Completion for object methods

[sidkshatriya]

This feels like a notable omission, or perhaps I'm missing something:

Let us say the OCaml class of an object is known e.g. abc : <some_class; ...>.

When I type abc# in the editor (with merlin or ocaml-lsp-server enabled) , I should be able to get an autocomplete on all the methods in some_class but I don't.

But it gets worse, supposing I have foo: < bar: int -> int; baz: int -> string > where the methods are explictly listed, even here when I do foo# I should be given an autocomplete which allows me to choose between bar and baz methods, but I dont...

I am using merlin 4.8-500. I've tried with ocaml-lsp-server and am aware that it uses merlin in the backend.

How can merlin help me select methods on an object?

[voodoos]

ack.

You should be write a one-file test similar to:

issue1575.t somewhere the tests sub-tree:

  $ cat >test.ml <<EOF
  > <your code here>
  > <maybe on several lines>
  > EOF
  
FIXME: we get ? but we would expect ?
  $ $MERLIN single complete-prefix -position <line>:<col> -prefix "<Prefix>." \
  > -filename test.ml < test.ml | jq

Then run the test:

dune build @issue1575

Promote the result:

dune promote

[sidkshatriya]

I added a PR for the test #1576

Interestingly this issue seems to be a regression. See #598 which was fixed (and was also a regression).

[sidkshatriya]

@voodoos Would greatly appreciate if you could have a look at this bug. It seems like a pretty big deal that autocomplete on object method names is broken. It seems like it used to work before. I tried to explore this issue in code but didn't make much progress.

[voodoos]

Yes, it looks like this issue is a bit subtle to fix because we cannot easily fold on methods available in the environment.
I will think about it.

[sidkshatriya]

If it is any help, even recent versionf of utop are able to autocomplete on object methods successfully. Check https://github.com/ocaml-community/utop/blob/master/src/lib/uTop_complete.ml -- hopefully it might be useful?

[voodoos]

That's an interesting pointer, thank you !