This version differs only in the changes required for Ruby 1.9 support. It is otherwise "bug compatible" with version 0.17. Users of version 0.17 can safely upgrade to 1.0.0. You will lose the ability to run the tests if you are using Ruby 1.6 (but who is doing that these days?).
Add a Rakefile. Add a gem hosted on rubyforge.org.
Handle parsing addresses with non-ASCII display names. We don't do anything intelligent with the non-ASCII data, but it is passed through in its original (invalid) non-encoded form.
Look for From_ lines only on the first line when parsing a message.
RMail::Header#subject= now sets the header field name to "Subject" instead of "subject" since some mail reading software is case sensitive and can't understand "subject: foo".
The RMail parser now requires only whitespace after MIME multipart boundary lines. This violates RFC2046's "NOTE TO IMPLEMENTORS" documented in testparsermultipart.rb's test_multipart_data_12, but it allows real-world messages created by Eudora to be parsed.
Ruby 1.8.1 compatibility fixes only.
Ruby 1.8.0 compatibility fixes only.
Improve RDoc documentation of the library.
Add file level comments.
No longer document RMail::Parser::PushbackReader and RMail::Parser::Multipart, as they are used only internally or by those who really want to dig deep into the library. I also think they might be changing soon.
Add ability to
require 'rmail'
and get all of RubyMail.
Switch to the BSD license, mainly so the license of RubyMail is not confusing.
Add convenience methods to RMail::Header that allow easy manipulation of the To, Cc, Bcc, Reply-To, Date, Subject and Message-Id fields. This includes unique Message-Id generation code and robust Date field parsing and formatting.
Add RMail::Header#set, to delete existing fields of the same name and then add a new one.
RMail::Address.parse now returns an RMail::Address::List instead of an Array (existing code still works, since RMail::Address::List inherits from Array).
Add RMail::Address#<=>, RMail::Address#hash and RMail::Address#eql? methods. This allows arrays of RMail::Address to be sorted as well as allowing RMail::Address to be placed in a hash (which in turn allows Array#uniq to work when holding RMail::Address objects).
Add an RMail::Mailbox::MBoxReader#each_message method.
Deleted the rmail/header/field.rb field and incorporated RMail::Header::Field into rmail/header.rb.
Improve uniqueness of the MIME boundary generation by including Time.now.tv_usec.
Correct broken docs for RMail::Header#match and RMail::Header#match?
Add an install.rb script.
parse->serialize "transparency" greatly improved. This means that when you parse a message and then serialize it back out you almost always get the exact same message back. This is true for all single part messages as well as all validly formatted multipart MIME messages (and even the most common invalidly formatted ones).
The result is that RMail can now be used safely in mail filters without risking damaging cryptographic signatures in the mails.
RMail::Header#add now uses to_str instead of to_s to convert arguments to strings. This makes it behave more like standard Ruby classes.
RMail::Mailbox::MBoxReader now always makes sure the last piece of data returned for each message is the end of line terminator, even if one isn't present in the input.
RMail::Parser::PushbackReader#read now takes nil argument to mean "read all available input." Derived classes should now override the read_chunk method instead of read.
Added rmail/mailbox/mboxreader.rb to the distribution.
New RMail::Mailbox.parse_mbox method that can be used to conveniently read raw messages out of a Unix mbox mailbox.
New RMail::Mailbox::MBoxReader class in rmail/mailbox/mboxreader.rb. This class can be used to easily read messages out of a file in Unix "mbox" format.
The RMail::Parser::PushbackReader class has been documented. It has moved out of rmail/parser/multipart.rb into rmail/parser/pushbackreader.rb
Various documentation fixes. E.g. RMail::Parser#parse can take a string as well as an IO object.
The RMail::Parser::PushbackReader has a setable chunk size. This is useful mostly for testing.
Fix an uncaught exception when parsing multipart MIME messages that contain only a preamble and an epilogue but no body parts.
Fix a bug where RMail::Parser.multipart? would not return true if the multipart message actually didn't have any parts.
The following has been removed from RubyMail and made part of the RubyFilter package:
All scripts that were in the RubyMail 0.7 bin directory.
Mail::LDA
Mail::Deliver
Mail::KeyedMailbox
Mail::MTA
Mail::AddressTagger
This keeps RubyMail a small and simple mail package, and provides RubyFilter as an example of how to use RubyMail to write a mail filter.
The Mail module has been renamed to RMail. I think "Mail" should be reserved for things included in the standard distribution of Ruby.
RMail::Header#match and match? don't require the name or value arguments to be a case insensitive Regexps. Also, when the value argument is converted to a string, it is passed through Regexp::escape first.
RMail::Parser#parse can now parse from a string.
Mail::Address#comments= can now take a simpple string to set just one comment.
A new chunked input scheme that makes parsing huge messages about 7 times faster in ruby 1.7 and 50 times faster in ruby 1.6.
