mdoc VS sbt-mima-plugin

Literally what scaladoc is, it comes with sbt. Although, it's better when enhanced with mdoc so that you get the standard microsite template like these. It would be nice to have an sbt serveDocs and if everyone would host their docs for external linking, but javadoc doesn't do that either.

The good news is that scaladoc is produced by default by sbt and published by default. So you can often pull it from the same repository your library jar came from, extract it with zip, and read the docs. But that's also totally unnecessary - javadoc.io allows you to put in your module info and serves the docs for you, so if there's an older version you can access the documentation this way. Rely on the type signatures, since they can't lie, whilst comments (including scaladoc comments) can. Honestly, library authors should be using mdoc and including examples on every public method, and that type of documentation is something you can almost always contribute to a project for a quick pr kudos.

I know it's not new but the "Snippet validation and results (mdoc)" features in mdoc are so cool. Really takes some of the tedium out of working with documentation since you can know that as you evolve your code the compiler will make sure you keep the docs in sync. Whole new level of Readme-Driven Development

In the Scala ecosystem, MiMa [1] has been in widespread use for years. It automatically checks compatibility for the binary API of a library. Every library with any amount of success uses it. One could say it's the foundation of a stable ecosystem. We also have sbt-version-policy [2] to set it up with minimal configuration (and directly relate it to SemVer).

More recently, we got tasty-mima [3], which checks compatibility at the type system level, rather than the binary level.

[1] https://github.com/lightbend/mima

[2] https://github.com/scalacenter/sbt-version-policy

[3] https://github.com/scalacenter/tasty-mima

Off the top of my head, a naive & approximate solution would be to use test coverage to find out which tests test which blocks of code. Then, when a binary, syntactic incompatibility is detected, re-run only these tests captured for that piece of code.