Implement target distance metrics

[alex-torok]

Add target distance metrics to measure how far away an impacted target is from a directly impacted target. I tried to keep the commits relatively concise, so reviewing them one-by-one from the beginning may be easier than looking at all of the changes at once.

Fixes #223

[CLAassistant]

All committers have signed the CLA.

[alex-torok]

@tinder-maxwellelliott - I added a new E2E test that better showcases target / package distances with a new test workspace.

I don't think there is anything else I need to do here.

[tinder-maxwellelliott]

@alex-torok There have been some changes in main for BCR support, once those are resolved I can merge this

[alex-torok]

@tinder-maxwellelliott Let me know if there is anything else I can do to help land this.

[tinder-maxwellelliott]

@tinder-maxwellelliott Let me know if there is anything else I can do to help land this.

Looks like there are some test failures

[alex-torok]

I think I fixed it, but won't be able to run it locally and see until later today.

[alex-torok]

@tinder-maxwellelliott - should be passing now (at least it is on my machine)

[erikkerber]

I don't know if it was an intentional side-effect, but the format changes mean hashes generated pre-8.0.1 will fail parsing on 8.0.1+. We noticed since we store digests per commit on CI to speed everything up.

It's pretty transient on our part because we fall back to no work avoidance when bazel-diff fails, but others might be a little surprised.

[honnix]

I don't know if it was an intentional side-effect, but the format changes mean hashes generated pre-8.0.1 will fail parsing on 8.0.1+. We noticed since we store digests per commit on CI to speed everything up.

It's pretty transient on our part because we fall back to no work avoidance when bazel-diff fails, but others might be a little surprised.

We are having the same problem. This change is indeed surprising. I'm wondering what could be a migration path.

[alex-torok]

This PR changed how hashes are calculated -- you'll get a different hash value for an old commit on the new version of bazel-diff compared to the old version of bazel-diff.

If you version bazel-diff outside of your repo (i.e use a new version of bazel-diff for calculate-hashes on older commits), then you'd hit cache consistency issues. Versioning it in the repo would just mean that when you bump it, you'd create an 'all targets changed' commit.

If you want the backwards compatibility, you'd need to update the TargetHash parsing to work when reading old hash values - https://github.com/alex-torok/bazel-diff/blob/50b1609cb7ec9db2050bad00e518d6abad3b82dd/cli/src/main/kotlin/com/bazel_diff/hash/TargetHash.kt#L37-L40

We discussed this, but I think I just missed getting it in with a testcase.