Plugin parsers¶
These serializers are available in default RDFLib, you can use them by
passing the name to graph’s parse()
method:
graph.parse(my_url, format='n3')
The html
parser will auto-detect RDFa, HTurtle or Microdata.
It is also possible to pass a mime-type for the format
parameter:
graph.parse(my_url, format='application/rdf+xml')
If you are not sure what format your file will be, you can use rdflib.util.guess_format()
which will guess based on the file extension.
Name |
Class |
---|---|
json-ld |
|
hext |
|
n3 |
|
nquads |
|
nt |
|
trix |
|
turtle |
|
xml |
Multi-graph IDs¶
Note that for correct parsing of multi-graph data, e.g. Trig, HexT, etc., into a ConjunctiveGraph
or a Dataset
,
as opposed to a context-unaware Graph
, you will need to set the publicID
of the ConjunctiveGraph
a
Dataset
to the identifier of the default_context
(default graph), for example:
d = Dataset()
d.parse(
data=""" ... """,
format="trig",
publicID=d.default_context.identifier
)
(from the file tests/test_serializer_hext.py)