global-set-key [?\ M-ö]

global-set-key [?\ M-ö]

[Martin aus Chemnitz]

Hi!

I'm a very new Emacs user and hope this problem is not too frequently
asked.

I use
  GNU Emacs 21.3.1 (i386-msvc-windows98.2222)
and tried to bind a function to a German Umlaut key in my .emacs file,
but
  (global-set-key  [?\M-ö] 'forward-word)
did not work. There was no error message, but no binding as well.

The possible FAQ solution
  (add-hook 'window-setup-hook
    (lambda ()
      (global-set-key  [?\M-ö] 'forward-word)))
did not change the behaviour.

Thanks for any help.

Martin

Re: global-set-key [? \M-ö]

[Jesper Harder]

Ole Laursen <olau@hardworking.dk> writes:

> Martin aus Chemnitz <MartinAusChemnitz@gmx.net> writes:
>
>> and tried to bind a function to a German Umlaut key in my .emacs file,
>> but
>>   (global-set-key  [?\M-ö] 'forward-word)
>> did not work. There was no error message, but no binding as well.
>
> I have the same problem with binding C-æ. It was discussed here
> recently - it is probably a bug in Emacs 21.3, Jesper Harder filed a
> bug report AFAIK.

Yup, and it's fixed in CVS now.  So you can either

* downgrade to Emacs 21.2 (where it works)
* build Emacs from CVS
* wait for the next release :-)

Re: global-set-key [?\ M-ö]

[Martin aus Chemnitz]

>> I use
>>   GNU Emacs 21.3.1 (i386-msvc-windows98.2222)
>> and tried to bind a function to a German Umlaut key in my .emacs
>> file, but
>>   (global-set-key  [?\M-ö] 'forward-word)
>> did not work. There was no error message, but no binding as well.

> I have the same problem with binding C-æ. It was discussed here
> recently - it is probably a bug in Emacs 21.3, Jesper Harder filed
> a bug report AFAIK.

Thank you (both).

It wasn't this bug, but your recent posting quickly led me to the
solution:
  ;-*-unibyte: t;-*-

I would suggest that an appropriate note should be added to the FAQ
question
  10.3 Why doesn't this [terminal or window-system setup] code work
  in my `.emacs' file, but it works just fine after Emacs starts up?

Martin

Re: global-set-key [? \M-ö]

[Stefan Monnier]

>>>>> "Martin" == Martin aus Chemnitz <MartinAusChemnitz@gmx.net> writes:
> It wasn't this bug, but your recent posting quickly led me to the
> solution:
>   ;-*-unibyte: t;-*-

I suspect that ;-*- coding: latin-1 -*-
or ;-*- coding: latin-9 -*-
would be better (if it works).

And now that I think about it, neither of them should be necessary,
provided that your locale is properly set.  Is your locale set ?


        Stefan

Re: global-set-key [?\ M-ö]

[Martin aus Chemnitz]

> > It wasn't this bug, but your recent posting quickly led me to the
> > solution:
> >   ;-*-unibyte: t;-*-
> 
> I suspect that ;-*- coding: latin-1 -*-
> or ;-*- coding: latin-9 -*-
> would be better (if it works).
> 
> And now that I think about it, neither of them should be necessary,
> provided that your locale is properly set.  Is your locale set ?

I have to admit, that I do not understand the concept of processing
the .emacs file at all.

But to replace
  ;-*-unibyte: t;-*-
with 
  ;-*- coding: latin-1 -*-
or 
  ;-*- coding: latin-9 -*-
or
  ;(set-language-environment 'German)
does not work. Only the unibyte:t line works correct.

My basic assuption is of course that Emacs stores files in the same
way as any other text editor, because I edited the file in an other
application, so the .emacs file could stay open while I frequently
exited and started Emacs.

Martin

Re: global-set-key [? \M-ö]

[Stefan Monnier]

>> And now that I think about it, neither of them should be necessary,
>> provided that your locale is properly set.  Is your locale set ?
> [...other things...]

And your locale is ?


        Stefan

Re: global-set-key [?\ M-ö]

[Martin aus Chemnitz]

X-No-Archive:Yes

> And your locale is ?

Sorry, I'm a very new user of emacs. The manual uses `locale' in many
different ways. Please be more specific.

If you mean  
  M-x getenv LANG <ret>
then my locale is `deu'.

Martin

Re: global-set-key [? \M-ö]

[Stefan Monnier]

> Sorry, I'm a very new user of emacs.

Locales are actually a POSIX thing, they're not specific to Emacs at all.

> The manual uses `locale' in many different ways. Please be more specific.
> If you mean
>   M-x getenv LANG <ret>
> then my locale is `deu'.

