QueryCache Improvement
----------------------
Key: OPENJPA-738
URL: https://issues.apache.org/jira/browse/OPENJPA-738
Project: OpenJPA
Issue Type: Improvement
Components: datacache
Affects Versions: 1.2.0, 1.2.1, 1.3.0
Reporter: Sandhya
Hi All,
I have changed the jpa code to eliminate the scan of the jpa query cache
to invalidate cached queries that is triggered by entity bean values changes.
The problem with the current code is that whenever an entity bean value
changes, the query cache needs to be scanned and all the queries that are
related to the changed entity bean should be invalidated. This is resulting in
poor performance when the querycache has too many queries in it.
The new approach is to track the timestamp of the query and the timestamp
of last update for each entity class and compare the timestamps when retrieving
the query for reuse. If the timestamp of the query result is older than the
last update time of any entity in the access path of the query, the query
result would not be reused and the query result would be evicted from the query
cache.
I introduced a configurable attribute in the persistence.xml in
QueryCache property "evictOnTimestamp=true". Therefore the QueryCache property
in persistence.xml file looks as follows:
<property name="openjpa.QueryCache" value = "CacheSize=1000,
evictOnTimestamp=true" />
If the "evictOnTimestamp" value is set to true in the xml file the new
code is invoked otherwise the old code. The default behaviour would be the old
code. That is the value for "evictOnTimestamp" is false by default so that if
this property is not set persistence.xml, old code gets invoked.
In the process of changing the code to evict the queries using the
timestamp, I had to introduce 4 methods in QueryCache interface. We want this
change to be applied to JPA trunk and 1.2.x branch. Anyone who is extending the
QueryCache that implements this interface would have to implement these 4
methods. Please respond as soon as possible if this interface change is an
issue.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
