How to cache results of a Spring Data JPA query method without using query cache?

Question

I have a Spring Boot app with Spring Data JPA (hibernate backend) repository classes. I've added a couple custom finder methods, some with specific @Query annotation to tell it how to get the data. I have already set up EhCache for the hibernate 2nd level cache, but so far, the only way I can get these results caching is to enable the hibernate query cache. I'd prefer to define a specific cache and store the actual domain objects there just as if it were a normal finder. Below is my repo code:

public interface PromotionServiceXrefRepository extends PagingAndSortingRepository<PromotionServiceXref, Integer> {

  @Query("SELECT psx FROM Customer c " +
         "JOIN c.customerProductPromotions cpp " +
         "JOIN cpp.productPromotion pp " +
         "JOIN pp.promotion p JOIN p.promotionServiceXrefs psx " +
         "WHERE c.customerId = ?1")
  @QueryHints(@QueryHint(name = "org.hibernate.cacheable", value = "true"))
  @Cache(usage = CacheConcurrencyStrategy.READ_ONLY, region = "promotionServiceXrefByCustomerId")
  Set<PromotionServiceXref> findByCustomerId(int customerId);
}

And here is the "promotionServiceXrefByCustomerId" cache I defined, that is NOT being used:

<cache name="promotionServiceXrefByCustomerId" overflowToDisk="true" diskPersistent="true"
       maxEntriesLocalHeap="3000000" eternal="true" diskSpoolBufferSizeMB="20" memoryStoreEvictionPolicy="LFU"
       transactionalMode="off" statistics="true">
</cache>

What am I doing wrong? If I enable StandardQueryCache then this data gets cached there and hibernate does not execute a query. But when I disable the query caching, this does not get cached. What am I doing wrong here? PLEASE HELP!

Answer 1

The reason the code you have is not working is that @Cache is not intended to work that way. If you want to cache the results of a query method execution, the easiest way is to use Spring's caching abstraction.

interface PromotionServiceXrefRepository extends PagingAndSortingRepository<PromotionServiceXref, Integer> {

  @Query("…")
  @Cacheable("servicesByCustomerId")
  Set<PromotionServiceXref> findByCustomerId(int customerId);

  @Override
  @CacheEvict(value = "servicesByCustomerId", key = "#p0.customer.id")
  <S extends PromotionServiceXref> S save(S service);
}

This setup will cause results of calls to findByCustomerId(…) be cached by the customer identifier. Note, that we added an @CacheEvict to the overridden save(…) method, so that the cache we populate with the query method is evicted, whenever an entity is saved. This probably has to be propagated to the delete(…) methods as well.

Now you can go ahead an configure a dedicated CacheManager (see the reference documentation for details) to plug in whichever caching solution you prefer (using a plain ConcurrentHashMap here).

 @Configuration
 @EnableCaching
 class CachingConfig {

   @Bean
   CacheManager cacheManager() {

     SimpleCacheManager cacheManager = new SimpleCacheManager();
     cacheManager.addCaches(Arrays.asList(new ConcurrentMapCache("servicesByCustomerId)));

     return cacheManager;
   }
 }

Answer 2

You need to be aware that by giving up on the Hibernate QueryCache your are responsible for invalidating the queries that become stale when saving, updating, deleting entities that influenced the query result(what Oliver is doing by setting CacheEvict on save) - which I think can be a pain- or at least you need to take into account and ignore it if it's not really a problem for your scenario.

Answer 3

First I quote your question:

What am I doing wrong?

The way you are trying to name the cache is not appropriate to how hibernate will use it. Check org.hibernate.engine.spi.CacheInitiator which uses org.hibernate.internal.CacheImpl which is based on:

if ( settings.isQueryCacheEnabled() ) {
    final TimestampsRegion timestampsRegion = regionFactory.buildTimestampsRegion(
            qualifyRegionName( UpdateTimestampsCache.REGION_NAME ),
            sessionFactory.getProperties()
    );
    updateTimestampsCache = new UpdateTimestampsCache( sessionFactory, timestampsRegion );
    ...
}

And UpdateTimestampsCache.REGION_NAME (equals to org.hibernate.cache.spi.UpdateTimestampsCache) is what you are missing as the cache name. For the query cache you'll have to use exactly that cache name and no other!

Now few other thoughts related to your problem:

• removing @Cache and setting cache name to org.hibernate.cache.spi.UpdateTimestampsCache will allow your query to be cached with ehcache by hibernate (spring cache abstraction is not involved here)
• setting a hardcoded cache name won't make you happy I'm sure but at least you know why this happens
• Balamaci Serban (the post just below) is painfully right

Below is the configuration from one of my projects where ehcache + @Query + @QueryHints work as expected (ehcache/ehcache-in-memory.xml file):

<?xml version="1.0" encoding="UTF-8"?>
<ehcache name="in-memory" xmlns="http://ehcache.org/ehcache.xsd">
    <!--<diskStore path="java.io.tmpdir"/>-->

    <!--
        30d = 3600×24×30 = 2592000
    -->

    <cache name="org.hibernate.cache.internal.StandardQueryCache"
           maxElementsInMemory="9999" eternal="false"
           timeToIdleSeconds="2592000" timeToLiveSeconds="2592000"
           overflowToDisk="false" overflowToOffHeap="false"/>

    <cache name="org.hibernate.cache.spi.UpdateTimestampsCache"
           maxElementsInMemory="9999" eternal="true"
           overflowToDisk="false" overflowToOffHeap="false"/>

    <defaultCache maxElementsInMemory="9999" eternal="false"
                  timeToIdleSeconds="2592000" timeToLiveSeconds="2592000"
                  overflowToDisk="false" overflowToOffHeap="false"/>
</ehcache>

and hibernate.properties:

hibernate.jdbc.batch_size=20
hibernate.show_sql=true
hibernate.format_sql=true
hibernate.validator.autoregister_listeners=false
hibernate.cache.use_second_level_cache=true
hibernate.cache.use_query_cache=true
hibernate.cache.region.factory_class=org.hibernate.cache.ehcache.EhCacheRegionFactory
hibernate.hbm2ddl.auto=update
net.sf.ehcache.configurationResourceName=ehcache/ehcache-in-memory.xml
hibernate.dialect=org.hibernate.dialect.H2Dialect

and some versions from pom.xml for which my explanation applies:

<springframework.version>5.0.6.RELEASE</springframework.version>
<spring-security.version>5.0.5.RELEASE</spring-security.version>
<spring-data-jpa.version>2.1.0.RELEASE</spring-data-jpa.version>
<hibernate.version>5.2.13.Final</hibernate.version>
<jackson-datatype-hibernate5.version>2.9.4</jackson-datatype-hibernate5.version>

And the full working test is image.persistence.repositories.ImageRepositoryTest.java found here: https://github.com/adrhc/photos-server/tree/how-to-cache-results-of-a-spring-data-jpa-query-method-without-using-query-cache
Yep, run mvn clean install or change my env.sh if you really want to use my shell scripts. Check then the number of sql queries on behalf of 3x imageRepository.count() call.