How to extract email body and attachment - email

Question

I am trying to extract a message rom multi-part email body or from attachment, so I used :0B to try each option like the following:
msgID=""
#extract message in the attachment if it's plain text
:0B
* ^Content-Disposition: *attachment.*(($)[a-z0-9].*)*($)($)\/[a-z0-9+]+=*
{msgID="$MATCH"}
#extract message in the body if it's there
:0EB
* ^()\/[a-z]+[0-9]+[^\+]
{msgID = "$MATCH"}
But msgID got the same message from the body which was inline image code, what's wrong with it, who know the better condition to filter it?
I also need to detect if the sub-header in the body is text and base64 encoded, then decode it, how to stipulate it with regex:
:0B
* ^Content-Type:text/html;
* ^Content-Location:text_0.txt
* ^Content-Transfer-Encoding:base64
* ^Content-Disposition: *attachment.*(($)[a-z0-9].*)*($)($)\/[a-z0-9+]+=*
{ msgID= msgId =`printf '%s' "$MATCH" | base64 -d` }
It always complains no match: ^Content-Type:text/html;

Answer 1

I'm guessing you are trying to say, there are two types of incoming messages. One looks something like this:
From: Sender <there#example.net>
To: You <AmyX#example.com>
Subject: plain text
ohmigod0
And the other is a complex MIME multipart with the same contents:
From: Sender <there#example.net>
To: Amy X <AmyX#example.com>
Subject: MIME complexity
MIME-Version: 1.0
Content-Type: multipart/related; boundary=12345
--12345
Content-type: text/plain; charset="us-ascii"
Content-transfer-encoding: base64
Content-disposition: attachment; filename="text_0.txt"
Content-location: text_0.txt
b2htaWdvZDA=
--12345--
If this is correct, you would want to create a recipe to handle the more complex case first, because it has more features -- if your regex hits, it's unlikely to be a false positive. If not, fall back to the simpler pattern, and assume there will never be any false positives on this (perhaps because this account only receives email from a single system).
# extract message in the attachment if this is a MIME message
:0B
* ^Content-Disposition: *attachment.*(($)[a-z0-9].*)*($))($)\/[a-z0-9+]+=*
{ msgID="$MATCH" } # hafta have spaces inside the braces
:0EB # else, do this: assume the first non-empty body line is msgID
* ^()\/[a-z]+[0-9]+[^\+]
{ msgID="$MATCH" } # still need spaces inside braces;
# ... and, as pointed out many times before, cannot have spaces
# around the equals sign
The regular expression for the attachment is an oversimplification, but I already showed you how to cope with a complex MIME message in a previous question of yours -- if you have multiple cases (for example, base64-encoded attachment, or just a plain-text attachment, or no MIME), I would arrange them from more-complex (meaning more features in the regex) and fall back successively to simpler regexes, with higher chance of false positives. You can chain :0E ("else") cases for as long as you like -- if a regex succeeds and the following recipes are :0E recipes, they will all be skipped.
In response to your update, there are two problems with your attempt. The first, as you note, is that the first regex doesn't match. You have no space after the colon, and I'm guessing there is one in the message you are matching against. You need to understand that every character in a regex needs to match exactly, with the exception of regex metacharacters, which have special meaning. You would typically see something like this in many Procmail recipes:
* ^Content-Type:[ ]*text/html;
where the spaces between the square brackets are a space and a tab. The character class (the stuff in the square brackets) matches either character once, and the asterisk * says to repeat this pattern zero or more times. This allows for arbitrary spacing after the colon. The square brackets and the star are metacharacters. (This is very basic stuff which should be in any Procmail introduction you may have read.)
Your other problem is that each regex is applied in isolation. So your recipe says, if the Content-Type header appears anywhere in the body, and the Content-Location header appears anywhere else (typically, in another MIME header somewhere) etc. In other words, your recipe is very prone to false positives. This is why the rule I proposed earlier is so complex: It looks for these headers in sequence, in a single block, that is, in a single MIME header (though there is nothing to actually make sure that the context is a MIME body part header; more on that in a bit).
