Possible implementation that casts rather than infers type?

[dhimmel]

I saw this repo mentioned in RDFLib/sparqlwrapper#125. Very cool. Here's the heart of the conversion:

sparql-dataframe/sparql_dataframe/sparql_dataframe.py

Lines 28 to 31 in 86b1e85

	sparql.setReturnFormat(CSV)
	results = sparql.query().convert()
	_csv = StringIO(results.decode('utf-8'))
	return pd.read_csv(_csv, sep=",")

The SPARQLWrapper.Wrapper.QueryResult object is converted to a CSV string, and then loaded into pandas.

I am guessing this works well most of the time! But it does discard type information and then lets Pandas infer types upon reading the CSV. This makes me a bit nervous.

Have you considered at all whether a type-casting rather than a type-inferring implementation is possible?

Also related is RDFLib/sparqlwrapper#160. Ideally this would be part of RDFLib/sparqlwrapper to reduce the dependencies required to use SPARQL in Python, but this library seems the furthest ahead at providing user-friendly outputs and API.

[WolfgangFahl]

https://github.com/WolfgangFahl/pyLoDStorage has a proper python type handling mechanism.