Emacs bindings

[Chris00]

After discussing with @monnier, @mattiase, and @erikmd about Tuareg documentation bindings, we would like to expand this discussion with the Merlin developers.

• There is unfortunately no standard Emacs' documentation keybinding but we are thinking to use C-c C-d (we cannot use C-c C-h as C-h is reserved for discoverability). This however conflicts with the key for m︭erlin-d︭estruct. May we suggest that C-c M-d be used for this (which is even more mnemonic 😀).
• Merlin shadows C-c C-r (bound in Tuareg and Caml-mode to ...-eval-region). What do you think of binding merlin-error-check to C-c M-r?
• Given that M-. (resp. M-,) are standard in Emacs (and already supported by Merlin), one can remove C-c C-l (resp. C-c &) — I'm thinking to do the same in Tuareg after making caml-types-show-ident available through xref.
• How about binding merlin-occurrences to M-? (through xref)?
• Contrarily to what is asserted on the wiki, C-c t is no longer bound to merlin-type-expr (a good thing as it is reserved to the user). May we suggest to bind it to C-c C-t with a prefix?

[mattiase]

And as an alternative binding for merlin-destruct, C-c C-e seems to be free -- E for Explode?

[erikmd]

And as an alternative binding for merlin-destruct, C-c C-e seems to be free -- E for Explode?

I don't think so: it is already bound (and AFAICT, C-c C-e is a very useful keybinding in Tuareg :)

[mattiase]

My mistake, sorry. I always use C-M-x because it's faster to type!

[Chris00]

It is also bound to C-x C-e (which is not harder to type than C-c C-e, contrarily to C-M-x for me), so we could free that binding...

[erikmd]

Hi @Chris00,

It is also bound to C-x C-e (which is not harder to type than C-c C-e, contrarily to C-M-x for me), so we could free that binding...

Indeed! so to sum up the suggestions regarding merlin-destruct, it might be bound, for example:

• to C-c M-d
• or to C-c C-e, keeping C-M-x and C-x C-e as the binding for tuareg-eval-phrase (albeit it can be noted that C-x C-e shadows the global command eval-last-sexp which may be handy to evaluate elisp fragments in any mode (including sexp's in OCaml comments for example 😅))
• or to something else, maybe C-c | (?)

[bbatsov]

I think that C-c M-d is a good option for merlin-destruct and C-c C-d would better for something documentation related. I would not mess with classic keybindings from evaluating stuff.

Btw, I've started playing with OCaml recently and I've noticed that Merlin's mode menu is almost empty:

This doesn't help with the feature discovery.

[erikmd]

@Chris00 to follow-up your first post:

Merlin shadows C-c C-r (bound in Tuareg and Caml-mode to ...-eval-region). What do you think of binding merlin-error-check to C-c M-r?

For C-c M-r: why not, but that's less convenient to type;

Actually I'd like to suggest binding C-c C-v to this function (merlin-error-check), because this keybinding is unbound in Tuareg/Merlin, and is already used in the standard Emacs modes for Python (to check the ambient script using linters):

• cf. python.el
• cf. elpy.el

[erikmd]

Dear all, I believe the discussion led to good solutions (requiring few unbindings, for the sake of backward-compatibility),
so I'm about to prepare two small PRs (for Tuareg and Merlin) that perform the various keybinding enhancements that were suggested in these two issues:

• Unbind C-c ? and C-c C-h (maybe) tuareg#273
• Emacs bindings #1386

[erikmd]

Here is an (exhaustive) list of the changes that the PRs involve:

Tuareg changes (issue ocaml/tuareg#273 | PR ocaml/tuareg#309, ocaml/caml-mode#18)

Old binding (to remove)	Function name	Current binding (already existing or new)	Comment	Author
C-c ?	tuareg-interactive-next-error-source	C-c ` (existing)	Emacs expects ? and C-h to be unbound in prefix maps (for self-documenting purpose, e.g., try C-x ?)	Suggested by @mattiase
C-c C-h	caml-help	C-h . (new)	same binding as Eglot	Suggested by @catern and @monnier

Merlin changes (issue #1386 | PR #1668)

Old binding (to remove)	Function name	Current binding (already existing or new)	Comment	Author
C-c C-r	merlin-error-check	C-c C-v (new)	to avoid shadowing Tuareg's C-c C-r = tuareg-eval-region, and C-c C-v is similar to python.el's and elpy's check commands	Suggested by @Chris00 and @erikmd
C-c t (cf. the wiki)	merlin-type-expr	C-u C-c C-t (new)	the binding was documented but not available actually	Suggested by @Chris00
C-c C-d	merlin-destruct	C-c M-d (new)	to be able to use C-c C-d for a documentation-related function	Suggested by @erikmd and @bbatsov
C-c C-d	merlin-destruct	C-c | (new)	to recall the | in pattern matching and provide a simpler keybinding than C-c M-d	Suggested by @erikmd
N/A (unbound)	merlin-document	C-c C-d (remap)	sounds better for a documentation-related function	Suggested by @mattiase and @bbatsov

Here is a list of possible future changes (more discussion/work is needed):

Old binding (to remove)	Function name	Current binding (already existing or new)	Comment	Author
N/A (unbound)	merlin-occurrences	M-? (existing)	may need new xref-related code	Suggested by @mattiase and @Chris00
C-c C-l	caml-types-show-ident and merlin-locate	M-. (existing)	may need new xref-related code	Suggested by @Chris00
C-c &	merlin-pop-stack	M-, (existing)	may need new xref-related code	Suggested by @Chris00