Re: issue with test filter-scope-1

Francis McCabe wrote:
> 
> I am having some trouble reconciling this test with the spec, the 
> reported results of the query.
> 
> The test reads:
> 
> FILTERs in an OPTIONAL do not extend to variables bound outside of the 
> LeftJoin(...) operation
> 
> 
> 
> however, the spec reads
> 
> Query variables in SPARQL queries have global scope; use of a given 
> variable name anywhere in a query identifies the same variable.

Hi Frank,

These two bits of text are compatible. The second piece of text 
indicates that anywhere that the lexical token ?v appears in a query 
refers to the same variable. The first piece of text indicates that a 
FILTER clause in an OPTIONAL only filters solutions with respect to the 
LeftJoin(...) algebra expression in which it appears.

More on this specific example:

> 
> The reported results of the test:
> 
> PREFIX :    <http://example/>
> 
> SELECT *
> {
>     :x :p ?v .
>     { :x :q ?w
>       OPTIONAL {  :x :p ?v2 FILTER(?v = 1) }
>     }
> }
> 
> include the results
> 
> v w
[snip]

To simplify, note that the results of this test are every possible pair 
of values for ?v (objects of :p) and ?w (objects of :q).

> (My formatting, the XML is illegible by mortals)

:-)

> However, I think that the test is wrong: v2 should be v. On my budding 
> implementation, I get the above results only if I change v2 to v. If I 
> don't, I get:
> 

I believe the test is correct. Here's what happens:

After following the algorithm in #convertGraphPattern, the above query 
corresponds to this algebra expression:

Join(BGP(:x :p ?v), LeftJoin(BGP(:x :q ?w), BGP(:x :p ?v2), ?v = 1)

The filter (?v = 1) is tied to the LeftJoin. Of course, all of the 
solutions to Join(BGP(:x :q ?w), BGP(:x :p ?v2)) fail that filter since 
none of the solutions bind ?v. Therefore, the filter on the LeftJoin 
rejects all of the "extended" solutions (the solutions composed from 
both the left-hand and right-hand sides of the OPTIONAL). That means 
that the solutions to the LeftJoin(...) are just the solutions of the 
left-hand side:

?w    ?v2
---------
1   unbound
2   unbound
3   unbound

The solutions to the other BGP are straightforward

?v
--
1
2
3
4

...and the conjunction of those two result sets are the results as given 
in the test suite (the XML result format omits unbound variables)

I hope this is helpful.

Lee

> v v2 w
> (4, (null), 3)
> 
> (4, (null), 2)
> 
> (4, (null), 1)
> 
> (3, (null), 3)
> 
> (3, (null), 2)
> 
> (3, (null), 1)
> 
> (2, (null), 3)
> 
> (2, (null), 2)
> 
> (2, (null), 1)
> 
> (1, 4, 3)
> 
> (1, 3, 3)
> 
> (1, 2, 3)
> 
> (1, 1, 3)
> 
> (1, 4, 2)
> 
> (1, 3, 2)
> 
> (1, 2, 2)
> 
> (1, 1, 2)
> 
> (1, 4, 1)
> 
> (1, 3, 1)
> 
> (1, 2, 1)
> 
> (1, 1, 1)
> 
> 
> Note that optional implies that there will be solutions that do not bind 
> v2.
> 
> I would appreciate clarification on this matter.
> 
> 
> 
> Frank
> 
> 
> 
> 
> 
>

Received on Thursday, 25 October 2007 20:09:26 UTC