- From: Gregg Kellogg <gregg@greggkellogg.com>
- Date: Tue, 5 Nov 2013 12:27:03 -0800
- To: Andy Seaborne <andy@apache.org>
- Cc: RDF WG <public-rdf-wg@w3.org>
On Nov 5, 2013, at 9:56 AM, Andy Seaborne <andy@apache.org> wrote: > On 02/11/13 22:02, Gregg Kellogg wrote: >> Hi Andy, thanks for the implementation reports. > > Report on making the reports: > > > Note that your test results all have a test URI starting with > > <file:///home/afs/W3C/rdf/trig/tests/manifest.ttl#>, which doesn’t > > match the actual manifest location. I fixed these, but in the future, > > please report against the actual manifest location, e.g., > > <https://dvcs.w3.org/hg/rdf/raw-file/default/trig/tests/manifest.ttl#>. > > 1/ The problem I encountered that lead me to run from disk turns out that there is no conneg - I thought it was the https, and a bug at my end, but in fact my problems were due to the manifest files coming back as text/plain from the web. > > When a file, Jena is using the file extension. It was less pushy about doing that from the web when there was an explicit (and wrong) content-type. I think in my case, I ignore the mime-type text/plain for the purposes of format detection, and use the extension from the URI path. > I'll rerun the tests to get a clean run. > > I have no idea how to fix the conneg issue except to note in the tests description though I think it's unfortunately that the RDF-WG who define the media type can't get it right! Agreed, given the importance W3C places on content-negotiation, at least for RDF, we should really try to get this right. Ideally, adding a .htaccess to the relevant directories would be sufficient, but the Mercurial web interface might pose some problems. > But unless there is admin magic for https://dvcs.w3.org/hg/, we're struck. > >> Note that your test results all have a test URI starting with > <file:///home/afs/W3C/rdf/trig/tests/manifest.ttl#>, which doesn’t match > the actual manifest location. I fixed these, but in the future, please > report against the actual manifest location, e.g., > <https://dvcs.w3.org/hg/rdf/raw-file/default/trig/tests/manifest.ttl#>. >> >> Also, the N-Quads tests should also include the N-Triples tests, as you can see from the report. > > (the report is now readable and Gregg's made sure the fix does not get lost). > > 2/ (feedback) I think the reports for N-Quads / N-Triples is going to catch people out and may lead to duff report data. The only reason not to consolodate tests was just to avoid replication. If that's the consensus, I'll just copy the N-Triples tests as .nq files into the N-Quads test directory and add all the entries to the manifest, so that there's a single manifest. Perhaps we can reserve 5 minutes on the call to talk about such test logistics. > One EARL test has one URI (object of earl:test) but nothing else. > > <#nt-syntax-file-01> rdf:type rdft:TestNTriplesPositiveSyntax, rdft:TestNQuadsPositiveSyntax > > but it has two types. Which is the right one? Both - this is RDF. The thinking here was that it was both an N-Triples Positive Syntax test and an N-Quads Positive Syntax test. > It's fortunately that the NT test suite does not have a negative test for N-quads. > > Unsurprisingly, my code processes the RDF manifest and reacts to the type. But as it is for EARL, "one test, one URI", it can only run and report the test once. The EARL report does not contain enough information to know which version of the test was run - they have the same URI - so running the test twice does not help. While I can produce a special version for these test manifests, it's an easy mistake to make. > > Gregg - if a report has the same test reported twice, what happens? If everything was the same, it would be like asserting duplicate triples; they'd be consolidated. Within each report, they should be consistent, even though when considered together it would be ambiguous. > (I point a program at the test manifest and say "run tests". One manifest, one report -- I'm guessing but that seems like a reasonable pattern for implementers.) JSON-LD uses many manifests, as does SPARQL and RDFa, so multiple manifests is not unusual. In the case of JSON-LD and SPARQL there is typicallly a single manifest that includes other manifests, but for SPARQL I believe you need to run most of the 1.0 and 1.1 tests, which are in different manifests. > 3/ (feedback) The NQ and TriG report description refers to > http://www.w3.org/2011/rdf-wg/wiki/Turtle_Test_Suite > but there is nothing about NQ and trig there. We could create different Wiki pages for the different formats, but that didn't seem to be too useful to me; Obviously, someone should have an action to set this up. I'll do it if the group comes to consensus. > It disagrees with the NT manifest as it shows only one type. > > <#nt-syntax-file-01> rdf:type rdft:TestNTriplesPositiveSyntax ; That looks like an inconsistency that should just be fixed, but depending on what we decide for manifest consolidation, it may be a moot point. Gregg > Andy > >> >> The consolidated EARL reports have been updated [1] [2] [3]. >> >> Gregg Kellogg >> gregg@greggkellogg.net >> >> [1] https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-turtle/reports-nt/index.html >> [2] https://dvcs.w3.org/hg/rdf/raw-file/default/nquads/reports/index.html >> [3] https://dvcs.w3.org/hg/rdf/raw-file/default/trig/reports/index.html >> >> On Nov 2, 2013, at 7:01 AM, Andy Seaborne <andy@apache.org> wrote: >> >>> On 31/10/13 03:59, Gregg Kellogg wrote: >>>> I've created some initial implementation reports for N-Triples [1], N-Quads [2], and TriG [3]. The files are in place to create an implementation report for Semantics; it just requires at least one submission to run the reporting software. >>>> >>>> Note that the boilerplate within the report may need to be updated depending on the particulars of each spec. >>>> >>>> Gregg Kellogg >>>> gregg@greggkellogg.net >>>> >>>> [1] https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-turtle/reports-nt/index.html >>>> [2] https://dvcs.w3.org/hg/rdf/raw-file/default/nquads/reports/index.html >>>> [3] https://dvcs.w3.org/hg/rdf/raw-file/default/trig/reports/index.html >>> >>> >>> Up-to-date reports are available from >>> >>> http://people.apache.org/~andy/RDF-EARL/ >>> >>> Andy >>> >>> == N-Quads >>> http://people.apache.org/~andy/RDF-EARL/Apache-Jena-nquads-earl-2013-11-02.ttl >>> >>> == N-Triples >>> http://people.apache.org/~andy/RDF-EARL/Apache-Jena-ntriples-earl-2013-11-02.ttl >>> >>> == TriG >>> http://people.apache.org/~andy/RDF-EARL/Apache-Jena-trig-earl-2013-11-02.ttl >>> >> > >
Received on Tuesday, 5 November 2013 20:27:33 UTC