Query Caching

JDO doesn't currently define a mechanism for caching of queries. DataNucleus provides 3 levels of caching

  • Generic Compilation : when a query is compiled it is initially compiled generically into expression trees. This generic compilation is independent of the datastore in use, so can be used for other datastores. This can be cached.
  • Datastore Compilation : after a query is compiled into expression trees (above) it is then converted into the native language of the datastore in use. For example with RDBMS, it is converted into SQL. This can be cached
  • Results : when a query is run and returns objects of the candidate type, you can cache the identities of the result objects.


Generic Query Compilation Cache

This cache is by default set to weak , meaning that the generic query compilation is cached using weak references. This is set using the persistence property datanucleus.cache.queryCompilation.type . You can also set it to strong meaning that strong references are used, or soft meaning that soft references are used, or finally to none meaning that there is no caching of generic query compilation information

You can turn caching on/off (default = on) on a query-by-query basis by specifying the query extension datanucleus.query.compilation.cached as true/false.



Datastore Query Compilation Cache

This cache is by default set to weak , meaning that the datastore query compilation is cached using weak references. This is set using the persistence property datanucleus.cache.queryCompilationDatastore.type . You can also set it to strong meaning that strong references are used, or soft meaning that soft references are used, or finally to none meaning that there is no caching of datastore-specific query compilation information

You can turn caching on/off (default = on) on a query-by-query basis by specifying the query extension datanucleus.query.compilation.cached as true/false.



Query Results Cache

This cache is by default set to weak , meaning that the datastore query results are cached using weak references. This is set using the persistence property datanucleus.cache.queryResult.type . You can also set it to strong meaning that strong references are used, or soft meaning that soft references are used, or finally to none meaning that there is no caching of query results information. You can also specify datanucleus.cache.queryResult.cacheName to define the name of the cache used for the query results cache.

You can turn caching on/off (default = off) on a query-by-query basis by specifying the query extension datanucleus.query.results.cached as true/false. As a finer degree of control, where cached results are used, you can omit the validation of object existence in the datastore by setting the query extension datanucleus.query.resultCache.validateObjects .

Obviously with a cache of query results, you don't necessarily want to retain this cached over a long period. In this situation you can evict results from the cache like this.

import org.datanucleus.jdo.JDOQueryCache;
import org.datanucleus.jdo.JDOPersistenceManagerFactory;

...
JDOQueryCache cache = ((JDOPersistenceManagerFactory)pmf).getQueryCache();

cache.evict(query);

which evicts the results of the specific query. The JDOQueryCache has more options available should you need them ... .