dropped 4th arg from all-completions

dropped 4th arg from all-completions

[Drew Adams]

I object to Emacs 23.2 dropping the 4th arg from `all-completions'.

I do not understand why the feature of ignoring candidates that start with a
space was removed. It is an optional arg - no call is forced to use it.

FWIW, I use this 4th arg. I even have a user option and a minibuffer toggle key
that lets users control the removal of such candidates. If you don't want to use
the feature, that's one thing, but why remove it from Emacs?

AFAIK there was no proposal to drop the 4th arg, no discussion here. If similar
things have also been done elsewhere, I have the same objection. This is not the
right way.

Re: dropped 4th arg from all-completions

[Stefan Monnier]

> I object to Emacs 23.2 dropping the 4th arg from `all-completions'.

My 23.2 code shows all-completions accepts 4 args.
Heck, even the 24 branch still accepts 4 args.

> I do not understand why the feature of ignoring candidates that start
> with a space was removed.

It doesn't have its place in `all-completions' (which is an internal
function, and hence not directly related to any UI feature).

> FWIW, I use this 4th arg.  I even have a user option and a minibuffer
> toggle key that lets users control the removal of such candidates.
> If you don't want to use the feature, that's one thing, but why remove
> it from Emacs?

You can get the same result by changing the completion table instead, or
by changing the `predicate'.  These approaches also have the advantage
of being a lot more flexible.


        Stefan

RE: dropped 4th arg from all-completions

[Drew Adams]

> > I object to Emacs 23.2 dropping the 4th arg from `all-completions'.
> 
> My 23.2 code shows all-completions accepts 4 args.
> Heck, even the 24 branch still accepts 4 args.

Your "acceptance" is hollow - bad faith. Emacs does not raise an error, but it
does not respect the arg. You have broken any code that uses a non-nil 4th arg.
And the doc shows only 3 args, and the byte compiler warns about #4.

> > I do not understand why the feature of ignoring candidates 
> > that start with a space was removed.
> 
> It doesn't have its place in `all-completions' (which is an internal
> function, and hence not directly related to any UI feature).

I disagree that it is an internal function.

> > FWIW, I use this 4th arg.  I even have a user option and a 
> > minibuffer toggle key that lets users control the removal
> > of such candidates. If you don't want to use the feature,
> > that's one thing, but why remove it from Emacs?
> 
> You can get the same result by changing the completion table 
> instead, or by changing the `predicate'.  These approaches
> also have the advantage of being a lot more flexible.

My code works with multiple versions of Emacs. There is no reason that
`all-completions' cannot continue to support this argument. It has worked and
been used by Emacs users since Day One.

No proposal for the change. *No reason* given. No discussion.
When will you change the name from Emacs to Monnier?

Re: dropped 4th arg from all-completions

[Chong Yidong]

"Drew Adams" <drew.adams@oracle.com> writes:

>> My 23.2 code shows all-completions accepts 4 args.
>> Heck, even the 24 branch still accepts 4 args.
>
> Your "acceptance" is hollow - bad faith. Emacs does not raise an
> error, but it does not respect the arg.

(all-completions "" '(" asdfgh") nil nil)
   => (" asdf")

(all-completions "" '(" asdfgh") nil t)
   => nil

> And the doc shows only 3 args, and the byte compiler warns about #4.

The byte compiler warns, but the code still works.

RE: dropped 4th arg from all-completions

[Drew Adams]

> > Emacs does not raise an error, but it does not respect the arg.
> 
> (all-completions "" '(" asdfgh") nil nil)
>    => (" asdf")
>
> (all-completions "" '(" asdfgh") nil t)
>    => nil

Mille excuses for that part. I apparently checked only with a string that starts
with a space, which is the exceptional case where matching occurs anyway
("ignored unless STRING itself starts with a space"):

(all-completions " abc" '((" abc") ("a")) nil t)
=> (" abc")

That doesn't change the rest, but I'm glad this is still supported under the
covers.

The 4th arg should not be deprecated and it should be documented normally. And
changes like this should not be made with no proposal, no reasons given (still
the case), and no discussion.

Re: dropped 4th arg from all-completions

[Stefan Monnier]

> Your "acceptance" is hollow - bad faith.  Emacs does not raise an
> error, but it does not respect the arg.

IIRC That would be a bug.  Please provide a recipe.

>> > I do not understand why the feature of ignoring candidates 
>> > that start with a space was removed.
>> It doesn't have its place in `all-completions' (which is an internal
>> function, and hence not directly related to any UI feature).
> I disagree that it is an internal function.

Internal in the sense that it's used by completion code to provide the
UI, so its effect does not directly affect the end user.

> No proposal for the change. *No reason* given. No discussion.

AFAIK it was introduced for buffer completion and just happened to have
pretty much no effect on other forms of completions (because completion
candidates that start with a space are extremely rare).  Nowadays,
buffer completion provides the same feature in a different way, so this
arg ends up being completely unused.

In any case, I did give you the reason in the previous message:

   You can get the same result by changing the completion table instead, or
   by changing the `predicate'.  These approaches also have the advantage
   of being a lot more flexible.

And indeed AFAIK it's completely unused inside Emacs right now (and was
barely used, in an inconsistent way, and mostly by accident only) in
versions before I messed with it.

> When will you change the name from Emacs to Monnier?

Any time now,


        Stefan

Re: dropped 4th arg from all-completions

[Jay Belanger]

Stefan Monnier <monnier@iro.umontreal.ca> writes:
...
>> When will you change the name from Emacs to Monnier?
>
> Any time now,

I thought it was already called  Emacs: Monnier And Chong's Software.

Re: dropped 4th arg from all-completions

[David Kastrup]

Jay Belanger <jay.p.belanger@gmail.com> writes:

> Stefan Monnier <monnier@iro.umontreal.ca> writes:
> ...
>>> When will you change the name from Emacs to Monnier?
>>
>> Any time now,
>
> I thought it was already called  Emacs: Monnier And Chong's Software.

It's a good thing Drew does not throw tantrums.

-- 
David Kastrup

Re: dropped 4th arg from all-completions

[Stefan Monnier]

>>>> When will you change the name from Emacs to Monnier?
>>> Any time now,
>> I thought it was already called  Emacs: Monnier And Chong's Software.
> It's a good thing Drew does not throw tantrums.

?? IAGTDDNTT ??
That's not even close to "Emacs"?


        Stefan "probably missing something here"

Re: dropped 4th arg from all-completions

[Jay Belanger]

Stefan Monnier <monnier@iro.umontreal.ca> writes:
...
>>> I thought it was already called  Emacs: Monnier And Chong's Software.
...
> That's not even close to "Emacs"?

I thought perhaps Emacs was a recursive acronym; my bad.

Re: dropped 4th arg from all-completions

[Stephen J. Turnbull]

Jay Belanger writes:

 > I thought it was already called  Emacs: Monnier And Chong's Software.

Dammit, Jay, it's four weeks past April 1.  Give some warning in the
Subject line!