Thomas Mueller created OAK-12100:
------------------------------------
Summary: Queries with nodeName don't run efficiently (estimated
cost is too high)
Key: OAK-12100
URL: https://issues.apache.org/jira/browse/OAK-12100
Project: Jackrabbit Oak
Issue Type: Improvement
Components: lucene, query
Reporter: Thomas Mueller
The following query:
{noformat}
SELECT * FROM [acme:Asset] AS asset
WHERE ISDESCENDANTNODE(asset, '/content')
AND (NAME(asset) = 'berne'
OR asset.[jcr:content/metadata/alias] = 'running'
OR asset.[jcr:uuid] = 'running')
{noformat}
with jcr:content/metadata/alias and jcr:uuid are both configured in the
definition with propertyIndex=true .
When run, it uses an index but it only uses the path filter (ancestor).
And when removing the filter on nodeName, the filter on alias is applied at
index level.
Removing the filter on alias and jcr:uuid : the filter on nodeName is applied
at index level.
Why jcr:uuid is never used in the query even if configured as property index?
The same behaviour happens on both elastic and lucene. It may be a problem is
in the query engine.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
