bug#65519: 30.0.50; [FR Eglot] keymaps for useful functions

[Gerd Möllmann]

João Távora <joaotavora@gmail.com> writes:

> Gerd Möllmann <gerd.moellmann@gmail.com> writes:
>
>> There are some useful eglot functions that can currently only invoked
>> with M-x, by default.  I'm currently defining a keymap
>>
>> (define-keymap :prefix 'my-eglot-bindings
>>   "a" 'eglot-code-actions
>>   "f" 'eglot-format
>>   "h" 'eglot-inlay-hints-mode
>>   "r" 'eglot-rename)
>>
>> that I bind to a a prefix key in c-mode-common-hook.
>
> I think this is a fine way to go about it.
>
>> Feature request: Could eglot-mode do something like that?
>
> This request comes up often.  The reason Eglot does this is partly
> answered in the top bullet of eglot.el's description:
>
> ;; * Eglot's main job is to hook up the information that language
> ;;   servers offer via LSP to Emacs's UI facilities: Xref for
> ;;   definition-chasing, Flymake for diagnostics, Eldoc for at-point
> ;;   documentation, etc.  Eglot's job is generally *not* to provide
> ;;   such a UI itself, though a small number of simple
> ;;   counter-examples do exist, e.g. in the `eglot-rename' command or
> ;;   the `eglot-inlay-hints-mode' minor mode.  When a new UI is
> ;;   evidently needed, consider adding a new package to Emacs, or
> ;;   extending an existing one.
>
> So this is how Eglot wants to sell itself: to be a low-profile
> middle-man between Emacs and LSP.  It doesn't always work and people
> understandibly want Eglot to be more of a front-man.
>
> I try to resist this temptation as much as possible because that leads
> to bloat and duplicated functionality with idiosyncractic interfaces.

Ok.

> So I always push for other "standard" packages to provide the UI.  But
> as you can see in those 4 examples, I ended up being pragmatic and
> putting the command in Eglot itself.
>
> But it's not unthinkable (in fact, it's desired I think and I've looked
> into it) that 'eglot-format' simply dissolves into Emacs's longstanding
> 'indent.el' machinery.  And there's some talk of a "refactoring
> interface" for Emacs in emacs-devel (I think).  So that's where
> 'eglot-code-actions' and 'eglot-rename' should really live.
>
> In short, I think it's OK to be pragmatic.  Why not make a
> eglot-bindings.el package?

You mean making an ELPA package? Sorry, that's too much effort for me.
Maybe someone else will do it.