Hmmm.... never heard of such a locale.
Emacs only knows about locales of the form `de_DE' or things like that.
Can you try to set LANG to de_DE for example ?

And since `deu' seems to work for you, I'd be interested to know what
kind of system you're using, to see whether `deu' (and probably a bunch
more) should be added to Emacs.

Also have you set LANG to `deu' yourself ?
Where did it tell you (or whoever set it) to use `deu' ?


        Stefan

Re: global-set-key [?\ M-ö]

[Martin aus Chemnitz]

X-No-Archive:Yes

> Locales are actually a POSIX thing, they're not specific to Emacs
> at all.

>From my first posting you can see that I am using Win98. 
Therefore my confusion.

> Can you try to set LANG to de_DE for example ?

Tell me how!

> And since `deu' seems to work for you, I'd be interested to know
> what kind of system you're using, to see whether `deu' (and 
> probably a bunch more) should be added to Emacs.
> 
> Also have you set LANG to `deu' yourself ?
> Where did it tell you (or whoever set it) to use `deu' ?

The three letter code depends on the settings in  "control panel" >
"country settings" (or whatever the correct translation).

Martin

Re: global-set-key [? \M-ö]

[Stefan Monnier]

> From my first posting you can see that I am using Win98. 
> Therefore my confusion.

I missed that, sorry.  Then I have no clue.


        Stefan

Re: global-set-key [?\ M-ö]

[Eli Zaretskii]

> From: Martin aus Chemnitz <MartinAusChemnitz@gmx.net>
> Newsgroups: gnu.emacs.help
> Date: Thu, 22 May 2003 20:01:34 +0100
> 
> > Can you try to set LANG to de_DE for example ?
> 
> Tell me how!

Set the environment variable LANG to the string "de_DE" (omitting the
quotes).  For Windows 9x, I think the usual way is to add to your
AUTOEXEC.BAT the appropriate SET commands.

However, I vaguely remember that the Windows port of Emacs already
converts the Windows locales to the appropriate Emacs locales, so I
don't think this will help you solve your problem.  It doesn't do any
harm to try, though.

Re: global-set-key [? \M-ö]

[Harald Maier]

Martin aus Chemnitz <MartinAusChemnitz@gmx.net> writes:

>> And your locale is ?
>
> Sorry, I'm a very new user of emacs. The manual uses `locale' in many
> different ways. Please be more specific.
>
> If you mean  
>   M-x getenv LANG <ret>
> then my locale is `deu'.

Can you try to set your LOCAL to 'C' under 
  'Einstellungen/Systemsteuerung/System/Erweitert/Umgebungsvariablen'

In the ".emacs" file please add the following:

  (set-locale-environment "de_DE")

This settings works fine for me, especially with emacs-21.3. You can
try too to set "de_DE@euro" as emacs local environment but then you
need a font that supports the Euro sign. The default font supports it.

Harald

Re: global-set-key [?\ M-ö]

[Eli Zaretskii]

> From: Martin aus Chemnitz <MartinAusChemnitz@gmx.net>
> Newsgroups: gnu.emacs.help
> Date: Wed, 21 May 2003 08:41:22 +0100
> 
> It wasn't this bug, but your recent posting quickly led me to the
> solution:
>   ;-*-unibyte: t;-*-
> 
> I would suggest that an appropriate note should be added to the FAQ
> question
>   10.3 Why doesn't this [terminal or window-system setup] code work
>   in my `.emacs' file, but it works just fine after Emacs starts up?

The Emacs manual already suggests to put -*-unibyte: t-*- cookie into
your .emacs if it binds commands to non-ASCII keys; see the node
"Non-ASCII Rebinding".  Using the wonderful `i' command, I typed the
following in the *info* buffer:

	i non-ascii keys RET

and was immediately placed in that section of the manual.  Given this,
why would one want to search the humongously large FAQ instead?

Re: global-set-key [? \ M-ö]

[Stefan Monnier]

> The Emacs manual already suggests to put -*-unibyte: t-*- cookie into
> your .emacs if it binds commands to non-ASCII keys; see the node

It had better be fixed, then, because it's not correct.
You'll end up binding characters of code 128..255 which will only
work if you set the keyboard-coding-system to nil (and will simply
never work under X).


        Stefan

Re: global-set-key [? \ M-ö]

[Eli Zaretskii]

> From: "Stefan Monnier" <monnier+gnu.emacs.help/news/@flint.cs.yale.edu>
> Newsgroups: gnu.emacs.help
> Date: 24 May 2003 11:10:06 -0400
> 
> > The Emacs manual already suggests to put -*-unibyte: t-*- cookie into
> > your .emacs if it binds commands to non-ASCII keys; see the node
> 
> It had better be fixed, then, because it's not correct.

