MIME-tools - modules for parsing (and creating!) MIME entities
Here's some pretty basic code for parsing a MIME message, and outputting its decoded components to a given directory:
use MIME::Parser;
### Create parser, and set some parsing options:
my $parser = new MIME::Parser;
$parser->output_under("$ENV{HOME}/mimemail");
### Parse input:
$entity = $parser->parse(\*STDIN) or die "parse failed\n";
### Take a look at the top-level entity (and any parts it has):
$entity->dump_skeleton;
Here's some code which composes and sends a MIME message containing three parts: a text file, an attached GIF, and some more text:
use MIME::Entity;
### Create the top-level, and set up the mail headers:
$top = MIME::Entity->build(Type =>"multipart/mixed",
From => "me\@myhost.com",
To => "you\@yourhost.com",
Subject => "Hello, nurse!");
### Part #1: a simple text document:
$top->attach(Path=>"./testin/short.txt");
### Part #2: a GIF file:
$top->attach(Path => "./docs/mime-sm.gif",
Type => "image/gif",
Encoding => "base64");
### Part #3: some literal text:
$top->attach(Data=>$message);
### Send it:
open MAIL, "| /usr/lib/sendmail -t -oi -oem" or die "open: $!";
$top->print(\*MAIL);
close MAIL;
For more examples, look at the scripts in the examples directory of the MIME-tools distribution.
MIME-tools is a collection of Perl5 MIME:: modules for parsing, decoding, and generating single- or multipart (even nested multipart) MIME messages. (Yes, kids, that means you can send messages with attached GIF files).
You will need the following installed on your system:
File::Path File::Spec IPC::Open2 (optional) IO::ScalarArray from the IO-stringy distribution MIME::Base64 MIME::QuotedPrint Net::SMTP Mail::Internet, ... from the MailTools distribution.
See the Makefile.PL in your distribution for the most-comprehensive list of prerequisite modules and their version numbers.
Here are the classes you'll generally be dealing with directly:
(START HERE) results() .-----------------.
\ .-------->| MIME:: |
.-----------. / | Parser::Results |
| MIME:: |--' `-----------------'
| Parser |--. .-----------------.
`-----------' \ filer() | MIME:: |
| parse() `-------->| Parser::Filer |
| gives you `-----------------'
| a... | output_path()
| | determines
| | path() of...
| head() .--------. |
| returns... | MIME:: | get() |
V .-------->| Head | etc... |
.--------./ `--------' |
.---> | MIME:: | |
`-----| Entity | .--------. |
parts() `--------'\ | MIME:: | /
returns `-------->| Body |<---------'
sub-entities bodyhandle() `--------'
(if any) returns... | open()
| returns...
|
V
.--------. read()
| IO:: | getline()
| Handle | print()
`--------' etc...
To illustrate, parsing works this way:
MIME::Parser.
You hand it an input stream (like a filehandle) to parse a message from:
if the parse is successful, the result is an "entity".
MIME::Entity (a subclass of Mail::Internet).
If the message had "parts" (e.g., attachments), then those parts
are "entities" as well, contained inside the top-level entity.
Each entity has a "head" and a "body".
MIME::Head (a subclass of Mail::Header).
It contains information from the message header: content type,
sender, subject line, etc.
A typical multipart message containing two parts -- a textual greeting and an "attached" GIF file -- would be a tree of MIME::Entity objects, each of which would have its own MIME::Head. Like this:
.--------.
| MIME:: | Content-type: multipart/mixed
| Entity | Subject: Happy Samhaine!
`--------'
|
`----.
parts |
| .--------.
|---| MIME:: | Content-type: text/plain; charset=us-ascii
| | Entity | Content-transfer-encoding: 7bit
| `--------'
| .--------.
|---| MIME:: | Content-type: image/gif
| Entity | Content-transfer-encoding: base64
`--------' Content-disposition: inline;
filename="hs.gif"
You usually start by creating an instance of MIME::Parser and setting up certain parsing parameters: what directory to save extracted files to, how to name the files, etc.
You then give that instance a readable filehandle on which waits a MIME message. If all goes well, you will get back a MIME::Entity object (a subclass of Mail::Internet), which consists of...
If the original message was a multipart document, the MIME::Entity object will have a non-empty list of "parts", each of which is in turn a MIME::Entity (which might also be a multipart entity, etc, etc...).
Internally, the parser (in MIME::Parser) asks for instances of MIME::Decoder whenever it needs to decode an encoded file. MIME::Decoder has a mapping from supported encodings (e.g., 'base64') to classes whose instances can decode them. You can add to this mapping to try out new/experiment encodings. You can also use MIME::Decoder by itself.
All message composition is done via the MIME::Entity class. For single-part messages, you can use the MIME::Entity/build constructor to create MIME entities very easily.
