New Feature CORE-159
1 vote

Provide an abstract ISparqlDataset implementation more suited to out of memory quad stores
Created by Rob Vesse on 1/12/2012 10:00 PM Last Updated by Rob Vesse on 2/3/2012 5:57 PM
%
 (hrs)
Logged: 0  (hrs)

 Description

Current implementations of ISparqlDataset are limited in that they require graphs to be loaded into memory for various operations.  It would be much nicer for some kinds of stores if we had a more quad oriented view where we just maintained the list of active and default graph URIs and made calls of the form GetQuadsWithXY(Uri graphUri, INode x, INode y) instead.

 

We could also then handle all the returning of graphs by wrapping the dataset in a triple collection of some sort.  The only tricky bits start to become handling the modification related operations so maybe in an initial release this dataset implementation would be read-only.
Skip Navigation Links

    Rob Vesse (Friday, February 03, 2012 5:57 PM) #

SVN revision 2052 fixes a bug in the DAWG test rig which accounted for the Update test failures when the InMemoryQuadDataset was used

    Rob Vesse (Friday, February 03, 2012 5:48 PM) #

Initial implementation is completed and mostly tested, provided an in-memory quad dataset implementation which works correctly

 

Confirmed fully correct for query for SPARQL DAWG Tests for Query but there are some issues with the Update tests still to be resolved

    Rob Vesse (Tuesday, January 31, 2012 10:22 PM) #

The basic implementation outline of this is now done as of revision 2045, still lots of work and testing to do but interested users can now review the design

    Rob Vesse (Sunday, January 29, 2012 5:08 AM) #

Some of the refactoring around this has been done but this is at an early stage, the work is not yet completed or heavily tested and should be considered unstable and unusable