The unibyte binding is only _one_ of the methods suggested by the
manual.  IIRC, none of the methods mentioned there works universally,
so having each one fail in some situation is something we should
expect, and this deficiency in itself should not IMHO be the reason to
delete this advice.  The fact is that it helped the OP in this case.

However, feel free to raise the issue on emacs-devel if you think the
manual should be changed in this regard.

Re: global-set-key [? \ M-ö]

[Kai Großjohann]

"Eli Zaretskii" <eliz@elta.co.il> writes:

> The unibyte binding is only _one_ of the methods suggested by the
> manual.  IIRC, none of the methods mentioned there works universally,
> so having each one fail in some situation is something we should
> expect, and this deficiency in itself should not IMHO be the reason to
> delete this advice.  The fact is that it helped the OP in this case.

But the keyboard has to send characters in *some* encodings, doesn't
it?  So wouldn't it be enough to ensure that the keyboard coding
system and the coding system used by Emacs to read ~/.emacs are the
same?

Did the OP try the above method, or did he just try the unibyte thing?

It seems to me if the make-codings-agree method works, it should be
used, since it is cleaner.  IMHO.
-- 
This line is not blank.

Re: global-set-key [? \ M-ö]

[Martin aus Chemnitz]

> But the keyboard has to send characters in *some* encodings, doesn't
> it?  So wouldn't it be enough to ensure that the keyboard coding
> system and the coding system used by Emacs to read ~/.emacs are the
> same?
> 
> Did the OP try the above method, or did he just try the unibyte
> thing?

I'm a little bit confused by your discussion.

As far as I can remember, I tried all of your suggestions including
(kbd "M-ö")
;-*- coding: latin-1 -*-
;-*- coding: latin-9 -*-
;(set-language-environment 'German)
comparing the encoding (C-u C-x C-=, its the same)

Tell me (fool-proof please) what to do and I'll try it.

Martin

Re: global-set-key [? \ M-ö]

[Kai Großjohann]

Martin aus Chemnitz <MartinAusChemnitz@gmx.net> writes:

> I'm a little bit confused by your discussion.
>
> As far as I can remember, I tried all of your suggestions including
> (kbd "M-ö")
> ;-*- coding: latin-1 -*-
> ;-*- coding: latin-9 -*-
> ;(set-language-environment 'German)
> comparing the encoding (C-u C-x C-=, its the same)
>
> Tell me (fool-proof please) what to do and I'll try it.

I created a file foo.el that looks like this (two lines):

;-*- coding: latin-1; -*-
(global-set-key (kbd "M-ö") 'forward-word)

I then started Emacs as follows:

    emacs -q -no-site-file -l foo.el

In the *scratch* buffer, I moved to the beginning of the buffer and
then I typed ESC <compose> " o.  (I don't have a German keyboard, so
<compose> " o is my way of typing the ö character in X11.  You would
hit your ö key directly.)  The cursor moved forward a word.

I then changed the file foo.el to say latin-9 instead of latin-1, and
then tried the same thing.  The cursor did not move forward.
Instead, I got a beep.

Also at the beginning of the *scratch* buffer, I typed <compose> "
o (you'd type ö directly).  Then I typed C-u C-x = and saw the
following:

/----
|   character: ö (04366, 2294, 0x8f6, U+00F6)
|     charset: latin-iso8859-1
| 	     (Right-Hand Part of Latin Alphabet 1 (ISO/IEC 8859-1): ISO-IR-100.)
|  code point: 118
|      syntax: w 	which means: word
|    category: l:Latin  
| buffer code: 0x81 0xF6
|   file code: F6 (encoded by coding system iso-latin-9)
|        font: -ETL-Fixed-Medium-R-Normal--16-160-72-72-C-80-ISO8859-1
\----

Note the second line -- the character is a latin-1 character.

Now I typed C-x C-f foo.el RET and positioned point on the ö and
typed C-u C-x = again, and got this output:

/----
|   character: ö (07566, 3958, 0xf76, U+00F6)
|     charset: latin-iso8859-15
| 	     (Right-Hand Part of Latin Alphabet 9 (ISO/IEC 8859-15): ISO-IR-203.)
|  code point: 118
|      syntax: w 	which means: word
|    category: l:Latin  
| buffer code: 0x8E 0xF6
|   file code: F6 (encoded by coding system iso-latin-9-unix)
|        font: -Adobe-Helvetica-Medium-R-Normal--17-120-100-100-P-88-ISO8859-15
\----

Note the second line -- the character is a latin-9 character.