For multipart messages, you can start by creating a top-level
multipart entity with MIME::Entity::build(), and then use
the similar MIME::Entity::attach() method to attach parts to
that message. Please note: what most people think of as
"a text message with an attached GIF file" is really a multipart
message with 2 parts: the first being the text message, and the
second being the GIF file.
When building MIME a entity, you'll have to provide two very important
pieces of information: the content type and the
content transfer encoding. The type is usually easy, as it is directly
determined by the file format; e.g., an HTML file is text/html.
The encoding, however, is trickier... for example, some HTML files are
7bit-compliant, but others might have very long lines and would need to be
sent quoted-printable for reliability.
See the section on encoding/decoding for more details, as well as "A MIME PRIMER".
Since MIME::Entity inherits directly from Mail::Internet, you can use the normal Mail::Internet mechanisms to send email. For example,
$entity->smtpsend;
The MIME::Decoder class can be used to encode as well; this is done when printing MIME entities. All the standard encodings are supported (see "A MIME PRIMER" for details):
Encoding: | Normally used when message contents are:
-------------------------------------------------------------------
7bit | 7-bit data with under 1000 chars/line, or multipart.
8bit | 8-bit data with under 1000 chars/line.
binary | 8-bit data with some long lines (or no line breaks).
quoted-printable | Text files with some 8-bit chars (e.g., Latin-1 text).
base64 | Binary files.
Which encoding you choose for a given document depends largely on (1) what you know about the document's contents (text vs binary), and (2) whether you need the resulting message to have a reliable encoding for 7-bit Internet email transport.
In general, only quoted-printable and base64 guarantee reliable
transport of all data; the other three "no-encoding" encodings simply
pass the data through, and are only reliable if that data is 7bit ASCII
with under 1000 characters per line, and has no conflicts with the
multipart boundaries.
I've considered making it so that the content-type and encoding can be automatically inferred from the file's path, but that seems to be asking for trouble... or at least, for Mail::Cap...
MIME-tools is a large and complex toolkit which tries to deal with a wide variety of external input. It's sometimes helpful to see what's really going on behind the scenes. There are several kinds of messages logged by the toolkit itself:
These are printed directly to the STDERR, with a prefix of
"MIME-tools: debug".
Debug message are only logged if you have turned /debugging on in the MIME::Tools configuration.
These are logged by the standard Perl warn() mechanism
to indicate an unusual situation.
They all have a prefix of "MIME-tools: warning".
Warning messages are only logged if $^W is set true
and MIME::Tools is not configured to be /quiet.
These are logged by the standard Perl warn() mechanism
to indicate that something actually failed.
They all have a prefix of "MIME-tools: error".
Error messages are only logged if $^W is set true
and MIME::Tools is not configured to be /quiet.
Unlike "typical" warnings above, which warn about problems processing data, usage-warnings are for alerting developers of deprecated methods and suspicious invocations.
Usage messages are currently only logged if $^W is set true
and MIME::Tools is not configured to be /quiet.
When a MIME::Parser (or one of its internal helper classes) wants to report a message, it generally does so by recording the message to the MIME::Parser::Results object immediately before invoking the appropriate function above. That means each parsing run has its own trace-log which can be examined for problems.
If you want to tweak the way this toolkit works (for example, to turn on debugging), use the routines in the MIME::Tools module.
Turn debugging on or off. Default is false (off).
MIME::Tools->debugging(1);
Turn the reporting of warning/error messages on or off. Default is true, meaning that these message are silenced.
MIME::Tools->quiet(1);
Return the toolkit version.
print MIME::Tools->version, "\n";
The MIME-Tools distribution comes with an "examples" directory. The scripts in there are basically just tossed-together, but they'll give you some ideas of how to use the parser.
Always run your Perl script with -w.
If you see a warning about a deprecated method, change your
code ASAP. This will ease upgrades tremendously.
Don't try to MIME-encode using the non-standard MIME encodings. It's just not a good practice if you want people to be able to read your messages.
For example, if your mail-handling code absolutely must not die, then perform mail parsing like this:
$entity = eval { $parser->parse(\*INPUT) };
Parsing is a complex process, and some components may throw exceptions
if seriously-bad things happen. Since "seriously-bad" is in the
eye of the beholder, you're better off catching possible exceptions
instead of asking me to propagate undef up the stack. Use of exceptions in
reusable modules is one of those religious issues we're never all
going to agree upon; thankfully, that's what eval{} is good for.
As of 5.3xx, the parser tries extremely hard to give you a MIME::Entity. If there were any problems, it logs warnings/errors to the underlying "results" object (see MIME::Parser::Results). Look at that object after each parse. Print out the warnings and errors, especially if messages don't parse the way you thought they would.
