translation-table-for-input

translation-table-for-input

[Eli Zaretskii]

Compiling simple.el curses thusly:

    In quoted-insert:
    simple.el:692:22:Warning: `translation-table-for-input' is an obsolete
	variable (as of Emacs 23.1).

    In zap-to-char:
    simple.el:3150:21:Warning: `translation-table-for-input' is an obsolete
	variable (as of Emacs 23.1).
    simple.el:3151:28:Warning: `translation-table-for-input' is an obsolete
	variable (as of Emacs 23.1).

I don't think we want to release Emacs 23 with these warnings.  But
what should references to `translation-table-for-input' be replaced
with?  I looked in etc/NEWS, and all it says is that this variable is
obsolete, but doesn't offer any advice about the alternatives.

Re: translation-table-for-input

[Stefan Monnier]

> Compiling simple.el curses thusly:
>     In quoted-insert:
>     simple.el:692:22:Warning: `translation-table-for-input' is an obsolete
> 	variable (as of Emacs 23.1).

>     In zap-to-char:
>     simple.el:3150:21:Warning: `translation-table-for-input' is an obsolete
> 	variable (as of Emacs 23.1).
>     simple.el:3151:28:Warning: `translation-table-for-input' is an obsolete
> 	variable (as of Emacs 23.1).

> I don't think we want to release Emacs 23 with these warnings.  But
> what should references to `translation-table-for-input' be replaced
> with?  I looked in etc/NEWS, and all it says is that this variable is
> obsolete, but doesn't offer any advice about the alternatives.

This code is part of the implementation of the feature accessible via
translation-table-for-input, so removing it would amount to removing
the feature.

There is no alternative to this feature, BTW.  The only use I've ever
seen for it, was the char-unification we did in Emacs-22, but in
Emacs-23 we do this unification differently, which is why it's obsolete.

These `obsolete' warnings are a general problem: we have no way to say
"yes, I know this variable is obsolete, but this is the implementation
rather than a use of the feature".  The closest we have is
with-no-warnings, which is a rather blunt tool.


        Stefan

Re: translation-table-for-input

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: Kenichi Handa <handa@m17n.org>,  emacs-devel@gnu.org
> Date: Fri, 30 Jan 2009 20:50:50 -0500
> 
> > Compiling simple.el curses thusly:
> >     In quoted-insert:
> >     simple.el:692:22:Warning: `translation-table-for-input' is an obsolete
> > 	variable (as of Emacs 23.1).
> 
> >     In zap-to-char:
> >     simple.el:3150:21:Warning: `translation-table-for-input' is an obsolete
> > 	variable (as of Emacs 23.1).
> >     simple.el:3151:28:Warning: `translation-table-for-input' is an obsolete
> > 	variable (as of Emacs 23.1).
> 
> > I don't think we want to release Emacs 23 with these warnings.  But
> > what should references to `translation-table-for-input' be replaced
> > with?  I looked in etc/NEWS, and all it says is that this variable is
> > obsolete, but doesn't offer any advice about the alternatives.
> 
> This code is part of the implementation of the feature accessible via
> translation-table-for-input, so removing it would amount to removing
> the feature.
> 
> There is no alternative to this feature, BTW.  The only use I've ever
> seen for it, was the char-unification we did in Emacs-22, but in
> Emacs-23 we do this unification differently, which is why it's obsolete.

IIUC, you are saying that translation-table-for-input was only ever
used by Emacs itself.  In that case, there should be no need to mark
this variable as ``obsolete'', since we already got rid of its
obsolete uses in Emacs.  For other uses, this is a useful feature
(according to you), so marking it ``obsolete'' is simply misleading.

> These `obsolete' warnings are a general problem: we have no way to say
> "yes, I know this variable is obsolete, but this is the implementation
> rather than a use of the feature".  The closest we have is
> with-no-warnings, which is a rather blunt tool.

We could say something in the doc string, but what would that be?
``don't use this variable for character unification''?

Re: translation-table-for-input

[Stefan Monnier]

> IIUC, you are saying that translation-table-for-input was only ever
> used by Emacs itself.  In that case, there should be no need to mark
> this variable as ``obsolete'', since we already got rid of its
> obsolete uses in Emacs.

Yes, that's basically the situation, except that this only true
"AFAIK".  I.e. I am not 100% sure that nobody else uses it.
Maybe we should just get rid of it.  I even seem to remember that
I suggested getting rid of it at some point and that someone
(Handa maybe?) favored keeping it as obsolete for now.


