[
https://issues.apache.org/jira/browse/JENA-587?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13820026#comment-13820026
]
Andy Seaborne edited comment on JENA-587 at 11/12/13 11:31 AM:
---------------------------------------------------------------
We should not reply on the ARQ join strategy.
* It may change and it does not apply to all storage systems.
* The join order is only sufficiently predicable for some cases like BGPs -
even adding union default graph may scramble the the order coming of the
{{WHERE}} clause depends on the pattern (e.g. some inner SELECTs mixed with
other things). We use hash tables for {{MINUS}}.
It is a legal optimization if the {{DISTINCT}} is of variables that are in
order due to {{ORDER BY}}
Legal:
1. {{DISTINCT ?v ORDER BY ?v}}
2. {{DISTINCT ?v ORDER BY ?v ?w}}
3. {{DISTINCT ?v ?w ORDER BY ?v ?w}}
4. {{DISTINCT ?v ?w ORDER BY ?w ?v}}
{{DISTINCT * ORDER BY ...}} is possible only if the ORDER BY is a total
ordering of the underlying pattern.
Not legal:
1. {{DISTINCT ?v ORDER BY ?w}}
2. {{DISTINCT ?v ORDER BY ?w ?v}} because not sorted by ?v first.
Maybe the first step is to just do some simple cases such as {{ORDER BY}}
exactly the variables of the project of the {{DISTINCT}} then expand the
intelligence of the transformation.
was (Author: andy.seaborne):
We should not reply on the ARQ join strategy.
* It may change.
* The join order is only predicable for BGPs - even adding union default graph
may
The order coming of the {{WHERE}} clause depends on the pattern (e.g. sub
SELECTs mixed with other things).
*
It is a legal optimization if the {{DISTINCT}} is of variables that are in
order due to {{ORDER BY}}
Legal:
1. {{DISTINCT ?v ORDER BY ?v}}
2. {{DISTINCT ?v ORDER BY ?v ?w}}
3. {{DISTINCT ?v ?w ORDER BY ?v ?w}}
4. {{DISTINCT ?v ?w ORDER BY ?w ?v}}
Not legal:
1. {{DISTINCT ?v ORDER BY ?w}}
2. {{DISTINCT ?v ORDER BY ?w ?v}} because not sorted by ?v first.
> SELECT DISTINCT returns duplicate results
> -----------------------------------------
>
> Key: JENA-587
> URL: https://issues.apache.org/jira/browse/JENA-587
> Project: Apache Jena
> Issue Type: Bug
> Components: ARQ
> Affects Versions: Jena 2.11.0
> Reporter: Veyriere
> Attachments: D.ttl, Q.rq, bug Jena2.11.0.zip, jena-587.zip
>
>
> SELECT DISTINCT returns duplicate results. Attaching a small quads dump and
> the query to reproduce with TDB
> Reproduced with Jena 2.11.0 and Jena 2.10.1 (was working with 2.7.4)
--
This message was sent by Atlassian JIRA
(v6.1#6144)
