Adding should check the new triplet

[jeroenh]

Consider the following:

>>> g = rdflib.Graph()
>>> g.add(("#foo","#bar","#foobar"))
>>> g.serialize()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/rdflib/graph.py", line 802, in serialize
    serializer.serialize(stream, base=base, encoding=encoding, **args)
  File "/opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/rdflib/plugins/serializers/rdfxml.py", line 62, in serialize
    bindings = list(self.__bindings())
  File "/opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/rdflib/plugins/serializers/rdfxml.py", line 31, in __bindings
    prefix, namespace, name = nm.compute_qname(predicate)
  File "/opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/rdflib/namespace.py", line 264, in compute_qname
    namespace, name = split_uri(uri)
  File "/opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/rdflib/namespace.py", line 395, in split_uri
    if not category(c) in NAME_CATEGORIES:
TypeError: must be unicode, not str

The error is only caught when serializing the graph, not before. It seems to me that the .add() method should have caught this already.

[gromgull]

This is the same as #200.

It's been open for long - but I wanted to do some performance checks before fixing - I'll have a look at your pull req, thanks!

[jeroenh]

I heard that performance might be an issue, which is why I used assert statements. Those will get removed if you're asking Python to do performance enhancements.
On the one hand this means that it's not a fool-proof fix, but it does make it a lot more userfriendly for 90% of the users. And the other 10% should run their code in normal mode first anyway ;).