        Stefan

Re: translation-table-for-input

[Richard M Stallman]

    These `obsolete' warnings are a general problem: we have no way to say
    "yes, I know this variable is obsolete, but this is the implementation
    rather than a use of the feature".  The closest we have is
    with-no-warnings, which is a rather blunt tool.

It is not blunt if you use it around a small piece of code
with a comment to explain why.

Re: translation-table-for-input

[Stefan Monnier]

>     These `obsolete' warnings are a general problem: we have no way to say
>     "yes, I know this variable is obsolete, but this is the implementation
>     rather than a use of the feature".  The closest we have is
>     with-no-warnings, which is a rather blunt tool.

> It is not blunt if you use it around a small piece of code
> with a comment to explain why.

It still silences any other warning that might occur (e.g. in the
future) in that same small piece of code.  So it's still blunt.


        Stefan

Re: translation-table-for-input

[Richard M Stallman]

    It still silences any other warning that might occur (e.g. in the
    future) in that same small piece of code.  So it's still blunt.

Yes, but it's rare that another warning occurs in the same small piece
of code.

Re: translation-table-for-input

[Eli Zaretskii]

> From: Richard M Stallman <rms@gnu.org>
> Date: Tue, 03 Feb 2009 04:59:07 -0500
> Cc: eliz@gnu.org, emacs-devel@gnu.org, handa@m17n.org
> 
>     It still silences any other warning that might occur (e.g. in the
>     future) in that same small piece of code.  So it's still blunt.
> 
> Yes, but it's rare that another warning occurs in the same small piece
> of code.

I ended up removing the make-obsolete-variable call for
translation-table-for-input, and instead added a blurb to its doc
string to discourage its use for unification purposes.

Re: translation-table-for-input

[Glenn Morris]

Eli Zaretskii wrote:

> I ended up removing the make-obsolete-variable call for
> translation-table-for-input,

I'm as anti-compilation-warnings as the next person, but I think that
was a bit much. It is obsolete and should be marked as such. I had no
problem just ignoring the warning.

Re: translation-table-for-input

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Cc: rms@gnu.org,  handa@m17n.org,  monnier@iro.umontreal.ca,  emacs-devel@gnu.org
> Date: Sat, 07 Feb 2009 21:00:20 -0500
> 
> Eli Zaretskii wrote:
> 
> > I ended up removing the make-obsolete-variable call for
> > translation-table-for-input,
> 
> I'm as anti-compilation-warnings as the next person, but I think that
> was a bit much. It is obsolete and should be marked as such.

No, it's not obsolete.  Its use for character code unification is
obsolete, but otherwise having a possibility to translate the output
of an input method is a valuable feature.  I understand Handa-san
objected to deleting it in the past.

Re: translation-table-for-input

[Stefan Monnier]

> I ended up removing the make-obsolete-variable call for
> translation-table-for-input, and instead added a blurb to its doc
> string to discourage its use for unification purposes.

Please don't.  If you really want to fix the warning then either use
with-no-warnings, or just remove the whole feature.
But don't just remove the call to make-obsolete-variable.

> No, it's not obsolete.

The "obsolete" thingy mostly means "slated for removal", AFAIC.  So,
yes, it is "obsolete" because I plan to remove it altogether.

> Its use for character code unification is obsolete, but otherwise
> having a possibility to translate the output of an input method is
> a valuable feature.

I see no such evidence.

> I understand Handa-san objected to deleting it in the past.

Yes, simply out of fear that someone somewhere might use it.  Hence the
`make-obsolete-variable' so aso to warn people that if they use it, they
should come and see us because the feature may disappear.


        Stefan

Re: translation-table-for-input

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: rms@gnu.org,  emacs-devel@gnu.org,  handa@m17n.org
> Date: Thu, 12 Feb 2009 16:58:54 -0500
> 
> > I ended up removing the make-obsolete-variable call for
> > translation-table-for-input, and instead added a blurb to its doc
> > string to discourage its use for unification purposes.
> 
> Please don't.

Too late, it's done.

> If you really want to fix the warning then either use
> with-no-warnings, or just remove the whole feature.
> But don't just remove the call to make-obsolete-variable.

Based on the discussion that followed my original question, I became
convinced that this variable is not obsolete at all.  Thus, what I did
eventually had nothing to do with eliminating the warning per se.