Does this make things clearer?
-- 
This line is not blank.

Re: global-set-key [? \ M-ö]

[Martin aus Chemnitz]

> I created a file foo.el that looks like this (two lines):
:
> I then started Emacs as follows:
:
> In the *scratch* buffer, I moved to the beginning of the buffer and

My *scratch* buffer is empty if I start with "-l foo.el", but I type
in some word and do the same.

> then I typed ESC <compose> " o.  (I don't have a German keyboard, so
> <compose> " o is my way of typing the ö character in X11.  You would
> hit your ö key directly.)  The cursor moved forward a word.

Do you mean M-ö? When I type M-ö nothing happens.

C-h k M-ö produces `M-ö is undefined'.

If I type in ö and C-u C-x C-= over it I get a similar result as you:
`latin-iso8859-1'.
,----
|  character: ö (04366, 2294, 0x8f6)
|    charset: latin-iso8859-1
| (Right-Hand Part of Latin Alphabet 1 (ISO/IEC 8859-1): ISO-IR-100)
| code point: 118
|     syntax: word
|   category: l:Latin  
|buffer code: 0x81 0xF6
|  file code: F6 (encoded by coding system iso-latin-1-dos)
:
`----

If I open foo.el and do the same I get latin-1 as well.

,----
|  character: ö (04366, 2294, 0x8f6)
|    charset: latin-iso8859-1
| (Right-Hand Part of Latin Alphabet 1 (ISO/IEC 8859-1): ISO-IR-100)
| code point: 118
|     syntax: word
|   category: l:Latin  
|buffer code: 0x81 0xF6
|  file code: F6 (encoded by coding system iso-latin-1-dos)
:
`----

M-x local-set-key M-ö forward-word
works well and assigns the key. The file foo.el does not.

But if my foo.el looks like this
,----
| ;-*- unibyte:t -*-
| (global-set-key (kbd "M-ö") 'forward-word) 
`----
and I start Emacs like above, the key is assigned.


I'll get the same behaviour as you for latin-9, but latin-1 file does
not work.

Martin

Re: global-set-key [? \ M-ö]

[Kai Großjohann]

Martin aus Chemnitz <MartinAusChemnitz@gmx.net> writes:

>> I created a file foo.el that looks like this (two lines):
> :
>> I then started Emacs as follows:
> :
>> In the *scratch* buffer, I moved to the beginning of the buffer and
>
> My *scratch* buffer is empty if I start with "-l foo.el", but I type
> in some word and do the same.

Ah, good.

>> then I typed ESC <compose> " o.  (I don't have a German keyboard, so
>> <compose> " o is my way of typing the ö character in X11.  You would
>> hit your ö key directly.)  The cursor moved forward a word.
>
> Do you mean M-ö? When I type M-ö nothing happens.

What happens with ESC ö?

> C-h k M-ö produces `M-ö is undefined'.

Hm.

> If I type in ö and C-u C-x C-= over it I get a similar result as you:
> `latin-iso8859-1'.
> ,----
> |  character: ö (04366, 2294, 0x8f6)
> |    charset: latin-iso8859-1
> | (Right-Hand Part of Latin Alphabet 1 (ISO/IEC 8859-1): ISO-IR-100)
> | code point: 118
> |     syntax: word
> |   category: l:Latin  
> |buffer code: 0x81 0xF6
> |  file code: F6 (encoded by coding system iso-latin-1-dos)
> :
> `----
>
> If I open foo.el and do the same I get latin-1 as well.
>
> ,----
> |  character: ö (04366, 2294, 0x8f6)
> |    charset: latin-iso8859-1
> | (Right-Hand Part of Latin Alphabet 1 (ISO/IEC 8859-1): ISO-IR-100)
> | code point: 118
> |     syntax: word
> |   category: l:Latin  
> |buffer code: 0x81 0xF6
> |  file code: F6 (encoded by coding system iso-latin-1-dos)
> :
> `----

Okay, so I guess that ESC ö would work.  Hm.  I haven't tried M-ö
because I don't know how to type that on my US keyboard.

> M-x local-set-key M-ö forward-word
> works well and assigns the key. The file foo.el does not.
>
> But if my foo.el looks like this
> ,----
> | ;-*- unibyte:t -*-
> | (global-set-key (kbd "M-ö") 'forward-word) 
> `----
> and I start Emacs like above, the key is assigned.

Fascinating.

> I'll get the same behaviour as you for latin-9, but latin-1 file does
> not work.

So latin-9 doesn't work, either, right?  You get a beep and C-h k
tells you the key is undefined?

Anyhow, thanks for all the testing.  I wonder what can be done at
this point, and how come that your keys are different.

