Prior to this commit, the support for SQL script execution via @Sql
provided an algorithm for looking up a required
PlatformTransactionManager to use to drive transactions. However, a
transaction manager is not actually required for all testing scenarios.
This commit improves the transaction management support for @Sql so
that SQL scripts can be executed without a transaction if a transaction
manger is not present in the ApplicationContext. The updated algorithm
now supports the following use cases.
- If a transaction manager and data source are both present (i.e.,
explicitly specified via the transactionManager and dataSource
attributes of @SqlConfig or implicitly discovered in the
ApplicationContext based on conventions), both will be used.
- If a transaction manager is not explicitly specified and not
implicitly discovered based on conventions, SQL scripts will be
executed without a transaction but requiring the presence of a data
source. If a data source is not present, an exception will be thrown.
- If a data source is not explicitly specified and not implicitly
discovered based on conventions, an attempt will be made to retrieve
it by using reflection to invoke a public method named
getDataSource() on the transaction manager. If this attempt fails,
an exception will be thrown.
- If a data source can be retrieved from the resolved transaction
manager using reflection, an exception will be thrown if the
resolved data source is not the data source associated with the
resolved transaction manager. This helps to avoid possibly
unintended configuration errors.
- If @SqlConfig.transactionMode is set to ISOLATED, an exception will
be thrown if a transaction manager is not present.
Issue: SPR-11911