Assorted feedback

[purcell]

Just been test-driving this — great job! Enjoying more detailed font lock and better indentation than with ocaml-ts-mode.

Wanted to share a few thoughts I had along the way:

• let* ... in should affect indentation the same as let ... in, e.g. when you let (let*) = Lwt.bind and then subsequently write let* foo = Bar.baz in .... (I wonder if there are test cases for something like ocp-indent so you could try to achieve the same general layout without needing to write lots yourself.)
• The regular and -interface modes feel like they only really differ in terms of the grammar: would it make things easier for key bindings and lsp setup etc. to make the -interface mode inherit from the regular mode?
• It might be worth deliberately separating the grammars to some extent, e.g. ensuring that .ml-only constructs get no font-lock/indent support in .mli files, to strongly indicate what's valid and what's not?
• Perhaps consider making the repl switch reversible, so that C-c C-z from .ml takes you to the repl, and then C-c C-z in the repl pops you back. You can save the calling buffer as a local var in the repl to make this easy. Quite a few repl modes have this.
• utop as an alternate neocaml-repl-program doesn't work super well, as the point isn't after the # prompt: probably due to terminal cleverness. Might be worth not suggesting it for now!

[bbatsov]

Thanks for another batch of great feedback! I forgot to mention it, but I addressed all the feedback you had on the MELPA PR, so let's see what I'll be able to do about this round.

let* ... in should affect indentation the same as let ... in, e.g. when you let (let*) = Lwt.bind and then subsequently write let* foo = Bar.baz in .... (I wonder if there are test cases for something like ocp-indent so you could try to achieve the same general layout without needing to write lots yourself.)

I'll see if I can get some inspiration from there. Working on the indentation logic is always my least favorite part in every major mode, as there are so many corner cases to consider. 😅 At least most OCamlers seems to stick to the same indentation conventions.

The regular and -interface modes feel like they only really differ in terms of the grammar: would it make things easier for key bindings and lsp setup etc. to make the -interface mode inherit from the regular mode?

Yeah, that's definitely on the table. At first I even thought of using the same mode for both, given the small grammar differences, but I opted to make things right. The code duplication right now is not a lot, but I think your suggestion makes sense and will simplify the code a bit.

It might be worth deliberately separating the grammars to some extent, e.g. ensuring that .ml-only constructs get no font-lock/indent support in .mli files, to strongly indicate what's valid and what's not?

I thought I had already separated them in this regard. Perhaps I don't quite understand what you mean, or I simply got something wrong in the implementation.

Perhaps consider making the repl switch reversible, so that C-c C-z from .ml takes you to the repl, and then C-c C-z in the repl pops you back. You can save the calling buffer as a local var in the repl to make this easy. Quite a few repl modes have this.

Yeah, definitely. A small oversight on my part!

utop as an alternate neocaml-repl-program doesn't work super well, as the point isn't after the # prompt: probably due to terminal cleverness. Might be worth not suggesting it for now!

I definitely have a bit of work left of the toplevel side of things, but I'll improve this for sure.

[bbatsov]

Sorry about closing this - GitHub is not very smart (or perhaps it's way too smart) when it comes to handling some issue references.