GitHub personal timeline

GitHub doesn’t seem to have a page that shows my comprehensive personal timeline. I notice this when I comment on a commit⁠—it’s hard to find and easy to lose track of, especially when the committer hasn’t responded yet.

The information is available from the user events API, though. This API is fairly primitive, with no ability to search or filter. Results are split into pages of 30 events each, with up to 10 pages available for the most recent 300 events.

There are a couple ways to approach such an API. One way is to write an all-in-one tool that fetches responses and runs queries against them. That’s complicated by paging, though, and slow to run because it needs to make the API calls each time. I could keep a cache between runs, but that trades off simplicity for speed.

Another approach is to separate concerns by writing a fetcher that stores the responses locally, then I can query against my local database. One advantage of this approach is that I can run the fetcher periodically to update the local database, so over time I can accumulate a history that’s longer than the 300 events available from the API.

Here’s the fetcher: fetch-github-events

If you’d like to try it yourself, you’ll need two things:

  1. GitHub’s hub command-line tool for making authenticated API calls. Sign in by making an API call with hub api user⁠—the tool will save your API token in ~/.config/hub for later use.

  2. Stephen Dolan’s jq JSON processor for working with the API responses. jq is available in most distributions, for example on Fedora it’s dnf install jq. If you haven’t heard of jq before, take a moment to read about it⁠—jq is as indispensable for working with structured data as grep and awk are for ad hoc text.

Once you have those, you can run the fetcher:

✸ fetch-github-events
Fetched 297 events

I don’t know why GitHub only returned 297 events. It seems like sometimes the last page isn’t fully populated. If you run the tool again later, it will fetch the recent events:

✸ fetch-github-events
Fetched 9 events

If there are no new events, or if the output isn’t a tty (when running from cron) then there will be no output.

The events are stored as JSONL in ~/Documents/github-events.jsonl. The JSON Lines format is convenient because we can continuously append records to the end. Unlike a JSON array, it doesn’t need leading and trailing brackets, nor are records separated by commas. That means it’s easy to see the oldest records with head or the newest records with tail.

I’m running the fetcher once an hour from cron:

0 * * * * fetch-github-events

I don’t have a separate tool yet for querying. In this case, separation of concerns meant that I could finish the first part without building the second part. But it’s not hard to make ad hoc queries using jq. Here’s the series of commits against the github-events repository:

✸ jq -r 'select(.repo.name == "agriffis/github-events") |
    .payload.commits[]? | .message' < github-events.jsonl
Initial commit
Initial commit
Initial commit
improvement: reduce deps and simplify with hub api command

What, doesn’t everybody start a repo with a series of forced pushes until the “Initial commit” settles?

Now I can track down those commit comments:

✸ jq -r 'select(.type == "CommitCommentEvent") |
    .payload.comment.html_url' < github-events.jsonl