Dave Beckett wrote:
> Looking at tests: > csv-tsv-res/manifest#csv01 > csv-tsv-res/manifest#csv02 > csv-tsv-res/manifest#csv03 > > Test #csv03 seems to be a comparison of syntaxes, not a result set comparison. > > To pass regular query tests you generally check two result sets match: > 1. query(query file, query data)) > 2. result-set(result-file, result-format) > > To pass this one you have to mostly compare bytes sequences (syntaxes): > 1. format(query(query file, query data), result-format) > 2. result-file > but not quite, the blank node ids don't match > > The result set comparisons in #csv03 fails because datatypes are lost. > > csv01 and csv02 could be seen as a plain literal / xsd:string issue and > I think the RDF WG changed the rules on that, so I might be catching up. > > So I'd ask that csv03 is adjusted to pass using the regular result set > comparisons unless the syntax detail is important. If that's the case then > please adjust the manifest to record that that the test is not a result-set > comparison.
The working group has agreed to update the CSV test manifest to indicate that these are result format syntax tests and should be treated differently from the query evaluation tests. The CSV tests in the manifest now have rdf:type mf:CSVResultFormatTest, and a description of these tests has been added to the draft version of the test suite overview.
> I was pointed at > http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2012Feb/0025.html > but didn't see a public reply
The reply to that comment is at .
We would be grateful if you would acknowledge that your comments have been answered by sending a reply to this mailing list.
Regards, Gregory Williams, on behalf of the SPARQL WG.