I was thinking actual PageRank would be interesting, ranking libraries by who uses what, transitively.
Might be nice to include weightings of app usage by users and downloads (because apps aren't used by other libraries), but just sticking to code-used-by-code could be interesting:
So, "one lib/app, one vote" (weighted by how much it is used by other libs/apps), measures coders' evaluation, not app-users'.
And not have the search part (which PageRank ranks), just the ranking - e.g. a top 10