[
https://issues.apache.org/jira/browse/OAK-419?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Thomas Mueller updated OAK-419:
-------------------------------
Description:
Currently the query plan for the following query
{code}
select * from [nt:base] as c
right outer join [nt:base] as p on p.id = c.p
where p.id is not null
and isdescendantnode(p, '/testRoot')
and isdescendantnode(c, '/testRoot')
{code}
is:
{code}
[nt:base] as [p]
/* traverse "/testRoot//*"
where (([p].[id] is not null)
and (isdescendantnode([p], [/testRoot])))
and ([p].[id] is not null) */
left outer join [nt:base] as [c]
/* traverse "//*"
where (isdescendantnode([c], [/testRoot]))
and ([c].[p] is not null) */
on [p].[id] = [c].[p]
{code}
Because of that, far too many nodes are traversed on "c".
Instead, the isdescendantnode condition should be pused to "c" so that
it only traverses "/testRoot//*".
The result of the query is still correct, but performance is potentially very
bad (even if there is an index).
was:
Currently the query plan for the following query
select * from [nt:base] as c
right outer join [nt:base] as p on p.id = c.p
where p.id is not null
and isdescendantnode(p, '/testRoot')
and isdescendantnode(c, '/testRoot')
is:
[nt:base] as [p]
/* traverse "/testRoot//*"
where (([p].[id] is not null)
and (isdescendantnode([p], [/testRoot])))
and ([p].[id] is not null) */
left outer join [nt:base] as [c]
/* traverse "//*"
where (isdescendantnode([c], [/testRoot]))
and ([c].[p] is not null) */
on [p].[id] = [c].[p]
Because of that, far too many nodes are traversed on [c].
Instead, the isdescendantnode condition should be pused to [c] so that
it only traverses "/testRoot//*".
The result of the query is still correct, but performance is potentially very
bad (even if there is an index).
> Query: Conditions are not pushed to "outer join" selectors
> ----------------------------------------------------------
>
> Key: OAK-419
> URL: https://issues.apache.org/jira/browse/OAK-419
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: core
> Reporter: Thomas Mueller
> Assignee: Thomas Mueller
> Priority: Minor
>
> Currently the query plan for the following query
> {code}
> select * from [nt:base] as c
> right outer join [nt:base] as p on p.id = c.p
> where p.id is not null
> and isdescendantnode(p, '/testRoot')
> and isdescendantnode(c, '/testRoot')
> {code}
> is:
> {code}
> [nt:base] as [p]
> /* traverse "/testRoot//*"
> where (([p].[id] is not null)
> and (isdescendantnode([p], [/testRoot])))
> and ([p].[id] is not null) */
> left outer join [nt:base] as [c]
> /* traverse "//*"
> where (isdescendantnode([c], [/testRoot]))
> and ([c].[p] is not null) */
> on [p].[id] = [c].[p]
> {code}
> Because of that, far too many nodes are traversed on "c".
> Instead, the isdescendantnode condition should be pused to "c" so that
> it only traverses "/testRoot//*".
> The result of the query is still correct, but performance is potentially very
> bad (even if there is an index).
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
