ktLint 0.34 changes not compatible with spotless

[Tolriq]

Last version of ktLint 0.34 seems to have changes that are not compatible with spotless.

Execution failed for task ':utils:spotlessKotlin'.
> java.lang.NoSuchMethodException: com.pinterest.ktlint.core.KtLint.format(java.lang.String, java.lang.Iterable, java.util.Map, kotlin.jvm.functions.Function2)

[nedtwigg]

Bummer :( Some relevant PR's for anyone who wants a guide for how to fix this:

• Upgrade Ktlint maven coordinate (com.pinterest), default version (0.32.0) #394
• New maven coordinates for 2.0.0 release of Scalafmt #416

[shashachu]

Sorry about this. I made a change to clean up the API for the lint and format APIs and moved editorconfig parsing out of Main.kt and into ktlint-core. I'll be sure to reach out the next time we make breaking API changes. pinterest/ktlint#503

[nedtwigg]

No biggie, I'll be happy to merge a PR whenever someone gets around to it. FWIW, Kotlin has some great tools to allow you to refactor without breaking downstream clients: https://dzone.com/articles/deprecated-annotation-in-kotlin

[abhinavnair]

@shashachu Any updates on this?

[nedtwigg]

Please refrain from "bump comments" unless you have new information to offer. Versions of ktlint >= 0.34 will not work with Spotless until either or both of the following happen:

1. someone updates Spotless to support the new API
2. ktlint adds the old methods back in a new version, perhaps with @Deprecated annotations

If you want to help with option 1, there are example PRs in my first comment that describe exactly how to do it. If you want to help with option 2, the PR which changed the KtLint API is also in this issue.

Spotless is about people scratching their own itch. At some point, someone wanted KtLint support, so they built it. If it works, and has a test case, then we will merge it. We'll make sure that those testcases keep working forever, which we have done. We don't promise to follow every API change of every formatter we support - this would force us to be much choosier about letting people merge new features in, which would be a bummer. It should be (and has been) easier to add things to Spotless over time, not harder.

[shashachu]

@shashachu Any updates on this?

Sorry, I wasn't aware that this was expected to be fixed on the ktlint side. I'd rather not add the old APIs back (in fact it'd be pretty difficult), so if this could be fixed on the Spotless side, that feels easier. The functionality is available, it's just been moved out of Main and into ktlint-core.

[nedtwigg]

I wasn't aware that this was expected to be fixed on the ktlint side

It's opensource, we're all volunteers, nobody is expected to do anything! We've had cases in the past where a formatter broke their public API on accident, and then put the old methods back with @Deprecated for compatibility in a follow-up release. You don't have to do this work, but someone who wants KtLint to maintain a stable API could do it, which is why I offered it up.

I'd rather not add the old APIs back

Fair enough, option 2 is off the table. Option 1 is probably 2 hrs of work, max, even if you haven't work on Spotless before. Happy to merge a PR from anyone. We support a ton of formatters, and we eagerly help people to merge changes they would like to see. We do not take proactive responsibility for adapting to breaking changes across the entire ecosystem that we support.

[ZacSweers]

I'm working on a PR for this

[ZacSweers]

Side note - I don't really understand the backward compatibility requirement for a formatter that doesn't have a stable API yet. Makes this quite brittle :/. Especially as the APIs being reflected are Kotlin, so much of reflected APIs are implementation details of the Kotlin compiler's outputs.