Re: master 38697a07c0 1/2: Add :safe property to ffap-machine-p-* variables

Re: master 38697a07c0 1/2: Add :safe property to ffap-machine-p-* variables

[Eli Zaretskii]

> branch: master
> commit 38697a07c0f2b99b76c41cb1096543681342a405
> Author: Stefan Kangas <stefan@marxist.se>
> Commit: Stefan Kangas <stefan@marxist.se>
> 
>     Add :safe property to ffap-machine-p-* variables
>     
>     * lisp/ffap.el (ffap--accept-or-reject-p): New predicate defun.
>     (ffap-machine-p-local, ffap-machine-p-known)
>     (ffap-machine-p-unknown): Add :safe property using above new
>     predicate.

This uses an internal function as the value of a property of a user
option, which I think is not a good idea.  How about making the
function public instead (and changing its name to
'ffap-accept-or-reject-p' while at that)?

Re: master 38697a07c0 1/2: Add :safe property to ffap-machine-p-* variables

[Stefan Monnier]

> This uses an internal function as the value of a property of a user
> option, which I think is not a good idea.

Why is that a problem?  Internal names means users shouldn't rely on the
fact that this function exists and will keep existing under that name or
that its behavior will be preserved, but here the users aren't expected
to call this function under this internal name, they're just expected to
call the function that happens to be stored in the
`safe-local-variable`, regardless of its name and even if
it's anonymous.


        Stefan

Re: master 38697a07c0 1/2: Add :safe property to ffap-machine-p-* variables

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: Stefan Kangas <stefan@marxist.se>,  emacs-devel@gnu.org
> Date: Fri, 08 Jul 2022 09:03:25 -0400
> 
> > This uses an internal function as the value of a property of a user
> > option, which I think is not a good idea.
> 
> Why is that a problem?

It's not nice.  If we are serious about internal things being
internal, we should avoid exposing them to users.

Re: master 38697a07c0 1/2: Add :safe property to ffap-machine-p-* variables

[Stefan Monnier]

Eli Zaretskii [2022-07-08 16:16:57] wrote:
>> From: Stefan Monnier <monnier@iro.umontreal.ca>
>> Cc: Stefan Kangas <stefan@marxist.se>,  emacs-devel@gnu.org
>> Date: Fri, 08 Jul 2022 09:03:25 -0400
>> > This uses an internal function as the value of a property of a user
>> > option, which I think is not a good idea.
>> Why is that a problem?
> It's not nice.  If we are serious about internal things being
> internal, we should avoid exposing them to users.

I don't see why.  AFAIK it's definitely no worse (and arguably better)
than having an anonymous function, and I've never heard anyone complain
about exposing an anonymous function on `safe-local-variable`.

To me, this is similar to

    (defalias 'foo-bar #'foo--bar-thingy)

The fact that the value stored is an internal function is not important
because the official API is "call `foo-bar`" and we "promise" that this
API will be preserved even if `foo--bar-thingy` disappears or changes.


        Stefan