent to + # sender_email_address !~ /^\// + name, email = props.sender_name, props.sender_email_address + if props.sender_addrtype == 'SMTP' + headers['From'] = if name and email and name != email + [%{"#{name}" <#{email}>}] + else + [email || name] + end + elsif !headers.has_key?('From') + # some messages were never sent, so that sender stuff isn't filled out. need to find another + # way to get something + # what about marking whether we thing the email was sent or not? or draft? + # for partition into an eventual Inbox, Sent, Draft mbox set? + # i've now seen cases where this stuff is missing, but exists in transport message headers, + # so maybe i should inhibit this in that case. + if email + # disabling this warning for now + #Log.warn "* no smtp sender email address available (only X.400). creating fake one" + # this is crap. though i've specially picked the logic so that it generates the correct + # email addresses in my case (for my organisation). + # this user stuff will give valid email i think, based on alias. + user = name ? name.sub(/(.*), (.*)/, "\\2.\\1") : email[/\w+$/].downcase + domain = (email[%r{^/O=([^/]+)}i, 1].downcase + '.com' rescue email) + headers['From'] = [name ? %{"#{name}" <#{user}@#{domain}>} : "<#{user}@#{domain}>" ] + elsif name + # we only have a name? thats screwed up. + # disabling this warning for now + #Log.warn "* no smtp sender email address available (only name). creating fake one" + headers['From'] = [%{"#{name}"}] + else + # disabling this warning for now + #Log.warn "* no sender email address available at all. FIXME" + end + # else we leave the transport message header version + end + + # for all of this stuff, i'm assigning in utf8 strings. + # thats ok i suppose, maybe i can say its the job of the mime class to handle that. + # but a lot of the headers are overloaded in different ways. plain string, many strings + # other stuff. what happens to a person who has a " in their name etc etc. encoded words + # i suppose. but that then happens before assignment. and can't be automatically undone + # until the header is decomposed into recipients. + recips_by_type = recipients.group_by { |r| r.type } + # i want to the the types in a specific order. + [:to, :cc, :bcc].each do |type| + # don't know why i bother, but if we can, we try to sort recipients by the numerical part + # of the ole name, or just leave it if we can't + recips = recips_by_type[type] + recips = (recips.sort_by { |r| r.obj.name[/\d{8}$/].hex } rescue recips) + # switched to using , for separation, not ;. see issue #4 + # recips.empty? is strange. i wouldn't have thought it possible, but it was right? + headers[type.to_s.sub(/^(.)/) { $1.upcase }] = [recips.join(', ')] unless recips.empty? + end + headers['Subject'] = [props.subject] if props.subject + + # fill in a date value. by default, we won't mess with existing value hear + if !headers.has_key?('Date') + # we want to get a received date, as i understand it. + # use this preference order, or pull the most recent? + keys = %w[message_delivery_time client_submit_time last_modification_time creation_time] + time = keys.each { |key| break time if time = props.send(key) } + time = nil unless Date === time + + # now convert and store + # this is a little funky. not sure about time zone stuff either? + # actually seems ok. maybe its always UTC and interpreted anyway. or can be timezoneless. + # i have no timezone info anyway. + # in gmail, i see stuff like 15 Jan 2007 00:48:19 -0000, and it displays as 11:48. + # can also add .localtime here if desired. but that feels wrong. + headers['Date'] = [Time.iso8601(time.to_s).rfc2822] if time + end + + # some very simplistic mapping between internet message headers and the + # mapi properties + # any of these could be causing duplicates due to case issues. the hack in #to_mime + # just stops re-duplication at that point. need to move some smarts into the mime + # code to handle it. + mapi_header_map = [ + [:internet_message_id, 'Message-ID'], + [:in_reply_to_id, 'In-Reply-To'], + # don't set these values if they're equal to the defaults anyway + [:importance, 'Importance', proc { |val| val.to_s == '1' ? nil : val }], + [:priority, 'Priority', proc { |val| val.to_s == '1' ? nil : val }], + [:sensitivity, 'Sensitivity', proc { |val| val.to_s == '0' ? nil : val }], + # yeah? + [:conversation_topic, 'Thread-Topic'], + # not sure of the distinction here + # :originator_delivery_report_requested ?? + [:read_receipt_requested, 'Disposition-Notification-To', proc { |val| from }] + ] + mapi_header_map.each do |mapi, mime, *f| + next unless q = val = props.send(mapi) or headers.has_key?(mime) + next if f[0] and !(val = f[0].call(val)) + headers[mime] = [val.to_s] + end + end + + # redundant? + def type + props.message_class[/IPM\.(.*)/, 1].downcase rescue nil + end + + # shortcuts to some things from the headers + %w[From To Cc Bcc Subject].each do |key| + define_method(key.downcase) { headers[key].join(' ') if headers.has_key?(key) } + end + + def body_to_tmail + # to create the body + # should have some options about serializing rtf. and possibly options to check the rtf + # for rtf2html conversion, stripping those html tags or other similar stuff. maybe want to + # ignore it in the cases where it is generated from incoming html. but keep it if it was the + # source for html and plaintext. + if props.body_rtf or props.body_html + # should plain come first? + part = TMail::Mail.new + # its actually possible for plain body to be empty, but the others not. + # if i can get an html version, then maybe a callout to lynx can be made... + part.parts << TMail::Mail.parse("Content-Type: text/plain\r\n\r\n" + props.body) if props.body + # this may be automatically unwrapped from the rtf if the rtf includes the html + part.parts << TMail::Mail.parse("Content-Type: text/html\r\n\r\n" + props.body_html) if props.body_html + # temporarily disabled the rtf. its just showing up as an attachment anyway. + #mime.parts << Mime.new("Content-Type: text/rtf\r\n\r\n" + props.body_rtf) if props.body_rtf + # its thus currently possible to get no body at all if the only body is rtf. that is not + # really acceptable FIXME + part['Content-Type'] = 'multipart/alternative' + part + else + # check no header case. content type? etc?. not sure if my Mime class will accept + Log.debug "taking that other path" + # body can be nil, hence the to_s + TMail::Mail.parse "Content-Type: text/plain\r\n\r\n" + props.body.to_s + end + end + + def to_tmail + # intended to be used for IPM.note, which is the email type. can use it for others if desired, + # YMMV + Log.warn "to_mime used on a #{props.message_class}" unless props.message_class == 'IPM.Note' + # we always have a body + mail = body = body_to_tmail + + # If we have attachments, we take the current mime root (body), and make it the first child + # of a new tree that will contain body and attachments. + unless attachments.empty? + raise NotImplementedError + mime = Mime.new "Content-Type: multipart/mixed\r\n\r\n" + mime.parts << body + # i don't know any better way to do this. need multipart/related for inline images + # referenced by cid: urls to work, but don't want to use it otherwise... + related = false + attachments.each do |attach| + part = attach.to_mime + related = true if part.headers.has_key?('Content-ID') or part.headers.has_key?('Content-Location') + mime.parts << part + end + mime.headers['Content-Type'] = ['multipart/related'] if related + end + + # at this point, mime is either + # - a single text/plain, consisting of the body ('taking that other path' above. rare) + # - a multipart/alternative, consiting of a few bodies (plain and html body. common) + # - a multipart/mixed, consisting of 1 of the above 2 types of bodies, and attachments. + # we add this standard preamble if its multipart + # FIXME preamble.replace, and body.replace both suck. + # preamble= is doable. body= wasn't being done because body will get rewritten from parts + # if multipart, and is only there readonly. can do that, or do a reparse... + # The way i do this means that only the first preamble will say it, not preambles of nested + # multipart chunks. + mail.quoted_body = "This is a multi-part message in MIME format.\r\n" if mail.multipart? + + # now that we have a root, we can mix in all our headers + headers.each do |key, vals| + # don't overwrite the content-type, encoding style stuff + next if mail[key] + # some new temporary hacks + next if key =~ /content-type/i and vals[0] =~ /base64/ + #next if mime.headers.keys.map(&:downcase).include? key.downcase + mail[key] = vals.first + end + # just a stupid hack to make the content-type header last, when using OrderedHash + #mime.headers['Content-Type'] = mime.headers.delete 'Content-Type' + + mail + end + end + + class Attachment + def to_tmail + # TODO: smarter mime typing. + mimetype = props.attach_mime_tag || 'application/octet-stream' + part = TMail::Mail.parse "Content-Type: #{mimetype}\r\n\r\n" + part['Content-Disposition'] = %{attachment; filename="#{filename}"} + part['Content-Transfer-Encoding'] = 'base64' + part['Content-Location'] = props.attach_content_location if props.attach_content_location + part['Content-ID'] = props.attach_content_id if props.attach_content_id + # data.to_s for now. data was nil for some reason. + # perhaps it was a data object not correctly handled? + # hmmm, have to use read here. that assumes that the data isa stream. + # but if the attachment data is a string, then it won't work. possible? + data_str = if @embedded_msg + raise NotImplementedError + mime.headers['Content-Type'] = 'message/rfc822' + # lets try making it not base64 for now + mime.headers.delete 'Content-Transfer-Encoding' + # not filename. rather name, or something else right? + # maybe it should be inline?? i forget attach_method / access meaning + mime.headers['Content-Disposition'] = [%{attachment; filename="#{@embedded_msg.subject}"}] + @embedded_msg.to_mime.to_s + elsif @embedded_ole + raise NotImplementedError + # kind of hacky + io = StringIO.new + Ole::Storage.new io do |ole| + ole.root.type = :dir + Ole::Storage::Dirent.copy @embedded_ole, ole.root + end + io.string + else + data.read.to_s + end + part.body = @embedded_msg ? data_str : Base64.encode64(data_str).gsub(/\n/, "\r\n") + part + end + end + + class Msg < Message + def populate_headers + super + if !headers.has_key?('Date') + # can employ other methods for getting a time. heres one in a similar vein to msgconvert.pl, + # ie taking the time from an ole object + time = @root.ole.dirents.map { |dirent| dirent.modify_time || dirent.create_time }.compact.sort.last + headers['Date'] = [Time.iso8601(time.to_s).rfc2822] if time + end + end + end +end + diff --git a/vendor/ruby-msg/lib/mapi/msg.rb b/vendor/ruby-msg/lib/mapi/msg.rb new file mode 100644 index 000000000..fc30a9170 --- /dev/null +++ b/vendor/ruby-msg/lib/mapi/msg.rb @@ -0,0 +1,440 @@ +require 'rubygems' +require 'ole/storage' +require 'mapi' +require 'mapi/rtf' + +module Mapi + # + # = Introduction + # + # Primary class interface to the vagaries of .msg files. + # + # The core of the work is done by the <tt>Msg::PropertyStore</tt> class. + # + class Msg < Message + # + # = Introduction + # + # A big compononent of +Msg+ files is the property store, which holds + # all the key/value pairs of properties. The message itself, and all + # its <tt>Attachment</tt>s and <tt>Recipient</tt>s have an instance of + # this class. + # + # = Storage model + # + # Property keys (tags?) can be either simple hex numbers, in the + # range 0x0000 - 0xffff, or they can be named properties. In fact, + # properties in the range 0x0000 to 0x7fff are supposed to be the non- + # named properties, and can be considered to be in the +PS_MAPI+ + # namespace. (correct?) + # + # Named properties are serialized in the 0x8000 to 0xffff range, + # and are referenced as a guid and long/string pair. + # + # There are key ranges, which can be used to imply things generally + # about keys. + # + # Further, we can give symbolic names to most keys, coming from + # constants in various places. Eg: + # + # 0x0037 => subject + # {00062002-0000-0000-C000-000000000046}/0x8218 => response_status + # # displayed as categories in outlook + # {00020329-0000-0000-C000-000000000046}/"Keywords" => categories + # + # Futher, there are completely different names, coming from other + # object models that get mapped to these things (CDO's model, + # Outlook's model etc). Eg "urn:schemas:httpmail:subject" + # I think these can be ignored though, as they aren't defined clearly + # in terms of mapi properties, and i'm really just trying to make + # a mapi property store. (It should also be relatively easy to + # support them later.) + # + # = Usage + # + # The api is driven by a desire to have the simple stuff "just work", ie + # + # properties.subject + # properties.display_name + # + # There also needs to be a way to look up properties more specifically: + # + # properties[0x0037] # => gets the subject + # properties[0x0037, PS_MAPI] # => still gets the subject + # properties['Keywords', PS_PUBLIC_STRINGS] # => gets outlook's categories array + # + # The abbreviated versions work by "resolving" the symbols to full keys: + # + # # the guid here is just PS_PUBLIC_STRINGS + # properties.resolve :keywords # => #<Key {00020329-0000-0000-c000-000000000046}/"Keywords"> + # # the result here is actually also a key + # k = properties.resolve :subject # => 0x0037 + # # it has a guid + # k.guid == Msg::Properties::PS_MAPI # => true + # + # = Parsing + # + # There are three objects that need to be parsed to load a +Msg+ property store: + # + # 1. The +nameid+ directory (<tt>Properties.parse_nameid</tt>) + # 2. The many +substg+ objects, whose names should match <tt>Properties::SUBSTG_RX</tt> + # (<tt>Properties#parse_substg</tt>) + # 3. The +properties+ file (<tt>Properties#parse_properties</tt>) + # + # Understanding of the formats is by no means perfect. + # + # = TODO + # + # * While the key objects are sufficient, the value objects are just plain + # ruby types. It currently isn't possible to write to the values, or to know + # which encoding the value had. + # * Update this doc. + # * Perhaps change from eager loading, to be load-on-demand. + # + class PropertyStore + include PropertySet::Constants + Key = PropertySet::Key + + # note that binary and default both use obj.open. not the block form. this means we should + # #close it later, which we don't. as we're only reading though, it shouldn't matter right? + # not really good though FIXME + # change these to use mapi symbolic const names + ENCODINGS = { + 0x000d => proc { |obj| obj }, # seems to be used when its going to be a directory instead of a file. eg nested ole. 3701 usually. in which case we shouldn't get here right? + 0x001f => proc { |obj| Ole::Types::FROM_UTF16.iconv obj.read }, # unicode + # ascii + # FIXME hack did a[0..