Hibernate查询缓存如何工作

介绍

既然我已经介绍了实体和集合缓存，现在该研究查询缓存的工作原理了。

查询缓存与实体严格相关，它在搜索条件和满足该特定查询过滤器的实体之间绘制关联。像其他Hibernate功能一样，查询缓存也不像人们想象的那么琐碎。

实体模型

对于我们的测试用例，我们将使用以下域模型：

Post实体与Author 具有多对一关联，并且两个实体都存储在第二级缓存中。

启用查询缓存

默认情况下，查询缓存处于禁用状态，要激活它，我们需要提供以下Hibernate属性：

properties.put("hibernate.cache.use_query_cache", Boolean.TRUE.toString());

为了使Hibernate缓存给定的查询结果，我们需要在创建Query时显式设置cachable查询属性。

直读缓存

查询缓存是只读的，就像NONSTRICT_READ_WRITE并发策略一样，它只能使陈旧的条目无效。

在下一个示例中，我们将缓存以下查询：

private List<Post> getLatestPosts(Session session) {return (List<Post>) session.createQuery("select p " +"from Post p " +"order by p.createdOn desc").setMaxResults(10).setCacheable(true).list();
}

首先，我们将使用以下测试案例来研究查询缓存的内部结构：

doInTransaction(session -> {LOGGER.info("Evict regions and run query");session.getSessionFactory().getCache().evictAllRegions();assertEquals(1, getLatestPosts(session).size());
});doInTransaction(session -> {LOGGER.info("Check get entity is cached");Post post = (Post) session.get(Post.class, 1L);
});doInTransaction(session -> {LOGGER.info("Check query result is cached");assertEquals(1, getLatestPosts(session).size());
});

该测试生成以下输出：

QueryCacheTest - Evict regions and run queryStandardQueryCache - Checking cached query results in region: org.hibernate.cache.internal.StandardQueryCache
EhcacheGeneralDataRegion - Element for key sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ order byquerycache0_.created_on desc;parameters: ; named parameters: {}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2
is null
StandardQueryCache - Query results were not found in cacheselectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_
fromPost querycache0_
order byquerycache0_.created_on desc limit 10StandardQueryCache - Caching query results in region: org.hibernate.cache.internal.StandardQueryCache; timestamp=5872026465492992
EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ order byquerycache0_.created_on desc;parameters: ; named parameters: {}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2
value: [5872026465492992, 1]JdbcTransaction - committed JDBC Connection------------------------------------------------------------QueryCacheTest - Check get entity is cachedJdbcTransaction - committed JDBC Connection------------------------------------------------------------QueryCacheTest - Check query is cachedStandardQueryCache - Checking cached query results in region: org.hibernate.cache.internal.StandardQueryCache
StandardQueryCache - Checking query spaces are up-to-date: [Post]EhcacheGeneralDataRegion - key: Post
UpdateTimestampsCache - [Post] last update timestamp: 5872026465406976, result set timestamp: 5872026465492992
StandardQueryCache - Returning cached query resultsJdbcTransaction - committed JDBC Connection

清除所有缓存区域，以确保缓存为空
运行后查询后，查询缓存将检查以前存储的结果
因为没有缓存条目，所以查询转到数据库
所选实体和查询结果均被缓存
然后，我们验证Post实体是否存储在二级缓存中
后续查询请求将从缓存中解决，而无需访问数据库

查询参数

查询参数嵌入在缓存条目键中，如以下示例所示。

基本类型

首先，我们将使用基本的类型过滤：

private List<Post> getLatestPostsByAuthorId(Session session) {return (List<Post>) session.createQuery("select p " +"from Post p " +"join p.author a " +"where a.id = :authorId " +"order by p.createdOn desc").setParameter("authorId", 1L).setMaxResults(10).setCacheable(true).list();
}

doInTransaction(session -> {LOGGER.info("Query cache with basic type parameter");List<Post> posts = getLatestPostsByAuthorId(session);assertEquals(1, posts.size());
});

查询缓存条目如下所示：

EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ inner joinAuthor querycache1_ on querycache0_.author_id=querycache1_.id wherequerycache1_.id=? order byquerycache0_.created_on desc;parameters: ; named parameters: {authorId=1}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2
value: [5871781092679680, 1]

该参数存储在高速缓存条目键中。缓存条目值的第一个元素始终是结果集的获取时间戳。以下元素是此查询返回的实体标识符。

实体类型

我们还可以使用实体类型作为查询参数：

private List<Post> getLatestPostsByAuthor(Session session) {Author author = (Author) session.get(Author.class, 1L);return (List<Post>) session.createQuery("select p " +"from Post p " +"join p.author a " +"where a = :author " +"order by p.createdOn desc").setParameter("author", author).setMaxResults(10).setCacheable(true).list();
}

doInTransaction(session -> {LOGGER.info("Query cache with entity type parameter");List<Post> posts = getLatestPostsByAuthor(session);assertEquals(1, posts.size());
});

