Easy way to truncate all tables, clear first and second level hibernate cache?

对于数据库,请使用 SchemaExport tool to recreate the schema:

Configuration cfg = ....;
new SchemaExport(cfg).create(false, true);

For the 2nd level cache, get access to the underlying cache regions from the SessionFactory and evict everything:

SessionFactory sf = ...;
Cache cache = sf.getCache();
cache.evictEntityRegions();
cache.evictCollectionRegions();
cache.evictQueryRegions();

For the 1st level cache, well, simply get a new Session 或呼叫 session.clear() .

采用上面的Pascal方法,我在Spring中寻找创建SchemaUpdate对象的正确方法,并意识到我不需要这样做。相反,我只需要为Hibernate获取一个Spring的sessionFactory对象,并要求它删除/创建模式。结合帕斯卡的其他解决方案,我们得到:

LocalSessionFactoryBean localSessionFactoryBean = (LocalSessionFactoryBean)appContext.getBean("&sessionFactory");
localSessionFactoryBean.dropDatabaseSchema();
localSessionFactoryBean.createDatabaseSchema();
Cache cache = sf.getCache();
cache.evictEntityRegions();
cache.evictCollectionRegions();
cache.evictQueryRegions();

这是相当有效的。唯一的缺点是(至少对我来说)它比在每个表名上调用“delete from obj1”、“delete from obj2”慢得多。不过,我还是喜欢不必重复自己的话。

看一看 Unitils . It has great support for database testing (使用dbunit)我们已经使用了很长一段时间。It's really flexible, so you can use it if you ever find the need to pre-load data into the database for particular unit tests.

With Unitils, you'll create a dataset file ( empty-db.xml )表示空数据库:

<?xml version='1.0' encoding='UTF-8'?>
<dataset>
  <obj1/>
  <obj2/>
</dataset>

在需要配置数据集的类或测试上

@DataSet("empty-db.xml")

对于所有持久性测试,我们都有一个公共的基类,因此我们能够将注释放在一个地方。

缺点是,每次将实体添加到休眠时,都必须向该文件添加行。您必须获得正确的顺序才能与外键约束保持一致。我们最后添加了一个单元测试来对照hibernate配置检查这个文件,以保持对它的检查。

The up side, especially for a large schema, is that it's way faster than rebuilding the schema.