W3C home > Mailing lists > Public > public-rdf-dawg-comments@w3.org > September 2006

about ISSUE: duplicate from UNION

From: Jorge Pérez <jperez@ing.puc.cl>
Date: Fri, 22 Sep 2006 19:24:11 -0400 (CLT)
Message-ID: <61740.146.155.4.12.1158967451.squirrel@mail.ing.puc.cl>
To: fred.zemke@oracle.com, public-rdf-dawg-comments@w3.org

Hi!, about "ISSUE: duplicate from UNION"

http://lists.w3.org/Archives/Public/public-rdf-dawg/2006JulSep/0265.html

I'm not sure but I think that the problem with duplicates is more complex
and from UNION it is carried to other graph patterns. For example consider
the data

:a :p :o1
:b :p :o2
:a :p :o3
:b :p :o4

and the pattern

{{ ?x :p :o1 } UNION { ?y :p :o2 }}

then, following the spec the solutions are

?x  ?y
:a  --
--  :b

(using -- for not bound). Consider now the group pattern with two UNION
patterns

{
{ { ?x :p :o1 } UNION { ?y :p :o2 } } .
{ { ?x :p :o3 } UNION { ?y :p :o4 } }
}

then I think that

?x  ?y
:a  :b

must be a solution, but, how many copies of this solution in the result
set? Trying the query in ARQ it gives me two copies of the same solution.
I don't know if the behavior of ARQ is ok, the spec says nothing about
duplicates, but I agree with ARQ that, if duplicates are considered, this
solution must have two copies in the result set. The problem here is that
the duplicates are generated in the group pattern *indirectly* from the
two UNION patterns.

Another thing that confuses me (and may be it is related) is, if

?x  ?y
:a  --
--  :b

are solutions for the pattern above. ARQ gives these solutions too, but in
this case I'm not so sure that I agree with ARQ :-)

- jorge
Received on Friday, 22 September 2006 23:24:49 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:14:50 GMT