Re: master 050b830 3/3: Do interactive mode tagging for finder.el

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Eli Zaretskii]

> @@ -423,7 +423,7 @@ FILE should be in a form suitable for passing to `locate-library'."
>  
>  (defun finder-mouse-select (event)
>    "Select item in a Finder buffer with the mouse."
> -  (interactive "e")
> +  (interactive "e" finder-mode)
>    (with-current-buffer (window-buffer (posn-window (event-start event)))
>      (goto-char (posn-point (event-start event)))
>      (finder-select)))

I think this part is dangerous: mouse commands are not necessarily
issued with the buffer on which they operate being current, so
conditioning such commands on the mode of the current buffer is
ill-advised.  Even if currently this marking doesn't produce any
visible problems (does it?), it could do so in the future, if and when
we decide to base more on this tagging than we do now.

I think this part should be reverted, and in the future commands bound
to mouse clicks should not be tagged with any major mode.

Thanks.

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> I think this part is dangerous: mouse commands are not necessarily
> issued with the buffer on which they operate being current, so
> conditioning such commands on the mode of the current buffer is
> ill-advised.

It's a command that is applicable only to finder-mode buffers, so I
think the tagging is correct?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: Stefan Kangas <stefan@marxist.se>,  emacs-devel@gnu.org
> Date: Fri, 19 Mar 2021 09:28:16 +0100
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > I think this part is dangerous: mouse commands are not necessarily
> > issued with the buffer on which they operate being current, so
> > conditioning such commands on the mode of the current buffer is
> > ill-advised.
> 
> It's a command that is applicable only to finder-mode buffers, so I
> think the tagging is correct?

It depends on what we do with the tag.

"Applicable only to finder-mode buffers" doesn't tell how we use this
information.  We use it to filter out commands as irrelevant in
certain contexts.  But this command _is_ relevant even when the major
mode of the current buffer is something other than finder-mode,
because the user can click on text in a non-selected window.

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Stefan Kangas]

Eli Zaretskii <eliz@gnu.org> writes:

>> It's a command that is applicable only to finder-mode buffers, so I
>> think the tagging is correct?
>
> It depends on what we do with the tag.
>
> "Applicable only to finder-mode buffers" doesn't tell how we use this
> information.  We use it to filter out commands as irrelevant in
> certain contexts.  But this command _is_ relevant even when the major
> mode of the current buffer is something other than finder-mode,
> because the user can click on text in a non-selected window.

Currently we use it to filter commands in `M-x', where you anyways can't
use it (it will signal an error as EVENT is not bound).  So this change
doesn't break anything.  What am I missing?

If the above is correct, then that also means that there is no point to
show such commands even in `finder-mode'.  So it would be even better to
add some form that marks a command as only relevant to the mouse and
then never show those commands in `M-x'.  ISTR we discussed something
like that but I'm not sure if we reached any conclusion.

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Eli Zaretskii]

> From: Stefan Kangas <stefan@marxist.se>
> Date: Fri, 19 Mar 2021 11:23:25 -0500
> Cc: emacs-devel@gnu.org
> 
> Currently we use it to filter commands in `M-x', where you anyways can't
> use it (it will signal an error as EVENT is not bound).  So this change
> doesn't break anything.  What am I missing?

You are missing what I wrote in my OP.

> If the above is correct, then that also means that there is no point to
> show such commands even in `finder-mode'.  So it would be even better to
> add some form that marks a command as only relevant to the mouse and
> then never show those commands in `M-x'.  ISTR we discussed something
> like that but I'm not sure if we reached any conclusion.

That's a separate issue.  My point is that we shouldn't tag commands
with a mode unless we are 110% sure this tagging is correct and cannot
do any harm.  If there's doubt, there's no doubt.

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Stefan Kangas]

Eli Zaretskii <eliz@gnu.org> writes:

>> If the above is correct, then that also means that there is no point to
>> show such commands even in `finder-mode'.  So it would be even better to
>> add some form that marks a command as only relevant to the mouse and
>> then never show those commands in `M-x'.  ISTR we discussed something
>> like that but I'm not sure if we reached any conclusion.
>
> That's a separate issue.  My point is that we shouldn't tag commands
> with a mode unless we are 110% sure this tagging is correct and cannot
> do any harm.  If there's doubt, there's no doubt.

Makes sense.  I'll revert that part of the change.

Should we take this opportunity to discuss the above, though?  I think
it would be useful to have something like it.  But I don't have any
great ideas for what it could look like.