-2] before, seems right sometimes, but for some others it chopped the text. chomp + 0x001e => proc { |obj| obj.read.chomp 0.chr }, + 0x0102 => proc { |obj| obj.open }, # binary? + :default => proc { |obj| obj.open } + } + + SUBSTG_RX = /^__substg1\.0_([0-9A-F]{4})([0-9A-F]{4})(?:-([0-9A-F]{8}))?$/ + PROPERTIES_RX = /^__properties_version1\.0$/ + NAMEID_RX = /^__nameid_version1\.0$/ + VALID_RX = /#{SUBSTG_RX}|#{PROPERTIES_RX}|#{NAMEID_RX}/ + + attr_reader :nameid + + def initialize + @nameid = nil + # not exactly a cache currently + @cache = {} + end + + #-- + # The parsing methods + #++ + + def self.load obj + prop = new + prop.load obj + prop + end + + # Parse properties from the +Dirent+ obj + def load obj + # we need to do the nameid first, as it provides the map for later user defined properties + if nameid_obj = obj.children.find { |child| child.name =~ NAMEID_RX } + @nameid = PropertyStore.parse_nameid nameid_obj + # hack to make it available to all msg files from the same ole storage object + # FIXME - come up with a neater way + class << obj.ole + attr_accessor :msg_nameid + end + obj.ole.msg_nameid = @nameid + elsif obj.ole + @nameid = obj.ole.msg_nameid rescue nil + end + # now parse the actual properties. i think dirs that match the substg should be decoded + # as properties to. 0x000d is just another encoding, the dir encoding. it should match + # whether the object is file / dir. currently only example is embedded msgs anyway + obj.children.each do |child| + next unless child.file? + case child.name + when PROPERTIES_RX + parse_properties child + when SUBSTG_RX + parse_substg(*($~[1..-1].map { |num| num.hex rescue nil } + [child])) + end + end + end + + # Read nameid from the +Dirent+ obj, which is used for mapping of named properties keys to + # proxy keys in the 0x8000 - 0xffff range. + # Returns a hash of integer -> Key. + def self.parse_nameid obj + remaining = obj.children.dup + guids_obj, props_obj, names_obj = + %w[__substg1.0_00020102 __substg1.0_00030102 __substg1.0_00040102].map do |name| + remaining.delete obj/name + end + + # parse guids + # this is the guids for named properities (other than builtin ones) + # i think PS_PUBLIC_STRINGS, and PS_MAPI are builtin. + guids = [PS_PUBLIC_STRINGS] + guids_obj.read.scan(/.{16}/mn).map do |str| + Ole::Types.load_guid str + end + + # parse names. + # the string ids for named properties + # they are no longer parsed, as they're referred to by offset not + # index. they are simply sequentially packed, as a long, giving + # the string length, then padding to 4 byte multiple, and repeat. + names_data = names_obj.read + + # parse actual props. + # not sure about any of this stuff really. + # should flip a few bits in the real msg, to get a better understanding of how this works. + props = props_obj.read.scan(/.{8}/mn).map do |str| + flags, offset = str[4..-1].unpack 'v2' + # the property will be serialised as this pseudo property, mapping it to this named property + pseudo_prop = 0x8000 + offset + named = flags & 1 == 1 + prop = if named + str_off = *str.unpack('V') + len = *names_data[str_off, 4].unpack('V') + Ole::Types::FROM_UTF16.iconv names_data[str_off + 4, len] + else + a, b = str.unpack('v2') + Log.debug "b not 0" if b != 0 + a + end + # a bit sus + guid_off = flags >> 1 + # missing a few builtin PS_* + Log.debug "guid off < 2 (#{guid_off})" if guid_off < 2 + guid = guids[guid_off - 2] + [pseudo_prop, Key.new(prop, guid)] + end + + #Log.warn "* ignoring #{remaining.length} objects in nameid" unless remaining.empty? + # this leaves a bunch of other unknown chunks of data with completely unknown meaning. + # pp [:unknown, child.name, child.data.unpack('H*')[0].scan(/.{16}/m)] + Hash[*props.flatten] + end + + # Parse an +Dirent+, as per <tt>msgconvert.pl</tt>. This is how larger properties, such + # as strings, binary blobs, and other ole sub-directories (eg nested Msg) are stored. + def parse_substg key, encoding, offset, obj + if (encoding & 0x1000) != 0 + if !offset + # there is typically one with no offset first, whose data is a series of numbers + # equal to the lengths of all the sub parts. gives an implied array size i suppose. + # maybe you can initialize the array at this time. the sizes are the same as all the + # ole object sizes anyway, its to pre-allocate i suppose. + #p obj.data.unpack('V*') + # ignore this one + return + else + # remove multivalue flag for individual pieces + encoding &= ~0x1000 + end + else + Log.warn "offset specified for non-multivalue encoding #{obj.name}" if offset + offset = nil + end + # offset is for multivalue encodings. + unless encoder = ENCODINGS[encoding] + Log.warn "unknown encoding #{encoding}" + #encoder = proc { |obj| obj.io } #.read }. maybe not a good idea + encoder = ENCODINGS[:default] + end + add_property key, encoder[obj], offset + end + + # For parsing the +properties+ file. Smaller properties are serialized in one chunk, + # such as longs, bools, times etc. The parsing has problems. + def parse_properties obj + data = obj.read + # don't really understand this that well... + pad = data.length % 16 + unless (pad == 0 || pad == 8) and data[0...pad] == "\000" * pad + Log.warn "padding was not as expected #{pad} (#{data.length}) -> #{data[0...pad].inspect}" + end + data[pad..-1].scan(/.{16}/mn).each do |data| + property, encoding = ('%08x' % data.unpack('V')).scan /.{4}/ + key = property.hex + # doesn't make any sense to me. probably because its a serialization of some internal + # outlook structure... + next if property == '0000' + case encoding + when '0102', '001e', '001f', '101e', '101f', '000d' + # ignore on purpose. not sure what its for + # multivalue versions ignored also + when '0003' # long + # don't know what all the other data is for + add_property key, *data[8, 4].unpack('V') + when '000b' # boolean + # again, heaps more data than needed. and its not always 0 or 1. + # they are in fact quite big numbers. this is wrong. +# p [property, data[4..-1].unpack('H*')[0]] + add_property key, data[8, 4].unpack('V')[0] != 0 + when '0040' # systime + # seems to work: + add_property key, Ole::Types.load_time(data[8..-1]) + else + #Log.warn "ignoring data in __properties section, encoding: #{encoding}" + #Log << data.unpack('H*').inspect + "\n" + end + end + end + + def add_property key, value, pos=nil + # map keys in the named property range through nameid + if Integer === key and key >= 0x8000 + if !@nameid + Log.warn "no nameid section yet named properties used" + key = Key.new key + elsif real_key = @nameid[key] + key = real_key + else + # i think i hit these when i have a named property, in the PS_MAPI + # guid + Log.warn "property in named range not in nameid #{key.inspect}" + key = Key.new key + end + else + key = Key.new key + end + if pos + @cache[key] ||= [] + Log.warn "duplicate property" unless Array === @cache[key] + # ^ this is actually a trickier problem. the issue is more that they must all be of + # the same type. + @cache[key][pos] = value + else + # take the last. + Log.warn "duplicate property #{key.inspect}" if @cache[key] + @cache[key] = value + end + end + + # delegate to cache + def method_missing name, *args, &block + @cache.send name, *args, &block + end + end + + # these 2 will actually be of the form + # 1\.0_#([0-9A-Z]{8}), where $1 is the 0 based index number in hex + # should i parse that and use it as an index, or just return in + # file order? probably should use it later... + ATTACH_RX = /^__attach_version1\.0_.*/ + RECIP_RX = /^__recip_version1\.0_.*/ + VALID_RX = /#{PropertyStore::VALID_RX}|#{ATTACH_RX}|#{RECIP_RX}/ + + attr_reader :root + attr_accessor :close_parent + + # Alternate constructor, to create an +Msg+ directly from +arg+ and +mode+, passed + # directly to Ole::Storage (ie either filename or seekable IO object). + def self.open arg, mode=nil + msg = new Ole::Storage.open(arg, mode).root + # we will close the ole when we are #closed + msg.close_parent = true + if block_given? + begin yield msg + ensure; msg.close + end + else msg + end + end + + # Create an Msg from +root+, an <tt>Ole::Storage::Dirent</tt> object + def initialize root + @root = root + @close_parent = false + super PropertySet.new(PropertyStore.load(@root)) + Msg.warn_unknown @root + end + + def self.warn_unknown obj + # bit of validation. not important if there is extra stuff, though would be + # interested to know what it is. doesn't check dir/file stuff. + unknown = obj.children.reject { |child| child.name =~ VALID_RX } + Log.warn "skipped #{unknown.length} unknown msg object(s)" unless unknown.empty? + end + + def close + @root.ole.close if @close_parent + end + + def attachments + @attachments ||= @root.children. + select { |child| child.dir? and child.name =~ ATTACH_RX }. + map { |child| Attachment.new child }. + select { |attach| attach.valid? } + end + + def recipients + @recipients ||= @root.children. + select { |child| child.dir? and child.name =~ RECIP_RX }. + map { |child| Recipient.new child } + end + + class Attachment < Mapi::Attachment + attr_reader :obj, :properties + alias props :properties + + def initialize obj + @obj = obj + @embedded_ole = nil + @embedded_msg = nil + + super PropertySet.new(PropertyStore.load(@obj)) + Msg.warn_unknown @obj + + @obj.children.each do |child| + # temp hack. PropertyStore doesn't do directory properties atm - FIXME + if child.dir? and child.name =~ PropertyStore::SUBSTG_RX and + $1 == '3701' and $2.downcase == '000d' + @embedded_ole = child + class << @embedded_ole + def compobj + return nil unless compobj = self["\001CompObj"] + compobj.read[/^.{32}([^\x00]+)/m, 1] + end + + def embedded_type + temp = compobj and return temp + # try to guess more + if children.select { |child| child.name =~ /__(substg|properties|recip|attach|nameid)/ }.length > 2 + return 'Microsoft Office Outlook Message' + end + nil + end + end + if @embedded_ole.embedded_type == 'Microsoft Office Outlook Message' + @embedded_msg = Msg.new @embedded_ole + end + end + end + end + + def valid? + # something i started to notice when handling embedded ole object attachments is + # the particularly strange case where there are empty attachments + not props.raw.keys.empty? + end + end + + # + # +Recipient+ serves as a container for the +recip+ directories in the .msg. + # It has things like office_location, business_telephone_number, but I don't + # think enough to make a vCard out of? + # + class Recipient < Mapi::Recipient + attr_reader :obj, :properties + alias props :properties + + def initialize obj + @obj = obj + super PropertySet.new(PropertyStore.load(@obj)) + Msg.warn_unknown @obj + end + end + end +end + diff --git a/vendor/ruby-msg/lib/mapi/property_set.rb b/vendor/ruby-msg/lib/mapi/property_set.rb new file mode 100644 index 000000000..199bca525 --- /dev/null +++ b/vendor/ruby-msg/lib/mapi/property_set.rb @@ -0,0 +1,269 @@ +require 'yaml' +require 'mapi/types' +require 'mapi/rtf' +require 'rtf' + +module Mapi + # + # The Mapi::PropertySet class is used to wrap the lower level Msg or Pst property stores, + # and provide a consistent and more friendly interface. It allows you to just say: + # + # properties.subject + # + # instead of: + # + # properites.raw[0x0037, PS_MAPI] + # + # The underlying store can be just a hash, or lazily loading directly from the file. A good + # compromise is to cache all the available keys, and just return the values on demand, rather + # than load up many possibly unwanted values. + # + class PropertySet + # the property set guid constants + # these guids are all defined with the macro DEFINE_OLEGUID in mapiguid.h. + # see http://doc.ddart.net/msdn/header/include/mapiguid.h.html + oleguid = proc do |prefix| + Ole::Types::Clsid.parse "{#{prefix}-0000-0000-c000-000000000046}" + end + + NAMES = { + oleguid['00020328'] => 'PS_MAPI', + oleguid['00020329'] => 'PS_PUBLIC_STRINGS', + oleguid['00020380'] => 'PS_ROUTING_EMAIL_ADDRESSES', + oleguid['00020381'] => 'PS_ROUTING_ADDRTYPE', + oleguid['00020382'] => 'PS_ROUTING_DISPLAY_NAME', + oleguid['00020383'] => 'PS_ROUTING_ENTRYID', + oleguid['00020384'] => 'PS_ROUTING_SEARCH_KEY', + # string properties in this namespace automatically get added to the internet headers + oleguid['00020386'] => 'PS_INTERNET_HEADERS', + # theres are bunch of outlook ones i think + # http://blogs.msdn.com/stephen_griffin/archive/2006/05/10/outlook-2007-beta-documentation-notification-based-indexing-support.aspx + # IPM.Appointment + oleguid['00062002'] => 'PSETID_Appointment', + # IPM.Task + oleguid['00062003'] => 'PSETID_Task', + # used for IPM.Contact + oleguid['00062004'] => 'PSETID_Address', + oleguid['00062008'] => 'PSETID_Common', + # didn't find a source for this name. it is for IPM.StickyNote + oleguid['0006200e'] => 'PSETID_Note', + # for IPM.Activity. also called the journal? + oleguid['0006200a'] => 'PSETID_Log', + } + + module Constants + NAMES.each { |guid, name| const_set name, guid } + end + + include Constants + + # +Properties+ are accessed by <tt>Key</tt>s, which are coerced to this class. + # Includes a bunch of methods (hash, ==, eql?) to allow it to work as a key in + # a +Hash+. + # + # Also contains the code that maps keys to symbolic names. + class Key + include Constants + + attr_reader :code, :guid + def initialize code, guid=PS_MAPI + @code, @guid = code, guid + end + + def to_sym + # hmmm, for some stuff, like, eg, the message class specific range, sym-ification + # of the key depends on knowing our message class. i don't want to store anything else + # here though, so if that kind of thing is needed, it can be passed to this function. + # worry about that when some examples arise. + case code + when Integer + if guid == PS_MAPI # and < 0x8000 ? + # the hash should be updated now that i've changed the process + TAGS['%04x' % code].first[/_(.