João Távora <joaotavora@gmail.com> writes:

> Elijah Gabe Pérez <eg642616@gmail.com> writes:
> Your ideas are good, but did you test your patch thoroughly?  I did, see
> comments below.
>
>>                 (overlay-buffer eglot--suggestion-overlay))
>> -      (overlay-get eglot--suggestion-overlay 'eglot--suggestion-tooltip)))
>> +      (eglot--mode-line-props
>> +       eglot-code-action-indicator
>> +       'help-echo
>> +       `((mouse-1
>> +          eglot-code-actions-at-mouse
>> +          ,(plist-get (aref (overlay-get eglot--suggestion-overlay 'eglot--actions) 0) :title)))
>> +       "Execute code actions at point\n")))
>
> This removes the face from the mode-line indication and uses 'help-echo.
> Probably should be eglot-code-action-indicator-face.

Oops, sorry, I thought that function was a kind of `propertize'.

> Also the blurb looks like.
>
>    Execute code actions at point. mouse-1: extract subexpression to variable
>
> Maybe it should be:
>
>    mouse-1: execute code actions at point
>
> since your second sentence is misleading, as it will actually pop a
> menu, not execute the actual action right away.

I agree, fixed.

>>  (add-to-list
>> @@ -4130,7 +4136,7 @@ eglot-code-action-suggestion
>>             (when (cl-plusp (length actions))
>>               (setq blurb
>>                     (substitute-command-keys
>> -                    (eglot--format "\\[eglot-code-actions]: %s"
>> +                    (eglot--format "\\[eglot-code-actions] or
>>  \\[eglot-code-actions-at-mouse]: %s"
>
> This isn't right, did you test this?  The second part shows the user
> instructions that when followed do not do what they say they do.  For me
> it says:
>
>    <f2> or M-x eglot-code-actions-at-mouse: extract subexpression to variable
>
> I have M-x eglot-code-actions bound to <f2> of course.  In my case it
> should say 'mouse-2' as that is the (default) button that pops up a
> menu.  So it should be
>
>   <f2>: extract subexpression to variable, mouse-2: execute code actions
>   at point
>
> Needless to say if I press 'M-x eglot-code-actions-at-mouse' it doesn't
> work, and I get an error saying there is no mouse event when invoking
> the command.  So substitute-command-keys might not be a solution here,
> as you can't easily get it to be aware of the special map and not say
> anything if nothing is bound there.
>
> Even more problematic is the fact that it shows the changed blurb in the
> ElDoc area, in situations where the mouse (if there even is one!) is not
> hovering on the margin or fringe (if there even are fringes).
>
> So this isn't a straightforward problem to solve, which probably
> explains why I left this minor detail out.
>
> I also feel that if the code to do this grows too complicated, it's
> better to add generic utility to Emacs to solve these problems and
> conditionally use it in eglot.el with a 'fboundp' check.

A way I found for achieve this is only (re)propertizing the blurb
help-echo for `margin' and `nearby' options (like the patch below).
This let eldoc area unchanged (i tested it).

I used `substitute-command-keys' only for get the mouse-1/2/3 button for
margin, but since i see this is problematic, I've changed it to only
display `mouse-2'.

The help-echo for `margin' and `nearby' now should be displayed as:
"M-x eglot-code-actions: remove #include directive (and 1 more actions),
mouse-2: execute code actions at point"

and for the mode-line: "mouse-1: Execute code actions at point."