Łukasz Gryglicki

I've saved all single events JSONs which refer to `torvalds/linux` into a big JSON file (it is about 600M in size). There are many SHAs that are not present in...

This is the query to count distinct SHAs (I'm taking all commits from push events, otherwise Linux only has about 700 pushes/year (while 75k commits for the same range -...

The error returning [] was for event types other than 'PushEvent' (so there was no `commits` field). The BQ number of commits and number of commits from DevStats (which parses...

No I didn't make such an analysis, sorry.

Will evaluate as soon as I can, currently a bit busy.

So the first problem is that `filter for sig/wg labels` - this means it can only be applied to issues/PR, not for everything - only issue/PR objects have label(s).

Is this OK for you? This will find a SIG for the following types of metrics: - Approves - Reviews - Comments - Issue comments - Issues - PRs -...

I can have one more option - trying to connect commit event (which has commit SHA) with a PR (by using either PR head SHA or PR merge commit SHA),...

let me know if I should continue with that approach, which means - Issue belongs to SIG if it has a `SIG/xyz` label. - PR belongs to SIG if its...

We have per-file repo groups for [kubernetes](https://github.com/cncf/devstats/blob/master/util_sql/postprocess_repo_groups.sql) - it means that repo group can be different depending on file paths within a single repository (this is for commits, obviously). But...