If a pull request closes an issue, check that issue's title / tags to categorize the contribution

[choldgraf]

Context

We currently check for tags or metadata in titles for the pull request that was merged within a given window:

github-activity/github_activity/github_activity.py

Lines 471 to 484 in 2851438

	# Separate out items by their tag types
	for kind, kindmeta in tags_metadata.items():
	# First find the PRs based on tag
	mask = closed_prs["labels"].map(
	lambda a: any(ii == jj for ii in kindmeta["tags"] for jj in a)
	)
	# Now find PRs based on prefix
	mask_pre = closed_prs["title"].map(
	lambda title: any(f"{ipre}:" in title for ipre in kindmeta["pre"])
	)
	mask = mask | mask_pre
	kindmeta["data"] = closed_prs.loc[mask]
	kindmeta["mask"] = mask

However, many projects put more labeling / tagging effort into their issues rather than the pull requests that close them. For example, a PR might have no labels itself, but might close an issue labeled documentation.

Proposal

If a Pull Request doesn't contain any information that can be used to categorize it, but it does have an issue that it closes, then we should try to infer the "category" of the PR by looking at the issue's information.

Note on implementation

We can grab the pointer to a closing issue via the closingIssuesReferences graphql field. We can then use this to return the title and list of labels for that issue. Here's a rough query that does this for a generic pull requests query:

{
  viewer {
    pullRequests(
      first: 100
      states: CLOSED
      orderBy: {field: UPDATED_AT, direction: DESC}
    ) {
      nodes {
        closingIssuesReferences(first: 5) {
          edges {
            node {
              labels(first:10) {
                edges {
                  node {
                    id
                    name
                    url
                  }
                }
              }
              title
              url
            }
          }
        }
      }
    }
  }
}

Tasks and updates

No response