QualifiedNameTest sets output attributes to `None`

[Olemis Lang <olemis+trac@…>]

Consider the following script

'''
Created on Jul 9, 2013

@author: olemis
'''

from pprint import pprint

from genshi.template.markup import MarkupTemplate

if __name__ == '__main__':
    template_source = """<?xml version="1.0" encoding="utf-8"?>
    <element xmlns:nspc1="http://foo.org"
        nspc1:attribute="value">
    </element>
    """

    data = {}

    # XML stream
    template = MarkupTemplate(template_source)
    stream = template.generate(**data)

    for ev in stream.select('./@nspc1:attribute', {'nspc1' : 'http://foo.org'}):
        pprint(ev)

By executing it it's possible to notice how XPath expressions will always evaluate to None when matching qualified attribute names.

$ ./test_select1.py
Attrs([('attribute', None)])

After modifying QualifiedNameTest.__call__ like shown below ...

    def __call__(self, kind, data, pos, namespaces, variables):
        qname = QName('%s}%s' % (namespaces.get(self.prefix), self.name))
        if kind is START:
            if self.principal_type is ATTRIBUTE and qname in data[1]:
                return Attrs([(self.name, data[1].get(qname))])
            else:
                return data[0] == qname

... the result looks better

$ ./test_select1.py
Attrs([('attribute', u'value')])

Moreover this version will preserve namespace information

    def __call__(self, kind, data, pos, namespaces, variables):
        qname = QName('%s}%s' % (namespaces.get(self.prefix), self.name))
        if kind is START:
            if self.principal_type is ATTRIBUTE and qname in data[1]:
                return Attrs([(self.name, data[1].get(qname))])
            else:
                return data[0] == qname

Results

$ ./test_select1.py
Attrs([(QName('http://foo.org}attribute'), u'value')])

IMO the last one is the most accurate considering that this may be used to insert attributes in elements (e.g. py:attrs="select('@*')" in Genshi tutorial).

BTW , I set version=develop in the ticket because this issue was detected using both 0.7 and 0.6, and there's no option to select the former.

priority=blocker because IMO this ticket only is a good reason to release 0.7.1

[Olemis Lang <olemis+trac@…>]

Replying to Olemis Lang <olemis+trac@…>:

Moreover this version will preserve namespace information

I pasted the same code snippet twice . The code preserving namespace prefix should have been written as follows

    def __call__(self, kind, data, pos, namespaces, variables):
        qname = QName('%s}%s' % (namespaces.get(self.prefix), self.name))
        if kind is START:
            if self.principal_type is ATTRIBUTE and qname in data[1]:
                return Attrs([(qname, data[1].get(qname))])
            else:
                return data[0] == qname

Results

$ ./test_select1.py
Attrs([(QName('http://foo.org}attribute'), u'value')])

[hodgestar]

Thanks for the detailed bug report and suggestion for fix. Fix with properly namespaced attribute name committed to trunk in r1269 and backported to 0.7.x (r1270) and 0.6.x (r1271).