Some points to consider when using Mybatis cache

When using Mybatis cache, there are some points to consider.

Local session cache is enabled with default option
Cache boundary is for all the queries within a SqlSession
An item is cached when querying a record
The item is reused when querying with the same parameter
Cache is flushed when insert/update/delete is executed or when the SqlSession is closed
When flushing cache, all cache items are deleted

Global cache is on by explicit option
Cache boundary is a mapper’s namespace (= all the sqls within the mapper file)
An item is cached for queried data
Cache items are cached when SqlSession.close(), commit() or rollback() (which is different from local cache)
Cache flushing happens when SqlSession.close(), commit() or rollback()
Cache flushing happens if insert/update/delete is executed, in that case, all queried data before insert/update/delete are deleted
For example, query record 1 -> query record 2 -> update record 1 -> query record 3 happens, only record 3 is cached (record 1 and 2 are not cached because update is executed)

If you want to use global cache, split mapper into cache boundary (Don’t put all sqls into one mapper.xml)
If insert/update/delete happens frequently, global cache is not recommended (It is diffecult to verify a cache’s lifecycle)
Use local cache only if concurrent access to the same record does not happen
If insert/update/delete happens frequently, local cache also is not recommended
To sum up, I recommend not to use Mybatis cache (local or global) if you don’t understand it’s caching mechanism. Rather than, explicit logic (explicit cache api call or Spring @Transactional annotation) could be better solution

Published by TK