RDFLib 4.2.1

This is a bug-fix release.

Minor enhancements:

• Added a Networkx connector
  #471,
  #507
• Added a graph_tool connector
  #473
• Added a graphs method to the Dataset object
  #504,
  #495
• Batch commits for SPARQLUpdateStore
  #486

Bug fixes:

• Fixed bnode collision bug
  #506,
  #496,
  #494
• fix util.from_n3() parsing Literals with datatypes and Namespace support
  #503,
  #502
• make Identifier.__hash__ stable wrt. multi processes
  #501,
  #500
• fix handling URLInputSource without content-type
  #499,
  #498
• no relative import in algebra when run as a script
  #497
• Duplicate option in armstrong theme.conf removed
  #491
• Variable.__repr__ returns a python representation string, not n3
  #488
• fixed broken example
  #482
• trig output fixes
  #480
• set PYTHONPATH to make rdfpipe tests use the right rdflib version
  #477
• fix RDF/XML problem with unqualified use of rdf:about
  #470,
  #468
• AuditableStore improvements
  #469,
  #463
• added asserts for graph.set([s,p,o]) so s and p aren't None
  #467
• threading.RLock instances are context managers
  #465
• SPARQLStore does not transform Literal('') into Literal('None') anymore
  #459,
  #457
• slight performance increase for graph.all_nodes()
  #458

Testing improvements:

• travis: migrate to docker container infrastructure
  #508
• test for narrow python builds (chars > 0xFFFF) (related to
  #453,
  #454
  )
  #456,
  #509
• dropped testing py3.2
  #448
• Running a local fuseki server on travis and making it failsafe
  #476,
  #475,
  #474,
  #466,
  #460
• exclude def main(): functions from test coverage analysis
  #472