*)/, 1].downcase.to_sym rescue code + else + # handle other guids here, like mapping names to outlook properties, based on the + # outlook object model. + NAMED_MAP[self].to_sym rescue code + end + when String + # return something like + # note that named properties don't go through the map at the moment. so #categories + # doesn't work yet + code.downcase.to_sym + end + end + + def to_s + to_sym.to_s + end + + # FIXME implement these + def transmittable? + # etc, can go here too + end + + # this stuff is to allow it to be a useful key + def hash + [code, guid].hash + end + + def == other + hash == other.hash + end + + alias eql? :== + + def inspect + # maybe the way to do this, would be to be able to register guids + # in a global lookup, which are used by Clsid#inspect itself, to + # provide symbolic names... + guid_str = NAMES[guid] || "{#{guid.format}}" rescue "nil" + if Integer === code + hex = '0x%04x' % code + if guid == PS_MAPI + # just display as plain hex number + hex + else + "#<Key #{guid_str}/#{hex}>" + end + else + # display full guid and code + "#<Key #{guid_str}/#{code.inspect}>" + end + end + end + + # duplicated here for now + SUPPORT_DIR = File.dirname(__FILE__) + '/../..' + + # data files that provide for the code to symbolic name mapping + # guids in named_map are really constant references to the above + TAGS = YAML.load_file "#{SUPPORT_DIR}/data/mapitags.yaml" + NAMED_MAP = YAML.load_file("#{SUPPORT_DIR}/data/named_map.yaml").inject({}) do |hash, (key, value)| + hash.update Key.new(key[0], const_get(key[1])) => value + end + + attr_reader :raw + + # +raw+ should be an hash-like object that maps <tt>Key</tt>s to values. Should respond_to? + # [], keys, values, each, and optionally []=, and delete. + def initialize raw + @raw = raw + end + + # resolve +arg+ (could be key, code, string, or symbol), and possible +guid+ to a key. + # returns nil on failure + def resolve arg, guid=nil + if guid; Key.new arg, guid + else + case arg + when Key; arg + when Integer; Key.new arg + else sym_to_key[arg.to_sym] + end + end + end + + # this is the function that creates a symbol to key mapping. currently this works by making a + # pass through the raw properties, but conceivably you could map symbols to keys using the + # mapitags directly. problem with that would be that named properties wouldn't map automatically, + # but maybe thats not too important. + def sym_to_key + return @sym_to_key if @sym_to_key + @sym_to_key = {} + raw.keys.each do |key| + sym = key.to_sym + unless Symbol === sym + Log.debug "couldn't find symbolic name for key #{key.inspect}" + next + end + if @sym_to_key[sym] + Log.warn "duplicate key #{key.inspect}" + # we give preference to PS_MAPI keys + @sym_to_key[sym] = key if key.guid == PS_MAPI + else + # just assign + @sym_to_key[sym] = key + end + end + @sym_to_key + end + + def keys + sym_to_key.keys + end + + def values + sym_to_key.values.map { |key| raw[key] } + end + + def [] arg, guid=nil + raw[resolve(arg, guid)] + end + + def []= arg, *args + args.unshift nil if args.length == 1 + guid, value = args + # FIXME this won't really work properly. it would need to go + # to TAGS to resolve, as it often won't be there already... + raw[resolve(arg, guid)] = value + end + + def method_missing name, *args + if name.to_s !~ /\=$/ and args.empty? + self[name] + elsif name.to_s =~ /(.*)\=$/ and args.length == 1 + self[$1] = args[0] + else + super + end + end + + def to_h + sym_to_key.inject({}) { |hash, (sym, key)| hash.update sym => raw[key] } + end + + def inspect + "#<#{self.class} " + to_h.sort_by { |k, v| k.to_s }.map do |k, v| + v = v.inspect + "#{k}=#{v.length > 32 ? v[0..29] + '..."' : v}" + end.join(' ') + '>' + end + + # ----- + + # temporary pseudo tags + + # for providing rtf to plain text conversion. later, html to text too. + def body + return @body if defined?(@body) + @body = (self[:body] rescue nil) + # last resort + if !@body or @body.strip.empty? + Log.warn 'creating text body from rtf' + @body = (::RTF::Converter.rtf2text body_rtf rescue nil) + end + @body + end + + # for providing rtf decompression + def body_rtf + return @body_rtf if defined?(@body_rtf) + @body_rtf = (RTF.rtfdecompr rtf_compressed.read rescue nil) + end + + # for providing rtf to html conversion + def body_html + return @body_html if defined?(@body_html) + @body_html = (self[:body_html].read rescue nil) + @body_html = (RTF.rtf2html body_rtf rescue nil) if !@body_html or @body_html.strip.empty? + # last resort + if !@body_html or @body_html.strip.empty? + Log.warn 'creating html body from rtf' + @body_html = (::RTF::Converter.rtf2text body_rtf, :html rescue nil) + end + @body_html + end + end +end + diff --git a/vendor/ruby-msg/lib/mapi/pst.rb b/vendor/ruby-msg/lib/mapi/pst.rb new file mode 100644 index 000000000..9ac64b097 --- /dev/null +++ b/vendor/ruby-msg/lib/mapi/pst.rb @@ -0,0 +1,1806 @@ +# +# = Introduction +# +# This file is mostly an attempt to port libpst to ruby, and simplify it in the process. It +# will leverage much of the existing MAPI => MIME conversion developed for Msg files, and as +# such is purely concerned with the file structure details. +# +# = TODO +# +# 1. solve recipient table problem (test4). +# this is done. turns out it was due to id2 clashes. find better solution +# 2. check parse consistency. an initial conversion of a 30M file to pst, shows +# a number of messages conveting badly. compare with libpst too. +# 3. xattribs +# 4. generalise the Mapi stuff better +# 5. refactor index load +# 6. msg serialization? +# + +=begin + +quick plan for cleanup. + +have working tests for 97 and 03 file formats, so safe. + +want to fix up: + +64 bit unpacks scattered around. its ugly. not sure how best to handle it, but am slightly tempted +to override String#unpack to support a 64 bit little endian unpack (like L vs N/V, for Q). one way or +another need to fix it. Could really slow everything else down if its parsing the unpack strings twice, +once in ruby, for every single unpack i do :/ + +the index loading process, and the lack of shared code between normal vs 64 bit variants, and Index vs Desc. +should be able to reduce code by factor of 4. also think I should move load code into the class too. then +maybe have something like: + +class Header + def index_class + version_2003 ? Index64 : Index + end +end + +def load_idx + header.index_class.load_index +end + +OR + +def initialize + @header = ... + extend @header.index_class::Load + load_idx +end + +need to think about the role of the mapi code, and Pst::Item etc, but that layer can come later. + +=end + +require 'mapi' +require 'enumerator' +require 'ostruct' +require 'ole/ranges_io' + +module Mapi +class Pst + class FormatError < StandardError + end + + # unfortunately there is no Q analogue which is little endian only. + # this translates T as an unsigned quad word, little endian byte order, to + # not pollute the rest of the code. + # + # didn't want to override String#unpack, cause its too hacky, and incomplete. + def self.unpack str, unpack_spec + return str.unpack(unpack_spec) unless unpack_spec['T'] + @unpack_cache ||= {} + t_offsets, new_spec = @unpack_cache[unpack_spec] + unless t_offsets + t_offsets = [] + offset = 0 + new_spec = '' + unpack_spec.scan(/([^\d])_?(\*|\d+)?/o) do + num_elems = $1.downcase == 'a' ? 1 : ($2 || 1).to_i + if $1 == 'T' + num_elems.times { |i| t_offsets << offset + i } + new_spec << "V#{num_elems * 2}" + else + new_spec << $~[0] + end + offset += num_elems + end + @unpack_cache[unpack_spec] = [t_offsets, new_spec] + end + a = str.unpack(new_spec) + t_offsets.each do |offset| + low, high = a[offset, 2] + a[offset, 2] = low && high ? low + (high << 32) : nil + end + a + end + + # + # this is the header and encryption encapsulation code + # ---------------------------------------------------------------------------- + # + + # class which encapsulates the pst header + class Header + SIZE = 512 + MAGIC = 0x2142444e + + # these are the constants defined in libpst.c, that + # are referenced in pst_open() + INDEX_TYPE_OFFSET = 0x0A + FILE_SIZE_POINTER = 0xA8 + FILE_SIZE_POINTER_64 = 0xB8 + SECOND_POINTER = 0xBC + INDEX_POINTER = 0xC4 + SECOND_POINTER_64 = 0xE0 + INDEX_POINTER_64 = 0xF0 + ENC_OFFSET = 0x1CD + + attr_reader :magic, :index_type, :encrypt_type, :size + attr_reader :index1_count, :index1, :index2_count, :index2 + attr_reader :version + def initialize data + @magic = data.unpack('N')[0] + @index_type = data[INDEX_TYPE_OFFSET] + @version = {0x0e => 1997, 0x17 => 2003}[@index_type] + + if version_2003? + # don't know? + # >> data1.unpack('V*').zip(data2.unpack('V*')).enum_with_index.select { |(c, d), i| c != d and not [46, 56, 60].include?(i) }.select { |(a, b), i| b == 0 }.map { |(a, b), i| [a / 256, i] } + # [8, 76], [32768, 84], [128, 89] + # >> data1.unpack('C*').zip(data2.unpack('C*')).enum_with_index.select { |(c, d), i| c != d and not [184..187, 224..227, 240..243].any? { |r| r === i } }.select { |(a, b), i| b == 0 and ((Math.log(a) / Math.log(2)) % 1) < 0.0001 } + # [[[2, 0], 61], [[2, 0], 76], [[2, 0], 195], [[2, 0], 257], [[8, 0], 305], [[128, 0], 338], [[128, 0], 357]] + # i have only 2 psts to base this guess on, so i can't really come up with anything that looks reasonable yet. not sure what the offset is. unfortunately there is so much in the header + # that isn't understood... + @encrypt_type = 1 + + @index2_count, @index2 = data[SECOND_POINTER_64 - 4, 8].unpack('V2') + @index1_count, @index1 = data[INDEX_POINTER_64 - 4, 8].unpack('V2') + + @size = data[FILE_SIZE_POINTER_64, 4].unpack('V')[0] + else + @encrypt_type = data[ENC_OFFSET] + + @index2_count, @index2 = data[SECOND_POINTER - 4, 8].unpack('V2') + @index1_count, @index1 = data[INDEX_POINTER - 4, 8].unpack('V2') + + @size = data[FILE_SIZE_POINTER, 4].unpack('V')[0] + end + + validate! + end + + def version_2003? + version == 2003 + end + + def encrypted? + encrypt_type != 0 + end + + def validate! + raise FormatError, "bad signature on pst file (#{'0x%x' % magic})" unless magic == MAGIC + raise FormatError, "only index types 0x0e and 0x17 are handled (#{'0x%x' % index_type})" unless [0x0e, 0x17].include?(index_type) + raise FormatError, "only encrytion types 0 and 1 are handled (#{encrypt_type.inspect})" unless [0, 1].include?(encrypt_type) + end + end + + # compressible encryption! :D + # + # simple substitution. see libpst.c + # maybe test switch to using a String#tr! + class CompressibleEncryption + DECRYPT_TABLE = [ + 0x47, 0xf1, 0xb4, 0xe6, 0x0b, 0x6a, 0x72, 0x48, + 0x85, 0x4e, 0x9e, 0xeb, 0xe2, 0xf8, 0x94, 0x53, # 0x0f + 0xe0, 0xbb, 0xa0, 0x02, 0xe8, 0x5a, 0x09, 0xab, + 0xdb, 0xe3, 0xba, 0xc6, 0x7c, 0xc3, 0x10, 0xdd, # 0x1f + 0x39, 0x05, 0x96, 0x30, 0xf5, 0x37, 0x60, 0x82, + 0x8c, 0xc9, 0x13, 0x4a, 0x6b, 0x1d, 0xf3, 0xfb, # 0x2f + 0x8f, 0x26, 0x97, 0xca, 0x91, 0x17, 0x01, 0xc4, + 0x32, 0x2d, 0x6e, 0x31, 0x95, 0xff, 0xd9, 0x23, # 0x3f + 0xd1, 0x00, 0x5e, 0x79, 0xdc, 0x44, 0x3b, 0x1a, + 0x28, 0xc5, 0x61, 0x57, 0x20, 0x90, 0x3d, 0x83, # 0x4f + 0xb9, 0x43, 0xbe, 0x67, 0xd2, 0x46, 0x42, 0x76, + 0xc0, 0x6d, 0x5b, 0x7e, 0xb2, 0x0f, 0x16, 0x29, # 0x5f + 0x3c, 0xa9, 0x03, 0x54, 0x0d, 0xda, 0x5d, 0xdf, + 0xf6, 0xb7, 0xc7, 0x62, 0xcd, 0x8d, 0x06, 0xd3, # 0x6f + 0x69, 0x5c, 0x86, 0xd6, 0x14, 0xf7, 0xa5, 0x66, + 0x75, 0xac, 0xb1, 0xe9, 0x45, 0x21, 0x70, 0x0c, # 0x7f + 0x87, 0x9f, 0x74, 0xa4, 0x22, 0x4c, 0x6f, 0xbf, + 0x1f, 0x56, 0xaa, 0x2e, 0xb3, 0x78, 0x33, 0x50, # 0x8f + 0xb0, 0xa3, 0x92, 0xbc, 0xcf, 0x19, 0x1c, 0xa7, + 0x63, 0xcb, 0x1e, 0x4d, 0x3e, 0x4b, 0x1b, 0x9b, # 0x9f + 0x4f, 0xe7, 0xf0, 0xee, 0xad, 0x3a, 0xb5, 0x59, + 0x04, 0xea, 0x40, 0x55, 0x25, 0x51, 0xe5, 0x7a, # 0xaf + 0x89, 0x38, 0x68, 0x52, 0x7b, 0xfc, 0x27, 0xae, + 0xd7, 0xbd, 0xfa, 0x07, 0xf4, 0xcc, 0x8e, 0x5f, # 0xbf + 0xef, 0x35, 0x9c, 0x84, 0x2b, 0x15, 0xd5, 0x77, + 0x34, 0x49, 0xb6, 0x12, 0x0a, 0x7f, 0x71, 0x88, # 0xcf + 0xfd, 0x9d, 0x18, 0x41, 0x7d, 0x93, 0xd8, 0x58, + 0x2c, 0xce, 0xfe, 0x24, 0xaf, 0xde, 0xb8, 0x36, # 0xdf + 0xc8, 0xa1, 0x80, 0xa6, 0x99, 0x98, 0xa8, 0x2f, + 0x0e, 0x81, 0x65, 0x73, 0xe4, 0xc2, 0xa2, 0x8a, # 0xef + 0xd4, 0xe1, 0x11, 0xd0, 0x08, 0x8b, 0x2a, 0xf2, + 0xed, 0x9a, 0x64, 0x3f, 0xc1, 0x6c, 0xf9, 0xec # 0xff + ] + + ENCRYPT_TABLE = [nil] * 256 + DECRYPT_TABLE.each_with_index { |i, j| ENCRYPT_TABLE[i] = j } + + def self.decrypt_alt encrypted + decrypted = '' + encrypted.length.times { |i| decrypted << DECRYPT_TABLE[encrypted[i]] } + decrypted + end + + def self.encrypt_alt decrypted + encrypted = '' + decrypted.length.times { |i| encrypted << ENCRYPT_TABLE[decrypted[i]] } + encrypted + end + + # an alternate implementation that is possibly faster.... + # TODO - bench + DECRYPT_STR, ENCRYPT_STR = [DECRYPT_TABLE, (0...256)].map do |values| + values.map { |i| i.chr }.join.gsub(/([\^\-\\])/, "\\\\\\1") + end + + def self.decrypt encrypted + encrypted.tr ENCRYPT_STR, DECRYPT_STR + end + + def self.encrypt decrypted + decrypted.tr DECRYPT_STR, ENCRYPT_STR + end + end + + class RangesIOEncryptable < RangesIO + def initialize io, mode='r', params={} + mode, params = 'r', mode if Hash === mode + @decrypt = !!params[:decrypt] + super + end + + def encrypted? + @decrypt + end + + def read limit=nil + buf = super + buf = CompressibleEncryption.decrypt(buf) if encrypted? + buf + end + end + + attr_reader :io, :header, :idx, :desc, :special_folder_ids + + # corresponds to + # * pst_open + # * pst_load_index + def initialize io + @io = io + io.pos = 0 + @header = Header.new io.read(Header::SIZE) + + # would prefer this to be in Header#validate, but it doesn't have the io size. + # should perhaps downgrade this to just be a warning... + raise FormatError, "header size field invalid (#{header.size} != #{io.size}}" unless header.size == io.size + + load_idx + load_desc + load_xattrib + + @special_folder_ids = {} + end + + def encrypted? + @header.encrypted? + end + + # until i properly fix logging... + def warn s + Mapi::Log.warn s + end + + # + # this is the index and desc record loading code + # ---------------------------------------------------------------------------- + # + + ToTree = Module.new + + module Index2 + BLOCK_SIZE = 512 + module RecursiveLoad + def load_chain + #... + end + end + + module Base + def read + #... + end + end + + class Version1997 < Struct.new(:a)#...) + SIZE = 12 + + include RecursiveLoad + include Base + end + + class Version2003 < Struct.new(:a)#...) + SIZE = 24 + + include RecursiveLoad + include Base + end + end + + module Desc2 + module Base + def desc + #... + end + end + + class Version1997 < Struct.new(:a)#...) + #include Index::RecursiveLoad + include Base + end + + class Version2003 < Struct.new(:a)#...) + #include Index::RecursiveLoad + include Base + end + end + + # more constants from libpst.c + # these relate to the index block + ITEM_COUNT_OFFSET = 0x1f0 # count byte + LEVEL_INDICATOR_OFFSET = 0x1f3 # node or leaf + BACKLINK_OFFSET = 0x1f8 # backlink u1 value + + # these 3 classes are used to hold various file records + + # pst_index + class Index < Struct.new(:id, :offset, :size, :u1) + UNPACK_STR = 'VVvv' + SIZE = 12 + BLOCK_SIZE = 512 # index blocks was 516 but bogus + COUNT_MAX = 41 # max active items (ITEM_COUNT_OFFSET / Index::SIZE = 41) + + attr_accessor :pst + def initialize data + data = Pst.unpack data, UNPACK_STR if String === data + super(*data) + end + + def type + @type ||= begin + if id & 0x2 == 0 + :data + else + first_byte, second_byte = read.unpack('CC') + if first_byte == 1 + raise second_byte unless second_byte == 1 + :data_chain_header + elsif first_byte == 2 + raise second_byte unless second_byte == 0 + :id2_assoc + else + raise FormatError, 'unknown first byte for block - %p' % first_byte + end + end + end + end + + def data? + (id & 0x2) == 0 + end + + def read decrypt=true + # only data blocks are every encrypted + decrypt = false unless data? + pst.pst_read_block_size offset, size, decrypt + end + + # show all numbers in hex + def inspect + super.gsub(/=(\d+)/) { '=0x%x' % $1.to_i }.sub(/Index /, "Index type=#{type.inspect}, ") + end + end + + # mostly guesses. + ITEM_COUNT_OFFSET_64 = 0x1e8 + LEVEL_INDICATOR_OFFSET_64 = 0x1eb # diff of 3 between these 2 as above... + + # will maybe inherit from Index64, in order to get the same #type function. + class Index64 < Index + UNPACK_STR = 'TTvvV' + SIZE = 24 + BLOCK_SIZE = 512 + COUNT_MAX = 20 # bit of a guess really. 512 / 24 = 21, but doesn't leave enough header room + + # this is the extra item on the end of the UNPACK_STR above + attr_accessor :u2 + + def initialize data + data = Pst.unpack data, UNPACK_STR if String === data + @u2 = data.pop + super data + end + + def inspect + super.sub(/>$/, ', u2=%p>' % u2) + end + + def self.load_chain io, header + load_idx_rec io, header.index1, 0, 0 + end + + # almost identical to load code for Index, just different offsets and unpack strings. + # can probably merge them, or write a generic load_tree function or something. + def self.load_idx_rec io, offset, linku1, start_val + io.seek offset + buf = io.read BLOCK_SIZE + idxs = [] + + item_count = buf[ITEM_COUNT_OFFSET_64] + raise "have too many active items in index (#{item_count})" if item_count > COUNT_MAX + + #idx = Index.new buf[BACKLINK_OFFSET, Index::SIZE] + #raise 'blah 1' unless idx.id == linku1 + + if buf[LEVEL_INDICATOR_OFFSET_64] == 0 + # leaf pointers + # split the data into item_count index objects + buf[0, SIZE * item_count].scan(/.{#{SIZE}}/mo).each_with_index do |data, i| + idx = new data + # first entry + raise 'blah 3' if i == 0 and start_val != 0 and idx.id != start_val + #idx.pst = self + break if idx.id == 0 + idxs << idx + end + else + # node pointers + # split the data into item_count table pointers + buf[0, SIZE * item_count].scan(/.{#{SIZE}}/mo).each_with_index do |data, i| + start, u1, offset = Pst.unpack data, 'T3' + # for the first value, we expect the start to be equal + raise 'blah 3' if i == 0 and start_val != 0 and start != start_val + break if start == 0 + idxs += load_idx_rec io, offset, u1, start + end + end + + idxs + end + end + + # pst_desc + class Desc64 < Struct.new(:desc_id, :idx_id, :idx2_id, :parent_desc_id, :u2) + UNPACK_STR = 'T3VV' + SIZE = 32 + BLOCK_SIZE = 512 # descriptor blocks was 520 but bogus + COUNT_MAX = 15 # guess as per Index64 + + include RecursivelyEnumerable + + attr_accessor :pst + attr_reader :children + def initialize data + super(*Pst.unpack(data, UNPACK_STR)) + @children = [] + end + + def desc + pst.idx_from_id idx_id + end + + def list_index + pst.idx_from_id idx2_id + end + + def self.load_chain io, header + load_desc_rec io, header.index2, 0, 0x21 + end + + def self.load_desc_rec io, offset, linku1, start_val + io.seek offset + buf = io.read BLOCK_SIZE + descs = [] + item_count = buf[ITEM_COUNT_OFFSET_64] + + # not real desc + #desc = Desc.new buf[BACKLINK_OFFSET, 4] + #raise 'blah 1' unless desc.desc_id == linku1 + + if buf[LEVEL_INDICATOR_OFFSET_64] == 0 + # leaf pointers + raise "have too many active items in index (#{item_count})" if item_count > COUNT_MAX + # split the data into item_count desc objects + buf[0, SIZE * item_count].scan(/.{#{SIZE}}/mo).each_with_index do |data, i| + desc = new data + # first entry + raise 'blah 3' if i == 0 and start_val != 0 and desc.desc_id != start_val + break if desc.desc_id == 0 + descs << desc + end + else + # node pointers + raise "have too many active items in index (#{item_count})" if item_count > Index64::COUNT_MAX + # split the data into item_count table pointers + buf[0, Index64::SIZE * item_count].scan(/.{#{Index64::SIZE}}/mo).each_with_index do |data, i| + start, u1, offset = Pst.unpack data, 'T3' + # for the first value, we expect the start to be equal note that ids -1, so even for the + # first we expect it to be equal. thats the 0x21 (dec 33) desc record. this means we assert + # that the first desc record is always 33... + # thats because 0x21 is the pst root itself... + raise 'blah 3' if i == 0 and start_val != -1 and start != start_val + # this shouldn't really happen i'd imagine + break if start == 0 + descs += load_desc_rec io, offset, u1, start + end + end + + descs + end + + def each_child(&block) + @children.each(&block) + end + end + + # _pst_table_ptr_struct + class TablePtr < Struct.new(:start, :u1, :offset) + UNPACK_STR = 'V3' + SIZE = 12 + + def initialize data + data = data.unpack(UNPACK_STR) if String === data + super(*data) + end + end + + # pst_desc + # idx_id is a pointer to an idx record which gets the primary data stream for the Desc record. + # idx2_id gets you an idx record, that when read gives you an ID2 association list, which just maps + # another set of ids to index values + class Desc < Struct.new(:desc_id, :idx_id, :idx2_id, :parent_desc_id) + UNPACK_STR = 'V4' + SIZE = 16 + BLOCK_SIZE = 512 # descriptor blocks was 520 but bogus + COUNT_MAX = 31 # max active desc records (ITEM_COUNT_OFFSET / Desc::SIZE = 31) + + include ToTree + + attr_accessor :pst + attr_reader :children + def initialize data + super(*data.unpack(UNPACK_STR)) + @children = [] + end + + def desc + pst.idx_from_id idx_id + end + + def list_index + pst.idx_from_id idx2_id + end + + # show all numbers in hex + def inspect + super.gsub(/=(\d+)/) { '=0x%x' % $1.to_i } + end + end + + # corresponds to + # * _pst_build_id_ptr + def load_idx + @idx = [] + @idx_offsets = [] + if header.version_2003? + @idx = Index64.load_chain io, header + @idx.each { |idx| idx.pst = self } + else + load_idx_rec header.index1, header.index1_count, 0 + end + + # we'll typically be accessing by id, so create a hash as a lookup cache + @idx_from_id = {} + @idx.each do |idx| + warn "there are duplicate idx records with id #{idx.id}" if @idx_from_id[idx.id] + @idx_from_id[idx.id] = idx + end + end + + # load the flat idx table, which maps ids to file ranges. this is the recursive helper + # + # corresponds to + # * _pst_build_id_ptr + def load_idx_rec offset, linku1, start_val + @idx_offsets << offset + + #_pst_read_block_size(pf, offset, BLOCK_SIZE, &buf, 0, 0) < BLOCK_SIZE) + buf = pst_read_block_size offset, Index::BLOCK_SIZE, false + + item_count = buf[ITEM_COUNT_OFFSET] + raise "have too many active items in index (#{item_count})" if item_count > Index::COUNT_MAX + + idx = Index.new buf[BACKLINK_OFFSET, Index::SIZE] + raise 'blah 1' unless idx.id == linku1 + + if buf[LEVEL_INDICATOR_OFFSET] == 0 + # leaf pointers + # split the data into item_count index objects + buf[0, Index::SIZE * item_count].scan(/.{#{Index::SIZE}}/mo).each_with_index do |data, i| + idx = Index.new data + # first entry + raise 'blah 3' if i == 0 and start_val != 0 and idx.id != start_val + idx.pst = self + # this shouldn't really happen i'd imagine + break if idx.id == 0 + @idx << idx + end + else + # node pointers + # split the data into item_count table pointers + buf[0, TablePtr::SIZE * item_count].scan(/.{#{TablePtr::SIZE}}/mo).each_with_index do |data, i| + table = TablePtr.new data + # for the first value, we expect the start to be equal + raise 'blah 3' if i == 0 and start_val != 0 and table.start != start_val + # this shouldn't really happen i'd imagine + break if table.start == 0 + load_idx_rec table.offset, table.u1, table.start + end + end + end + + # most access to idx objects will use this function + # + # corresponds to + # * _pst_getID + def idx_from_id id + @idx_from_id[id] + end + + # corresponds to + # * _pst_build_desc_ptr + # * record_descriptor + def load_desc + @desc = [] + @desc_offsets = [] + if header.version_2003? + @desc = Desc64.load_chain io, header + @desc.each { |desc| desc.pst = self } + else + load_desc_rec header.index2, header.index2_count, 0x21 + end + + # first create a lookup cache + @desc_from_id = {} + @desc.each do |desc| + desc.pst = self + warn "there are duplicate desc records with id #{desc.desc_id}" if @desc_from_id[desc.desc_id] + @desc_from_id[desc.desc_id] = desc + end + + # now turn the flat list of loaded desc records into a tree + + # well, they have no parent, so they're more like, the toplevel descs. + @orphans = [] + # now assign each node to the parents child array, putting the orphans in the above + @desc.each do |desc| + parent = @desc_from_id[desc.parent_desc_id] + # note, besides this, its possible to create other circular structures. + if parent == desc + # this actually happens usually, for the root_item it appears. + #warn "desc record's parent is itself (#{desc.inspect})" + # maybe add some more checks in here for circular structures + elsif parent + parent.children << desc + next + end + @orphans << desc + end + + # maybe change this to some sort of sane-ness check. orphans are expected +# warn "have #{@orphans.length} orphan desc record(s)." unless @orphans.empty? + end + + # load the flat list of desc records recursively + # + # corresponds to + # * _pst_build_desc_ptr + # * record_descriptor + def load_desc_rec offset, linku1, start_val + @desc_offsets << offset + + buf = pst_read_block_size offset, Desc::BLOCK_SIZE, false + item_count = buf[ITEM_COUNT_OFFSET] + + # not real desc + desc = Desc.new buf[BACKLINK_OFFSET, 4] + raise 'blah 1' unless desc.desc_id == linku1 + + if buf[LEVEL_INDICATOR_OFFSET] == 0 + # leaf pointers + raise "have too many active items in index (#{item_count})" if item_count > Desc::COUNT_MAX + # split the data into item_count desc objects + buf[0, Desc::SIZE * item_count].scan(/.{#{Desc::SIZE}}/mo).each_with_index do |data, i| + desc = Desc.new data + # first entry + raise 'blah 3' if i == 0 and start_val != 0 and desc.desc_id != start_val + # this shouldn't really happen i'd imagine + break if desc.desc_id == 0 + @desc << desc + end + else + # node pointers + raise "have too many active items in index (#{item_count})" if item_count > Index::COUNT_MAX + # split the data into item_count table pointers + buf[0, TablePtr::SIZE * item_count].scan(/.