缓存条目与我们之前的示例相似，因为Hibernate仅将实体标识符存储在缓存条目键中。这很有意义，因为Hibernate已经缓存了Author实体。

EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ inner joinAuthor querycache1_ on querycache0_.author_id=querycache1_.id wherequerycache1_.id=? order byquerycache0_.created_on desc;parameters: ; named parameters: {author=1}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2
value: [5871781092777984, 1]

一致性

HQL / JPQL查询无效

Hibernate二级缓存偏向于强一致性，而查询缓存也不例外。与刷新一样，只要关联的表空间发生更改，查询缓存就可以使其条目无效。每次我们持久/删除/更新实体时 ，使用该特定表的所有查询缓存条目都将失效。

doInTransaction(session -> {Author author = (Author) session.get(Author.class, 1L);assertEquals(1, getLatestPosts(session).size());LOGGER.info("Insert a new Post");Post newPost = new Post("Hibernate Book", author);session.persist(newPost);session.flush();LOGGER.info("Query cache is invalidated");assertEquals(2, getLatestPosts(session).size());
});doInTransaction(session -> {LOGGER.info("Check Query cache");assertEquals(2, getLatestPosts(session).size());
});

该测试将添加一个新的Post ，然后重新运行可缓存的查询。运行此测试将给出以下输出：

QueryCacheTest - Insert a new Postinsert
intoPost(id, author_id, created_on, name)
values(default, 1, '2015-06-06 17:29:59.909', 'Hibernate Book')UpdateTimestampsCache - Pre-invalidating space [Post], timestamp: 5872029941395456
EhcacheGeneralDataRegion - key: Post value: 5872029941395456QueryCacheTest - Query cache is invalidated
StandardQueryCache - Checking cached query results in region: org.hibernate.cache.internal.StandardQueryCache
EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ order byquerycache0_.created_on desc;parameters: ; named parameters: {}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2StandardQueryCache - Checking query spaces are up-to-date: [Post]
EhcacheGeneralDataRegion - key: Post
UpdateTimestampsCache - [Post] last update timestamp: 5872029941395456, result set timestamp: 5872029695619072
StandardQueryCache - Cached query results were not up-to-dateselectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_
fromPost querycache0_
order byquerycache0_.created_on desc limit 10StandardQueryCache - Caching query results in region: org.hibernate.cache.internal.StandardQueryCache; timestamp=5872029695668224
EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ order byquerycache0_.created_on desc;parameters: ; named parameters: {}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2
value: [5872029695668224, 2, 1]JdbcTransaction - committed JDBC ConnectionUpdateTimestampsCache - Invalidating space [Post], timestamp: 5872029695680512
EhcacheGeneralDataRegion - key: Post value: 5872029695680512------------------------------------------------------------QueryCacheTest - Check Query cacheStandardQueryCache - Checking cached query results in region: org.hibernate.cache.internal.StandardQueryCache
EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ order byquerycache0_.created_on desc;parameters: ; named parameters: {}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2StandardQueryCache - Checking query spaces are up-to-date: [Post]
EhcacheGeneralDataRegion - key: Post
UpdateTimestampsCache - [Post] last update timestamp: 5872029695680512, result set timestamp: 5872029695668224
StandardQueryCache - Cached query results were not up-to-dateselectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_
fromPost querycache0_
order byquerycache0_.created_on desc limit 10StandardQueryCache - Caching query results in region: org.hibernate.cache.internal.StandardQueryCache; timestamp=5872029695705088
EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ order byquerycache0_.created_on desc;parameters: ; named parameters: {}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2
value: [5872029695705088, 2, 1]JdbcTransaction - committed JDBC Connection

一旦Hibernate检测到Entity状态转换，它将使受影响的查询缓存区域预先失效
不会删除查询缓存条目，但会更新其关联的时间戳
查询缓存始终检查条目键时间戳，如果键时间戳比结果集加载时间戳新，它将跳过读取其值。
如果当前会话重新运行此查询，则结果将再次被缓存
当前数据库事务的提交和更改从会话级隔离传播到常规读取一致性
发生实际的无效，并且缓存条目时间戳再次更新

