How to handle UTF-8 email headers (like Subject:) using Ruby?

Question

I\'m an email n00b but I am working on an application that sends HTML email with Unicode characters (as my friend noted \"enjoy encoding hell\").

The Subject:<

Answer 1

Ahah! ActionMailer::Quoting has a quoted_printable method.

So here's what I did:

def my_email(foo)
  ...
  @subject = quoted_printable(foo.some_subject_with_accented_chars, 'utf-8')
  ...
end

Doing this convinced Mail.app to display the rest of the email using UTF-8. Now to test the rest!

Answer 2

Since none of the answers tells about whole message with pure Ruby, here it is.

Net::SMTP.start("localhost") do |smtp|
    smtp.open_message_stream opts[:sender_address], opts[:receiver_address] do |f|

        f.puts "Content-type: text/plain; charset=UTF-8"
        f.puts from
        f.puts to
        f.puts subject
        f.puts message
    end
end

Here you open connection to localhost. Using external SMTP server is also possible, refer to documentation of net/smtp.

First line sets character set used in the message. Rest of the lines are variables defined separately:

from is an address in form of From: Name here <address@here.fi>. If no name is wanted, only address can be specified, like From: address@here.fi.

to uses same syntax, with exception From: changed to To:.

subject is in form of Subject: subject here. For UTF-8, it needs to be Base64-encoded to be shown correctly for clients.

subject = "Subject: =?UTF-8?B?" + Base64.strict_encode64(subject) + "?="

Message is plain text message encoded in UTF-8, without any prefix. net/smtp will take care of forming proper mail.

Answer 3

Your can optionally do the same using Base64 encoding:

require "base64"

value  = Base64.encode64("Your UTF-8 string")
header = "=?UTF-8?B?" + value + "?="

Note the "B", that marks Base64 encoded payload, as opposed to the "Q" which marks "Q-encoded" payload. The latter could be faked by URL-encoding the string and replacing all "%" characters with "=".

By "faking" I mean this: It would produce a valid result, but maybe more characters are encoded than would be necessary. The spec allows for encoding every character there is with "=" + ByteCodeAsHex, it merely impairs human readability of the raw headers. UrlEncode is + .gsub(/%/, "=") not a bad compromise when nothing else is available.