RDFPandas, providing RDF support for Pandas

[ghost]

RdfPandas

RdfPandas is a module providing RDF support for Pandas. It consists of two simple functions for Graph to DataFrame conversion and DataFrame to Graph conversion.

Creating RDF from DataFrame

As of version 1.1.0 NamespaceManager can be supplied to rdflib.to_graph for conversion to Graph.

from rdfpandas.graph import to_graph
import pandas as pd
import rdflib

df = pd.read_csv('to_graph_test.csv', index_col = '@id', keep_default_na = False)
namespace_manager = NamespaceManager(Graph())
namespace_manager.bind('skos', SKOS)
namespace_manager.bind('rdfpandas', Namespace('http://github.com/cadmiumkitty/rdfpandas/'))
g = to_graph(df, namespace_manager)
s = g.serialize(format = 'turtle')

Creating DataFrame from RDF

from rdfpandas.graph import to_dataframe
import pandas as pd
import rdflib

g = rdflib.Graph()
g.parse('to_df_test.ttl', format = 'ttl')
df = to_dataframe(g)
df.to_csv('test.csv', index = True, index_label = "@id")

Gotchas

No special effort is made for dealing with types, so please be aware of Pandas features such as https://pandas.pydata.org/pandas-docs/stable/user_guide/gotchas.html#support-for-integer-na that may result in surprising RDF statements like "10.0"^^<xsd:integer>.

[davidshumway]

Very interesting. I'm curious what the equivalent functionality is for sending the triples in a graph to a DataFrame? As a first try, I'd think something like this?

g = Graph()
g.parse('test.ttl')
p = pd.DataFrame()
for stmt in g:
  p = pd.concat([p, pd.DataFrame({'s': [stmt[0]], 'p': [stmt[1]], 'o': [stmt[2]]})])

For a graph with 10,000 triples this is pretty efficient but for larger graphs it seems to take longer.