这种方法可能会破坏READ COMMITTED一致性保证，因为可能进行脏读，因为在提交数据库事务之前，当前隔离的更改会传播到Cache。

本机查询无效

正如我前面提到，本机查询休眠留在黑暗中，因为它可以不知道本地查询最终可能会修改其表。在以下测试中，我们将更新Author表，同时检查它对当前Post Query Cache的影响：

doInTransaction(session -> {assertEquals(1, getLatestPosts(session).size());LOGGER.info("Execute native query");assertEquals(1, session.createSQLQuery("update Author set name = '\"'||name||'\"' ").executeUpdate());LOGGER.info("Check query cache is invalidated");assertEquals(1, getLatestPosts(session).size());
});

测试生成以下输出：

QueryCacheTest - Execute native queryUpdateTimestampsCache - Pre-invalidating space [Author], timestamp: 5872035446091776
EhcacheGeneralDataRegion - key: Author value: 5872035446091776
UpdateTimestampsCache - Pre-invalidating space [Post], timestamp: 5872035446091776
EhcacheGeneralDataRegion - key: Post value: 5872035446091776updateAuthor
setname = '"'||name||'"'QueryCacheTest - Check query cache is invalidatedStandardQueryCache - Checking cached query results in region: org.hibernate.cache.internal.StandardQueryCache
EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ order byquerycache0_.created_on desc;parameters: ; named parameters: {}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2StandardQueryCache - Checking query spaces are up-to-date: [Post]
EhcacheGeneralDataRegion - key: Post
UpdateTimestampsCache - [Post] last update timestamp: 5872035446091776, result set timestamp: 5872035200290816
StandardQueryCache - Cached query results were not up-to-dateselectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_
fromPost querycache0_
order byquerycache0_.created_on desc limit 10StandardQueryCache - Caching query results in region: org.hibernate.cache.internal.StandardQueryCache; timestamp=5872035200364544
EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ order byquerycache0_.created_on desc;parameters: ; named parameters: {}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2
value: [5872035200364544, 1]JdbcTransaction - committed JDBC ConnectionUpdateTimestampsCache - Invalidating space [Post], timestamp: 5872035200372736
EhcacheGeneralDataRegion - key: Post value: 5872035200372736
UpdateTimestampsCache - Invalidating space [Author], timestamp: 5872035200372736
EhcacheGeneralDataRegion - key: Author value: 5872035200372736

即使仅修改了Author表， Author和Post缓存区域也都无效。为了解决这个问题，我们需要让Hibernate知道我们要更改哪些表。

本机查询缓存区域同步

Hibernate允许我们通过查询同步提示来定义查询表空间。提供此信息时，Hibernate可使请求的缓存区域无效：

doInTransaction(session -> {assertEquals(1, getLatestPosts(session).size());LOGGER.info("Execute native query with synchronization");assertEquals(1, session.createSQLQuery("update Author set name = '\"'||name||'\"' ").addSynchronizedEntityClass(Author.class).executeUpdate());LOGGER.info("Check query cache is not invalidated");assertEquals(1, getLatestPosts(session).size());
});

正在生成以下输出：

QueryCacheTest - Execute native query with synchronizationUpdateTimestampsCache - Pre-invalidating space [Author], timestamp: 5872036893995008
EhcacheGeneralDataRegion - key: Author value: 5872036893995008updateAuthor
setname = '"'||name||'"'QueryCacheTest - Check query cache is not invalidatedStandardQueryCache - Checking cached query results in region: org.hibernate.cache.internal.StandardQueryCache
EhcacheGeneralDataRegion -
key: sql: selectquerycache0_.id as id1_1_,querycache0_.author_id as author_i4_1_,querycache0_.created_on as created_2_1_,querycache0_.name as name3_1_ fromPost querycache0_ order byquerycache0_.created_on desc;parameters: ; named parameters: {}; max rows: 10; transformer: org.hibernate.transform.CacheableResultTransformer@110f2StandardQueryCache - Checking query spaces are up-to-date: [Post]
EhcacheGeneralDataRegion - key: Post
UpdateTimestampsCache - [Post] last update timestamp: 5872036648169472, result set timestamp: 5872036648226816
StandardQueryCache - Returning cached query resultsJdbcTransaction - committed JDBC ConnectionUpdateTimestampsCache - Invalidating space [Author], timestamp: 5872036648263680
EhcacheGeneralDataRegion - key: Author value: 5872036648263680

只有提供的表空间无效，离开了邮政查询缓存不变。可以将本机查询和查询缓存混合在一起，但是需要一些努力。

结论

查询缓存可以提高频繁执行的实体查询的应用程序性能，但这不是免费的。它容易受到一致性问题的影响，并且如果没有适当的内存管理控制机制，它很容易变得很大。

代码可在GitHub上获得。

翻译自: https://www.javacodegeeks.com/2015/06/how-does-hibernate-query-cache-work.html