Parsing is a (slightly) lossy operation. Because of things like ambiguities in base64-encoding, the following is not going to spit out its input unchanged in all cases:
$entity = $parser->parse(\*STDIN);
$entity->print(\*STDOUT);
If you're using MIME::Tools to process email, remember to save the data you parse if you want to send it on unchanged. This is vital for things like PGP-signed email.
The MIME standard allows for text strings in headers to contain characters from any character set, by using special sequences which look like this:
=?ISO-8859-1?Q?Keld_J=F8rn_Simonsen?=
To be consistent with the existing Mail::Field classes, MIME::Tools does not automatically unencode these strings, since doing so would lose the character-set information and interfere with the parsing of fields (see MIME::Parser/decode_headers for a full explanation). That means you should be prepared to deal with these encoded strings.
The most common question then is, how do I decode these encoded strings? The answer depends on what you want to decode them to: ASCII, Latin1, UTF-8, etc. Be aware that your "target" representation may not support all possible character sets you might encounter; for example, Latin1 (ISO-8859-1) has no way of representing Big5 (Chinese) characters. A common practice is to represent "untranslateable" characters as "?"s, or to ignore them completely.
To unencode the strings into some of the more-popular Western byte
representations (e.g., Latin1, Latin2, etc.), you can use the decoders
in MIME::WordDecoder (see MIME::WordDecoder).
The simplest way is by using unmime(), a function wrapped
around your "default" decoder, as follows:
use MIME::WordDecoder;
...
$subject = unmime $entity->head->get('subject');
One place this is done automatically is in extracting the recommended filename for a part while parsing. That's why you should start by setting up the best "default" decoder if the default target of Latin1 isn't to your liking.
RFC 2045 dictates that MIME streams have lines terminated by CRLF
("\r\n"). However, it is extremely likely that folks will want to
parse MIME streams where each line ends in the local newline
character "\n" instead.
An attempt has been made to allow the parser to handle both CRLF and newline-terminated input.
The "7bit" and "8bit" decoders will decode both
a "\n" and a "\r\n" end-of-line sequence into a "\n".
The "binary" decoder (default if no encoding specified)
still outputs stuff verbatim... so a MIME message with CRLFs
and no explicit encoding will be output as a text file
that, on many systems, will have an annoying ^M at the end of
each line... but this is as it should be.
TODO FIXME
All encoders currently output the end-of-line sequence as a "\n",
with the assumption that the local mail agent will perform
the conversion from newline to CRLF when sending the mail.
However, there probably should be an option to output CRLF as per RFC 2045
Let's get something straight: this is an evil, EVIL practice. If your mailer creates multipart boundary strings that contain newlines, give it two weeks notice and find another one. If your mail robot receives MIME mail like this, regard it as syntactically incorrect, which it is.
People like to hand the parser raw messages straight from POP3 or from a mailbox. There is often predictable non-header information in front of the real headers; e.g., the initial "From" line in the following message:
From - Wed Mar 22 02:13:18 2000
Return-Path: <eryq@zeegee.com>
Subject: Hello
The parser simply ignores such stuff quietly. Perhaps it shouldn't, but most people seem to want that behavior.
Please note that there is currently an ambiguity in the way
preambles are parsed in. The following message fragments both
are regarded as having an empty preamble (where \n indicates a
newline character):
Content-type: multipart/mixed; boundary="xyz"\n
Subject: This message (#1) has an empty preamble\n
\n
--xyz\n
...
Content-type: multipart/mixed; boundary="xyz"\n
Subject: This message (#2) also has an empty preamble\n
\n
\n
--xyz\n
...
In both cases, the first completely-empty line (after the "Subject") marks the end of the header.
But we should clearly ignore the second empty line in message #2, since it fills the role of "the newline which is only there to make sure that the boundary is at the beginning of a line". Such newlines are never part of the content preceding the boundary; thus, there is no preamble "content" in message #2.
However, it seems clear that message #1 also has no preamble "content", and is in fact merely a compact representation of an empty preamble.
Why not do everything in core? Although the amount of core available on even a modest home system continues to grow, the size of attachments continues to grow with it. I wanted to make sure that even users with small systems could deal with decoding multi-megabyte sounds and movie files. That means not being core-bound.
As of the released 5.3xx, MIME::Parser gets by with only one temp file open per parser. This temp file provides a sort of infinite scratch space for dealing with the current message part. It's fast and lightweight, but you should know about it anyway.
Achim Bohnet once pointed out that MIME headers do nothing more than store a collection of attributes, and thus could be represented as objects which don't inherit from Mail::Header.
