performance issue with JDBCCacheLoader

[gmeroz]

i use version 2.0.0CR1 running inside JBoss AS with PojoCache.

i have a senario where i need to check for a list of nodes which of them are in the cache already (so i'll know if i need to add them to the cache latter).

to check if a node is in the cache i use cache.find(...), but those calls are expensive since they load a Blob from the DB.

Is there a way to do something like Map.containsKey() ? which will run a cheaper query to the DB?

[manik]

A find() will load the Pojo, potentially from the cache loader. Nodes are stored in the DB (when using the jdbc cache loader) as blobs, so that blob lookup is necessary to recreate the node in the cache, and then the Pojo before returning the Pojo when calling find().

Perhaps what you need is something like pojoExists() - which may perform a number of cache.hasChild() calls (which in turn calls a number of loader.exists() calls) on the components that make up the Pojo?

[gmeroz]

yes, that's exactly what i'm looking for.

It may perform queries to DB as long as it does not access the Blob column. (like select count(*)...).

[manik]

Why don't you create a JIRA for this as a feature request for the PojoCache API?

[gmeroz]

http://jira.jboss.com/jira/browse/JBCACHE-1158

[jason.greene]

As a workaround, you can use:

Fqn fqn = Fqn.fromString("/path/to/pojo");
pojoCache.getCache().getRoot().getChild(fqn.getParent()).getChildrenNames().contains(fqn.getLastElementAsString());