Because we want to ensure that there are four different headers, in any order, the regex for this is going to be huge: ABCD|ACDB|ACDB|ABDC|ADCB|BACD|... where A is the Content-Type header regex, B is the Content-Location regex, etc. You could cheat a little bit and craft a single regex which matches a sequence of four matches of the same header-identifying regex -- this is unlikely to cause any false positives (there is no sane reason to have two copies of the same header) and simplifies the code significantly, though it's still complex. Pay attention here: We want to create a single regex which matches any one out of these four headers.
^Content-(Type:[ ]text/plain;|\
Location:[ ]*text_0\.txt|\
Transfer-Encoding:[ ]*base64|\
Disposition:[ ]*attachment)
... followed by any header, repeated four times, followed by the MIME body part (which you had after the Content-Disposition header, slightly out of context, but not incorrectly per se).
(Your code has text/html but if the attachment isn't HTML, as suggested by the format and the filename, it should be text/plain; so I'm going with that instead.)
Before we go there, I'll point out that MIME parsing in Procmail is not done a lot, precisely because it tends to explode into enormously complex regular expressions. MIME has a lot of options, and you need each regex to allow for omission or inclusion of each optional element. There are options for how to encode things (base64, or quoted-printable, or not encoded at all) and options to include or omit quotes around many elements, and options to use a multipart message with one or more body parts or just put the data in the body, like in my constructed first example message (which is still technically a MIME message; its implied content type is text/plain; charset="us-ascii" and the default content transfer encoding is 7bit, which conveniently happens to be what email before MIME always had to look like).
So unless you are in this because (a) you really, really want to learn the deepest secrets of Procmail or (b) you are on a very constrained system where you have to because there is nothing else you can use, I would seriously suggest that you move to a language with a proper MIME parser. A Python script which decodes this would be just half a dozen lines or so, and you get everything normalized and decoded nicely for you with no need for you to reinvent quoted-printable decoding or character set translation. (You can still call the Python script from Procmail if you like.)
I'll also point out here that a proper MIME parser would extract the boundary= parameter from the top-level headers in a multipart message, and make sure any matching on body part headers only occurs immediately after a boundary separator. The following Procmail code does not do that, so we could get a false positive if a message contains a match somewhere else than in the MIME body part headers (such as, for example, if a bounce message contains a fragment of the MIME headers of the bounced message; in this case, you would like for the recipe not to match, but it will).
:0B
* ^(Content-(Type:[ ]text/plain;|\
Location:[ ]*text_0\.txt|\
Transfer-Encoding:[ ]*base64|\
Disposition:[ ]*attachment).*(($)[a-z0-9].*)*)($)\
(Content-(Type:[ ]text/plain;|\
Location:[ ]*text_0\.txt|\
Transfer-Encoding:[ ]*base64|\
Disposition:[ ]*attachment).*(($)[a-z0-9].*)*)($)\
(Content-(Type:[ ]text/plain;|\
Location:[ ]*text_0\.txt|\
Transfer-Encoding:[ ]*base64|\
Disposition:[ ]*attachment).*(($)[a-z0-9].*)*)($)\
(Content-(Type:[ ]text/plain;|\
Location:[ ]*text_0\.txt|\
Transfer-Encoding:[ ]*base64|\
Disposition:[ ]*attachment).*(($)[a-z0-9].*)*)($)\
($)\/[a-z0-9/+]+=*
{ msgid=`printf '%s' "$MATCH" | base64 -d` }
:0BE
* ^^\/[a-z]+[0-9]*[^\+]
{ msgid="$MATCH" }
(Unfortunately, Procmail's regex engine doesn't have the {4} repetition operator, so we have to repeat the regex literally four times!)
As noted before, Procmail, unfortunately, doesn't know anything about MIME. As far as Procmail is concerned, the top-level headers are headers, and everything else is body. There have been attempts to write MIME libraries or extensions for Procmail, but they don't tend to reduce complexity, just shuffle it around.