[Fwd: Re: more on mumamo/php mode]

[Fwd: Re: more on mumamo/php mode]

[Lennart Borgman (gmail)]

I got this report some days ago. Could please someone who understands 
what utf-translate-cjk-mode does comment on this?


-------- Original Message --------
Subject: Re: more on mumamo/php mode
Date: Thu, 19 Jul 2007 16:47:21 +0200
From: Claus <claus.klingberg@gmail.com>
To: Lennart Borgman (gmail) <lennart.borgman@gmail.com>

Hi Lennart,

some status update and a question on nxhtml-mode:

1. Somebody reported an error in nxml-mode upon validation/parsing
("Invalid regexp: Range striding over charsets") with Emacs 22 on the
nxml-mailing list (see e.g.
http://osdir.com/ml/emacs.nxml.general/2006-01/msg00039.html). Shortly
after Darkman posted a patch that seemed to work (at least for me):

http://drkm-lib.sourceforge.net/nxml/xsd-regexp.el.2006-01-26.patch

Re: [Fwd: Re: more on mumamo/php mode]

[Claus]

Hello,

all I could find about this issue was the explanation in the thread
linked from my message. The description of the cause goes like this:

"I found why this error happens. In the recent CVS Emacs, as
utf-translate-cjk-mode is on by default, some Unicode characters will
be decoded into one of CJK character set. But, as
xsdre-range-list-to-char-alternative (xsd-regexp.el) simply do
something like this to generate a character range: (format "%c-%c"
(decode-char 'ucs FROM) (decode-char 'ucs TO)) the resulting regexp
causes the above error."

The patch below is supposingly fixing this issue (and is working for
me). However, I guess we need someone to verify this since the
original maintainer is currently not available. Could someone look
into this and suggest whether this is the right solution?

--- xsd-regexp.el.orig	2006-01-27 00:31:24.000000000 +0100
+++ xsd-regexp.el	2006-01-27 00:32:20.328529600 +0100
@@ -290,7 +290,8 @@
 (defun xsdre-compile-single-char (ch)
   (if (memq ch '(?. ?* ?+ ?? ?\[ ?\] ?^ ?$ ?\\))
       (string ?\\ ch)
-    (string (decode-char 'ucs ch))))
+    (let ((utf-translate-cjk-mode nil))
+      (string (decode-char 'ucs ch)))))

 (defun xsdre-char-class-to-range-list (cc)
   "Return a range-list for a symbolic char-class."
@@ -403,7 +404,8 @@
       (setq range-list (cdr range-list)))
     (setq chars
 	  (mapcar (lambda (c)
-		    (decode-char 'ucs c))
+                    (let ((utf-translate-cjk-mode nil))
+                      (decode-char 'ucs c)))
 		  chars))
     (when caret
       (setq chars (cons ?^ chars)))

Thanks,
Claus



On 7/22/07, Lennart Borgman (gmail) <lennart.borgman@gmail.com> wrote:
> I got this report some days ago. Could please someone who understands
> what utf-translate-cjk-mode does comment on this?
>
>
> -------- Original Message --------
> Subject: Re: more on mumamo/php mode
> Date: Thu, 19 Jul 2007 16:47:21 +0200
> From: Claus <claus.klingberg@gmail.com>
> To: Lennart Borgman (gmail) <lennart.borgman@gmail.com>
>
> Hi Lennart,
>
> some status update and a question on nxhtml-mode:
>
> 1. Somebody reported an error in nxml-mode upon validation/parsing
> ("Invalid regexp: Range striding over charsets") with Emacs 22 on the
> nxml-mailing list (see e.g.
> http://osdir.com/ml/emacs.nxml.general/2006-01/msg00039.html). Shortly
> after Darkman posted a patch that seemed to work (at least for me):
>
> http://drkm-lib.sourceforge.net/nxml/xsd-regexp.el.2006-01-26.patch
>
>
>
> _______________________________________________
> Emacs-devel mailing list
> Emacs-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/emacs-devel
>

Re: [Fwd: Re: more on mumamo/php mode]

[Stefan Monnier]

> "I found why this error happens. In the recent CVS Emacs, as
> utf-translate-cjk-mode is on by default, some Unicode characters will
> be decoded into one of CJK character set. But, as
> xsdre-range-list-to-char-alternative (xsd-regexp.el) simply do
> something like this to generate a character range: (format "%c-%c"
> (decode-char 'ucs FROM) (decode-char 'ucs TO)) the resulting regexp
> causes the above error."

What it should do instead is to loop over all chars between FROM and TO and
add them one by one.  Character-ranges in regexps work rather poorly in
such cases.

> The patch below is supposingly fixing this issue (and is working for
> me).

It works around the problem, but will most likely break the use of those
CJK chars.


        Stefan

Re: [Fwd: Re: more on mumamo/php mode]

[Lennart Borgman (gmail)]

Stefan Monnier wrote:
>> "I found why this error happens. In the recent CVS Emacs, as
>> utf-translate-cjk-mode is on by default, some Unicode characters will
>> be decoded into one of CJK character set. But, as
>> xsdre-range-list-to-char-alternative (xsd-regexp.el) simply do
>> something like this to generate a character range: (format "%c-%c"
>> (decode-char 'ucs FROM) (decode-char 'ucs TO)) the resulting regexp
>> causes the above error."
> 
> What it should do instead is to loop over all chars between FROM and TO and
> add them one by one.  Character-ranges in regexps work rather poorly in
> such cases.
> 
>> The patch below is supposingly fixing this issue (and is working for
>> me).
> 
> It works around the problem, but will most likely break the use of those
> CJK chars.


There are too much things I do not understand here. Could please you or 
someone else who understands this write the needed code?

Re: [Fwd: Re: more on mumamo/php mode]

[Stefan Monnier]

>>> "I found why this error happens. In the recent CVS Emacs, as
>>> utf-translate-cjk-mode is on by default, some Unicode characters will
>>> be decoded into one of CJK character set. But, as
>>> xsdre-range-list-to-char-alternative (xsd-regexp.el) simply do
>>> something like this to generate a character range: (format "%c-%c"
>>> (decode-char 'ucs FROM) (decode-char 'ucs TO)) the resulting regexp
>>> causes the above error."
>> 
>> What it should do instead is to loop over all chars between FROM and TO and
>> add them one by one.  Character-ranges in regexps work rather poorly in
>> such cases.
>> 
>>> The patch below is supposingly fixing this issue (and is working for
>>> me).
>> 
>> It works around the problem, but will most likely break the use of those
>> CJK chars.


> There are too much things I do not understand here. Could please you or
> someone else who understands this write the needed code?

As I said above, you just need to enumerate the chars individually rather
than use char-ranges in regexps.  I.e. you ned to replace things like "a-d"
with "abcd".  It may result in absurdly large regexps and it may be
a problem in itself, but in that case the only better solution would be to
restructure the code so as not to use regexps, or use category classes or
something like that.


        Stefan