What's keyboard-coding-system, btw?  M-x describe-coding-system RET
RET should tell you.
-- 
This line is not blank.

Re: global-set-key [? \ M-ö]

[Martin aus Chemnitz]

> Okay, so I guess that ESC ö would work.  

No, it doesn't.

> So latin-9 doesn't work, either, right?  You get a beep and C-h k
> tells you the key is undefined?

I don't get a beep, but I think this is just a setting. The key is
undefined.

> Anyhow, thanks for all the testing.  I wonder what can be done at
> this point, and how come that your keys are different.

> What's keyboard-coding-system, btw?  M-x describe-coding-system RET
> RET should tell you.

* Coding system for keyboard input:
*   nil

I tried setting LANG to de_DE, but this does not affect the keyboard
coding.

Martin

Re: global-set-key [? \ M-ö]

[Kai Großjohann]

Martin aus Chemnitz <MartinAusChemnitz@gmx.net> writes:

>> What's keyboard-coding-system, btw?  M-x describe-coding-system RET
>> RET should tell you.
>
> * Coding system for keyboard input:
> *   nil

Ah, it should be whatever your locale says, so either latin-1 or
latin-9.  Please try C-x RET k to set it, and then try ESC ö and M-ö
again.

Hm.  OTOH, my kbd coding system is also nil.  Weird.

Note that the keyboard coding system should match your locale, and
both should match the default file coding.  So either set all three
to latin-1, or all three to latin-9, but no mixture.
-- 
This line is not blank.

Re: global-set-key [? \ M-ö]

[Stefan Monnier]

> What's keyboard-coding-system, btw?  M-x describe-coding-system RET

IIRC, on W32, w32-system-coding-system is used instead of
keyboard-coding-system (and in Emacs-21.3.50 that was changed to use
locale-coding-system, to align it with the X11 case).


        Stefan

Re: global-set-key [? \ M-ö]

[Stefan Monnier]

> The unibyte binding is only _one_ of the methods suggested by the
> manual.  IIRC, none of the methods mentioned there works universally,
> so having each one fail in some situation is something we should
> expect, and this deficiency in itself should not IMHO be the reason to
> delete this advice.  The fact is that it helped the OP in this case.

As I said somewhere else, the use of unibyte is generally a bug.
It might superficially fix things sometimes, but then it's because of a bug
somewhere else.  In the OP's case, I'm pretty sure the problem is that
Emacs does not properly set the keyboard coding-system.

I don't think the manual should encourage to use workarounds (e.g. set
the unibyte:t cookie in your .emacs) rather than real fixes (set the
keyboard coding system properly).


        Stefan

Re: global-set-key [? \ M-ö]

[Eli Zaretskii]

> From: "Stefan Monnier" <monnier+gnu.emacs.help/news/@flint.cs.yale.edu>
> Newsgroups: gnu.emacs.help
> Date: 28 May 2003 10:38:32 -0400
> 
> As I said somewhere else, the use of unibyte is generally a bug.

Yes, but the bug might be a design bug.  If that is so, we cannot tell
users "please wait for a few years until we redesign the darn thing."

> In the OP's case, I'm pretty sure the problem is that Emacs does not
> properly set the keyboard coding-system.

IIRC, there's more here than meets the eye.  Binding non-ASCII keys in
a .emacs requires that (a) Emacs reads the key correctly from the init
file and converts it to the internal representation that the user
meant, and (b) that keyboard decoding produces a code that matches
what was read from the init file.  It could be a bit tricky to satisfy
both in a given language environment, since .emacs files generally
don't have coding cookies.  Also, if I'm not mistaken, non-ASCII keys
with modifiers are very hard to express unless you go unibyte.

> I don't think the manual should encourage to use workarounds (e.g. set
> the unibyte:t cookie in your .emacs) rather than real fixes (set the
> keyboard coding system properly).

Unless setting keyboard coding system doesn't always solve the
problem, that is.

Re: global-set-key [? \ M-ö]

[Kai Großjohann]

"Eli Zaretskii" <eliz@elta.co.il> writes:

> IIRC, there's more here than meets the eye.  Binding non-ASCII keys in
> a .emacs requires that (a) Emacs reads the key correctly from the init
> file and converts it to the internal representation that the user
> meant, and (b) that keyboard decoding produces a code that matches
> what was read from the init file.  It could be a bit tricky to satisfy
> both in a given language environment, since .emacs files generally
> don't have coding cookies.

I've changed the Emacs manual to advise people to put coding cookies
into the .emacs file.  WDYT?

(I notice that I forgot to also tell them to save .emacs with the
right encoding.  Excuse me while I fix this...)

