Why all the Alt bindings by default?

Why all the Alt bindings by default?

[Drew Adams]

emacs -Q in Emacs 24, visit an Emacs-Lisp buffer, `C-h b'.
Do the same in Emacs 23.3.

In Emacs 24 (but not in 23.3) I see lots of Alt bindings that presumably (?)
insert characters:

A-SPC		 
A-!		¡
A-"		Prefix Command
A-$		¤
A-'		Prefix Command
A-*		Prefix Command
A-+		±
A-,		Prefix Command
A--		­
A-.		·
A-/		Prefix Command
A-1		Prefix Command
A-3		Prefix Command
A-<		«
A-=		¯
A->		»
A-?		¿
A-C		©
A-L		£
A-P		¶
A-R		®
A-S		§
A-Y		¥
...

Actually, I don't know what these keys do.  The fancy characters are shown in
the `binding' column.  What does it mean for a key to be bound to a character
(as opposed to command `self-insert-command')?

I don't even know how to hit such keys to see what `C-h k' says.  By default,
Alt is Meta, at least on most platforms.  I tried `(describe-key (kbd "A-/"))'
etc., but that just says "A-/ is undefined".

What's this all about?  

Searching for `Alt', or `A-', or even `Unicode' turns up nothing in NEWS.

Also, (emacs) Modifier Keys says this explicitly:

"Emacs supports three other modifier keys.  These are called
 <Super>, <Hyper> and <Alt>.  Few terminals provide ways to use these
 modifiers; the key labeled <Alt> on most keyboards usually issues the
 <Meta> modifier, not <Alt>.  The standard key bindings in Emacs do not
 include any characters with these modifiers.  However, you can
 customize Emacs to assign meanings to them.  The modifier bits are
 labelled as `s-', `H-' and `A-' respectively."

Apparently it is no longer true that "The standard key bindings...do not include
any characters with these modifiers." (?)

Re: Why all the Alt bindings by default?

[Andreas Schwab]

These are key translations.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

RE: Why all the Alt bindings by default?

[Drew Adams]

> These are key translations.

Doesn't help.

What is the new feature?  Why?  What does it mean for users?  Where is it
documented?...

Re: Why all the Alt bindings by default?

[Stefan Monnier]

> A-SPC		 
> A-!		¡
> A-"		Prefix Command
> A-$		¤
> A-'		Prefix Command
> A-*		Prefix Command
> A-+		±
> A-,		Prefix Command
> A--		­
> A-.		·
> A-/		Prefix Command
> A-1		Prefix Command
> A-3		Prefix Command
> A-<		«
> A-=		¯
> A->		»
> A-?		¿
> A-C		©
> A-L		£
> A-P		¶
> A-R		®
> A-S		§
> A-Y		¥
> ...

These are key remappings implemented in key-translation-map.  They are
available under C-x 8 as well as under the Alt modifier (I didn't
remember this and had to track it down to re-learn it).

> Actually, I don't know what these keys do.  The fancy characters are shown in
> the `binding' column.  What does it mean for a key to be bound to a character
> (as opposed to command `self-insert-command')?

These are not key bindings, but key remappings.

> I don't even know how to hit such keys to see what `C-h k' says.  By default,
> Alt is Meta, at least on most platforms.

Yes.  I think that's why it doesn't get in the way.

> I tried `(describe-key (kbd "A-/"))' etc., but that just says "A-/ is
> undefined".

Indeed, it's not bound.

BTW in my tests "emacs -Q" followed by "C-h b" does not show those Alt
mappings.  Only after loading iso-transl (which is triggered by hitting
C-x 8 but also by C-h b, so a second C-h b will show the Alt mappings
oddly enough).


        Stefan

RE: Why all the Alt bindings by default?

[Drew Adams]

> > A-Y		¥
> > ...
> 
> These are key remappings implemented in key-translation-map.  They are
> available under C-x 8 as well as under the Alt modifier (I didn't
> remember this and had to track it down to re-learn it).