{#{TablePtr::SIZE}}/mo).each_with_index do |data, i| + table = TablePtr.new data + # for the first value, we expect the start to be equal note that ids -1, so even for the + # first we expect it to be equal. thats the 0x21 (dec 33) desc record. this means we assert + # that the first desc record is always 33... + raise 'blah 3' if i == 0 and start_val != -1 and table.start != start_val + # this shouldn't really happen i'd imagine + break if table.start == 0 + load_desc_rec table.offset, table.u1, table.start + end + end + end + + # as for idx + # + # corresponds to: + # * _pst_getDptr + def desc_from_id id + @desc_from_id[id] + end + + # corresponds to + # * pst_load_extended_attributes + def load_xattrib + unless desc = desc_from_id(0x61) + warn "no extended attributes desc record found" + return + end + unless desc.desc + warn "no desc idx for extended attributes" + return + end + if desc.list_index + end + #warn "skipping loading xattribs" + # FIXME implement loading xattribs + end + + # corresponds to: + # * _pst_read_block_size + # * _pst_read_block ?? + # * _pst_ff_getIDblock_dec ?? + # * _pst_ff_getIDblock ?? + def pst_read_block_size offset, size, decrypt=true + io.seek offset + buf = io.read size + warn "tried to read #{size} bytes but only got #{buf.length}" if buf.length != size + encrypted? && decrypt ? CompressibleEncryption.decrypt(buf) : buf + end + + # + # id2 + # ---------------------------------------------------------------------------- + # + + class ID2Assoc < Struct.new(:id2, :id, :table2) + UNPACK_STR = 'V3' + SIZE = 12 + + def initialize data + data = data.unpack(UNPACK_STR) if String === data + super(*data) + end + end + + class ID2Assoc64 < Struct.new(:id2, :u1, :id, :table2) + UNPACK_STR = 'VVT2' + SIZE = 24 + + def initialize data + if String === data + data = Pst.unpack data, UNPACK_STR + end + super(*data) + end + + def self.load_chain idx + buf = idx.read + type, count = buf.unpack 'v2' + unless type == 0x0002 + raise 'unknown id2 type 0x%04x' % type + #return + end + id2 = [] + count.times do |i| + assoc = new buf[8 + SIZE * i, SIZE] + id2 << assoc + if assoc.table2 != 0 + id2 += load_chain idx.pst.idx_from_id(assoc.table2) + end + end + id2 + end + end + + class ID2Mapping + attr_reader :list + def initialize pst, list + @pst = pst + @list = list + # create a lookup. + @id_from_id2 = {} + @list.each do |id2| + # NOTE we take the last value seen value if there are duplicates. this "fixes" + # test4-o1997.pst for the time being. + warn "there are duplicate id2 records with id #{id2.id2}" if @id_from_id2[id2.id2] + next if @id_from_id2[id2.id2] + @id_from_id2[id2.id2] = id2.id + end + end + + # TODO: fix logging + def warn s + Mapi::Log.warn s + end + + # corresponds to: + # * _pst_getID2 + def [] id + #id2 = @list.find { |x| x.id2 == id } + id = @id_from_id2[id] + id and @pst.idx_from_id(id) + end + end + + def load_idx2 idx + if header.version_2003? + id2 = ID2Assoc64.load_chain idx + else + id2 = load_idx2_rec idx + end + ID2Mapping.new self, id2 + end + + # corresponds to + # * _pst_build_id2 + def load_idx2_rec idx + # i should perhaps use a idx chain style read here? + buf = pst_read_block_size idx.offset, idx.size, false + type, count = buf.unpack 'v2' + unless type == 0x0002 + raise 'unknown id2 type 0x%04x' % type + #return + end + id2 = [] + count.times do |i| + assoc = ID2Assoc.new buf[4 + ID2Assoc::SIZE * i, ID2Assoc::SIZE] + id2 << assoc + if assoc.table2 != 0 + id2 += load_idx2_rec idx_from_id(assoc.table2) + end + end + id2 + end + + class RangesIOIdxChain < RangesIOEncryptable + def initialize pst, idx_head + @idxs = pst.id2_block_idx_chain idx_head + # whether or not a given idx needs encrypting + decrypts = @idxs.map do |idx| + decrypt = (idx.id & 2) != 0 ? false : pst.encrypted? + end.uniq + raise NotImplementedError, 'partial encryption in RangesIOID2' if decrypts.length > 1 + decrypt = decrypts.first + # convert idxs to ranges + ranges = @idxs.map { |idx| [idx.offset, idx.size] } + super pst.io, :ranges => ranges, :decrypt => decrypt + end + end + + class RangesIOID2 < RangesIOIdxChain + def self.new pst, id2, idx2 + RangesIOIdxChain.new pst, idx2[id2] + end + end + + # corresponds to: + # * _pst_ff_getID2block + # * _pst_ff_getID2data + # * _pst_ff_compile_ID + def id2_block_idx_chain idx + if (idx.id & 0x2) == 0 + [idx] + else + buf = idx.read + type, fdepth, count = buf[0, 4].unpack 'CCv' + unless type == 1 # libpst.c:3958 + warn 'Error in idx_chain - %p, %p, %p - attempting to ignore' % [type, fdepth, count] + return [idx] + end + # there are 4 unaccounted for bytes here, 4...8 + if header.version_2003? + ids = buf[8, count * 8].unpack("T#{count}") + else + ids = buf[8, count * 4].unpack('V*') + end + if fdepth == 1 + ids.map { |id| idx_from_id id } + else + ids.map { |id| id2_block_idx_chain idx_from_id(id) }.flatten + end + end + end + + # + # main block parsing code. gets raw properties + # ---------------------------------------------------------------------------- + # + + # the job of this class, is to take a desc record, and be able to enumerate through the + # mapi properties of the associated thing. + # + # corresponds to + # * _pst_parse_block + # * _pst_process (in some ways. although perhaps thats more the Item::Properties#add_property) + class BlockParser + include Mapi::Types::Constants + + TYPES = { + 0xbcec => 1, + 0x7cec => 2, + # type 3 is removed. an artifact of not handling the indirect blocks properly in libpst. + } + + PR_SUBJECT = PropertySet::TAGS.find { |num, (name, type)| name == 'PR_SUBJECT' }.first.hex + PR_BODY_HTML = PropertySet::TAGS.find { |num, (name, type)| name == 'PR_BODY_HTML' }.first.hex + + # this stuff could maybe be moved to Ole::Types? or leverage it somehow? + # whether or not a type is immeidate is more a property of the pst encoding though i expect. + # what i probably can add is a generic concept of whether a type is of variadic length or not. + + # these lists are very incomplete. think they are largely copied from libpst + + IMMEDIATE_TYPES = [ + PT_SHORT, PT_LONG, PT_BOOLEAN + ] + + INDIRECT_TYPES = [ + PT_DOUBLE, PT_OBJECT, + 0x0014, # whats this? probably something like PT_LONGLONG, given the correspondence with the + # ole variant types. (= VT_I8) + PT_STRING8, PT_UNICODE, # unicode isn't in libpst, but added here for outlook 2003 down the track + PT_SYSTIME, + 0x0048, # another unknown + 0x0102, # this is PT_BINARY vs PT_CLSID + #0x1003, # these are vector types, but they're commented out for now because i'd expect that + #0x1014, # there's extra decoding needed that i'm not doing. (probably just need a simple + # # PT_* => unpack string mapping for the immediate types, and just do unpack('V*') etc + #0x101e, + #0x1102 + ] + + # the attachment and recipient arrays appear to be always stored with these fixed + # id2 values. seems strange. are there other extra streams? can find out by making higher + # level IO wrapper, which has the id2 value, and doing the diff of available id2 values versus + # used id2 values in properties of an item. + ID2_ATTACHMENTS = 0x671 + ID2_RECIPIENTS = 0x692 + + attr_reader :desc, :data, :data_chunks, :offset_tables + def initialize desc + raise FormatError, "unable to get associated index record for #{desc.inspect}" unless desc.desc + @desc = desc + #@data = desc.desc.read + if Pst::Index === desc.desc + #@data = RangesIOIdxChain.new(desc.pst, desc.desc).read + idxs = desc.pst.id2_block_idx_chain desc.desc + # this gets me the plain index chain. + else + # fake desc + #@data = desc.desc.read + idxs = [desc.desc] + end + + @data_chunks = idxs.map { |idx| idx.read } + @data = @data_chunks.first + + load_header + + @index_offsets = [@index_offset] + @data_chunks[1..-1].map { |chunk| chunk.unpack('v')[0] } + @offset_tables = [] + @ignored = [] + @data_chunks.zip(@index_offsets).each do |chunk, offset| + ignore = chunk[offset, 2].unpack('v')[0] + @ignored << ignore +# p ignore + @offset_tables.push offset_table = [] + # maybe its ok if there aren't to be any values ? + raise FormatError if offset == 0 + offsets = chunk[offset + 2..-1].unpack('v*') + #p offsets + offsets[0, ignore + 2].each_cons 2 do |from, to| + #next if to == 0 + raise FormatError, [from, to].inspect if from > to + offset_table << [from, to] + end + end + + @offset_table = @offset_tables.first + @idxs = idxs + + # now, we may have multiple different blocks + end + + # a given desc record may or may not have associated idx2 data. we lazily load it here, so it will never + # actually be requested unless get_data_indirect actually needs to use it. + def idx2 + return @idx2 if @idx2 + raise FormatError, 'idx2 requested but no idx2 available' unless desc.list_index + # should check this can't return nil + @idx2 = desc.pst.load_idx2 desc.list_index + end + + def load_header + @index_offset, type, @offset1 = data.unpack 'vvV' + raise FormatError, 'unknown block type signature 0x%04x' % type unless TYPES[type] + @type = TYPES[type] + end + + # based on the value of offset, return either some data from buf, or some data from the + # id2 chain id2, where offset is some key into a lookup table that is stored as the id2 + # chain. i think i may need to create a BlockParser class that wraps up all this mess. + # + # corresponds to: + # * _pst_getBlockOffsetPointer + # * _pst_getBlockOffset + def get_data_indirect offset + return get_data_indirect_io(offset).read + + if offset == 0 + nil + elsif (offset & 0xf) == 0xf + RangesIOID2.new(desc.pst, offset, idx2).read + else + low, high = offset & 0xf, offset >> 4 + raise FormatError if low != 0 or (high & 0x1) != 0 or (high / 2) > @offset_table.length + from, to = @offset_table[high / 2] + data[from...to] + end + end + + def get_data_indirect_io offset + if offset == 0 + nil + elsif (offset & 0xf) == 0xf + if idx2[offset] + RangesIOID2.new desc.pst, offset, idx2 + else + warn "tried to get idx2 record for #{offset} but failed" + return StringIO.new('') + end + else + low, high = offset & 0xf, offset >> 4 + if low != 0 or (high & 0x1) != 0 +# raise FormatError, + warn "bad - #{low} #{high} (1)" + return StringIO.new('') + end + # lets see which block it should come from. + block_idx, i = high.divmod 4096 + unless block_idx < @data_chunks.length + warn "bad - block_idx to high (not #{block_idx} < #{@data_chunks.length})" + return StringIO.new('') + end + data_chunk, offset_table = @data_chunks[block_idx], @offset_tables[block_idx] + if i / 2 >= offset_table.length + warn "bad - #{low} #{high} - #{i / 2} >= #{offset_table.length} (2)" + return StringIO.new('') + end + #warn "ok - #{low} #{high} #{offset_table.length}" + from, to = offset_table[i / 2] + StringIO.new data_chunk[from...to] + end + end + + def handle_indirect_values key, type, value + case type + when PT_BOOLEAN + value = value != 0 + when *IMMEDIATE_TYPES # not including PT_BOOLEAN which we just did above + # no processing current applied (needed?). + when *INDIRECT_TYPES + # the value is a pointer + if String === value # ie, value size > 4 above + value = StringIO.new value + else + value = get_data_indirect_io(value) + end + # keep strings as immediate values for now, for compatability with how i set up + # Msg::Properties::ENCODINGS + if value + if type == PT_STRING8 + value = value.read + elsif type == PT_UNICODE + value = Ole::Types::FROM_UTF16.iconv value.read + end + end + # special subject handling + if key == PR_BODY_HTML and value + # to keep the msg code happy, which thinks body_html will be an io + # although, in 2003 version, they are 0102 already + value = StringIO.new value unless value.respond_to?(:read) + end + if key == PR_SUBJECT and value + ignore, offset = value.unpack 'C2' + offset = (offset == 1 ? nil : offset - 3) + value = value[2..-1] +=begin + index = value =~ /^[A-Z]*:/ ? $~[0].length - 1 : nil + unless ignore == 1 and offset == index + warn 'something wrong with subject hack' + $x = [ignore, offset, value] + require 'irb' + IRB.start + exit + end +=end +=begin +new idea: + +making sense of the \001\00[156] i've seen prefixing subject. i think its to do with the placement +of the ':', or the ' '. And perhaps an optimization to do with thread topic, and ignoring the prefixes +added by mailers. thread topic is equal to subject with all that crap removed. + +can test by creating some mails with bizarre subjects. + +subject="\001\005RE: blah blah" +subject="\001\001blah blah" +subject="\001\032Out of Office AutoReply: blah blah" +subject="\001\020Undeliverable: blah blah" + +looks like it + +=end + + # now what i think, is that perhaps, value[offset..-1] ... + # or something like that should be stored as a special tag. ie, do a double yield + # for this case. probably PR_CONVERSATION_TOPIC, in which case i'd write instead: + # yield [PR_SUBJECT, ref_type, value] + # yield [PR_CONVERSATION_TOPIC, ref_type, value[offset..-1] + # next # to skip the yield. + end + + # special handling for embedded objects + # used for attach_data for attached messages. in which case attach_method should == 5, + # for embedded object. + if type == PT_OBJECT and value + value = value.read if value.respond_to?(:read) + id2, unknown = value.unpack 'V2' + io = RangesIOID2.new desc.pst, id2, idx2 + + # hacky + desc2 = OpenStruct.new(:desc => io, :pst => desc.pst, :list_index => desc.list_index, :children => []) + # put nil instead of desc.list_index, otherwise the attachment is attached to itself ad infinitum. + # should try and fix that FIXME + # this shouldn't be done always. for an attached message, yes, but for an attached + # meta file, for example, it shouldn't. difference between embedded_ole vs embedded_msg + # really. + # note that in the case where its a embedded ole, you actually get a regular serialized ole + # object, so i need to create an ole storage object on a rangesioidxchain! + # eg: +=begin +att.props.display_name # => "Picture (Metafile)" +io = att.props.attach_data +io.read(32).unpack('H*') # => ["d0cf11e0a1b11ae100000.... note the docfile signature. +# plug some missing rangesio holes: +def io.rewind; seek 0; end +def io.flush; raise IOError; end +ole = Ole::Storage.open io +puts ole.root.to_tree + +- #<Dirent:"Root Entry"> + |- #<Dirent:"\001Ole" size=20 data="\001\000\000\002\000..."> + |- #<Dirent:"CONTENTS" size=65696 data="\327\315\306\232\000..."