Re: Need to bind a "default" type when no type matched in a VALUES set

[Mark Feblowitz <Ma...@comcast.net>]

Your example is interesting and likely a very useful application of VALUES and COALESCE.

In thinking more about my specific case, I might be thinking about the problem incorrectly, or perhaps I’m overloading the VALUES  mechanism.

To state the problem succinctly, I’d like to match a particular resource E to any T in  { T1, T2, … },  such that  ?E a ?T. 
Failing a match, I’d like to bind a default for T,  e.g. owl:Thing. But I don’t want to bind to owl:thing if there’s a match in the set.


The VALUES mechanism lets me do the basic matching, but assigning a default presents problems.
	
	VALUES ?T { :Person :Place :Organization }
	?E a ?T.
 
If ?E is of one of the types in ?T, then ?T is bound, as far as subsequent references to ?T are concerned. And it’s bound for each solution (although I can’t think of any things that have multiple type matches in   { :Person :Place :Organization }  ). 

The trouble comes when trying to determine a match failure and act on ?T. 

In a few tests (at least with DBPedia Virtuoso), testing for BOUND(?T) when there’s no match to the VALUES set results in a “true”. But in the resultset the value of ?T appears to be undefined.

I confirmed this using:

	VALUES ?T { :Person :Place :Organization }
	?E a ?T.
	BIND(IF(BOUND(?T), :argh, owl:thing) as ?TB)

If ?E  is  or is not  of a type ?T in   { :Person :Place :Organization },  ?TB is bound to :argh (!)

In a way that makes sense - I’m thinking of ?T as both a variable to be bound and as a value generator. I can see where that might present confusion for implementers.

I tried binding ?T to another variable, but that just kicked the same can down the road, so to speak.

So how could I handle a non-match of I can’t determine whether a binding was found?



> On Sep 29, 2015, at 7:42 AM, Neubert, Joachim <J....@zbw.eu> wrote:
> 
> Hi Mark,
> 
> I had a similar problem in the past: I wanted to have the possiblity to set an input variable via VALUES, and use some default derived from the data if the variable is not set. I solved it this way:
> 
> values ( ?input ) ( {undef} ) 
> ?x ?y ?default 
> bind(coalesce(?input, $default) as ?actualVariableUsed )
> 
> For an example, see https://github.com/jneubert/skos-history/blob/master/sparql/added_concepts.rq, where it is possible to pass two version numbers to the query via VALUES. If not set, the latest and the version before are derived from the data. (The query is executable against Fuseki via a link on the readme page).
> 
> Hope that some of that could be helpful in your case - cheers, Joachim
> 
> -----Ursprüngliche Nachricht-----
> Von: Mark Feblowitz [mailto:MarkFeblowitz@comcast.net] 
> Gesendet: Dienstag, 29. September 2015 01:54
> An: users@jena.apache.org
> Betreff: Need to bind a "default" type when no type matched in a VALUES set
> 
> I have need to identify 0 or more type resources from a given set  as having been asserted as types of a given subject entity;  if none matches, I’d like to bind some default resource to symbol a non-match.
> 
> So, I have a set of “whitelisted” types
> 
> 	VALUES ?entType { ex:T1 ex:T2 ex:T3 }
> 
> and  I’m looking to match an entity against those types in this way:
> 
> 	?E a ?entType.
> 
> It’s fine to match more than one (although I’d rather not, but that’s a separate discussion).
> 
> In cases where there’s no match, I’d then like to force bind ?entType to, e.g., ex:T0 or owl:Thing.
> 
> To survive a non-match, the type match would likely need to be wrapped in an OPTIONAL block, likely with the VALUES statement in there too(?):
> 
> 	OPTIONAL {
> 		VALUES ?entTypeS { ex:T1 ex:T2 ex:T3 }
> 		?E a ?entTypeS.
> 		}
> 
> (Note the “S” in ?entTypeS - that’s for the next bit).
> 
> I’ve tried a conditional BIND
> 	BIND(IF(BOUND(?entType), ?entType, ex:T0) as ?entType) 
> 
> and also s COALESCE
> 	BIND(COALESCE(?entTypeS, ex:T0) as ?entType)
> 
> 
> But it seems that neither works as I’d like.  I appears that, even though optional, the test against the VALUES set acts as a constraint.
> 
> I’ve also tried a FILTER, with 
> 
> 	FILTER ( ?E1T in ( ex:T1, ex:T2, ex:T3 )) 
> 	BIND(COALESCE(?E1T, ex:T0) as ?E1Type)
> 
> That’s not even getting to executing (yet) - some non-parsing error on DBPedia Live. 
> 
> Have I stated the problem clearly enough? What might be other ways to address this?
> 
> Thanks,
> 
> Mark
>