letop-punning for extension nodes

[WardBrian]

Follow-on to #2746, this makes it so the letop-punning setting also affects nodes like let%foo bar = bar in ...

The trickiest piece here is that the ands for these nodes don't have the extension, so I needed to do this at the level of value_bindings, not value_binding

This also fixes an issue with comments when a pun is introduced. As far as I can tell, this was also an issue back in 0.26, when the equivalent of letop-punning=always was the default behavior.

[WardBrian]

(note I commented in #2746 that I thought the preserve feature would require parser changes, but it turns out the parser already was tracking this, my tests were just lacking any examples of it)

[WardBrian]

Hi @EmileTrotignon - I think this is ready for review. The let%ext change is very similar to the code from the previous PR and should be an easy review.

A bit trickier is I noticed that letop-punning=always could lead to errors if the right hand side had associated comments. I believe I found a fix (for both this new code and the existing), but I'm less confident in that part and would appreciate any pointers

[EmileTrotignon]

This looks good. Where is the fix about comments ?

[WardBrian]

The way I tackled the comment issue is by copying the location of the rhs into the pattern (starting in efb3931). I’m not sure if this is the recommended way or not, but it does appear to work on the added test cases

[Julow]

Awesome :)

[WardBrian]

@Julow I split out the testing and CONTRIBUTING changes into #2748 and then rebased this PR off of that one to try
to make review easier.

I'm now calling Cmts.relocate in Sugar.ml, which seems to be working well

[EmileTrotignon]

Thanks a lot for the contributing documentation, its a very big improvement.

[EmileTrotignon]

Sorry I reviewed the contribution.md changes in the wrong PR

[WardBrian]

No worries @EmileTrotignon. I've pushed updates to #2748

(stacked PRs still need a better workflow for github -- the diffs quickly diverge without a lot of ugly force pushing)

[Julow]

Looks perfect :) Let's merge just after #2748

[EmileTrotignon]

This can be merged after a rebase on main

[WardBrian]

Rebased :)

[EmileTrotignon]

Waiting for CI and merging

[WardBrian]

CI has passed. Thanks for the hand-holding on this feature!

[EmileTrotignon]

@WardBrian You're very welcome ! If you are interested in contributing further to ocamlformat, maybe we could have a chat about the project.

[WardBrian]

Sure @EmileTrotignon! I've been a happy user of ocamlformat for years, and this is the first time I found a gap in my needs for it that prompted me to contribute, but I have some spare cycles to do more here or there