Hi Paweł, first thing to note: lxml uses UTF-8 internally, also for filenames, as libxml2 requires a char sequence for their representation. If your system can't handle that, we'll have to figure out a way to make it work. This part of lxml is not much tested, so it would be nice if you could help us in getting this straight. Paweł Pałucha wrote:

I'm using lxml version 1.0.3 (from Debian package). My default system encoding is ISO-8859-2. I have a simple program:

#!/usr/bin/python # -*- coding: iso-8859-2 -*-

import lxml.etree d = lxml.etree.parse('/tmp/ó.xml')

Ok, so you're using 8-bit encoded filenames.

Traceback (most recent call last): File "./p.py", line 6, in ? d = lxml.etree.parse('/tmp/łódź.xml') File "etree.pyx", line 1615, in etree.parse File "parser.pxi", line 687, in etree._parseDocument File "apihelpers.pxi", line 343, in etree._utf8 AssertionError: All strings must be Unicode or ASCII

Right, I guess that treating the filename with the _utf8() function is not the right thing to do for 8-bit strings. We should have a separate way of treating filenames. I'll look into it.

If I try to write path using UTF-8 encoding, the file cannot be found (because the path with UTF-8 encoded name does not exists).

I assume you get the same file-not-found error if you pass the filename as unicode string? d = lxml.etree.parse(u'/tmp/ó.xml')

This did not happen with python 2.3 - the problem is only with python2.4. There's a workaround - reading file to StringIO object and then parsing XML from that object. It works fine but it's silly.

You can also pass an opened file object. Stefan