Jim,
If you want to extend the execution of SPARQL, then look OpExecutor.
This is the place that SPARQL algebra become a stream of results.
TDB2 extends this for its join support - TDB execution works on internal
64 bit ids, not the full RDF term representation.
TDB2 also delays turning ids back into RDF terms. For an operation like
"count" all that is need is the row (BindingTDB), not the values of each
cell of the row.
Andy
On 16/05/2025 21:41, Jim Balhoff wrote:
Hi,
I would like to provide a SPARQL endpoint over a virtual dataset. For the IRIs
in the dataset, I can compute anything known about them based on the IRI
itself. So rather than materializing the whole thing, I want to just respond to
triple pattern queries and generate the data on the fly. I can see how I might
implement the Model or Graph interfaces to support this, but wondering if there
is another API I should use that would be most efficient for backing a SPARQL
endpoint. For example, I could calculate counts for particular triple patterns
without iterating through all the results. Or as another example I could
imagine implementing certain joins involving two triple patterns more
efficiently than by generating the data for both.
Thank you,
Jim
