Do you ever find yourself urgently searching for a file that you know you received as an email attachment but do not remember who sent it or when? Has your company saved all the important documents received via email somewhere easily retrievable? Would you like to save the content of all your email messages automatically as separate, plain text files?

Being able to copy automatically, into one folder and as separate files, all the email attachments and message bodies hidden in your email archives might save your day in situations like these. This tutorial explains how to do it with one relatively simple shell script and tools available from the standard repositories of most Linux distributions. Only basic knowledge of shell scripts is necessary. Additionally, patching the script to make it save just the attachment is also very easy.

MIME and Mailbox Formats

To process email messages, you need to know how files are attached to email and how email messages are archived inside digital mailboxes. To extract attachments from one email message, you need a MIME-aware processor that can split all the email's parts into separate files. Multipurpose Internet Mail Extensions (MIME) [1] is the open standard that describes how to:

• encode non-ASCII character sets in email bodies and headers,
• format message bodies with multiple parts (e.g., content in both HTML and plain text formats), and
• encode and attach to email any non-textual content (e.g., images to generic binary files).

To archive multiple messages as one mailbox, the simplest format is MBOX, which just concatenates all messages into one plain text file. Because MBOX is as simple as it is inefficient, other formats that store each message in a separate file inside a folder were developed [2]. One of the most common formats, and the one used in this tutorial, is called Maildir, which has the internal structure shown in Figure 1.

Figure 1: Maildir mailboxes have three subfolders: new for new email, cur for already read messages, and tmp, which is used by email software for processing.

Preliminary Work

Unless the mailbox you want to scan for attachments already is in Maildir format (or in any other format that puts every email into a separate file), you have to convert it. You can manually, but quickly, convert any mailbox to Maildir format in any Linux terminal using the Mutt email client:

1. Type:

mutt -R -m Maildir -f <YOUR MAILBOX>

2. Type T followed by the Enter key.

3. Type a semicolon ( ; ) and then s.

The -R switch opens in read-only mode the mailbox defined with -f, and -m sets Maildir as the default mailbox format. By doing the second step inside Mutt, you tag all the messages in the mailbox. The third step makes Mutt ask you in which other mailbox all those messages should be saved, (i.e., copied, because the original mailbox is in read-only mode). Pass Mutt any mailbox name you want; when it has finished copying the messages, type q to exit.

If you have many mailboxes to convert, you can run Mutt from a script, as explained in an article online [3]. In any case, I recommend only working with copies of your mailbox, just to be on the safe side (not to mention that changing the access times of email files may confuse some email clients).

A Redundant, Overcautious Extractor

By default, the email extractor script that I introduce here extracts everything from email messages, not just attachments: message bodies, digital signatures, embedded images, and so on. With some very minor modifications, however, you can make it extract attachments only.

The code for this article is a shell script that contains three parts. For readability, I present them in three separate listings. The first part (Listing 1) defines all the necessary variables and folders, plus a shell function that generates unique folder names for each message.

01 #! /bin/bash
02
03 MAILBOX="$1"
04 TARGET="$2"
05 TMPDIR="$HOME/extractor-tmp"
06 CNT=0
07 declare -A CHECKSUMS
08
09 emaildirname() {
10   MSGTS=`stat -c %Y $EMAIL`
11   ORIGTS=`date -d @$MSGTS '+%Y%m%d%H%M.%S'`
12   FILENAMETS=`date -d @$MSGTS '+%Y%m%d%H%M%S'`
13   MSGDATE=`grep '^Date: ' $EMAIL | cut -c7- | head -1`
14   MSGSUBJ=`grep '^Subject: ' $EMAIL | cut -c10-210 | head -1`
15   DIR=`echo $MSGDATE-$MSGSUBJ | sed -r 's/[^a-zA-Z0-9]+/-/g' | sed -r 's/-+$//'`
16   DIR="$FILENAMETS-$DIR-$CNT"
17   let "CNT=CNT+1"
18 }
19
20 rm -rf   $TMPDIR $TARGET
21 mkdir -p $TMPDIR $TARGET/tmp
22
23 echo   "EXTR mbox      : $MAILBOX"
24 printf "EXTR contains  : %s messages ( %s KBytes )\n" `find $MAILBOX -type f | wc -l ` `du -sk $MAILBOX | cut  -f1`
25 echo   "EXTR target    : $TARGET"

I call the script "redundant" because its second part (Listing 2) does the actual job of saving all attachments and email bodies as separate files a total of four times, with four different tools. The final section (Listing 3) removes any empty or duplicate files produced by that process.

