| Key: |
SES-694
|
| Type: |
Bug
|
| Status: |
Open
|
| Priority: |
Major
|
| Assignee: |
Unassigned
|
| Reporter: |
Olaf Goerlitz
|
| Votes: |
0
|
| Watchers: |
0
|
|
If you were logged in you would be able to see more operations.
|
|
The transition from binary to n-ary join operators in Sesame 3.0 was not adapted in the logic of org.openrdf.query.algebra.evaluation.impl.FilterOptimizer.
For example, the query
SELECT *
WHERE {
?document dc:creator ?erdoes.
?document dc:creator ?author.
?author foaf:name ?name
FILTER (?author!=?erdoes)
}
will give following operator tree:
Filter
Compare (!=)
Var (name=author)
Var (name=erdoes)
Join
StatementPattern
Var (name=document)
Var (name=-const-10, value=http://purl.org/dc/elements/1.1/creator, anonymous)
Var (name=erdoes)
StatementPattern
Var (name=document)
Var (name=-const-11, value=http://purl.org/dc/elements/1.1/creator, anonymous)
Var (name=author)
StatementPattern
Var (name=author)
Var (name=-const-12, value=http://xmlns.com/foaf/0.1/name, anonymous)
Var (name=name)
The FilterOptimizer will not further improve this since it only checks if a contained tuple expression satisfies *all* filter expressions with its bindings. It does not consider that the first two statement patterns in combination do actually provide all the bindings names for the filter. However, with the binary join in Sesame 2.2 this optimization was possible.
I don't know if other optimizers are affected as well.
|
|
I assume that you mean Sesame 3.0 instead of 2.3?
Sorry, my mistake (first time diving into the source code). I'm working on the current trunk version. So it's Sesame 3.0 of course.
However, I did not find a way to alter the affected version in the bug report.
OK, thanks for mentioning this issue. I've changed the version numbers to 3.0.
|
|
FilterOptimizer is broken (ineffective) for new N-ary join operators of Sesame 3.0