feat(repl): evaluate Scala worksheets with the REPL

[warcholjakub]

Adds a worksheet evaluator that's built on the REPL.

Have you relied on LLM-based tools in this contribution?

Yes, and I checked the output by reading and understanding the code.

How was the solution tested?

New automated tests (including the issue's reproducer, if applicable)

[warcholjakub]

Wasn't sure whether to add MiMa here.

[SolalPirelli]

Do we have a spec or documentation somewhere of what worksheets are and how they should behave?

[warcholjakub]

Do we have a spec or documentation somewhere of what worksheets are and how they should behave?

Not sure. The main goal of this was to replace mdoc in Metals, so I designed it as closely to mdoc as I could. With a few exceptions ofc. Yeah, but besides that I don't think there's any spec.

[tgodzik]

Do we have a spec or documentation somewhere of what worksheets are and how they should behave?

I would say the spec will be the interfaces that we brought over from mdoc. We can iterate on them for sure, but the main goal is to have an easy way of evaluating whole code snippets. Would also reduce the reliance on compiler internals we have in mdoc.

[SolalPirelli]

In that case I'll let someone familiar with mdoc review this... the concept looks OK to me, the idea of doing it on top of the REPL is fine, but I've never used worksheets (my bad!) so I don't know what the code should or shouldn't do in the first place.

[tgodzik]

In that case I'll let someone familiar with mdoc review this... the concept looks OK to me, the idea of doing it on top of the REPL is fine, but I've never used worksheets (my bad!) so I don't know what the code should or shouldn't do in the first place.

I will review it next week (I am a bit sick this week, so trying to avoid anything that requires too much thinking.).

Overall, we also wanted to have it on top of REPL to allow for incremental reevaluation, so we would also only evaluate statements starting from the one that changes. With mdoc it would be quite hard and repl is perfect for that.

[bishabosha]

this seems to hand a lot of the responsibilities to the scala 3 repo. Is it better to provide a minimal interface that stays binary compatible (or even a java interface) and the rest is kept in metals?

having said that, it would be interesting to see if there is a common core that works for both worksheet and something like a notebook (preferrably reactive like https://github.com/bishabosha/notebook.sc)

[tgodzik]

this seems to hand a lot of the responsibilities to the scala 3 repo. Is it better to provide a minimal interface that stays binary compatible (or even a java interface) and the rest is kept in metals?

That would defeat the purpose of the PR. Currently maintaining worksheet is problematic because it heavily interacts with compiler internals. Any larger changes cause problems. Releasing for each version of the compiler separately is also causing a large maintenance burden.

having said that, it would be interesting to see if there is a common core that works for both worksheet and something like a notebook (preferrably reactive like https://github.com/bishabosha/notebook.sc)

That was also an idea we had to make it easier to implement notebook functionality.

[tgodzik]

Also it's not new to have worksheets in the compiler, that was already done in the previous language server.