Re: match specific blank terms

> On Nov 24, 2017, at 7:05 AM, Anthony Durity <a.durity@umail.ucc.ie> wrote:
> 
> Ok, having looked at the code for RDF::Node I guess my question is. Why is #eql? different to #== for RDF::Node ? That gives me the situation where terms.last == n is true but terms.include?(n) gives me false.

This is because, unless they’re defined in the same document, _:foo.eql?(_:foo). Blank Nodes are simple place holders, and the names have no meaning. The only reason they have labels, is so that, within a serialization, the same node can be referenced. If we were to change the sense of #eql? to allow for this, a lot of code (e.g., SPARQL and BGP matching) would break.

Node#== (and Node#===) do allow equivalence on Blank Node label, and the specific semantics derive from SPARQL::Algebra::Operator::Equal, including the ability to raise an exception if the node types are not the same.

The differences are where you might have two graphs both using _:foo, and a pattern (_:foo <p> “bar”) that needs these to be treated as separate statements, and not the same statement.

Node#== comes into play for a query such as SELECT  ?x WHERE { ?x1 :p ?v1 .  ?x2 :p ?v2 . FILTER ( ?v1 = ?v2 ) . }

Working with Blank Nodes is complicated, and best to avoid because of the nature of Blank Nodes not being independently identifiable, by design.

Gregg

> On 24 November 2017 at 09:14, Ian Dickinson <i.j.dickinson@gmail.com <mailto:i.j.dickinson@gmail.com>> wrote:
> Hi Anthony,
> On Thu, Nov 23, 2017 at 10:13 PM, Anthony Durity <a.durity@umail.ucc.ie <mailto:a.durity@umail.ucc.ie>> wrote:
> > I was working under the impression that 'local scope' meant local to the
> > repo.
> >
> > Given that I have:
> > [
> > #<RDF::URI:0x2ac3bff25014 URI:https://dh.ucc.ie/entity/α54 <https://dh.ucc.ie/entity/%CE%B154>>,
> > #<RDF::URI:0x2ac3be8bc444 URI:http://www.wikidata.org/prop/direct/P31 <http://www.wikidata.org/prop/direct/P31>>,
> > #<RDF::URI:0x2ac3bff2c5d0 URI:http://www.wikidata.org/entity/Q187685 <http://www.wikidata.org/entity/Q187685>>,
> > #<RDF::URI:0x2ac3bff32b60 URI:https://dh.ucc.ie/entity/α55 <https://dh.ucc.ie/entity/%CE%B155>>,
> > #<RDF::URI:0x2ac3bff3f9f0 URI:https://dh.ucc.ie/entity/α56 <https://dh.ucc.ie/entity/%CE%B156>>,
> > #<RDF::URI:0x2ac3be892504 URI:https://dh.ucc.ie/entity/α57 <https://dh.ucc.ie/entity/%CE%B157>>,
> > #<RDF::Node:0x2ac3c025570c(_:g47253167721100)>,
> > #<RDF::Node:0x2ac3bf67d7b8(_:59)>,
> > #<RDF::Node:0x2ac3c0265d50(_:α60)>
> > ]
> > what's the best way to match the blank node 'α60' ? I tried “intern” just
> > there and it didn't work :/
> 
> I think it would be easier to offer advice if you could back up a bit
> and say what you're
> trying to achieve? As Gregg said, bNodes can be funny things, so the 'best
> way' to handle them tends to be a bit "well, it depends ..."!
> 
> Ian
>