- From: Gregg Kellogg <gregg@greggkellogg.net>
- Date: Fri, 16 Aug 2013 07:53:44 -0700
- To: Markus Lanthaler <markus.lanthaler@gmx.net>
- Cc: Andy Seaborne <andy@apache.org>, "<public-rdf-wg@w3.org>" <public-rdf-wg@w3.org>
On Aug 16, 2013, at 3:17 AM, "Markus Lanthaler" <markus.lanthaler@gmx.net> wrote: > On Friday, August 16, 2013 11:48 AM, Andy Seaborne wrote: >> What defines passing a test with bNode named graphs? After reading >> Markus's position on isomorphism, I don't know what it takes to pass a > > That was not really a position. It was more asking whether I understand our > intention correctly (which I'm still not sure) and showing another way to > look at it. > > >> test (well, if the output is structurally isomorphic it passes, the >> question is more about negative tests and about weaker equivalences). > > Currently we take a rather pragmatic approach given that all the algorithms > produce deterministic bnode identifiers. A test passes if the resulting > N-Quads are the same (after being sorted). Most of us compare the result > directly, i.e., require that the same blank node identifiers are used; Gregg > (I believe) checks whether the result is isomorphic to the expected result. Actually, I do a textual comparison too, although it would be reasonable to test for isomorphism. Gregg >> PS there is a link in >> http://json-ld.org/test-suite/reports/ >> to >> http://json-ld.org/tests/ > > Fixed: > https://github.com/json-ld/json-ld.org/commit/4fdd5d1a4dcb273897074005eda386 > 837f97dda6 > > I changed it to point to http://json-ld.org/test-suite/. I'm not sure though > that that's the right target. Gregg? > > > -- > Markus Lanthaler > @markuslanthaler >
Received on Friday, 16 August 2013 14:54:12 UTC