> Also, if I'm not mistaken, non-ASCII keys with modifiers are very
> hard to express unless you go unibyte.

I don't have such a kbd.  Doesn't (kbd "C-M-ß"), say, work, if the
.emacs encoding and the kbd encoding match?

-- 
This line is not blank.

Re: Re: global-set-key [? \ M-ö]

[Eli Zaretskii]

> From: kai.grossjohann@gmx.net (=?iso-8859-1?q?Kai_Gro=DFjohann?=)
> Newsgroups: gnu.emacs.help
> Date: Wed, 28 May 2003 18:51:03 +0200
> 
> I've changed the Emacs manual to advise people to put coding cookies
> into the .emacs file.  WDYT?

I think it was not always working in the past, but perhaps that's not
so any longer, especially in the development sources.

> I don't have such a kbd.  Doesn't (kbd "C-M-SS"), say, work, if the
> .emacs encoding and the kbd encoding match?

Why don't you try it, with and without different cookies and in
different language environments, and see what you get?

Re: global-set-key [? \ M-ö]

[Kai Großjohann]

"Eli Zaretskii" <eliz@elta.co.il> writes:

>> From: kai.grossjohann@gmx.net (=?iso-8859-1?q?Kai_Gro=DFjohann?=)
>
>> I don't have such a kbd.  Doesn't (kbd "C-M-SS"), say, work, if the
                                               ^^
This should be a ß (es-zett, German sharp s).

>> .emacs encoding and the kbd encoding match?
>
> Why don't you try it, with and without different cookies and in
> different language environments, and see what you get?

I don't have the kbd...
-- 
This line is not blank.

Re: Re: global-set-key [? \ M-ö]

[Eli Zaretskii]

> From: kai.grossjohann@gmx.net (=?iso-8859-1?q?Kai_Gro=DFjohann?=)
> Newsgroups: gnu.emacs.help
> Date: Thu, 29 May 2003 10:23:41 +0200
> >
> >> I don't have such a kbd.  Doesn't (kbd "C-M-SS"), say, work, if the
>                                                ^^
> This should be a an es-zett, German sharp s.

Yes, I know.  I tried to avoid non-ASCII characters in my reply.

Re: global-set-key [? \ M-ö]

[Stefan Monnier]

> Yes, but the bug might be a design bug.  If that is so, we cannot tell
> users "please wait for a few years until we redesign the darn thing."

Of course, but I haven't seen such a thing w.r.t binding-non-ASCII-keys
and unibyte.

>> In the OP's case, I'm pretty sure the problem is that Emacs does not
>> properly set the keyboard coding-system.

> IIRC, there's more here than meets the eye.  Binding non-ASCII keys in
> a .emacs requires that (a) Emacs reads the key correctly from the init
> file and converts it to the internal representation that the user
> meant, and (b) that keyboard decoding produces a code that matches
> what was read from the init file.  It could be a bit tricky to satisfy
> both in a given language environment, since .emacs files generally
> don't have coding cookies.

We're talking about adding a unibyte-cookie, right ?
So we can assume that adding a coding-cookie is an acceptable cost
if it saves us from a unibyte-cookie.

> Also, if I'm not mistaken, non-ASCII keys
> with modifiers are very hard to express unless you go unibyte.

I've never heard of any such difficulty.  [?\M-é] works fine here.
Oh wait, you're probably referring to "\M-é", which probably won't work,
indeed.  But the "keys in a string" thingy should best be forgotten
anyway: either use a vector or use `kbd'.

>> I don't think the manual should encourage to use workarounds (e.g. set
>> the unibyte:t cookie in your .emacs) rather than real fixes (set the
>> keyboard coding system properly).

> Unless setting keyboard coding system doesn't always solve the
> problem, that is.

Setting the unibyte cookie doesn't either always solve the problem.
Right now I know of no case where setting the unibyte cookie solves
the problem while setting the keyboard-coding-system doesn't, which
is why I suggest we recommend setting the keyboard-coding-system.
Of course, there might be cases where a unibyte-keyboard with
unibyte-.emacs works better, but I haven't seen them yet.


        Stefan


PS: BTW, when I say keyboard-coding-system I also mean locale-coding-system
(since that's what is used in X and in W32 for the keyboard events).

Re: global-set-key [? \ M-ö]

[Eli Zaretskii]

> From: "Stefan Monnier" <monnier+gnu.emacs.help/news/@flint.cs.yale.edu>
> Newsgroups: gnu.emacs.help
> Date: 28 May 2003 13:53:12 -0400
> 
> > Yes, but the bug might be a design bug.  If that is so, we cannot tell
> > users "please wait for a few years until we redesign the darn thing."
> 
> Of course, but I haven't seen such a thing w.r.t binding-non-ASCII-keys
> and unibyte.

