[Expat-discuss] Expat with UTF-32 support (as implemented in 4Suite.org)

[Jeremy Kloth]

On Monday 03 October 2005 11:41 pm, Raphael Bossek wrote:
> Without the UTF-32 support in libexpat it will be mutch harder for
> me to maintain 4suite.org package for Debian GNU/Linux. The usage
> for libexpat instead of staticly linking the 4suite.org extension
> into the Python modules is mandatory for mod_python environments.
> We have to make sure only one version of expat symbols is loaded
> into memory.

It will make no difference if the platform's libexpat would have this support 
unless it was compiled with the same settings for XML_Char as Python does for 
Py_UNICODE.  By doing that, no transcoding is required to go from Expat 
strings to Python unicode objects, a large performance boost.

Linking to libexpat is only required for ELF systems not using symbol 
visibility (GCC 4 or 3.4+patch).  However, until Python gets updated to 
support that, it is in apache/mod_python's hands to do so.

> Any idears are welcome. In the meantime I'l investigate the patch
> done by the 4suite.org team and make a patch agains latest CVS.
> I've also see MSVC 6.0 support in it. Is there any interest in it?

One idea to solve this constant PITA wrt mod_python/apache would be a way to 
define a "prefix" for all exported symbols.  So, for example in expat.h, 
instead of:

XMLPARSEAPI(XML_Parser)
XML_ParserCreate(const XML_Char *encoding);

it could be something like:

XMLPARSEAPI(XML(Parser))
XML(ParserCreate)(const XML_Char *encoding);

where:

#define XML_PREFIX XML
#define XML(name) XML_PREFIX#_#name

Just a thought.

-- 
Jeremy Kloth
Fourthought, Inc.
http://fourthought.com/
http://4suite.org/