Today, I was over at a friend’s house and got sidetracked talking about music we liked. I mentioned that I’d recently discovered Jonathan Coulton, really liked his music. and played her a few songs of his. She liked them and asked whether she could have a copy. Since his songs are all licensed under the Creative Commons, that was no problem.

Unfortunately, the only copy of the songs that I had were on my iPod. As everyone knows by now, Apple makes it very difficult to copy songs off an iPod due to piracy concerns. On a Mac, at least, it’s easy to see the song files—they’re just under /Volumes/[iPod name]—but because their names are completely randomized, it can be hard to find any single song. Since ID3 tags are preserved, you can traverse the entire file hierarchy, scan the tags, and copy those files that match, but that solution doesn’t work either if all your files aren’t MP3s or if you have a prohibitively large number of files. In this case, I needed to extract about 40 songs out of several thousand, so I wanted something that would work faster.

The good news is that this problem is blissfully easy to solve with a little shell script and the magical tool lsof. Simply get iTunes displaying all the songs you want to copy, either with a search or a playlist, hit “Select All,” and then execute the following script:


COUNT=`osascript <<EOF
set num to 0
tell application "iTunes"
    repeat with song in selection
        set num to num + 1
    end repeat
end tell
return num

mkdir ~/copies
for ((n=0;n<$COUNT;n+=1)); do
    lsof | grep iTunes | grep mp3 | awk '{print $9}' \
         | xargs -J % cp % ~/copies
    osascript -e 'tell application "iTunes" to next track'

Most of the script is self-explanatory, but some notes on the more interesting parts:

  1. AppleScript supports a count element on most collections, but for whatever reason, count of selection is always 0. I still have a hunch this is a PEBKAC issue on my end, but I don’t mind kluging around it
  2. Bash actually supports C-style looping, which for some reason seems to catch a lot of people by surprise.
  3. lsof is an incredibly useful command. It stands for “list open files,” and lists all files (and sockets) that are open on your machine, along with which program opened them. We filter down to MP3 files opened by iTunes.
  4. awk has largely been replaced by more powerful scripting languages, yet it still has its uses—here, serving as a quicker and more powerful cut to extract the ninth column of whitespace-delimited text, which happens to be the full path to the open file.
  5. BSD’s xargs command allows you to specify an explicit marker for inserting text with the -J option, which we use here so that we can manipulate what’s being copied while holding the destination constant.

So, basically, we just make iTunes play each song long enough to find where on the iPod it’s located, copy it to a safe location, and then tell iTunes to advance to the next track. It’s not especially fast, but it’s far faster than walking the entire iPod looking for the relevant songs. On my system, it took iTunes about twenty seconds to run for forty songs, which is much better than the alternative.

There are relatively few times I honestly find myself wanting to copy songs from my iPod, but the next time you find yourself in a similar situation, you can use the script above to get your copies done quickly.