- From: Gregory Williams <greg@evilfunhouse.com>
- Date: Mon, 28 May 2012 10:41:19 -0400
- To: Andy Seaborne <andy.seaborne@epimorphics.com>
- Cc: public-rdf-dawg@w3.org
On May 28, 2012, at 9:27 AM, Andy Seaborne wrote: >> Question to greg here: is this because bnode-labels in greg's implementations are graph-specific? If so, and if these label >> renamings across graphs are round-trippable, then update 2 should probably also be idempotent, i.e. still the result would be >> >> GS'''' = {} >> <g1> { _:b :p :o } >> <g2> { _:c :p :o } >> >> Is that so? > > No (I hope). > > Greg - can you confirm that the problem is the way the results are specified, not test per se: > > > mf:result [ > ut:graphData [ ut:graph <insert-05-g1-pre.ttl> ; > rdfs:label "http://example.org/g1" ] ; > ut:graphData [ ut:graph <c> ; > rdfs:label "http://example.org/g2" ] ; > ] ; > > means read the file insert-05-g1-pre.ttl a second and third time, the first being in the mf:action. > > So it generates different blanks nodes each time it's read, hence no shared bank node *in creating the results* -- nothing to do with the operation. Correct. As I said, and as you describe, the problem is the multiple parsing of the same file into the expected dataset, not in the update evaluation. > Hence either specify results in TriG/N_quads (but these are under-defined in this area) or make a conclusion that records the intended result and test for that (my long update request suggestion). Yes. I think your proposed solution (inserting the statement count back into the dataset) is the only sensible path forward on this. .greg
Received on Monday, 28 May 2012 14:41:47 UTC