> + \- #<Dirent:"\003MailStream" size=12 data="\001\000\000\000[..."> +=end + # until properly fixed, i have disabled this code here, so this will break + # nested messages temporarily. + #value = Item.new desc2, RawPropertyStore.new(desc2).to_a + #desc2.list_index = nil + value = io + end + # this is PT_MV_STRING8, i guess. + # should probably have the 0x1000 flag, and do the or-ring. + # example of 0x1102 is PR_OUTLOOK_2003_ENTRYIDS. less sure about that one. + when 0x101e, 0x1102 + # example data: + # 0x802b "\003\000\000\000\020\000\000\000\030\000\000\000#\000\000\000BusinessCompetitionFavorites" + # this 0x802b would be an extended attribute for categories / keywords. + value = get_data_indirect_io(value).read unless String === value + num = value.unpack('V')[0] + offsets = value[4, 4 * num].unpack("V#{num}") + value = (offsets + [value.length]).to_enum(:each_cons, 2).map { |from, to| value[from...to] } + value.map! { |str| StringIO.new str } if type == 0x1102 + else + name = Mapi::Types::DATA[type].first rescue nil + warn '0x%04x %p' % [key, get_data_indirect_io(value).read] + raise NotImplementedError, 'unsupported mapi property type - 0x%04x (%p)' % [type, name] + end + [key, type, value] + end + end + +=begin +* recipients: + + affects: ["0x200764", "0x2011c4", "0x201b24", "0x201b44", "0x201ba4", "0x201c24", "0x201cc4", "0x202504"] + +after adding the rawpropertystoretable fix, all except the second parse properly, and satisfy: + + item.props.display_to == item.recipients.map { |r| r.props.display_name if r.props.recipient_type == 1 }.compact * '; ' + +only the second still has a problem + +#[#<struct Pst::Desc desc_id=0x2011c4, idx_id=0x397c, idx2_id=0x398a, parent_desc_id=0x8082>] + +think this is related to a multi block #data3. ie, when you use @x * rec_size, and it +goes > 8190, or there abouts, then it stuffs up. probably there is header gunk, or something, +similar to when #data is multi block. + +same problem affects the attachment table in test4. + +fixed that issue. round data3 ranges to rec_size. + +fix other issue with attached objects. + +all recipients and attachments in test2 are fine. + +only remaining issue is test4 recipients of 200044. strange. + +=end + + # RawPropertyStore is used to iterate through the properties of an item, or the auxiliary + # data for an attachment. its just a parser for the way the properties are serialized, when the + # properties don't have to conform to a column structure. + # + # structure of this chunk of data is often + # header, property keys, data values, and then indexes. + # the property keys has value in it. value can be the actual value if its a short type, + # otherwise you lookup the value in the indicies, where you get the offsets to use in the + # main data body. due to the indirect thing though, any of these parts could actually come + # from a separate stream. + class RawPropertyStore < BlockParser + include Enumerable + + attr_reader :length + def initialize desc + super + raise FormatError, "expected type 1 - got #{@type}" unless @type == 1 + + # the way that offset works, data1 may be a subset of buf, or something from id2. if its from buf, + # it will be offset based on index_offset and offset. so it could be some random chunk of data anywhere + # in the thing. + header_data = get_data_indirect @offset1 + raise FormatError if header_data.length < 8 + signature, offset2 = header_data.unpack 'V2' + #p [@type, signature] + raise FormatError, 'unhandled block signature 0x%08x' % @type if signature != 0x000602b5 + # this is actually a big chunk of tag tuples. + @index_data = get_data_indirect offset2 + @length = @index_data.length / 8 + end + + # iterate through the property tuples + def each + length.times do |i| + key, type, value = handle_indirect_values(*@index_data[8 * i, 8].unpack('vvV')) + yield key, type, value + end + end + end + + # RawPropertyStoreTable is kind of like a database table. + # it has a fixed set of columns. + # #[] is kind of like getting a row from the table. + # those rows are currently encapsulated by Row, which has #each like + # RawPropertyStore. + # only used for the recipients array, and the attachments array. completely lazy, doesn't + # load any of the properties upon creation. + class RawPropertyStoreTable < BlockParser + class Column < Struct.new(:ref_type, :type, :ind2_off, :size, :slot) + def initialize data + super(*data.unpack('v3CC')) + end + + def nice_type_name + Mapi::Types::DATA[ref_type].first[/_(.*)/, 1].downcase rescue '0x%04x' % ref_type + end + + def nice_prop_name + Mapi::PropertyStore::TAGS['%04x' % type].first[/_(.*)/, 1].downcase rescue '0x%04x' % type + end + + def inspect + "#<#{self.class} name=#{nice_prop_name.inspect}, type=#{nice_type_name.inspect}>" + end + end + + include Enumerable + + attr_reader :length, :index_data, :data2, :data3, :rec_size + def initialize desc + super + raise FormatError, "expected type 2 - got #{@type}" unless @type == 2 + + header_data = get_data_indirect @offset1 + # seven_c_blk + # often: u1 == u2 and u3 == u2 + 2, then rec_size == u3 + 4. wtf + seven_c, @num_list, u1, u2, u3, @rec_size, b_five_offset, + ind2_offset, u7, u8 = header_data[0, 22].unpack('CCv4V2v2') + @index_data = header_data[22..-1] + + raise FormatError if @num_list != schema.length or seven_c != 0x7c + # another check + min_size = schema.inject(0) { |total, col| total + col.size } + # seem to have at max, 8 padding bytes on the end of the record. not sure if it means + # anything. maybe its just space that hasn't been reclaimed due to columns being + # removed or something. probably should just check lower bound. + range = (min_size..min_size + 8) + warn "rec_size seems wrong (#{range} !=== #{rec_size})" unless range === rec_size + + header_data2 = get_data_indirect b_five_offset + raise FormatError if header_data2.length < 8 + signature, offset2 = header_data2.unpack 'V2' + # ??? seems a bit iffy + # there's probably more to the differences than this, and the data2 difference below + expect = desc.pst.header.version_2003? ? 0x000404b5 : 0x000204b5 + raise FormatError, 'unhandled block signature 0x%08x' % signature if signature != expect + + # this holds all the row data + # handle multiple block issue. + @data3_io = get_data_indirect_io ind2_offset + if RangesIOIdxChain === @data3_io + @data3_idxs = + # modify ranges + ranges = @data3_io.ranges.map { |offset, size| [offset, size / @rec_size * @rec_size] } + @data3_io.instance_variable_set :@ranges, ranges + end + @data3 = @data3_io.read + + # there must be something to the data in data2. i think data2 is the array of objects essentially. + # currently its only used to imply a length + # actually, at size 6, its just some auxiliary data. i'm thinking either Vv/vV, for 97, and something + # wider for 03. the second value is just the index (0...length), and the first value is + # some kind of offset i expect. actually, they were all id2 values, in another case. + # so maybe they're get_data_indirect values too? + # actually, it turned out they were identical to the PR_ATTACHMENT_ID2 values... + # id2_values = ie, data2.unpack('v*').to_enum(:each_slice, 3).transpose[0] + # table[i].assoc(PR_ATTACHMENT_ID2).last == id2_values[i], for all i. + @data2 = get_data_indirect(offset2) rescue nil + #if data2 + # @length = (data2.length / 6.0).ceil + #else + # the above / 6, may have been ok for 97 files, but the new 0x0004 style block must have + # different size records... just use this instead: + # hmmm, actually, we can still figure it out: + @length = @data3.length / @rec_size + #end + + # lets try and at least use data2 for a warning for now + if data2 + data2_rec_size = desc.pst.header.version_2003? ? 8 : 6 + warn 'somthing seems wrong with data3' unless @length == (data2.length / data2_rec_size) + end + end + + def schema + @schema ||= index_data.scan(/.{8}/m).map { |data| Column.new data } + end + + def [] idx + # handle funky rounding + Row.new self, idx * @rec_size + end + + def each + length.times { |i| yield self[i] } + end + + class Row + include Enumerable + + def initialize array_parser, x + @array_parser, @x = array_parser, x + end + + # iterate through the property tuples + def each + (@array_parser.index_data.length / 8).times do |i| + ref_type, type, ind2_off, size, slot = @array_parser.index_data[8 * i, 8].unpack 'v3CC' + # check this rescue too + value = @array_parser.data3[@x + ind2_off, size] +# if INDIRECT_TYPES.include? ref_type + if size <= 4 + value = value.unpack('V')[0] + end + #p ['0x%04x' % ref_type, '0x%04x' % type, (Msg::Properties::MAPITAGS['%04x' % type].first[/^.._(.*)/, 1].downcase rescue nil), + # value_orig, value, (get_data_indirect(value_orig.unpack('V')[0]) rescue nil), size, ind2_off, slot] + key, type, value = @array_parser.handle_indirect_values type, ref_type, value + yield key, type, value + end + end + end + end + + class AttachmentTable < BlockParser + # a "fake" MAPI property name for this constant. if you get a mapi property with + # this value, it is the id2 value to use to get attachment data. + PR_ATTACHMENT_ID2 = 0x67f2 + + attr_reader :desc, :table + def initialize desc + @desc = desc + # no super, we only actually want BlockParser2#idx2 + @table = nil + return unless desc.list_index + return unless idx = idx2[ID2_ATTACHMENTS] + # FIXME make a fake desc. + @desc2 = OpenStruct.new :desc => idx, :pst => desc.pst, :list_index => desc.list_index + @table = RawPropertyStoreTable.new @desc2 + end + + def to_a + return [] if !table + table.map do |attachment| + attachment = attachment.to_a + #p attachment + # potentially merge with yet more properties + # this still seems pretty broken - especially the property overlap + if attachment_id2 = attachment.assoc(PR_ATTACHMENT_ID2) + #p attachment_id2.last + #p idx2[attachment_id2.last] + @desc2.desc = idx2[attachment_id2.last] + RawPropertyStore.new(@desc2).each do |a, b, c| + record = attachment.assoc a + attachment << record = [] unless record + record.replace [a, b, c] + end + end + attachment + end + end + end + + # there is no equivalent to this in libpst. ID2_RECIPIENTS was just guessed given the above + # AttachmentTable. + class RecipientTable < BlockParser + attr_reader :desc, :table + def initialize desc + @desc = desc + # no super, we only actually want BlockParser2#idx2 + @table = nil + return unless desc.list_index + return unless idx = idx2[ID2_RECIPIENTS] + # FIXME make a fake desc. + desc2 = OpenStruct.new :desc => idx, :pst => desc.pst, :list_index => desc.list_index + @table = RawPropertyStoreTable.new desc2 + end + + def to_a + return [] if !table + table.map { |x| x.to_a } + end + end + + # + # higher level item code. wraps up the raw properties above, and gives nice + # objects to work with. handles item relationships too. + # ---------------------------------------------------------------------------- + # + + def self.make_property_set property_list + hash = property_list.inject({}) do |hash, (key, type, value)| + hash.update PropertySet::Key.new(key) => value + end + PropertySet.new hash + end + + class Attachment < Mapi::Attachment + def initialize list + super Pst.make_property_set(list) + + @embedded_msg = props.attach_data if Item === props.attach_data + end + end + + class Recipient < Mapi::Recipient + def initialize list + super Pst.make_property_set(list) + end + end + + class Item < Mapi::Message + class EntryID < Struct.new(:u1, :entry_id, :id) + UNPACK_STR = 'VA16V' + + def initialize data + data = data.unpack(UNPACK_STR) if String === data + super(*data) + end + end + + include RecursivelyEnumerable + + attr_accessor :type, :parent + + def initialize desc, list, type=nil + @desc = desc + super Pst.make_property_set(list) + + # this is kind of weird, but the ids of the special folders are stored in a hash + # when the root item is loaded + if ipm_wastebasket_entryid + desc.pst.special_folder_ids[ipm_wastebasket_entryid] = :wastebasket + end + + if finder_entryid + desc.pst.special_folder_ids[finder_entryid] = :finder + end + + # and then here, those are used, along with a crappy heuristic to determine if we are an + # item +=begin +i think the low bits of the desc_id can give some info on the type. + +it seems that 0x4 is for regular messages (and maybe contacts etc) +0x2 is for folders, and 0x8 is for special things like rules etc, that aren't visible. +=end + unless type + type = props.valid_folder_mask || ipm_subtree_entryid || props.content_count || props.subfolders ? :folder : :message + if type == :folder + type = desc.pst.special_folder_ids[desc.desc_id] || type + end + end + + @type = type + end + + def each_child + id = ipm_subtree_entryid + if id + root = @desc.pst.desc_from_id id + raise "couldn't find root" unless root + raise 'both kinds of children' unless @desc.children.empty? + children = root.children + # lets look up the other ids we have. + # typically the wastebasket one "deleted items" is in the children already, but + # the search folder isn't. + extras = [ipm_wastebasket_entryid, finder_entryid].compact.map do |id| + root = @desc.pst.desc_from_id id + warn "couldn't find root for id #{id}" unless root + root + end.compact + # i do this instead of union, so as not to mess with the order of the + # existing children. + children += (extras - children) + children + else + @desc.children + end.each do |desc| + item = @desc.pst.pst_parse_item(desc) + item.parent = self + yield item + end + end + + def path + parents, item = [], self + parents.unshift item while item = item.parent + # remove root + parents.shift + parents.map { |item| item.props.display_name or raise 'unable to construct path' } * '/' + end + + def children + to_enum(:each_child).to_a + end + + # these are still around because they do different stuff + + # Top of Personal Folder Record + def ipm_subtree_entryid + @ipm_subtree_entryid ||= EntryID.new(props.ipm_subtree_entryid.read).id rescue nil + end + + # Deleted Items Folder Record + def ipm_wastebasket_entryid + @ipm_wastebasket_entryid ||= EntryID.new(props.ipm_wastebasket_entryid.read).id rescue nil + end + + # Search Root Record + def finder_entryid + @finder_entryid ||= EntryID.new(props.finder_entryid.read).id rescue nil + end + + # all these have been replaced with the method_missing below +=begin + # States which folders are valid for this message store + #def valid_folder_mask + # props[0x35df] + #end + + # Number of emails stored in a folder + def content_count + props[0x3602] + end + + # Has children + def subfolders + props[0x360a] + end +=end + + # i think i will change these, so they can inherit the lazyness from RawPropertyStoreTable. + # so if you want the last attachment, you can get it without creating the others perhaps. + # it just has to handle the no table at all case a bit more gracefully. + + def attachments + @attachments ||= AttachmentTable.new(@desc).to_a.map { |list| Attachment.new list } + end + + def recipients + #[] + @recipients ||= RecipientTable.new(@desc).to_a.map { |list| Recipient.new list } + end + + def each_recursive(&block) + #p :self => self + children.each do |child| + #p :child => child + block[child] + child.each_recursive(&block) + end + end + + def inspect + attrs = %w[display_name subject sender_name subfolders] +# attrs = %w[display_name valid_folder_mask ipm_wastebasket_entryid finder_entryid content_count subfolders] + str = attrs.map { |a| b = props.send a; " #{a}=#{b.inspect}" if b }.compact * ',' + + type_s = type == :message ? 