N-triples parser: reading a file fails without binary mode on Python 3.6

[kouralex]

Whilst trying to read an N-triples (.nt) file, I get this stack trace:

Traceback (most recent call last):
    Graph().parse(f, format='nt')
  File "[redacted]/.local/lib/python3.6/site-packages/rdflib/graph.py", line 1078, in parse
    parser.parse(source, self, **args)
  File "[redacted]/.local/lib/python3.6/site-packages/rdflib/plugins/parsers/nt.py", line 26, in parse
    parser.parse(f)
  File "[redacted]/.local/lib/python3.6/site-packages/rdflib/plugins/parsers/ntriples.py", line 150, in parse
    self.line = self.readline()
  File "[redacted]/.local/lib/python3.6/site-packages/rdflib/plugins/parsers/ntriples.py", line 173, in readline
    buffer = self.file.read(bufsiz)
  File "/usr/lib/python3.6/codecs.py", line 499, in read
    data = self.bytebuffer + newdata
TypeError: can't concat str to bytes

where f = open(file, 'r').

RDFLib Version: 5.0.0

It seems to be working on Python 2.7 as is, OR on Python 3.6 by adding the rb mode to the open(), however, that is not required for xml nor n3 format values. One can, of course, use the n3 option for .nt files but that will probably cause a small overhead.

[ashleysommer]

@kouralex
Thanks for the bug report.
I think in this case most people probably either open with 'rb' as you suggested, or use the n3 parser, as you mentioned. Both of these are not ideal solutions, and we definitely need to get the ntriples parser updated to work properly in Python 3.6.

[ashleysommer]

This should be fixed by #1145