26-28 <these lines only contained comments...>
29 for TOOL in mu uudeview munpack ripmime
30 do
31   mkdir -p $TMPDIR/$TOOL/
32   printf "EXTR\nEXTR %-10s: %-9s  start\n"  $TOOL `date +%H:%M:%S`
33   CNT=0
34
35   for EMAIL in `find $MAILBOX/ -type f`
36   do
37   emaildirname
38   mkdir $TMPDIR/$TOOL/$DIR
39   cd $TMPDIR/$TOOL/$DIR
40   case $TOOL in
41
42     ripmime)
43     TOOLNUM=3
44     ripmime -i $EMAIL --paranoid ;;
45
46     munpack)
47       TOOLNUM=2
48     munpack -t -q  $EMAIL > /dev/null ;;
49
50     uudeview)
51     TOOLNUM=1
52     uudeview  +a -m -n -q -i $EMAIL  ;;
53
54     mu)
55     TOOLNUM=0
56     mu extract -a $EMAIL  ;;
57   esac
58
59   NUMFILES=`find . -type f | wc -l`
60   if [[ "$NUMFILES" -gt "0" ]]
61   then
62     find . -type f | cut -c3- > /tmp/file_list.txt
63     while IFS= read -r file
64       do
65     NEWNAME=`echo ${file%.*}`
66     NEWNAME=`echo $NEWNAME | sed -r 's/[^a-zA-Z0-9]+/-/g' | sed -r 's/-+$//'`
67     EXT=`echo $file | awk -F . '{print $NF}'`
68
69     if [ "$EXT" == "$NEWNAME" ]
70     then
71       EXT='probablyemailbody.txt'
72     fi
73     mv -- "$file"    $FILENAMETS-$TOOLNUM-$CNT-$NEWNAME.$EXT
74     touch -t $ORIGTS $FILENAMETS-$TOOLNUM-$CNT-$NEWNAME.$EXT
75       done < /tmp/file_list.txt
76   fi
77
78   done
79   printf "EXTR %-10s: %-9s  end\n"  $TOOL `date +%H:%M:%S`
80   printf "EXTR %-10s: %-9s  files extracted (%7s empty)\n"  $TOOL `find $TMPDIR/$TOOL/ -type f | wc -l` `find $TMPDIR/$TOOL/ -type f -empty| wc -l`
81
82   find $TMPDIR/$TOOL/ -type f -exec mv -i {} $TARGET/tmp \;
83
84 done

 85
 86 find $TARGET/tmp -type f -empty -exec rm {} \;
 87
 88 printf "EXTR\nEXTR cleaning  : %-9s  start\n" `date +%H:%M:%S`
 89
 90 CNT=0
 91
 92 for F in `find $TARGET/tmp -type f  | sort`
 93 do
 94   CK=`md5sum $F | sed 's/ .*$//' `
 95   if [ "${CHECKSUMS[$CK]}" == "found" ]
 96   then
 97     echo "DUPX: removing $F"
 98     let "CNT=CNT+1"
 99   else
100     echo "DUPX: keeping  $F"
101     mv $F $TARGET/
102     CHECKSUMS[$CK]='found'
103   fi
104 done
105
106 mv $TARGET/tmp $TARGET-tmp
107
108 printf "EXTR    total : %-9s files found, after removing %s duplicates\n"  `find $TARGET -type f | wc -l` $CNT
109 printf "EXTR cleaning : %-9s  end\n"  `date +%H:%M:%S`
110 >