'Message' : type == :folder ? 'Folder' : type.to_s.capitalize + 'Folder' + str2 = 'desc_id=0x%x' % @desc.desc_id + + !str.empty? ? "#<Pst::#{type_s} #{str2}#{str}>" : "#<Pst::#{type_s} #{str2} props=#{props.inspect}>" #\n" + props.transport_message_headers + ">" + end + end + + # corresponds to + # * _pst_parse_item + def pst_parse_item desc + Item.new desc, RawPropertyStore.new(desc).to_a + end + + # + # other random code + # ---------------------------------------------------------------------------- + # + + def dump_debug_info + puts "* pst header" + p header + +=begin +Looking at the output of this, for blank-o1997.pst, i see this part: +... +- (26624,516) desc block data (overlap of 4 bytes) +- (27136,516) desc block data (gap of 508 bytes) +- (28160,516) desc block data (gap of 2620 bytes) +... + +which confirms my belief that the block size for idx and desc is more likely 512 +=end + if 0 + 0 == 0 + puts '* file range usage' + file_ranges = + # these 3 things, should account for most of the data in the file. + [[0, Header::SIZE, 'pst file header']] + + @idx_offsets.map { |offset| [offset, Index::BLOCK_SIZE, 'idx block data'] } + + @desc_offsets.map { |offset| [offset, Desc::BLOCK_SIZE, 'desc block data'] } + + @idx.map { |idx| [idx.offset, idx.size, 'idx id=0x%x (%s)' % [idx.id, idx.type]] } + (file_ranges.sort_by { |idx| idx.first } + [nil]).to_enum(:each_cons, 2).each do |(offset, size, name), next_record| + # i think there is a padding of the size out to 64 bytes + # which is equivalent to padding out the final offset, because i think the offset is + # similarly oriented + pad_amount = 64 + warn 'i am wrong about the offset padding' if offset % pad_amount != 0 + # so, assuming i'm not wrong about that, then we can calculate how much padding is needed. + pad = pad_amount - (size % pad_amount) + pad = 0 if pad == pad_amount + gap = next_record ? next_record.first - (offset + size + pad) : 0 + extra = case gap <=> 0 + when -1; ["overlap of #{gap.abs} bytes)"] + when 0; [] + when +1; ["gap of #{gap} bytes"] + end + # how about we check that padding + @io.pos = offset + size + pad_bytes = @io.read(pad) + extra += ["padding not all zero"] unless pad_bytes == 0.chr * pad + puts "- #{offset}:#{size}+#{pad} #{name.inspect}" + (extra.empty? ? '' : ' [' + extra * ', ' + ']') + end + end + + # i think the idea of the idx, and indeed the idx2, is just to be able to + # refer to data indirectly, which means it can get moved around, and you just update + # the idx table. it is simply a list of file offsets and sizes. + # not sure i get how id2 plays into it though.... + # the sizes seem to be all even. is that a co-incidence? and the ids are all even. that + # seems to be related to something else (see the (id & 2) == 1 stuff) + puts '* idx entries' + @idx.each { |idx| puts "- #{idx.inspect}" } + + # if you look at the desc tree, you notice a few things: + # 1. there is a desc that seems to be the parent of all the folders, messages etc. + # it is the one whose parent is itself. + # one of its children is referenced as the subtree_entryid of the first desc item, + # the root. + # 2. typically only 2 types of desc records have idx2_id != 0. messages themselves, + # and the desc with id = 0x61 - the xattrib container. everything else uses the + # regular ids to find its data. i think it should be reframed as small blocks and + # big blocks, but i'll look into it more. + # + # idx_id and idx2_id are for getting to the data. desc_id and parent_desc_id just define + # the parent <-> child relationship, and the desc_ids are how the items are referred to in + # entryids. + # note that these aren't unique! eg for 0, 4 etc. i expect these'd never change, as the ids + # are stored in entryids. whereas the idx and idx2 could be a bit more volatile. + puts '* desc tree' + # make a dummy root hold everything just for convenience + root = Desc.new '' + def root.inspect; "#<Pst::Root>"; end + root.children.replace @orphans + # this still loads the whole thing as a string for gsub. should use directo output io + # version. + puts root.to_tree.gsub(/, (parent_desc_id|idx2_id)=0x0(?!\d)/, '') + + # this is fairly easy to understand, its just an attempt to display the pst items in a tree form + # which resembles what you'd see in outlook. + puts '* item tree' + # now streams directly + root_item.to_tree STDOUT + end + + def root_desc + @desc.first + end + + def root_item + item = pst_parse_item root_desc + item.type = :root + item + end + + def root + root_item + end + + # depth first search of all items + include Enumerable + + def each(&block) + root = self.root + block[root] + root.each_recursive(&block) + end + + def name + @name ||= root_item.props.display_name + end + + def inspect + "#<Pst name=#{name.inspect} io=#{io.inspect}>" + end +end +end + diff --git a/vendor/ruby-msg/lib/mapi/rtf.rb b/vendor/ruby-msg/lib/mapi/rtf.rb new file mode 100644 index 000000000..9fa133fac --- /dev/null +++ b/vendor/ruby-msg/lib/mapi/rtf.rb @@ -0,0 +1,169 @@ +require 'stringio' +require 'strscan' +require 'rtf' + +module Mapi + # + # = Introduction + # + # The +RTF+ module contains a few helper functions for dealing with rtf + # in mapi messages: +rtfdecompr+, and <tt>rtf2html</tt>. + # + # Both were ported from their original C versions for simplicity's sake. + # + module RTF + RTF_PREBUF = + "{\\rtf1\\ansi\\mac\\deff0\\deftab720{\\fonttbl;}" \ + "{\\f0\\fnil \\froman \\fswiss \\fmodern \\fscript " \ + "\\fdecor MS Sans SerifSymbolArialTimes New RomanCourier" \ + "{\\colortbl\\red0\\green0\\blue0\n\r\\par " \ + "\\pard\\plain\\f0\\fs20\\b\\i\\u\\tab\\tx" + + # Decompresses compressed rtf +data+, as found in the mapi property + # +PR_RTF_COMPRESSED+. Code converted from my C version, which in turn + # I wrote from a Java source, in JTNEF I believe. + # + # C version was modified to use circular buffer for back references, + # instead of the optimization of the Java version to index directly into + # output buffer. This was in preparation to support streaming in a + # read/write neutral fashion. + def rtfdecompr data + io = StringIO.new data + buf = RTF_PREBUF + "\x00" * (4096 - RTF_PREBUF.length) + wp = RTF_PREBUF.length + rtf = '' + + # get header fields (as defined in RTFLIB.H) + compr_size, uncompr_size, magic, crc32 = io.read(16).unpack 'V*' + #warn "compressed-RTF data size mismatch" unless io.size == data.compr_size + 4 + + # process the data + case magic + when 0x414c454d # "MELA" magic number that identifies the stream as a uncompressed stream + rtf = io.read uncompr_size + when 0x75465a4c # "LZFu" magic number that identifies the stream as a compressed stream + flag_count = -1 + flags = nil + while rtf.length < uncompr_size and !io.eof? + # each flag byte flags 8 literals/references, 1 per bit + flags = ((flag_count += 1) % 8 == 0) ? io.getc : flags >> 1 + if 1 == (flags & 1) # each flag bit is 1 for reference, 0 for literal + rp, l = io.getc, io.getc + # offset is a 12 byte number. 2^12 is 4096, so thats fine + rp = (rp << 4) | (l >> 4) # the offset relative to block start + l = (l & 0xf) + 2 # the number of bytes to copy + l.times do + rtf << buf[wp] = buf[rp] + wp = (wp + 1) % 4096 + rp = (rp + 1) % 4096 + end + else + rtf << buf[wp] = io.getc + wp = (wp + 1) % 4096 + end + end + else # unknown magic number + raise "Unknown compression type (magic number 0x%08x)" % magic + end + + # not sure if its due to a bug in the above code. doesn't seem to be + # in my tests, but sometimes there's a trailing null. we chomp it here, + # which actually makes the resultant rtf smaller than its advertised + # size (+uncompr_size+). + rtf.chomp! 0.chr + rtf + end + + # Note, this is a conversion of the original C code. Not great - needs tests and + # some refactoring, and an attempt to correct some inaccuracies. Hacky but works. + # + # Returns +nil+ if it doesn't look like an rtf encapsulated rtf. + # + # Some cases that the original didn't deal with have been patched up, eg from + # this chunk, where there are tags outside of the htmlrtf ignore block. + # + # "{\\*\\htmltag116 <br />}\\htmlrtf \\line \\htmlrtf0 \\line {\\*\\htmltag84 <a href..." + # + # We take the approach of ignoring all rtf tags not explicitly handled. A proper + # parse tree would be nicer to work with. will need to look for ruby rtf library + # + # Some of the original comment to the c code is excerpted here: + # + # Sometimes in MAPI, the PR_BODY_HTML property contains the HTML of a message. + # But more usually, the HTML is encoded inside the RTF body (which you get in the + # PR_RTF_COMPRESSED property). These routines concern the decoding of the HTML + # from this RTF body. + # + # An encoded htmlrtf file is a valid RTF document, but which contains additional + # html markup information in its comments, and sometimes contains the equivalent + # rtf markup outside the comments. Therefore, when it is displayed by a plain + # simple RTF reader, the html comments are ignored and only the rtf markup has + # effect. Typically, this rtf markup is not as rich as the html markup would have been. + # But for an html-aware reader (such as the code below), we can ignore all the + # rtf markup, and extract the html markup out of the comments, and get a valid + # html document. + # + # There are actually two kinds of html markup in comments. Most of them are + # prefixed by "\*\htmltagNNN", for some number NNN. But sometimes there's one + # prefixed by "\*\mhtmltagNNN" followed by "\*\htmltagNNN". In this case, + # the two are equivalent, but the m-tag is for a MIME Multipart/Mixed Message + # and contains tags that refer to content-ids (e.g. img src="cid:072344a7") + # while the normal tag just refers to a name (e.g. img src="fred.jpg") + # The code below keeps the m-tag and discards the normal tag. + # If there are any m-tags like this, then the message also contains an + # attachment with a PR_CONTENT_ID property e.g. "072344a7". Actually, + # sometimes the m-tag is e.g. img src="http://outlook/welcome.html" and the + # attachment has a PR_CONTENT_LOCATION "http://outlook/welcome.html" instead + # of a PR_CONTENT_ID. + # + def rtf2html rtf + scan = StringScanner.new rtf + # require \fromhtml. is this worth keeping? apparently you see \\fromtext if it + # was converted from plain text. + return nil unless rtf["\\fromhtml"] + html = '' + ignore_tag = nil + # skip up to the first htmltag. return nil if we don't ever find one + return nil unless scan.scan_until /(?=\{\\\*\\htmltag)/ + until scan.empty? + if scan.scan /\{/ + elsif scan.scan /\}/ + elsif scan.scan /\\\*\\htmltag(\d+) ?/ + #p scan[1] + if ignore_tag == scan[1] + scan.scan_until /\}/ + ignore_tag = nil + end + elsif scan.scan /\\\*\\mhtmltag(\d+) ?/ + ignore_tag = scan[1] + elsif scan.scan /\\par ?/ + html << "\r\n" + elsif scan.scan /\\tab ?/ + html << "\t" + elsif scan.scan /\\'([0-9A-Za-z]{2})/ + html << scan[1].hex.chr + elsif scan.scan /\\pntext/ + scan.scan_until /\}/ + elsif scan.scan /\\htmlrtf/ + scan.scan_until /\\htmlrtf0 ?/ + # a generic throw away unknown tags thing. + # the above 2 however, are handled specially + elsif scan.scan /\\[a-z-]+(\d+)? ?/ + #elsif scan.scan /\\li(\d+) ?/ + #elsif scan.scan /\\fi-(\d+) ?/ + elsif scan.scan /[\r\n]/ + elsif scan.scan /\\([{}\\])/ + html << scan[1] + elsif scan.scan /(.)