[ 
https://issues.apache.org/jira/browse/JENA-628?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13974427#comment-13974427
 ] 

ASF subversion and git services commented on JENA-628:
------------------------------------------------------

Commit 1588554 from [email protected] in branch 'jena/trunk'
[ https://svn.apache.org/r1588554 ]

Added additional test cases that pertain to JENA-628.  This is a sanitized 
version of a customer query in our regression suite where the changed filter 
placement behaviour broke the query when using the don't break BGPs mode.  As 
part of adding this test case I also found and fixed a case where filter 
pushing was not respecting the includeBGPs flag when considering whether to 
push into quadpatterns

> TransformFilterPlacement does not always push down through assign/extend when 
> it could
> --------------------------------------------------------------------------------------
>
>                 Key: JENA-628
>                 URL: https://issues.apache.org/jira/browse/JENA-628
>             Project: Apache Jena
>          Issue Type: Bug
>          Components: ARQ
>    Affects Versions: Jena 2.11.1
>            Reporter: Rob Vesse
>            Assignee: Rob Vesse
>             Fix For: Jena 2.11.2
>
>
> I've noticed a bug in the improved TransformFilterPlacement where it doesn't 
> always do the correct thing when pushing through assign/extend in cases where 
> the placement into BGPs is disabled.
> Consider the following example algebra:
> {noformat}
> (filter (= ?x 123) (assign ((?z 123)) (bgp (?s ?p ?x)) ))
> {noformat}
> With optFilterPlacementBGP enabled this produces the following improved 
> algebra:
> {noformat}
> (assign ((?z 123)) (filter (= ?x 123) (bgp (?s ?p ?x)) ))
> {noformat}
> However with optFilterPlacementBGP disabled this leaves the algebra as is 
> even though it could safely push it through the bind without pushing it into 
> the BGP itself.
> I think the fix for this is trivial and I will apply it shortly.  I will also 
> look and see if there are other cases where it fails to push down as far as 
> it could.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to