[BEAM-252] Make Regex Transform

[eljefe6a]

Jira issue BEAM-252 Make Regex Transform. Adding a pre-built transform for Regex.

[kennknowles]

Tests that run a pipeline now must be annotated with @NeedsRunner, since there is no runner in the core SDK.

[eljefe6a]

@kennknowles All passing now.

[eljefe6a]

@kennknowles Anything you need from on this PR?

[kennknowles]

Nope, just a bit backlogged. Thanks for the ping!

[dhalperi]

[ignore me, metadata only] R: @kennknowles

[eljefe6a]

@kennknowles I've updated the PR to the latest DoFN interface. Should make for an easier merge.

[eljefe6a]

@kennknowles I was looking at the WordCount code and realized that should be in RegexTransform. I added a split method that will allow the word splitting portion of WordCount to be done as words.apply(RegexTransform.split("\W*"));.

[kennknowles]

@eljefe6a indeed, very appropriate. This is still on my radar FYI.

[eljefe6a]

@kennknowles Are you ok with me merging this?

[kennknowles]

Actually, I had some more feedback for tightening up the API, and was working on a pull-request-for-your-pull-request. We've been considering moving non-core transforms to extensions. I'll put something together.

Sorry this didn't get prompt attention, but we really need to stick to waiting for an LGTM comment or GitHub "approval".

[eljefe6a]

I thought this was approved. My mistake.

[dhalperi]

The other thing I would add is that for very old branches the precommits need to be run.

Specifically, if you're the author: rebase and push so that Jenkins re-runs.

If not, it's trickier -- the merging committer can run the precommit manually (e.g., mvn clean verify -Prelease) before merging.

This merge immediately broke the build. I think Kenn has pushed fixes, but we should prevent these build breaks.