I've made the 4 private methods protected and added a protected
"createCache".
By this rational, I'd consider DatasetImpl creating models (in instances
where the client code is not requesting a created model) a bug. There is no
reason for an abstraction, whose only purpose is to relate a collection of
models to the collection of their backing graphs, to create new models to
then represent those graphs.
There is a cache for the case of databases with very large numbers of
graphs persistently on disk. If the DatasetImpl is churning through
models, they will slowly fill up memory in a long-running program.
Models aren't particularly cheap currently because of their own internal
caches and state EnhGraph). And just creating a model each call through
would not be good for code that looks like
loop:
dataset.getNamedModel("fixed name"). .. model operations ...
Architecturally, maybe a reworking where the usual Model works
(statelessly) over a Graph and the EnhGraph machinary only appears for
OntModels would be good.
Andy
