On 12/08/14 04:29, Wang Darrell wrote:
I just try to rewrite the builtin - print()
NoValue would be a better starting point.
public boolean bodyCall(final Node[] args, int length, RuleContext context)
{
Triple t = new Triple(args[0], args[1], args[2]);
See the NoValue code.
You should check the length before getting arguments.
To get the Node values in a suitable form use getArg(i, args, context)
GraphBase graphBase = new GraphBase() {
@Override
protected ExtendedIterator<Triple> graphBaseFind(TripleMatch m) {
return null;
}
};
?? GraphBase isn't relevant here, don't do this.
ExtendedIterator<Triple> ts = graphBase.find(t);
if (ts!=null) {
System.out.println(ts.toList().size());
}
else{
System.out.println("ts is null");
}
return true;
}
I look into the GraphBase class and want to use the graphbaseFind() , I
think it's the method I want to use.
If I don't think it in wrong way , I can get a match iterator and get the
size of it.
But it says I need to override the graphBaseFind method .
I just confused about the override. Don't know how can I continue deal with
it.
The override is your IDE telling you are trying instantiate an abstract
class.
For your case you want to use context.find(s, p, o) to search for
triples that match your arguments.
Dave
2014-08-11 15:39 GMT+08:00 Dave Reynolds <[email protected]>:
On 11/08/14 07:39, Wang Darrell wrote:
so I need to figure out how the bulitin work in code and built a new one
that kind of countMatch?
If you really need to do this in rules then yes.
What is that mean you said API would be eaiser than SPARQL?
Just that, given the info you have available in a builtin, it will be easy
to create a Triple pattern than you can pass to Graph.find and then just
count the number of results. You could create and run a full SPARQL query
but it's not necessary.
and should I
know about register the new builtin to somewhere ?
Yes, there's a brief mention of the details at:
https://jena.apache.org/documentation/inference/#extensions
Looking at the code for RuleMap (which defines a registers a "Deduce"
builtin) would probably help.
Dave
2014-08-08 15:14 GMT+08:00 Dave Reynolds <[email protected]>:
On 08/08/14 03:08, Wang Darrell wrote:
here is part of my ontology
- ( Stevie_Wonder - hasMusicalRole - wordnet_synthesist_110687231 )
- ( Stevie_Wonder - hasMusicalRole - wordnet_bass_guitar_102804123 )
- ( Stevie_Wonder - hasMusicalRole - wordnet_piano_103928116 )
Can I use something like
(?x :hasMusicalRole ?y)
and get a count like 3 , then produce a new fact like
( Stevie_Wonder - can - playMutipleInstrument )
while we know he can play 3 kind of instrument.
No, as I said in my last reply the rules can't count the number of times
they fire or would fire without a lot of messing with non-monotonic rules
and arithmetic.
Or is there any way to use Sparql in rules?
like
[ rule:
SELECT count(?y) WHERE { ?s yago:hasMusicalRole ?y . }
->
( Stevie_Wonder - can - playMutipleInstrument )
]
There isn't a way to do that directly at the moment. There is a Google
Summer of Code project exploring the possibility of this but it is a
separate experiment and not part of the main Jena codebase.
However, for just counting it would be pretty easy to create a rules
builtin, something like:
countMatches(?s, ?p, ?o, ?count)
which would do this. The countMatches implementation could use SPARQL but
since it only needs to count triples then just doing it using the API
would
be easier. Remember to mark the builtin as non-monotonic.
Dave
