9 comments

[ 0.25 ms ] story [ 31.2 ms ] thread
how big is the dataset ? I am wondering that if this is a month worth of bots comment that's a lot of data and the interface feels snappy.
90 days only we will backfill at some point but we need to test a bit more
the query is faster is written is follows

``` ['github-issue-comment-event'] | where actor contains "-bot" or actor contains "[bot]" | summarize count() by bin_auto(_time), actor | take 20 ```

You can run the same query with a query builder here https://play.axiom.co/axiom-play-qf1k/datasets/github-issues...

I changed it to 90 days :D

that reminds of that presentation about Google's monorepo, in which a graph makes the distinction of the commits made by humans versus done by bots. The bots there were really active.
Google makes use of its main version control system in a really unusual way. Because the system can handle both very large volumes of commits and large binary files, you get some systems that make use of it in ways you wouldn't see with Git. Some examples:

- Build pipelines that build an artifact based on a build number in a file in version control, then check in a commit that increments that build number for the next job, and finally check in the built artifact in another commit somewhere else.

- An oncall scheduler that uses a big config file to keep track of everything. At the top is a description of the rotation and people involved. At the bottom is a list of names in order for the next, say, 10 cycles. At the beginning of each OnCall cycle, a job deletes the name at the top of the second list, moving the next name to the top. That's who is OnCall. Another job occasionally "refills" the bottom list by generating new rotation cycles from the names at the top and appending them.

Very cool, thanks for sharing!