When parsing a huge message that has a 2 megabyte attachment, RubyMail running under ruby 1.7 is now faster than any email package for ruby, perl or python.
I wrote a benchmark that reads a 2 megabyte email from a file and writes it out again, doing this 100 times. The results are:
ruby 1.7.2 w/rubymail (100 times)
5.96s user 7.98s system 13.94s total
ruby 1.6.6 w/rubymail (100 times)
76.91s user 8.62s system 85.53s total
ruby 1.7.2 w/tmail 0.10.1 (100 times)
9.85s user 24.21s system 34.06s total
ruby 1.6.6 w/tmail 0.10.0 (100 times)
201.89s user 15.75s system 217.64s total
python 2.2 w/email (100 times)
76.73 user 15.16s system 91.89s total
perl 5.006001 w/mimetools 5.411 (parsing on disk) (100 times)
190.11s user 25.25s system 215.36s total
perl 5.006001 w/mimetools 5.411 (parsing in memory) (100 times)
962.69s user 6.77s system 969.46s total
This change also paves the way for streaming large messages to disk when they start to get huge, so RAM isn't needlessly used up.
Delivery to mbox files improved.
won't sleep forever waiting for the mailbox lock (flock)
won't delivery to files that don't look like an mbox (e.g. it is not a file, not in the right format).
delivering to '/dev/null' is now a simple nop.
Use the new File#fsync method when available in all of the mail delivery functions.
Add bin/rsendmail.rb as another example of how I'm using RubyMail.
A new Mail::AddressTagger class, included for fun. Requires the hmac-sha1 class (available in the ruby-hmac package on RAA).
Multipart MIME support.
A new Mail::AddressTagger class, for tagging addresses with cryptographically verifiable extensions akin to TMDA. Requires the hmac-sha1 module from the RAA. (experimental, I don't currently use this)
A new Mail::Message#== method.
A new Mail::Serialize class that can serialize a Mail::Message.
The rdeliver.rb script is now fully documented.
The rdeliver.rb script now evaluates the .rdeliver file in the context of a Deliver class (as opposed to simply loading it). The .rdeliver file must now define a Deliver#main method, where the simplest .rdeliver file would be:
def main lda.save("INBOX") end
Add a KeyedMailbox class that can be used to implement simple mailing list style confirmations.
Add a message= method to Mail::DeliveryAgent. This lets delivery scripts change the message being delivered.
Re-wrote the RFC2822 address parser. It is now more strict about what it considers a valid address.
The bin/rdeliver.rb script is now tested. Got rid of bin/rdeliver-wrap.rb.
Mail::DeliveryAgent::DeliveryStatus renamed to Mail::DeliveryAgent::DeliveryComplete. Mail::DeliveryAgent::DeliverFailure renamed to Mail::DeliveryAgent::DeliveryPipeFailure.
Mail::Deliver.deliver_mbox now uses File::SYNC to write the message.
Mail::Header and Mail::Message no longer understand how to parse messages. Message parsing has been moved to Mail::Parser. Mail::Parser uses the public API of Mail::Header and Mail::Message to build up the message.
The Mail::Header API has been greatly changed. It is now more like Array and Hash.
Mail::Deliver supports delivery to qmail style Maildir mailboxes. Mail::DeliveryAgent#save will now deliver to a Maildir if the folder's name ends in a slash. E.g. "/home/user/Maildir/".
Mail::DeliveryAgent no longer logs an abstract of the message being delivered. All logging is up to the users of Mail::DeliveryAgent.
Add Mail::Header.length and Mail::Header.size methods. Add Mail::Header.match and Mail::Header.match? methods.
Move deliver.rb to bin/rdeliver-wrap.rb and main.rb to bin/rdeliver.rb. These are workable local delivery agent scripts (suitable for running from .forward).
New Mail::MTA module that provides constants for exit codes.
New features for Mail::DeliveryAgent. Now Mail::DeliveryAgent never calls exit, instead it throws DeliveryStatus exceptions. There is also a new Mail::DeliveryAgent.process method that allows you to use Mail::DeliveryAgent in block form. Mail:DeliveryAgent.exitcode will return the correct exit code for a given DeliveryStatus exception.
HTML API documentation is now available in the doc subdirectory.
Mail::DeliveryAgent::strip_tag is now Mail::DeliveryAgent::strip_field_name.
Mail::Deliver::deliver_pipe implemented. Mail::DeliveryAgent#pipe implemented.
Mail::DeliveryAgent#pipe and Mail::DeliveryAgent#save now report success or failure with DeliverySuccess and DeliveryFailure exceptions. Mail::DeliveryAgent#reject and Mail::DeliveryAgent#defer do not yet use the exceptions (they still call exit).
Now runs clean under "ruby -w"
Now works with newer rubyunit versions.