Dear List, Tonight I did some refactoring:
SparqlRdfQueryParser: --------------------- I introduced 2 of the collection types (Namespaces and Triples) for easier handling of this stuff so there is a better separation of concerns. Regarding the Sparql query rendering. I now moved all the rendering stuff into the main method (getQuery) so that is consistently there. Its also possible to add a toSparql-method to Triples,Triple,Namespaces to let them be responsible for their own formatting. What do you think about this kind of responsibility stuff? Where to put such concerns? In a single class accessing the state of multiple objects and doing the stuff or rather having the objects doing it for themselves and just collecting the results? I replaced most of the StringBuilder calls (which were actually just a single StringBuilder call and not collecting parameters or such) with String.format which is much more concise and readable. I also redid the large "(if then else)+" construct in processFilter with return-early. I found also an inconsistency: in methods like processMatchesPredicate there was an addTuple call upfront which was unneccessary when the method failed with an exception. There seems to be a GenericAssociationInfo missing? (sister class to GenericPropertyInfo) ? RdfEntityFinderMixin: --------------------- I added the promised IoC callbacks reducing the size of the class by 2/3. Namely: import org.qi4j.entity.index.rdf.callback.CollectingQualifiedIdentityResultCallback; import org.qi4j.entity.index.rdf.callback.SingleQualifiedIdentityResultCallback; the interfaces are: import org.qi4j.entity.index.rdf.callback.TupleQueryResultCallback; import org.qi4j.entity.index.rdf.callback.QualifiedIdentityResultCallback all except the TupleQueryResultCallback,QualifiedIdentityTupleQueryResultCallback belonging rather to the general entity finder package. First I also had a Counting*Callback but resorted to having performQuery return the number of rows processed. There is/was also an inconsistency. For findEntities/findEntity null-Identities are ignored and for counting the results they are included! At last I replaced inheritance with delegation within the callbacks and now the tuplequery stuff could be re-inlined into the RdfEntityFinderMixin (as a separate method processRow(QICallback, row, QualifiedIdentity). Shall I do this?? SPARQLEntityFinderMixin: ------------------------ I also redid the SPARQLEntityFinderMixin with the new IoC callbacks, so implementing the two missing methods (findEntity/countEntities). It was just a breeze. RdfEntityFinderTest: -------------------- I also added assertions to RdfEntityFinderTest, storing the identity-name pairs when populating the network and using this information on the identities returned by the 22 finder tests to assert that the correct entities were found. The only test I didn't have the strength adding assertions to was the "dump the world as rdf" test. I'd really like to make this test the base for all EntityFinders, much like the AbstractEntityStoreTest for all EntityStores. WDYT ? Pulling the test methods up and just having the assembly setup and the concrete entity finder in the test subclasses would be easy. Now its 5 am and I'm going to get 2 hours of sleep before leaving for work :) Michael _______________________________________________ qi4j-dev mailing list [email protected] http://lists.ops4j.org/mailman/listinfo/qi4j-dev
