Protect repositories

[Crocmagnon]

Hello! Thank you for this nice tool, I wanted to do some cleanup in my forks but unfortunately I'd need some more features to use it.

There are many reasons a fork may need to stay even if it's old. Maybe a branch is unmerged and we're still working on it, maybe a PR is open waiting for a review. It also appears that the tool only filters forks by their createdAt date. I'm actively working on a fork and I regularly push commits & branches and they get merged regularly, so I don't want to delete this fork.

I suggest adding several features to make the tool useful to more people:

• [x] protecting some repositories manually, because we may want to keep them even if all other "protections" fail
• [x] protecting repositories based on last activity instead of creation date
• [ ] protecting repositories based on open PRs in the upstream repo
• [ ] protecting repositories based on unmerged branches (with recent activity?)

The output should mention clearly why a repo is being protected or if it's being deleted (or would be, in case the -delete flag is not provided).

[rednafi]

Thank you for the ideas; they are great!

I really like the idea of deleting repositories based on the last activity date, rather than the creation date. This can be done without changing the external flag.

Also, a --protect flag that will take a list of repositories sounds useful.

I'll start tackling tasks from the lowest level of complexity to the highest. Also, feel free to open a PR if you have strong opinions about the implementation.

[rednafi]

Version 0.2.0 now:

• Fetches forked repos based on pushed_at || updated_at || created_at date.

• Lists the forked repositories that will and won't be deleted:

  fork-sweeper --owner rednafi --token $GITHUB_TOKEN --older-than-days 0
  

  This returns:

  Fetching forked repositories for rednafi...
  
  Guarded forked repos [won't be deleted]:
  
  Unguarded forked repos [will be deleted]:
      - https://github.com/rednafi/cpython
      - https://github.com/rednafi/dysconfig
      - https://github.com/rednafi/pydantic
  

• Explicitly filter out forked repositories with fuzzy name matching:

  fork-sweeper --owner "rednafi" --token $GITHUB_TOKEN --older-than-days 0 --guard 'py'
  

  This prints:

    Fetching forked repositories for rednafi...
  
    Guarded forked repos [won't be deleted]:
        - https://github.com/rednafi/cpython
        - https://github.com/rednafi/pydantic
  
    Unguarded forked repos [will be deleted]:
        - https://github.com/rednafi/dysconfig
  

[Crocmagnon]

Nice ! 😀 I think however the biggest value would come from an "open PR" guard. It may require some more work though as I don't know if this information can easily be obtained with GitHub's API. Also, listing why repos are protected would be a good addition too IMO.

[rednafi]

Yeah, a PR guard would be nice I agree. But at the same time, it's hard to track that without consuming some event.

Listing why a repo was filtered out seems like an easy low-hanging fruit.

[Crocmagnon]

After doing some exploration, it looks like the open PR feature can be implemented quite easily with the GraphQL API but would require several calls with the REST API. There’s a discussion on the refined-github repo about this, specifically this comment: https://github.com/refined-github/refined-github/issues/2268#issuecomment-562495007