> > No, it's not obsolete.
> 
> The "obsolete" thingy mostly means "slated for removal", AFAIC.  So,
> yes, it is "obsolete" because I plan to remove it altogether.

Declaring something obsolete requires that we name which other feature
to use in its stead.  Do we have such an alternative in place?

> > Its use for character code unification is obsolete, but otherwise
> > having a possibility to translate the output of an input method is
> > a valuable feature.
> 
> I see no such evidence.

My evidence is that there's no alternative to translating the output
of an input method.

> > I understand Handa-san objected to deleting it in the past.
> 
> Yes, simply out of fear that someone somewhere might use it.  Hence the
> `make-obsolete-variable' so aso to warn people that if they use it, they
> should come and see us

I put the same information into the doc string, precisely for that
reason.

Re: translation-table-for-input

[Stefan Monnier]

> > > I ended up removing the make-obsolete-variable call for
> > > translation-table-for-input, and instead added a blurb to its doc
> > > string to discourage its use for unification purposes.
> > Please don't.
> Too late, it's done.

I know.  What I meant was "please revert it".

>> The "obsolete" thingy mostly means "slated for removal", AFAIC.  So,
>> yes, it is "obsolete" because I plan to remove it altogether.
> Declaring something obsolete requires that we name which other feature
> to use in its stead.  Do we have such an alternative in place?

Just as with make-local-hook, the replacement is to not put anything at
all, because the trasnlation is now automatically done.

>> > Its use for character code unification is obsolete, but otherwise
>> > having a possibility to translate the output of an input method is
>> > a valuable feature.
>> I see no such evidence.
> My evidence is that there's no alternative to translating the output
> of an input method.

That's not an evidence that it's "valuable".  The absence of any code
that uses that feature tells me it's not valuable.


        Stefan

Re: translation-table-for-input

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: rms@gnu.org,  emacs-devel@gnu.org,  handa@m17n.org
> Date: Fri, 13 Feb 2009 09:27:39 -0500
> 
> >> > Its use for character code unification is obsolete, but otherwise
> >> > having a possibility to translate the output of an input method is
> >> > a valuable feature.
> >> I see no such evidence.
> > My evidence is that there's no alternative to translating the output
> > of an input method.
> 
> That's not an evidence that it's "valuable".  The absence of any code
> that uses that feature tells me it's not valuable.

Look, if someone would suggest today adding such a feature, I'd
probably object on the grounds of lack of use.  But what on Earth
could be a good reason to remove something that is already there, have
been there for several Emacs releases, and is documented in the
manual??

And after you already agreed to leave it, what is pushing you so hard
to insist on declaring it obsolete?  Don't we have more important
things to do and argue about?

Re: translation-table-for-input

[Stefan Monnier]

> Look, if someone would suggest today adding such a feature, I'd
> probably object on the grounds of lack of use.  But what on Earth
> could be a good reason to remove something that is already there, have
> been there for several Emacs releases, and is documented in the
> manual??

It has various quirks, and doesn't provide any useful feature anymore.

> And after you already agreed to leave it, what is pushing you so hard
> to insist on declaring it obsolete?

Keeping the code maintainable is something I consider important.

> Don't we have more important things to do and argue about?

Indeed.  So just leave the make-variable-obsolete thing where it was,
and move on.


        Stefan

Re: translation-table-for-input

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: rms@gnu.org,  emacs-devel@gnu.org,  handa@m17n.org
> Date: Fri, 13 Feb 2009 10:54:51 -0500
> 
> > Look, if someone would suggest today adding such a feature, I'd
> > probably object on the grounds of lack of use.  But what on Earth
> > could be a good reason to remove something that is already there, have
> > been there for several Emacs releases, and is documented in the
> > manual??
> 
> It has various quirks, and doesn't provide any useful feature anymore.
> 
> > And after you already agreed to leave it, what is pushing you so hard
> > to insist on declaring it obsolete?
> 
> Keeping the code maintainable is something I consider important.
> 
> > Don't we have more important things to do and argue about?
> 
> Indeed.  So just leave the make-variable-obsolete thing where it was,
> and move on.

Done, under protest.

Re: translation-table-for-input

[Stefan Monnier]

>> Indeed.  So just leave the make-variable-obsolete thing where it was,
>> and move on.
> Done, under protest.

Thank you.  I note the vote of protest.


        Stefan