feat: implement Convert to collect code action

[scarf005]

collect.webm

resolves: scalameta/metals-feature-requests#288

[tgodzik]

One thing I would want to improve is to make sure that the collect actually exist, for that I think we could reshuffle things a bit and use codeAction/resolve.

I haven't added it yet, but I can:

• change your PR to use codeAction/resolve and only verify that collect exists on resolving
• add the codeAction/resolve separately to use this week and let you use it
• if you are interested I can explain exactly what can be done and let you work on that

What do you think?

Overall, I think we can check the symbol with semanticdb and the check for collect using compilers.info() <- if that returns some sensible information we should be fine.

There is also an option of rewriting it to the presentation compiler, but that's a much bigger rewrite.

What do you think?

[scarf005]

sorry for late response, i accidentally marked notification to this as None.

i'm fine with either approach; (to be honest, doesn't know which way is the best)

if you are interested I can explain exactly what can be done and let you work on that

that would be great!

[tgodzik]

I will try to merge #6978 tomorrow, but it adds the possibility of using codeAction/resolve.

We can add something similar to FilterMapChain class to data and no edits. If not edits are provided codeAction/resolve will be invoked, where we can additionally check if collect exists by:

1. First send go to compilers.definition on filter or map
2. Replace last map or filter with collect in the semanticdb symbol
3. Use compilers.info to see if that symbol exists and if it exists if it's in an expected shape.
4. If that's known we just create the edits.

[tgodzik]

Ok, the other PR is merged, let me know if you have any problems with what I suggested, I can also help finish the PR

[scarf005]

currently stuck on NPE while converting to codeAction/resolve.

==> X tests.codeactions.FilterMapToCollectCodeActionSuite.cursor-on-filter  4.928s java.lang.NullPointerException: Cannot invoke "org.eclipse.lsp4j.Range.getStart()" because "range$1" is null

let me know if you have any problems with what I suggested, I can also help finish the PR

I'd be grateful!

[scarf005]

@tgodzik i've rebased and migrated to codeAction/resolve. still stuck on semanticDB part.

strangely, while it doesn't happen on my local machine the NPE still seems to happen in CI:

tests.codeactions.FilterMapToCollectCodeActionSuite:
  + cursor-on-filter 8.24s
  + cursor-on-filter-newline 2.663s
  + cursor-on-map 2.07s
  + higher-order-function 1.943s
  + higher-order-function-rename-filter-argument 1.924s
  + higher-order-function-rename-map-argument 1.516s
  + multiple-map-calls 1.744s
  + complex-predicate 1.779s
  + multiline-predicate 1.406s
  + multiline-predicate-block 1.843s
  + with-type-annotations 1.632s
[info] Passed: Total 11, Failed 0, Errors 0, Passed 11

https://github.com/scalameta/metals/actions/runs/12094357358/job/33726319967?pr=6969#step:5:1816

[tgodzik]

if data is null then this will throw an exception