Auto label node PR's

[phillipj]

Very much a work in progress ATM, just wanted to get this out there for public scrutiny and work on it the next couple of days.

Goals

• lists appropriate labels for any PR on /labels/#PR-NUM
• adds labels to created PR's

UPDATE In the discussion below, we've decided to only resolve c++ and test labels at first, then add more labels in later iterations of this.

Refs #1

[Fishrock123]

I don't mind if it over-labels a bit. We can probably do a trial of sorts by just regexing against tag names with some sort of priority, falling back to lib / src if there are more than, say, 5.

[Fishrock123]

the refactoring bits lgtm

[phillipj]

Pushed some improvements. Now it does some very basic analysis on the file paths changed and returns a list of labels.

Obvious things still missing:

• doesn't handle src changes
• changes to lib/_*.js and lib/internal has to be mapped manually. Currently changes to lib/_stream_readable.js would be resolved to a "_stream_readable" label

[Fishrock123]

I think it is probably more reliable to just hardcode a bunch of regex's. I'm willing to figure out what those should be.

We should also try to fallback to lib / src if there is > 5 subsystem labels

[williamkapke]

I need the github client extracted out as you did here in lib/githubClient.js. Maybe that can go in a PR of its own?

For now I'll use a copy of that file but won't commit it.

[Fishrock123]

@phillipj How about we make this only check against regex's and only use c++ from the start, then we can expand the map.

This will also make sure the bot never uses labels that people willy-nilly add/remove.

[phillipj]

@Fishrock123 sure, we could start off with a limited set of labels.

What about keeping the current test label functionality in here? Adding test when the changeset only contains ./test/* files..

[Fishrock123]

@phillipj Also seems good to me, but I think it should be cleaned up a bit to be a bit more portable.. We probably need some to check isOnly and other to check is.

Another example would be a purely benchmarks change. I don't think we should add that yet, just let it be easier to build on after, if possible. :)

[Fishrock123]

@phillipj lgtm minus a nit, although I'm not sure we should deploy this yet. We'll want to have a go-ahead from collaborators with some example in a test repo I think.

If you think we should go straight to deploy though, I'm willing to back that.

[phillipj]

@Fishrock123 PTAL

This branch is currently deployed and have done some initial tests in https://github.com/TestOrgPleaseIgnore/node/pulls

[phillipj]

@Fishrock123 @williamkapke Thanks for reviewing! Just pushed an update which merges the resolved labels with the PR's existing labels, so that we don't accidentally remove any labels.

If it looks good to you, merge at will.

[phillipj]

Forgot to mention I tested it with three PRs (9 - 11) in the test repo https://github.com/TestOrgPleaseIgnore/node/pulls

[Fishrock123]

Seems good to me, let's try it imo.