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Dmitry Gutov]

On 19.03.2021 18:23, Stefan Kangas wrote:
> So it would be even better to
> add some form that marks a command as only relevant to the mouse and
> then never show those commands in `M-x'.

Perhaps that could be inferred from the presence of 'e' in the 
interactive spec?

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Stefan Monnier]

>> So it would be even better to
>> add some form that marks a command as only relevant to the mouse and
>> then never show those commands in `M-x'.
>
> Perhaps that could be inferred from the presence of 'e' in the
> interactive spec?

`@` in the interactive spec seems even more relevant.


        Stefan

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Eli Zaretskii]

> From: Stefan Kangas <stefan@marxist.se>
> Date: Fri, 19 Mar 2021 12:08:00 -0500
> Cc: larsi@gnus.org, emacs-devel@gnu.org
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> If the above is correct, then that also means that there is no point to
> >> show such commands even in `finder-mode'.  So it would be even better to
> >> add some form that marks a command as only relevant to the mouse and
> >> then never show those commands in `M-x'.  ISTR we discussed something
> >> like that but I'm not sure if we reached any conclusion.
> >
> > That's a separate issue.  My point is that we shouldn't tag commands
> > with a mode unless we are 110% sure this tagging is correct and cannot
> > do any harm.  If there's doubt, there's no doubt.
> 
> Makes sense.  I'll revert that part of the change.

Thanks.

> Should we take this opportunity to discuss the above, though?  I think
> it would be useful to have something like it.  But I don't have any
> great ideas for what it could look like.

Sure, let's discuss.  For starters, how about not showing in "M-x"
completions any commands whose interactive spec expects a mouse event?

RE: [External] : Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Drew Adams]

> Sure, let's discuss.  For starters, how about not showing in "M-x"
> completions any commands whose interactive spec expects a mouse event?

Once again, I hope you're talking about optional
behavior.  And hopefully opt-in.

(Cue the recent discussion about using `M-x' for
more than just invoking a command.)

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> That's a separate issue.  My point is that we shouldn't tag commands
> with a mode unless we are 110% sure this tagging is correct and cannot
> do any harm.  If there's doubt, there's no doubt.

There is no doubt.  This is a mode-specific command and should be tagged
as such.

Now, what we do with the tagging later is a different issue.  The
combination of a mode tagging + "e" spec suggests that we can do
something more.

But Stefan's tagging is 100% correct, and should not be reverted.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Lars Ingebrigtsen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> There is no doubt.  This is a mode-specific command and should be tagged
> as such.

(If there is something missing in the manual to clarify why this tagging
is correct, perhaps that can be added, so that no doubt remains for
anybody?)

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: Stefan Kangas <stefan@marxist.se>,  emacs-devel@gnu.org
> Date: Sat, 20 Mar 2021 08:51:33 +0100
> 
> This is a mode-specific command

But it isn't: I can issue this command from a buffer that is under
_any_ mode.

I guess the disagreement is about what exactly does "mode-specific"
mean here.

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> I guess the disagreement is about what exactly does "mode-specific"
> mean here.

If there are unclear points in the manual under

  21.2.4 Specifying Modes For Commands

I can try to clarify.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: stefan@marxist.se,  emacs-devel@gnu.org
> Date: Sat, 20 Mar 2021 10:05:48 +0100
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > I guess the disagreement is about what exactly does "mode-specific"
> > mean here.
> 
> If there are unclear points in the manual under
> 
>   21.2.4 Specifying Modes For Commands
> 
> I can try to clarify.

Just what I wrote above: that subsection talks about commands "tied to
a mode", or commands that can be used "in pretty much any mode", or
make no sense "outside" of a mode, or commands that "belong to a
mode", without clarifying what that means.

My interpretation of the notion of "being in the FOO mode" is that the
current buffer has FOO as its major mode.  Under that interpretation,
commands that can only be invoked by mouse gestures cannot be
"mode-specific", because such commands normally can be invoked while
the current buffer is any buffer under any mode.  The relevant buffer
is provided by the mouse event, not by the current buffer.

Re: master 050b830 3/3: Do interactive mode tagging for finder.el

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> My interpretation of the notion of "being in the FOO mode" is that the
> current buffer has FOO as its major mode.  Under that interpretation,
> commands that can only be invoked by mouse gestures cannot be
> "mode-specific", because such commands normally can be invoked while
> the current buffer is any buffer under any mode.  The relevant buffer
> is provided by the mouse event, not by the current buffer.

OK, I'll try to clarify that.  The "being in the FOO mode" in a mouse
context is ambiguous (and in that case, the "being" is inherent in the
mouse event, not by `current-buffer'.)

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no