|
If you were logged in you would be able to see more operations.
|
|
Issue reported on the forum: http://www.openrdf.org/forum/mvnforum/viewthread?thread=1440
--------------------------------------------------------------------------------
As part of a crosswalk from an ontology that holds people as strings to an ontology that has people as objects, we would like to create people objects for people that do not exist as objects yet. This almost works, i.e.
CONSTRUCT DISTINCT {} rdf:type {iribib:Person}; iribib:name {nam}
FROM {doc} enlbib:author {nam} ,
[ {aut} rdf:type {iribib:Person} ;iribib:name {nam} ]
WHERE aut=NULL
using namespace
iribib = <http://iridl.ldeo.columbia.edu/ontologies/iribib.owl#>,
enlbib = <http://iridl.ldeo.columbia.edu/ontologies/john/enlbib.owl#>
does make the blank nodes with the iribib:name property for the missing iribib:Person objects. The only problem is that repeats are not detected (apparently the presence of a blank node means every instance is distinct). While that might be the case from one perspective, I would argue that, since blank nodes only have meaning from their properties, that is not ever what is wanted from CONSTRUCT DISTINCT (and if you did want that you could leave out the DISTINCT request).
--------------------------------------------------------------------------------
This is a problem of the SeRQL parser, which translates the above query to the following query model:
Distinct
MultiProjection
Projection
ProjectionElem "-anon-1" AS "subject"
ProjectionElem "-const-5" AS "predicate"
ProjectionElem "-const-6" AS "object"
Projection
ProjectionElem "-anon-1" AS "subject"
ProjectionElem "-const-7" AS "predicate"
ProjectionElem "nam" AS "object"
Extension
ExtensionElem (-const-6)
ValueConstant (value=http://iridl.ldeo.columbia.edu/ontologies/iribib.owl#Person)
ExtensionElem (-anon-1)
BNodeGenerator
ExtensionElem (-const-7)
ValueConstant (value=http://iridl.ldeo.columbia.edu/ontologies/iribib.owl#name)
ExtensionElem (-const-5)
ValueConstant (value=http://www.w3.org/1999/02/22-rdf-syntax-ns#type)
Filter
Compare (=)
Var (name=aut)
Null
LeftJoin
StatementPattern
Var (name=doc)
Var (name=-const-1, value=http://iridl.ldeo.columbia.edu/ontologies/john/enlbib.owl#author)
Var (name=nam)
Join
StatementPattern
Var (name=aut)
Var (name=-const-2, value=http://www.w3.org/1999/02/22-rdf-syntax-ns#type)
Var (name=-const-3, value=http://iridl.ldeo.columbia.edu/ontologies/iribib.owl#Person)
StatementPattern
Var (name=aut)
Var (name=-const-4, value=http://iridl.ldeo.columbia.edu/ontologies/iribib.owl#name)
Var (name=nam)
The DISTINCT operator is applied after the bnodes are generated, which causes all projections to be unique. To fix this issue, the DISTINCT operator should be applied directly after the FILTER operation (all operations after that are related to the constructor).
|
|
Addendum: the proposed solution also requires an additional PROJECTION to restrict the binding sets to the variables that are used in the constructor. Without the PROJECTION, multiple combinations of 'doc' and 'nam' would otherwise still result in the generation of duplicates.
|
|
Bug
Closed
Minor
SeRQL-"construct distinct" queries generate duplicates when bnode generation is involved