Yes, I guessed that they were the same as `C-x 8'.

1. I think that in both cases we need to identify them somehow as key remappings
in the output.

2. What is the new feature that causes the `A-' bindings to now show up in
addition to the `C-x 8' bindings?  Shouldn't that feature be called out in NEWS,
even if it is just a new set of key remappings?  What was the rationale behind
the change?

Re: Why all the Alt bindings by default?

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Date: Wed, 26 Oct 2011 09:50:34 -0700
> 
> emacs -Q in Emacs 24, visit an Emacs-Lisp buffer, `C-h b'.
> Do the same in Emacs 23.3.
> 
> In Emacs 24 (but not in 23.3) I see lots of Alt bindings that presumably (?)
> insert characters:
> 
> A-SPC		 
> A-!		¡
> A-"		Prefix Command
> A-$		¤
> A-'		Prefix Command
> A-*		Prefix Command
> A-+		±

I cannot reproduce this, FWIW, not with the recipe shown above.

Re: Why all the Alt bindings by default?

[Andreas Schwab]

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

> What is the new feature?

There is nothing new here.  This is almost as old as Emacs.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

RE: Why all the Alt bindings by default?

[Drew Adams]

> > emacs -Q in Emacs 24, visit an Emacs-Lisp buffer, `C-h b'.
> > Do the same in Emacs 23.3.
> > 
> > In Emacs 24 (but not in 23.3) I see lots of Alt bindings 
> > that presumably (?) insert characters:
> > 
> > A-SPC		 
> > A-!		¡
> > A-"		Prefix Command
> > A-$		¤
> > A-'		Prefix Command
> > A-*		Prefix Command
> > A-+		±
> 
> I cannot reproduce this, FWIW, not with the recipe shown above.

Read Stefan's replu.  He provided the necessary clarification: hit `C-h b'
twice.

RE: Why all the Alt bindings by default?

[Drew Adams]

> > What is the new feature?
> 
> There is nothing new here.  This is almost as old as Emacs.

I thought that seeing `A-/' etc. listed as a binding was new in Emacs 24.  But
following the same recipe in Emacs 23.2 and 23.3 (but not 23.1) shows the same
thing.  So this was new with Emacs 23.2, not Emacs 24, and it belongs in the
Emacs 23.2 NEWS.

Or maybe you feel that it is a bug rather than a feature or the consequence of a
feature?

Re: Why all the Alt bindings by default?

[Tim Cross]

On Thu, Oct 27, 2011 at 7:20 AM, Drew Adams <drew.adams@oracle.com> wrote:
>> > What is the new feature?
>>
>> There is nothing new here.  This is almost as old as Emacs.
>
> I thought that seeing `A-/' etc. listed as a binding was new in Emacs 24.  But
> following the same recipe in Emacs 23.2 and 23.3 (but not 23.1) shows the same
> thing.  So this was new with Emacs 23.2, not Emacs 24, and it belongs in the
> Emacs 23.2 NEWS.
>
> Or maybe you feel that it is a bug rather than a feature or the consequence of a
> feature?
>
>
>

Regardless of whether it appeared in emacs 23 or 24 or whether it
requires more than 1 C-h b for the Alt 'bindings' to be shown with C-h
b, the bottom line is that these key translations are shown as key
bindings when they are not. Apart from being misleading for users, I
find it a pain having to scroll past all of these misleading entries
to get to what I'm after - information on key bindings.

Something is needed that at the very least will filter these entries
out of C-h b.

One question I do have - what happens if you do have a keyboard that
is setup to have both meta and alt (as well as super and hypa) and you
want to do a real alt binding using one of these keys? It is fairly
trivial to setup your keyboard to have hypa nd supa under X, so I
suspect adding alt is not that hard given the number of keyboards with
extra/dead keys available.

Tim


-- 
Tim Cross

Re: Why all the Alt bindings by default?