I'm not sure, but my observation was of a more general nature, since
the part of your message to which I was referring seemed to state a
very general principle.

> We're talking about adding a unibyte-cookie, right ?
> So we can assume that adding a coding-cookie is an acceptable cost
> if it saves us from a unibyte-cookie.

My failing memory keeps telling me that a coding cookie was not always
enough, e.g. if the various language-environment specific aspects are
set up _after_ the non-ASCII keybinding is read.  Again, I might be
out of touch with the current codebase.

> Setting the unibyte cookie doesn't either always solve the problem.

That's true.  IIRC, that is why the manual mentions both.

> Right now I know of no case where setting the unibyte cookie solves
> the problem while setting the keyboard-coding-system doesn't, which
> is why I suggest we recommend setting the keyboard-coding-system.

What about the order of settings Mule-related things, as I mentioned
above?

Re: global-set-key [? \ M-ö]

[Stefan Monnier]

> My failing memory keeps telling me that a coding cookie was not always
> enough, e.g. if the various language-environment specific aspects are
> set up _after_ the non-ASCII keybinding is read.  Again, I might be
> out of touch with the current codebase.

But assuming the coding-system of the file is the same as the one of the
keyboard, the only possible problem is if reading the .emacs (and other
things) end up changing the behavior of that coding-system.
Most coding-systems have an immutable behavior.  The only exception I know
of is coding-systems that do some form of unification upon decoding
and might unify into different charsets at different times.

I guess I could build a setup where the problem manifests itself, but
I don't know how common it is and I suspect that any such setup would
require significant changes to make it work with a unibyte .emacs.

...

Hmmm.... it seems that there's one case where a unibyte answer might
make sense: the user wants to use latin-9 but also has some latin-1 files
and doesn't want to suffer from the corner cases of encode-only
unification, so he turns on decode-unification in his .emacs.  His keyboard
coding system is set to latin-9, but unification translates most of those
to latin-1 chars, so his .emacs should be read as latin-1 for most
non-ASCII keybindings.  A unibyte cookie in the .emacs won't directly help,
but if you additionally set the keyboard-coding-system to nil, it might
end up working.


        Stefan

Re: global-set-key [? \ M-ö]

[Kai Großjohann]

"Stefan Monnier" <monnier+gnu.emacs.help/news/@flint.cs.yale.edu> writes:

> Hmmm.... it seems that there's one case where a unibyte answer might
> make sense: the user wants to use latin-9 but also has some latin-1 files
> and doesn't want to suffer from the corner cases of encode-only
> unification, so he turns on decode-unification in his .emacs.  His keyboard
> coding system is set to latin-9, but unification translates most of those
> to latin-1 chars, so his .emacs should be read as latin-1 for most
> non-ASCII keybindings.  A unibyte cookie in the .emacs won't directly help,
> but if you additionally set the keyboard-coding-system to nil, it might
> end up working.

What happens if people use iso-2022-7bit as the encoding for the
.emacs file?  Then they can put latin-1 chars for the latin-1 cases
and latin-9 chars for the other cases.

-- 
This line is not blank.

Re: global-set-key [?\ M-ö]

[Martin aus Chemnitz]

X-No-Archive:Yes

> The Emacs manual already suggests to put -*-unibyte: t-*- cookie
> into your .emacs if it binds commands to non-ASCII keys; see the
> node "Non-ASCII Rebinding".  Using the wonderful `i' command, I
> typed the following in the *info* buffer:
> 
>         i non-ascii keys RET
> 
> and was immediately placed in that section of the manual.  Given
> this, why would one want to search the humongously large FAQ
> instead?

Because I did not find this particular help page.

Because I don't know about the `i' command yet.

Because I thought `ö' to be an ASCII character. 
`7-bit-ASCII' would have been more precise, I guess.

Because I looked at the help page `Rebinding Keys in Your Init File'
that did not mention `unibyte' at all. Not even with a link to the
topic.

Because the example at the same page `[?\C-=]' worked.
(There is a bug. `=' is considered non-ASCII, which is wrong.)

Because the FAQ answers frequently asked questions.

Because the description
  10.3 Why doesn't this [terminal or window-system setup] code work
  in my `.emacs' file, but it works just fine after Emacs starts up?
was pretty much what I was expecting.


Just some reasons.

Thanks for all the help.
(I hope this posting does not sound angry to anyone.)

Martin

Re: global-set-key [? \ M-ö]

[Stefan Monnier]

> Because I thought `ö' to be an ASCII character.
> `7-bit-ASCII' would have been more precise, I guess.

