foreign language patch for build-index+cl-ppcre branch


On Thu, 24 Feb 2011, Raymond Toy wrote:

< On 2/24/11 4:10 PM, Leo Butler wrote:
< >  I've pushed a patch that fixes this issue for 
< >  sbcl, clisp and cmucl (debian testing). All the
< >  maxima-index.lisp files are made, and I don't see
< >  any non-printable characters.
< >
< Mostly works for me with cmucl, but I had to apply the following patch,
< basically adding with-standard-io-syntax, which is a good idea for all
< Lisps.  For some reason I have non-NIL values for *print-length* and
< *print-level*, so maxima-index.lisp had ... characters in there.

Ok, this is good to know. I will patch the slurp function, too.

< 
< One other thing that needs to be solved for cmucl is that when I used
< LANG=de_DE, maxima somehow decides to use utf8 encoding.  That's fine,
< but when the documentation is printed, cmucl uses it's default encoding
< of iso8859-1, so ? is used in quite a few places.  For example, ?
< additive gives question marks in various places.

This is built from doc/info/de/maxima.info, correct? Could you check
what this file says on the line

coding: ...

near the bottom? The coding should be iso-8859-1 and it should be
picked up by Maxima (it is for me).

I have chosen the default external format to be utf-8, but the default
could be set at build time, too.

< 
< I'll need to look around to see where to hook  these together.
< 
< If it still matters, I don't think gcl has any kind of external format
< support.  
< 
< In many ways, it would be good to treat the info files as binary files
< of octets and use octet offsets.  Then when we read and display the
< documentation, we don't need any special support.  Just output the
< octets assuming the user has done the right magic so that the
< terminal/wxmaxima/etc. have the correct encoding.  (The terminal setting
< has to be set correctly no matter whether we use octets or characters.)
 
 Yes, this was my preconception, too. We do not need to use an encoding
 to read, categorise or search the info files. *But* when it comes to
 to outputting the stuff, I think you are wrong: the de.utf8 online
 help is displayed correctly in emacs when output as a utf8 encoded
 string, but not without the encoding.
 
 
 I can patch build-index.lisp so
 that you can compare the results with encodings vs. octets.

 Leo

-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.