One way to switch out your data source is to define a "runtime" repository that is configured with the "runtime" data source. But this will make client code aware of the different repos:
package com...runtime.repository;
public interface RuntimeRepo extends JpaRepository<OBJECT, ID> { ... }
@Configuration
@EnableJpaRepositories(
transactionManagerRef="runtimeTransactionManager",
entityManagerFactoryRef="runtimeEmfBean")
@EnableTransactionManagement
public class RuntimeDatabaseConfig {
@Bean public DataSource runtimeDataSource() {
DriverManagerDataSource rds = new DriverManagerDataSource();
// setup driver, username, password, url
return rds;
}
@Bean public LocalContainerEntityManagerFactoryBean runtimeEmfBean() {
LocalContainerEntityManagerFactoryBean factoryBean = new LocalContainerEntityManagerFactoryBean();
factoryBean.setDataSource(runtimeDataSource());
// setup JpaVendorAdapter, jpaProperties,
return factoryBean;
}
@Bean public PlatformTransactionManager runtimeTransactionManager() {
JpaTransactionManager jtm = new JpaTransactionManager();
jtm.setEntityManagerFactory(runtimeEmfBean());
return jtm;
}
}
I have combined the code to save space; you would define the javaconfig and the repo interface in separate files, but within the same package.
To make client code agnostic of the repo type, implement your own repo factory, autowire the repo factory into client code and have your repo factory check application state before returning the particular repo implementation.