[rdf4j-dev] BindingSet

[Håvard Ottestad <hmottestad@xxxxxxxxx>]

Hi

As part of optimizing sorting in RDF4J I discovered that we are spending a lot of time instantiating HashMaps within BindingSet and then a lot of time hashing strings. 

I was considering looking into having internal variables names that allow for lookup directly in arrays instead of maps. Essentially having variables be bytes. Then mapping those bytes back into the variable names for the enduser. This would support up to 256 variables per BindingSet. So we would have to fall back to the old one if we need more. 

For things like group by or count I think this will make a massive difference. Also for queries that do a lot of internal joining of data, but actually return very few results. 

Håvard