I’ve used Debian and Ubuntu for a long while now, and managing its list of software sources has never been that much of a joy. It’s easy enough to add things in, but removing them is not always a great thing to do. If you have software installed from one of these sources, you probably want the updates, for security reasons if nothing else.

Still, I managed to go on a tear, installing all sorts of applications, and when the smoke cleared I was pretty sure I didn’t need everything that I had. It was time for a bit of cleaning up. Problem is, the APT tools don’t give you a lot of insight into how to proceed.

Disclaimer

I should add that there’s a good reason for this obscurity: if you have a package, and you drop the source, again, you run a good risk of having that software installed and never able to be updated.

Generate list

With your safety hat on, you can now proceed to running the following command:

for i in `dpkg --get-selections | cut -f1`; do \
  grep -lr "^Filename: .*/${i}" \
    /var/lib/apt/lists/ 2>/dev/null >> /tmp/list; \
done

The first line gets a list of the software which you have either installed or half installed (“deinstall”). For each of these packages, then, we perform a grep to look for that package in all the files in /var/lib/apt/lists/.

Process list

This list which you created is a bit of a mess. Each line is the filename from which you found the package, which corresponds to one of your software sources. Luckily, two stock utilities can clean this up:

sed 's/-[a-z0-9]\+_Packages$//' list | sort | uniq

You can omit the sed command if you’d like, but it will filter out duplicates for each architecture — well, in my case anyway, for amd64 and i386. The list still needs room for a bit of interpretation, but it should be easy enough to tie each line to a record in the /etc/apt/sources.list file, or in one of potentially many files in /etc/apt/sources.list.d/.