Forks of a repo may get higher score than the main one

[ferki]

As a follow-up from #122 (comment):

I noticed a strange thing though. I get score for the forks of a repo as well, often more than I get for the canonical one. Not sure if that's normal or expected (or even related).

I noticed it with the RexOps/Rex repo, which is scored as 22.3 experience points as of this moment.

It has an sdondley/Rex fork, which is a community fork scored as 52.2. I expected ~0 since it's not my fork/repo.

[peti2001]

Hi @ferki,

We will look into this and come back in a few days.

[ferki]

@peti2001: thanks!

Looks like the situation has changed somewhat after updating the rankings today on the site:

The mains RexOps/Rex is now scored as 33.7, and forks VVelox/Rex is 29.9, while mestia/Rex is 17.1. Previously scored ferki/Rex and sdondley/Rex are scored 0 now. 🤔

Let me know if I can provide further info that helps!

[peti2001]

Sure, thanks for the update

[ferki]

It still seems to affect how the points are awarded across similar repositories where the repo_info_extractor sees one canonical repository and one or more forks of it.

For example the overwhelming majority of my open source effort is concentrated in the canonical RexOps/Rex repo which gets only 11.5 points today. In contrast, it has an uralm1/Rex fork where I never contributed directly, but that has 110.5 points 😕

Of course, all/most commits from the main repo ends up in any forked repo, but awarding most of the points to the fork still feel confusing. I would expect all points awarded to the canonical repo, and no points awarded to the forks in this case.

GitHub exposes the list of forks of a repository, and also from where a given repository was forked from. Perhaps it would be possible to use that information in the repo_info_extractor, and process the top-level repos first for a more balanced distribution of points?