I agree in principle, but RFC 2045 says otherwise. RFC 2045 [MIME] headers are a syntactic subset of RFC-822 [email] headers. Perhaps a better name for these modules would have been RFC1521:: instead of MIME::, but we're a little beyond that stage now.
When I originally wrote these modules for the CPAN, I agonized for a long time about whether or not they really should subclass from Mail::Internet (then at version 1.17). Thanks to Graham Barr, who graciously evolved MailTools 1.06 to be more MIME-friendly, unification was achieved at MIME-tools release 2.0. The benefits in reuse alone have been substantial.
So you need to parse (or create) MIME, but you're not quite up on the specifics? No problem...
Here are some definitions adapted from RFC 1521 (predecessor of the current RFC 204[56789] defining MIME) explaining the terminology we use; each is accompanied by the equivalent in MIME:: module terms...
An "attachment" is common slang for any part of a multipart message -- except, perhaps, for the first part, which normally carries a user message describing the attachments that follow (e.g.: "Hey dude, here's that GIF file I promised you.").
In our system, an attachment is just a MIME::Entity under the top-level entity, probably one of its parts.
The "body" of an entity is that portion of the entity which follows the header and which contains the real message content. For example, if your MIME message has a GIF file attachment, then the body of that attachment is the base64-encoded GIF file itself.
A body is represented by an instance of MIME::Body. You get the body of an entity by sending it a bodyhandle() message.
One of the parts of the body of a multipart /entity. A body part has a /header and a /body, so it makes sense to speak about the body of a body part.
Since a body part is just a kind of entity, it's represented by an instance of MIME::Entity.
An "entity" means either a /message or a /body part. All entities have a /header and a /body.
An entity is represented by an instance of MIME::Entity. There are instance methods for recovering the header (a MIME::Head) and the body (a MIME::Body).
A "message" generally means the complete (or "top-level") message being transferred on a network.
There currently is no explicit package for "messages"; under MIME::, messages are streams of data which may be read in from files or filehandles. You can think of the MIME::Entity returned by the MIME::Parser as representing the full message.
This indicates what kind of data is in the MIME message, usually as majortype/minortype. The standard major types are shown below. A more-comprehensive listing may be found in RFC-2046.
application/octet-stream, application/gzip, application/postscript...
audio/basic...
image/gif, image/jpeg...
message/rfc822...
multipart/mixed, multipart/alternative...
text/plain, text/html...
video/mpeg...
This is how the message body is packaged up for safe transit. There are the 5 major MIME encodings. A more-comprehensive listing may be found in RFC-2045.
MIME::Parser, MIME::Head, MIME::Body, MIME::Entity, MIME::Decoder, Mail::Header, Mail::Internet
At the time of this writing, the MIME-tools homepage was http://www.mimedefang.org/static/mime-tools.php. Check there for updates and support.
The MIME format is documented in RFCs 1521-1522, and more recently in RFCs 2045-2049.
The MIME header format is an outgrowth of the mail header format documented in RFC 822.
Please file support requests via rt.cpan.org.
Released as MIME-parser (1.0): 28 April 1996. Released as MIME-tools (2.0): Halloween 1996. Released as MIME-tools (4.0): Christmas 1997. Released as MIME-tools (5.0): Mother's Day 2000.
See ChangeLog file for full details.
Eryq (eryq@zeegee.com), ZeeGee Software Inc (http://www.zeegee.com). David F. Skoll (dfs@roaringpenguin.com) http://www.roaringpenguin.com.
Copyright (c) 1998, 1999 by ZeeGee Software Inc (www.zeegee.com). Copyright (c) 2004 by Roaring Penguin Software Inc (www.roaringpenguin.com)
This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.
See the COPYING file in the distribution for details.
This kit would not have been possible but for the direct contributions of the following:
Gisle Aas The MIME encoding/decoding modules.
Laurent Amon Bug reports and suggestions.
Graham Barr The new MailTools.
Achim Bohnet Numerous good suggestions, including the I/O model.
Kent Boortz Initial code for RFC-1522-decoding of MIME headers.
Andreas Koenig Numerous good ideas, tons of beta testing,
and help with CPAN-friendly packaging.
Igor Starovoitov Bug reports and suggestions.
Jason L Tibbitts III Bug reports, suggestions, patches.
Not to mention the Accidental Beta Test Team, whose bug reports (and comments) have been invaluable in improving the whole:
Phil Abercrombie
Mike Blazer
Brandon Browning
Kurt Freytag
Steve Kilbane
Jake Morrison
Rolf Nelson
Joel Noble
Michael W. Normandin
Tim Pierce
Andrew Pimlott
Dragomir R. Radev
Nickolay Saukh
Russell Sutherland
Larry Virden
Zyx
Please forgive me if I've accidentally left you out. Better yet, email me, and I'll put you in.