[Stephen J. Turnbull]

Tim Cross writes:

 > One question I do have - what happens if you do have a keyboard that
 > is setup to have both meta and alt (as well as super and hypa) and you
 > want to do a real alt binding using one of these keys?

Most PC-based systems do define Alt as well as Meta, but good luck
using them if you have a desktop installed: the desktops typically
grab a couple dozen for their own use, leaving Meta free for the
programs.  Also, accessibility features often use Alt to access the
menus (XEmacs does this).  So I would say "binding Alt is not
something that Emacs devs should worry about, although users are
welcome to do it themselves".

As for what happens, I believe translations happen at a lower level
(ie, earlier in the event-processing pipeline) than binding, so to use
those Alt-KEY combinations you would need to disable their
translations as well as bind the keys.

Re: Why all the Alt bindings by default?

[Tim Cross]

On Thu, Oct 27, 2011 at 12:48 PM, Stephen J. Turnbull
<stephen@xemacs.org> wrote:
> Tim Cross writes:
>
>  > One question I do have - what happens if you do have a keyboard that
>  > is setup to have both meta and alt (as well as super and hypa) and you
>  > want to do a real alt binding using one of these keys?
>
> Most PC-based systems do define Alt as well as Meta, but good luck
> using them if you have a desktop installed: the desktops typically
> grab a couple dozen for their own use, leaving Meta free for the
> programs.  Also, accessibility features often use Alt to access the
> menus (XEmacs does this).  So I would say "binding Alt is not
> something that Emacs devs should worry about, although users are
> welcome to do it themselves".

Yes, the very first thing I do whith a new system is remove most of
the desktop shortcuts. 80% of my work is done inside emacs and it is
inside emacs I need as many key binding options as possible, plus most
of the desktop shortucts I never use, so they are wasted.

>
> As for what happens, I believe translations happen at a lower level
> (ie, earlier in the event-processing pipeline) than binding, so to use
> those Alt-KEY combinations you would need to disable their
> translations as well as bind the keys.
>

I think the main issue here is that if I do have real alt bindings and
I want to use C-h b to verify or remind me of what bindings are
available in a mode, I now have no easy way of knowing which are real
bindings and which are 'fake' bindings i.e. key translations. While I
personally am aware of the translation entries, many other users would
not. Less experienced users are also likely to be the more frequent
users of C-h b.

For me, its really just a minor annoyance as I have to scroll past
lots of 'fake' alt bindings to see the real ones.

Tim



-- 
Tim Cross

Re: Why all the Alt bindings by default?

[Eli Zaretskii]

> Date: Thu, 27 Oct 2011 08:35:47 +1100
> From: Tim Cross <theophilusx@gmail.com>
> Cc: Andreas Schwab <schwab@linux-m68k.org>, emacs-devel@gnu.org
> 
> One question I do have - what happens if you do have a keyboard that
> is setup to have both meta and alt (as well as super and hypa) and you
> want to do a real alt binding using one of these keys?

Are you saying that you can actually get these "bindings" to work?
That is, if you press Alt-R, you get the registered trademark?
Because that's not what happens to me.  In my case, Emacs says "A-R is
undefined".

Re: Why all the Alt bindings by default?

[Tim Cross]

On Thu, Oct 27, 2011 at 3:07 PM, Eli Zaretskii <eliz@gnu.org> wrote:
>> Date: Thu, 27 Oct 2011 08:35:47 +1100
>> From: Tim Cross <theophilusx@gmail.com>
>> Cc: Andreas Schwab <schwab@linux-m68k.org>, emacs-devel@gnu.org
>>
>> One question I do have - what happens if you do have a keyboard that
>> is setup to have both meta and alt (as well as super and hypa) and you
>> want to do a real alt binding using one of these keys?
>
> Are you saying that you can actually get these "bindings" to work?
> That is, if you press Alt-R, you get the registered trademark?
> Because that's not what happens to me.  In my case, Emacs says "A-R is
> undefined".
>
>

No, what I was asking is what happens if I do want to bind alt-R? I
think this was answered in another post i.e. you have to 'undo' the
key translation first. The other question is if I have defined some
real alt bindings, how do you tell those form key translations?

