Superfast email header parser in Ruby, using regular expressions. This solution is 250 times faster than using the "Mail" gem. :O Time with my regex: 0.063965 seconds Time with Mail gem: 16.327161 seconds Note that I included some encoding-fix code. YMMV and encoding fixes are all debatable or fail in some corner case.

email_header_parser.rb

require 'ap'
require 'mail'

# String monkeypatch
# This is one of many possible "encoding problem" solutions. It's actually an intractable problem
# but you'd have to read "Gödel, Escher, Bach" to understand why...
class String
  def clean_utf8
    # self.force_encoding("UTF-8").encode("UTF-16BE", :invalid=>:replace, :replace=>"?").encode("UTF-8")
    unpack('C*').pack('U*') if !valid_encoding?
  end
end

module Parsing
  module Email
    class Header
      EMAIL_HEADER_PARSER_REGEX = /
        ([A-Za-z-]+):\s       # Find a header key, which ends in a colon and a space. Capture the hyphenated word portion.
        (                     # Now start capturing the value.
          [^\r\n]+            # First, match everything that is not a line ending char.
          (?:                 # Then start a non-capturing repeating match which first consists of...
            \r?\n             # A line ending combo...
            (?![A-Za-z-]+:\s) # But first, do a negative lookahead to make sure the next line does not start with another header-key-looking string
            [^\r\n]+          # Then match all text that is not a line ending char.
          )*                  # Repeat this non-capturing group match 0 or more times.
        )                     # End capturing the value.
      /mx                     # Allow matches to cross line endings (m) and allow whitespace and comments in this regex (x).
      EMAIL_MULTILINE_HEADER_VALUE_REGEX = /\r?\n\s*/m

      attr_reader :headers
      def initialize(opts = {})
        @headers = (Hash === opts ? opts[:headers] : opts)
      end

      def call
        return {} unless @headers
        clean_header_values_hash(headers_hash)
      end
      alias to_h call

      private

      def headers_hash
        h = {}
        @headers.scan(EMAIL_HEADER_PARSER_REGEX).map do |k,v|
          if k && v
            if h[k]
              # if this key already has a value, wrap it in an array and append the new value, otherwise just set the key value
              h[k] = [h[k]] unless Array === h[k]
              h[k] << v
            else
              h[k] = v
            end
          end
        end
        h
      end

      # Clean up runs of a newline followed by whitespace in header values by replacing with a space
      def clean_header_values_hash(h)
        h.each do |k,v|
          if Array === v
            v.map! do |val|
              val =~ EMAIL_MULTILINE_HEADER_VALUE_REGEX ? val.gsub!(EMAIL_MULTILINE_HEADER_VALUE_REGEX,' ') : val
            end
          else
            v.gsub!(EMAIL_MULTILINE_HEADER_VALUE_REGEX, ' ') if v =~ EMAIL_MULTILINE_HEADER_VALUE_REGEX
          end
        end
        h
      end

    end
  end
end


headers = <<-HEADERS
Delivered-To: [email protected]
Return-Path: <[email protected]>
Received: from smtp36.gate.dfw1a (gate36.gate.dfw.mlsrvr.com [172.20.100.36])
       by mail138a.mail.dfw.mlsrvr.com (SMTP Server) with ESMTP id 107434FCCD
       for <[email protected]>; Wed, 29 Dec 2010 05:46:02 -0500 (EST)
X-Spam-Threshold: 95
X-Spam-Score: 0
X-Spam-Flag: NO
X-Virus-Scanned: OK
X-MessageSniffer-Scan-Result: 0
X-MessageSniffer-Rules: 0-0-0-32767-c
X-CMAE-Scan-Result: 0
X-CNFS-Analysis: v=1.0 c=1 a=410GYZBmXEQA:10 a=8nJEP1OIZ-IA:10 a=jrTSQGAaAAAA:8 a=zr8WG4425YMt8zt5NHsA:9 a=9yu0GlTtA_AyZ2ttE_-TzX-Oml0A:4 a=wPNLvfGTeEIA:10 a=vUvjSO8KtUCdM2_P:21 a=bi41oqnuo1lMp9yW:21
X-Orig-To: [email protected]
X-Originating-Ip: [173.203.22.3]
Received: from [173.203.22.3] ([173.203.22.3:54314] helo=287315-web1.comixology.com)
       by smtp36.gate.dfw1a.rsapps.net (envelope-from <[email protected]>)
       (ecelerity 2.2.3.46 r(37554)) with ESMTP
       id 4A/C8-28011-8611B1D4; Wed, 29 Dec 2010 05:46:01 -0500
Received: from comixology.com (localhost [127.0.0.1])
       by 287315-web1.comixology.com (Postfix) with ESMTP id E41991188DCD;
       Wed, 29 Dec 2010 05:45:59 -0500 (EST)
Date: Wed, 29 Dec 2010 05:45:59 -0500
To: [email protected]
From: comiXology <[email protected]>
Reply-to: [email protected]
Subject: Feedback from the Marvel Mobile Comics App
Message-ID: <[email protected]>
X-Priority: 3
X-Mailer: PHPMailer 5.0.2 (phpmailer.codeworxtech.com)
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Content-Type: text/plain; charset="ISO-8859-1"
HEADERS

t = Time.now
500.times do
  begin
    email = Parsing::Email::Header.new(headers)
    email.call
  rescue ArgumentError
    email = Parsing::Email::Header.new(headers.clean_utf8)
    email.call
  end
end
puts "Time with my regex: #{Time.now - t} seconds"

t = Time.now
500.times do
  email = Mail.new(headers)
  email.received
end
puts "Time with Mail gem: #{Time.now - t} seconds"



# Time with my regex: 0.063965 seconds
# Time with Mail gem: 16.327161 seconds
# [Finished in 16.7s]

@synaq Glad if it helped you! (Just saw this comment, lol.) My boss at the time ended up not going with this solution (some people hate even an extensively-commented regex...), but I felt pretty proud of it so I at least put it out there.

The bit dealing with encoding issues is basically taking a sledgehammer to an otherwise-intractable doorknob; it'll get you in most of the time, but if it doesn't, things could get ugly...

If anyone ever uses any bit of it (go nuts), would suggest much more test coverage, especially taking actual malformed examples extracted from real-world emails.