/ + html << scan[1] + else + p :wtf + end + end + html.strip.empty? ? nil : html + end + + module_function :rtf2html, :rtfdecompr + end +end + diff --git a/vendor/ruby-msg/lib/mapi/types.rb b/vendor/ruby-msg/lib/mapi/types.rb new file mode 100644 index 000000000..71416afd5 --- /dev/null +++ b/vendor/ruby-msg/lib/mapi/types.rb @@ -0,0 +1,51 @@ +require 'rubygems' +require 'ole/types' + +module Mapi + Log = Logger.new_with_callstack + + module Types + # + # Mapi property types, taken from http://msdn2.microsoft.com/en-us/library/bb147591.aspx. + # + # The fields are [mapi name, variant name, description]. Maybe I should just make it a + # struct. + # + # seen some synonyms here, like PT_I8 vs PT_LONG. seen stuff like PT_SRESTRICTION, not + # sure what that is. look at `grep ' PT_' data/mapitags.yaml | sort -u` + # also, it has stuff like PT_MV_BINARY, where _MV_ probably means multi value, and is + # likely just defined to | in 0x1000. + # + # Note that the last 2 are the only ones where the Mapi value differs from the Variant value + # for the corresponding variant type. Odd. Also, the last 2 are currently commented out here + # because of the clash. + # + # Note 2 - the strings here say VT_BSTR, but I don't have that defined in Ole::Types. Should + # maybe change them to match. I've also seen reference to PT_TSTRING, which is defined as some + # sort of get unicode first, and fallback to ansii or something. + # + DATA = { + 0x0001 => ['PT_NULL', 'VT_NULL', 'Null (no valid data)'], + 0x0002 => ['PT_SHORT', 'VT_I2', '2-byte integer (signed)'], + 0x0003 => ['PT_LONG', 'VT_I4', '4-byte integer (signed)'], + 0x0004 => ['PT_FLOAT', 'VT_R4', '4-byte real (floating point)'], + 0x0005 => ['PT_DOUBLE', 'VT_R8', '8-byte real (floating point)'], + 0x0006 => ['PT_CURRENCY', 'VT_CY', '8-byte integer (scaled by 10,000)'], + 0x000a => ['PT_ERROR', 'VT_ERROR', 'SCODE value; 32-bit unsigned integer'], + 0x000b => ['PT_BOOLEAN', 'VT_BOOL', 'Boolean'], + 0x000d => ['PT_OBJECT', 'VT_UNKNOWN', 'Data object'], + 0x001e => ['PT_STRING8', 'VT_BSTR', 'String'], + 0x001f => ['PT_UNICODE', 'VT_BSTR', 'String'], + 0x0040 => ['PT_SYSTIME', 'VT_DATE', '8-byte real (date in integer, time in fraction)'], + #0x0102 => ['PT_BINARY', 'VT_BLOB', 'Binary (unknown format)'], + #0x0102 => ['PT_CLSID', 'VT_CLSID', 'OLE GUID'] + } + + module Constants + DATA.each { |num, (mapi_name, variant_name, desc)| const_set mapi_name, num } + end + + include Constants + end +end + diff --git a/vendor/ruby-msg/lib/mime.rb b/vendor/ruby-msg/lib/mime.rb new file mode 100644 index 000000000..4340e4901 --- /dev/null +++ b/vendor/ruby-msg/lib/mime.rb @@ -0,0 +1,165 @@ +# +# = Introduction +# +# A *basic* mime class for _really_ _basic_ and probably non-standard parsing +# and construction of MIME messages. +# +# Intended for two main purposes in this project: +# 1. As the container that is used to build up the message for eventual +# serialization as an eml. +# 2. For assistance in parsing the +transport_message_headers+ provided in .msg files, +# which are then kept through to the final eml. +# +# = TODO +# +# * Better streaming support, rather than an all-in-string approach. +# * Add +OrderedHash+ optionally, to not lose ordering in headers. +# * A fair bit remains to be done for this class, its fairly immature. But generally I'd like +# to see it be more generally useful. +# * All sorts of correctness issues, encoding particular. +# * Duplication of work in net/http.rb's +HTTPHeader+? Don't know if the overlap is sufficient. +# I don't want to lower case things, just for starters. +# * Mime was the original place I wrote #to_tree, intended as a quick debug hack. +# +class SimpleMime + Hash = begin + require 'orderedhash' + OrderedHash + rescue LoadError + Hash + end + + attr_reader :headers, :body, :parts, :content_type, :preamble, :epilogue + + # Create a SimpleMime object using +str+ as an initial serialization, which must contain headers + # and a body (even if empty). Needs work. + def initialize str, ignore_body=false + headers, @body = $~[1..-1] if str[/(.*?\r?\n)(?:\r?\n(.*))?\Z/m] + + @headers = Hash.new { |hash, key| hash[key] = [] } + @body ||= '' + headers.to_s.scan(/^\S+:\s*.*(?:\n\t.*)*/).each do |header| + @headers[header[/(\S+):/, 1]] << header[/\S+:\s*(.*)/m, 1].gsub(/\s+/m, ' ').strip # this is kind of wrong + end + + # don't have to have content type i suppose + @content_type, attrs = nil, {} + if content_type = @headers['Content-Type'][0] + @content_type, attrs = SimpleMime.split_header content_type + end + + return if ignore_body + + if multipart? + if body.empty? + @preamble = '' + @epilogue = '' + @parts = [] + else + # we need to split the message at the boundary + boundary = attrs['boundary'] or raise "no boundary for multipart message" + + # splitting the body: + parts = body.split(/--#{Regexp.quote boundary}/m) + unless parts[-1] =~ /^--/; warn "bad multipart boundary (missing trailing --)" + else parts[-1][0..1] = '' + end + parts.each_with_index do |part, i| + part =~ /^(\r?\n)?(.*?)(\r?\n)?\Z/m + part.replace $2 + warn "bad multipart boundary" if (1...parts.length-1) === i and !($1 && $3) + end + @preamble = parts.shift + @epilogue = parts.pop + @parts = parts.map { |part| SimpleMime.new part } + end + end + end + + def multipart? + @content_type && @content_type =~ /^multipart/ ? true : false + end + + def inspect + # add some extra here. + "#<SimpleMime content_type=#{@content_type.inspect}>" + end + + def to_tree + if multipart? + str = "- #{inspect}\n" + parts.each_with_index do |part, i| + last = i == parts.length - 1 + part.to_tree.split(/\n/).each_with_index do |line, j| + str << " #{last ? (j == 0 ? "\\" : ' ') : '|'}" + line + "\n" + end + end + str + else + "- #{inspect}\n" + end + end + + def to_s opts={} + opts = {:boundary_counter => 0}.merge opts + if multipart? + boundary = SimpleMime.make_boundary opts[:boundary_counter] += 1, self + @body = [preamble, parts.map { |part| "\r\n" + part.to_s(opts) + "\r\n" }, "--\r\n" + epilogue]. + flatten.join("\r\n--" + boundary) + content_type, attrs = SimpleMime.split_header @headers['Content-Type'][0] + attrs['boundary'] = boundary + @headers['Content-Type'] = [([content_type] + attrs.map { |key, val| %{#{key}="#{val}"} }).join('; ')] + end + + str = '' + @headers.each do |key, vals| + vals.each { |val| str << "#{key}: #{val}\r\n" } + end + str << "\r\n" + @body + end + + def self.split_header header + # FIXME: haven't read standard. not sure what its supposed to do with " in the name, or if other + # escapes are allowed. can't test on windows as " isn't allowed anyway. can be fixed with more + # accurate parser later. + # maybe move to some sort of Header class. but not all headers should be of it i suppose. + # at least add a join_header then, taking name and {}. for use in SimpleMime#to_s (for boundary + # rewrite), and Attachment#to_mime, among others... + attrs = {} + header.scan(/;\s*([^\s=]+)\s*=\s*("[^"]*"|[^\s;]*)\s*/m).each do |key, value| + if attrs[key]; warn "ignoring duplicate header attribute #{key.inspect}" + else attrs[key] = value[/^"/] ? value[1..-2] : value + end + end + + [header[/^[^;]+/].strip, attrs] + end + + # +i+ is some value that should be unique for all multipart boundaries for a given message + def self.make_boundary i, extra_obj = SimpleMime + "----_=_NextPart_#{'%03d' % i}_#{'%08x' % extra_obj.object_id}.#{'%08x' % Time.now}" + end +end + +=begin +things to consider for header work. +encoded words: +Subject: =?iso-8859-1?q?p=F6stal?= + +and other mime funkyness: +Content-Disposition: attachment; + filename*0*=UTF-8''09%20%D7%90%D7%A5; + filename*1*=%20%D7%A1%D7%91-; + filename*2*=%D7%A7%95%A5.wma +Content-Transfer-Encoding: base64 + +and another, doing a test with an embedded newline in an attachment name, I +get this output from evolution. I get the feeling that this is probably a bug +with their implementation though, they weren't expecting new lines in filenames. +Content-Disposition: attachment; filename="asdf'b\"c +d efgh=i: ;\\j" +d efgh=i: ;\\j"; charset=us-ascii +Content-Type: text/plain; name="asdf'b\"c"; charset=us-ascii + +=end + diff --git a/vendor/ruby-msg/lib/orderedhash.rb b/vendor/ruby-msg/lib/orderedhash.rb new file mode 100644 index 000000000..16a4f5860 --- /dev/null +++ b/vendor/ruby-msg/lib/orderedhash.rb @@ -0,0 +1,218 @@ +# = OrderedHash +# +# == Version +# 1.2006.07.13 (change of the first number means Big Change) +# +# == Description +# Hash which preserves order of added items (like PHP array). +# +# == Usage +# +# (see examples directory under the ruby gems root directory) +# +# require 'rubygems' +# require 'ordered_hash' +# +# hsh = OrderedHash.new +# hsh['z'] = 1 +# hsh['a'] = 2 +# hsh['c'] = 3 +# p hsh.keys # ['z','a','c'] +# +# == Source +# http://simplypowerful.1984.cz/goodlibs/1.2006.07.13 +# +# == Author +# jan molic (/mig/at_sign/1984/dot/cz/) +# +# == Thanks to +# Andrew Johnson for his suggestions and fixes of Hash[], merge, to_a, inspect and shift +# Desmond Dsouza for == fixes +# +# == Licence +# You can redistribute it and/or modify it under the same terms of Ruby's license; +# either the dual license version in 2003, or any later version. +# + +class OrderedHash < Hash + + attr_accessor :order + + class << self + + def [] *args + hsh = OrderedHash.new + if Hash === args[0] + hsh.replace args[0] + elsif (args.size % 2) != 0 + raise ArgumentError, "odd number of elements for Hash" + else + hsh[args.shift] = args.shift while args.size > 0 + end + hsh + end + + end + + def initialize(*a, &b) + super + @order = [] + end + + def store_only a,b + store a,b + end + + alias orig_store store + + def store a,b + @order.push a unless has_key? a + super a,b + end + + alias []= store + + def == hsh2 + return hsh2==self if !hsh2.is_a?(OrderedHash) + return false if @order != hsh2.order + super hsh2 + end + + def clear + @order = [] + super + end + + def delete key + @order.delete key + super + end + + def each_key + @order.each { |k| yield k } + self + end + + def each_value + @order.each { |k| yield self[k] } + self + end + + def each + @order.each { |k| yield k,self[k] } + self + end + + alias each_pair each + + def delete_if + @order.clone.each { |k| + delete k if yield + } + self + end + + def values + ary = [] + @order.each { |k| ary.push self[k] } + ary + end + + def keys + @order + end + + def invert + hsh2 = Hash.new + @order.each { |k| hsh2[self[k]] = k } + hsh2 + end + + def reject &block + self.dup.delete_if( &block ) + end + + def reject! &block + hsh2 = reject( &block ) + self == hsh2 ? nil : hsh2 + end + + def replace hsh2 + @order = hsh2.keys + super hsh2 + end + + def shift + key = @order.first + key ? [key,delete(key)] : super + end + + def unshift k,v + unless self.include? k + @order.unshift k + orig_store(k,v) + true + else + false + end + end + + def push k,v + unless self.include? k + @order.push k + orig_store(k,v) + true + else + false + end + end + + def pop + key = @order.last + key ? [key,delete(key)] : nil + end + + def first + self[@order.first] + end + + def last + self[@order.last] + end + + def to_a + ary = [] + each { |k,v| ary << [k,v] } + ary + end + + def to_s + self.to_a.to_s + end + + def inspect + ary = [] + each {|k,v| ary << k.inspect + "=>" + v.inspect} + '{' + ary.join(", ") + '}' + end + + def update hsh2 + hsh2.each { |k,v| self[k] = v } + self + end + + alias :merge! update + + def merge hsh2 + self.dup update(hsh2) + end + + def select + ary = [] + each { |k,v| ary << [k,v] if yield k,v } + ary + end + +end + +#=end diff --git a/vendor/ruby-msg/lib/rtf.rb b/vendor/ruby-msg/lib/rtf.rb new file mode 100755 index 000000000..3afac68a8 --- /dev/null +++ b/vendor/ruby-msg/lib/rtf.rb @@ -0,0 +1,109 @@ +require 'stringio' + +# this file is pretty crap, its just to ensure there is always something readable if +# there is an rtf only body, with no html encapsulation. + +module RTF + class Tokenizer + def self.process io + while true do + case c = io.getc + when ?{; yield :open_group + when ?}; yield :close_group + when ?\\ + case c = io.getc + when ?{, ?}, ?\\; yield :text, c.chr + when ?'; yield :text, [io.read(2)].pack('H*') + when ?a..?z, ?A..?Z + # read control word + str = c.chr + str << c while c = io.read(1) and c =~ /[a-zA-Z]/ + neg = 1 + neg = -1 and c = io.read(1) if c == '-' + num = if c =~ /[0-9]/ + num = c + num << c while c = io.read(1) and c =~ /[0-9]/ + num.to_i * neg + end + raise "invalid rtf stream" if neg == -1 and !num # ???? \blahblah- some text + io.seek(-1, IO::SEEK_CUR) if c != ' ' + yield :control_word, str, num + when nil + raise "invalid rtf stream" # \EOF + else + # other kind of control symbol + yield :control_symbol, c.chr + end + when nil + return + when ?\r, ?\n + # ignore + else yield :text, c.chr + end + end + end + end + + class Converter + # crappy + def self.rtf2text str, format=:text + group = 0 + text = '' + text << "<html>\n<body>" if format == :html + group_type = [] + group_tags = [] + RTF::Tokenizer.process(StringIO.new(str)) do |a, b, c| + add_text = '' + case a + when :open_group; group += 1; group_type[group] = nil; group_tags[group] = [] + when :close_group; group_tags[group].reverse.each { |t| text << "</#{t}>" }; group -= 1; + when :control_word; # ignore + group_type[group] ||= b + # maybe change this to use utf8 where possible + add_text = if b == 'par' || b == 'line' || b == 'page'; "\n" + elsif b == 'tab' || b == 'cell'; "\t" + elsif b == 'endash' || b == 'emdash'; "-" + elsif b == 'emspace' || b == 'enspace' || b == 'qmspace'; " " + elsif b == 'ldblquote'; '"' + else '' + end + if b == 'b' || b == 'i' and format == :html + close = c == 0 ? '/' : '' + text << "<#{close}#{b}>" + if c == 0 + group_tags[group].delete b + else + group_tags[group] << b + end + end + # lot of other ones belong in here.\ +=begin +\bullet Bullet character. +\lquote Left single quotation mark. +\rquote Right single quotation mark. +\ldblquote Left double quotation mark. +\rdblquote +=end + when :control_symbol; # ignore + group_type[group] ||= b + add_text = ' ' if b == '~' # non-breakable space + add_text = '-' if b == '_' # non-breakable hypen + when :text + add_text = b if group <= 1 or group_type[group] == 'rtlch' && !group_type[0...group].include?('*') + end + if format == :html + text << add_text.gsub(/([<>&"'])/) do + ent = { '<' => 'lt', '>' => 'gt', '&' => 'amp', '"' => 'quot', "'" => 'apos' }[$1] + "&#{ent};" + end + text << '<br>' if add_text == "\n" + else + text << add_text + end + end + text << "</body>\n</html>\n" if format == :html + text + end + end +end + |