Counting packages in Fedora with Modularity

I wrote an article about Modularity for the Fedora Magazine and I used some numbers there: “At the time of writing — out of the 49,464 binary RPM packages in Fedora 30 — 1,119 (2.26%) come from a module.” How did I get to those numbers?

The first step was to list all binary packages in Fedora. Since we have Modularity now, using dnf repoquery as it’s currently implemented wouldn’t show me RPM packages in modules that are not enabled or default. So I used my modular-repoquery script I hacked together a few months ago. That gave me a list of all binary packages in the repositories.

What I want to count? I want to cound every package (by package name) just once — just the lates version. That’s one for the non-modular repository, and then one for every module stream. So for example, I counted four postgresql packages:

postgresql
postgresql (postgresql:10)
postgresql (postgresql:11)
postgresql (postgresql:9.6)

To get that, I wrote a series of terrible sed commands and put them into a git repository along with the generated files (it took about an hour to generate them, the modular-repoquery script is not very efficient, but gets the job done).

Then I just counted all the lines in the all-names-unique.txt in that repo to get the total number of packages, and then lines containing the ( character to get the modular ones.

Not sure how much useful this is, but I figured I share it anyway!