Tim


-- 
Tim Cross

Re: Why all the Alt bindings by default?

[Stephen J. Turnbull]

Tim Cross writes:

 > The other question is if I have defined some real alt bindings, how
 > do you tell those form key translations?

In what context?  If you're a user looking at the C-h b buffer,
heuristically: non-ASCII single characters are not command bindings,
they're keyboard translations.  In a program, you'll use different
APIs to access them, so one can hardly be mistaken for the other.

Re: Why all the Alt bindings by default?

[Tim Cross]

On Thu, Oct 27, 2011 at 3:52 PM, Stephen J. Turnbull <stephen@xemacs.org> wrote:
> Tim Cross writes:
>
>  > The other question is if I have defined some real alt bindings, how
>  > do you tell those form key translations?
>
> In what context?  If you're a user looking at the C-h b buffer,
> heuristically: non-ASCII single characters are not command bindings,
> they're keyboard translations.  In a program, you'll use different
> APIs to access them, so one can hardly be mistaken for the other.
>

Yes, but is that reasonable in the context of a new user who is likely
using C-h b to identify and learn about available key bindings? It
would seem a reasonable assumption that when you use a command which
is supposed to describe key bindings, what that command shows is key
bindings and not key bindings and key translations.

The doc string for the command doesn't mention anything about key
translations. Nor does there seem to be much in the emacs manual.
Perhaps all that is required is for this to be added to the doc
string. I think this would then largely reduce the issue to minor
annoyance at the list of key bindings including data I'm probably not
interested in and didn't ask for.

Tim


-- 
Tim Cross

Re: Why all the Alt bindings by default?

[Eli Zaretskii]

> Date: Thu, 27 Oct 2011 15:14:22 +1100
> From: Tim Cross <theophilusx@gmail.com>
> Cc: schwab@linux-m68k.org, drew.adams@oracle.com, emacs-devel@gnu.org
> 
> No, what I was asking is what happens if I do want to bind alt-R? I
> think this was answered in another post i.e. you have to 'undo' the
> key translation first.

That's false.  Just try it.  I tried this:

  M-x global-set-key RET
  Alt-R
  ignore RET

After that, "C-h c Alt-R" says

  ® (translated from A-R) runs the command ignore

> The other question is if I have defined some
> real alt bindings, how do you tell those form key translations?

Why would you need to?

Re: Why all the Alt bindings by default?

[Tim Cross]

On Thu, Oct 27, 2011 at 7:24 PM, Eli Zaretskii <eliz@gnu.org> wrote:
>> The other question is if I have defined some
>> real alt bindings, how do you tell those form key translations?
>
> Why would you need to?
>

I'm old and forgetful. sometimes I need a reminder about what key
bindings are available in a mode. As pointed out by SJT,
heuristically, I can distinguish key translations from key bindings by
the thing they are bound to, but maybe I forget about this (did I
mention I'm old and forgetful?)

Tim

P.S. In all seriousness, not a big issue, though I do think it is
unfortunate and a possible source of confusion to have both key
bindings and key translations displayed in a buffer created by a
command to describe key bindings.  At the very least, updated doc
string informing users about the key translations would probably be
worthwhile.



-- 
Tim Cross

Re: Why all the Alt bindings by default?

[Stefan Monnier]

>> No, what I was asking is what happens if I do want to bind alt-R? I
>> think this was answered in another post i.e. you have to 'undo' the
>> key translation first.

> That's false.  Just try it.  I tried this:

>   M-x global-set-key RET
>   Alt-R
>   ignore RET

> After that, "C-h c Alt-R" says

>   ® (translated from A-R) runs the command ignore

But then if you have some other key that generates ®, it will also be
rebound to `ignore' since you have really just done
(global-set-key [?®] 'ignore)

I think these Alt bindings should either be removed, or turned into
function-key-map bindings (so they don't take precedence over real key
bindings).
Also we should address the problem that those key-bindings are only
present after loading iso-transl.el.