ASCII tends to be use loosely to mean "plain text", but really ASCII is
a character set with 128 elements coded in 7bits.  We should probably be
more careful in the doc to make it clear that ASCII is not just "plain
text".

> Because I looked at the help page `Rebinding Keys in Your Init File'
> that did not mention `unibyte' at all. Not even with a link to the

Good.  It shouldn't mention unibyte.  If something requires you to use
unibyte, it's probably because there's a bug (or because you've erroneously
used unibyte elsewhere).

> Because the example at the same page `[?\C-=]' worked.
> (There is a bug. `=' is considered non-ASCII, which is wrong.)

I think if you reread closely, you'll note that what is meant is that C-=
is not ASCII (whereas C-m, for example, *is* the ASCII character with code
13 usually spelled "carriage return").


        Stefan

Re: global-set-key [?\ M-ö]

[Kai Großjohann]

Martin aus Chemnitz <MartinAusChemnitz@gmx.net> writes:

>> and was immediately placed in that section of the manual.  Given
>> this, why would one want to search the humongously large FAQ
>> instead?
>
> Because I did not find this particular help page.
>
> Because I don't know about the `i' command yet.

The FAQ should mention the i command, then.  The i command is your
friend.

> Because I thought `ö' to be an ASCII character. 
> `7-bit-ASCII' would have been more precise, I guess.

Ascii is always 7bit: it defines characters 0 through 127.  `ö' is
Latin-1 or Latin-9.

> Because I looked at the help page `Rebinding Keys in Your Init File'
> that did not mention `unibyte' at all. Not even with a link to the
> topic.

That's a good point.  I'll try to add some wording to the development
version.  Thanks for the hint.

> Because the example at the same page `[?\C-=]' worked.
> (There is a bug. `=' is considered non-ASCII, which is wrong.)

`=' is ascii, `C-=' is not.  (Note that both `a' and `C-a' are ascii!)

> Because the FAQ answers frequently asked questions.

It's been a while that I last saw this question.  It used to be more
common, I think, but for some reason it isn't asked often anymore.
Even when it was common, that meant maybe a question every couple of
months...  Hm.

> Because the description
>   10.3 Why doesn't this [terminal or window-system setup] code work
>   in my `.emacs' file, but it works just fine after Emacs starts up?
> was pretty much what I was expecting.

Hm.
-- 
This line is not blank.

Re: global-set-key [?\ M-ö]

[Kai Großjohann]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

> Martin aus Chemnitz <MartinAusChemnitz@gmx.net> writes:
>
>> Because I looked at the help page `Rebinding Keys in Your Init File'
>> that did not mention `unibyte' at all. Not even with a link to the
>> topic.
>
> That's a good point.  I'll try to add some wording to the development
> version.  Thanks for the hint.

I've now added an xref.
-- 
This line is not blank.

Re: global-set-key [?\ M-ö]

[Eli Zaretskii]

> From: Martin aus Chemnitz <MartinAusChemnitz@gmx.net>
> Newsgroups: gnu.emacs.help
> Date: Mon, 26 May 2003 00:22:57 +0100
> 
> Because the FAQ answers frequently asked questions.
> 
> Because the description
>   10.3 Why doesn't this [terminal or window-system setup] code work
>   in my `.emacs' file, but it works just fine after Emacs starts up?
> was pretty much what I was expecting.

For some reason I refuse to understand, people nowadays expect to
find any piece of useful information in a FAQ.  While this might be a
good policy for many a software out there (since not many developers
bother to write good documentation for their programs), Emacs (and
GNU software in general) are not part of this plot.  Emacs developers
try very hard to put user-level information in the manual, and they
index that information in a way that is supposed to make it easy for
users to find it quickly.

Therefore, I suggest to always look in the manual first using the `i'
command.

Re: global-set-key [?\M-ö]

[Kai Großjohann]

Martin aus Chemnitz <MartinAusChemnitz@gmx.net> writes:

>   (global-set-key  [?\M-ö] 'forward-word)

The problem is that there are many characters in Emacs that look like
ö: one from Latin-1, one from Latin-2, one from Latin-9, and probably
others.  Emacs considers them all to be different.

Now, let's say the kbd sends a Latin-9 ö to Emacs, but the file
contains a Latin-1 ö.

You might open the .emacs file with C-x C-f and hit C-u C-x = on the ö
character.  Then you might wish to type ö into a buffer and also hit
C-u C-x = on it.  (Please do not use C-x =, use C-u C-x =.)  Then you
know whether it's the same ö, or whether they just look the same.

What happens if you write (kbd "M-ö") instead of [?\M-ö]?
-- 
This line is not blank.