issue with ordering of /resource endpoint not returning some predicates

[jyucsiro]

This reorders the union query to get around the issue of missing predicates for Locality.

[jyucsiro]

Screenshot of change applied to the API. Additional properties returned:

[jyucsiro]

Current API test endpoint: https://test-api.loci.cat/api/v1/resource?uri=http%3A%2F%2Flinked.data.gov.au%2Fdataset%2Fgnaf-2016-05%2Flocality%2FVIC1736

[benjaminleighton]

In general I'm wondering why this failed in the first place. I've been reading https://www.w3.org/TR/2013/REC-sparql11-query-20130321/ and my interpretation of that is that a UNION operation should return the same result if the right and left sides are swapped. The only thing I can think of is that this is something to do with using variables ?p ?o in both sides of the union. Possibly this should be rewritten so variables aren't recycled on both sides?

[jyucsiro]

Could be the way the query engine is implemented? i didn't think it would make a difference but swapping it around did make a difference!

[benjaminleighton]

I've had a bit more of a look into it and I think swapping it and getting the right answer possibly isn't addressing some underlying bugs. One of which is that they original query which I think is http://3.104.29.122/sparql?name=&infer=true&sameAs=true&query=PREFIX+rdf%3A+%3Chttp%3A%2F%2Fwww.w3.org%2F1999%2F02%2F22-rdf-syntax-ns%23%3E%0ASELECT+DISTINCT+%3Fp+%3Fo+%3Fp1+%3Fo1+%3Fp2+%3Fo2%0AWHERE+%7B%0A++++%7B%0A++++++++++++++++%3Fs+rdf%3Asubject+%3Chttp%3A%2F%2Flinked.data.gov.au%2Fdataset%2Fgnaf-2016-05%2Flocality%2FVIC1736%3E+%3B%0A+++++++++++rdf%3Apredicate+%3Fp%3B%0A+++++++++++rdf%3Aobject+%3Fo+.%0A++++++++OPTIONAL+%7B+FILTER+(isBlank(%3Fo))%0A++++++++++++%7B%0A++++++++++++++++%3Fs2+rdf%3Asubject+%3Fo+%3B%0A++++++++++++++++rdf%3Apredicate+%3Fp1%3B%0A++++++++++++++++rdf%3Aobject+%3Fo1+.%0A++++++++++++%7D%0A++++++++++++UNION%0A++++++++++++%7B+%3Fo+%3Fp1+%3Fo1+.+%7D%0A++++++++++++OPTIONAL+%7B+FILTER+(isBlank(%3Fo1))%0A++++++++++++++++%3Fo1+%3Fp2+%3Fo2+.%0A++++++++++++%7D%0A++++++++%7D%0A++++++++%0A++++++%0A++++%7D%0A++++UNION%0A++++%7B%0A++%0A++++++++%3Chttp%3A%2F%2Flinked.data.gov.au%2Fdataset%2Fgnaf-2016-05%2Flocality%2FVIC1736%3E+%3Fp+%3Fo+.%0A++++++++OPTIONAL+%7B+FILTER+(isBlank(%3Fo))%0A++++++++++++%7B%0A++++++++++++++++%3Fs3+rdf%3Asubject+%3Fo+%3B%0A++++++++++++++++rdf%3Apredicate+%3Fp1%3B%0A++++++++++++++++rdf%3Aobject+%3Fo1+.%0A++++++++++++%7D%0A++++++++++++UNION%0A++++++++++++%7B+%3Fo+%3Fp1+%3Fo1+.+%7D%0A++++++++++++OPTIONAL+%7B+FILTER+(isBlank(%3Fo1))%0A++++++++++++++++%3Fo1+%3Fp2+%3Fo2+.%0A++++++++++++%7D%0A++++++++%7D%0A++++%7D%0A%7D does actually return the resource information, eventually, on page 9 right near the end of ~8000 results. Another confusing thing is why only one result is returned. I think part of the issue is that query_graphdb_endpoint is called without a count argument and it defaults to 1000 so this result will never be seen.

[jyucsiro]

ah i see. it's at the end.

[jyucsiro]

regardless, it does make sense for this content to be ordered first before the reified triples.

[benjaminleighton]

Perhaps, I'm not sure how safe it is to depend on UNION structure providing a consistent ordering, maybe that is in the spec I haven't checked. Nevertheless there are at least two other bugs in the code, why is only one result returned and paging issues.

[dr-shorthair]

https://www.w3.org/TR/sparql11-query/#solutionModifiers