        Stefan

RE: Why all the Alt bindings by default?

[Drew Adams]

> it is unfortunate and a possible source of confusion to have both key
> bindings and key translations displayed in a buffer created by a
> command to describe key bindings.  At the very least, updated doc
> string informing users about the key translations would probably be
> worthwhile.

I suggested distuishing bindings from key translations in the *Help* buffer.
There are several ways that could be done.  Perhaps the simplest for the user
would be to simply list key translations separately, with their own header.

RE: Why all the Alt bindings by default?

[Drew Adams]

> I think these Alt bindings should either be removed, or turned into
> function-key-map bindings (so they don't take precedence over real key
> bindings). Also we should address the problem that those key-bindings
> are only present after loading iso-transl.el.

That would be good, I think.  Or else list them separately, with their own
header, at the end of the buffer (so they don't get in the way).

Likewise, C-x 8 bindings?

Re: Why all the Alt bindings by default?

[Andreas Schwab]

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

> There are several ways that could be done.  Perhaps the simplest for the user
> would be to simply list key translations separately, with their own header.

Which they are already.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

RE: Why all the Alt bindings by default?

[Drew Adams]

d> There are several ways that could be done.  Perhaps the 
d> simplest for the user would be to simply list key
d> translations separately, with their own header.

> Which they are already.

Read further:

d> That would be good, I think.  Or else list them separately,
d> with their own header, at the end of the buffer (so they
d> don't get in the way).
d>
d> Likewise, C-x 8 bindings?

Which part of "at the end of the buffer" wasn't clear?

Currently, they are at the _start_ of the buffer, in the way.

Re: Why all the Alt bindings by default?

[Andreas Schwab]

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

> Which part of "at the end of the buffer" wasn't clear?

The bindings are listed in order of precedence.

> Currently, they are at the _start_ of the buffer, in the way.

Which is the correct place.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

RE: Why all the Alt bindings by default?

[Drew Adams]

> The bindings are listed in order of precedence.

Interesting.  I wasn't aware of that.  There is nothing in the buffer that tells
users that (bug #1), nothing in the doc string of `describe-bindings' about it
either (bug #2), nothing in (emacs) `Help Summary' or (emacs) `Misc Help' either
(bug #3, bug #4), and nothing in (elisp) `Scanning Keymaps' either (bug #6).

So is this by design or accident?  It's apparently documented _nowhere_.

Even if by design, perhaps the design needs to be revisited.  We did something
similar (at my request) a few years back for `C-h m', when we moved the
minor-mode stuff after the major-mode description.  Prior to that, like key
translation listings here, the minor-mode stuff was _in the way_.

IF we decide to keep showing key translations in `C-h b', then we should get
them out of the way, one way or another.

It would be fine if, in the new intro text that you will please add to explain
that bindings are listed in order of precedence, you also say that key
translations (which are not bindings) are listed after all of the bindings, even
though they take precedence.

> > Currently, they are at the _start_ of the buffer, in the way.
> 
> Which is the correct place.

"Correct"?  It's probably correct that that the correct place is _nowhere_ in a
listing of key bindings (Stefan's suggestion).  If they do belong there then
they certainly don't belong at the beginning of the buffer.

Another alternative would be to show them only on demand, e.g. using a
particular prefix arg.  The point is that they should not be front-and-center,
in the way of seeing the key bindings.

Re: Why all the Alt bindings by default?

[Stefan Monnier]

> That would be good, I think.  Or else list them separately, with their own
> header, at the end of the buffer (so they don't get in the way).

The display in C-x b is a separate issue.

> Likewise, C-x 8 bindings?

Not sure what "likewise" means here: C-x 8 bindings don't suffer
from this "only enabled after you load iso-transl.el", and should never
conflict with anything else, so there's no reason to change this part.


        Stefan