Data migration / Legacy type mapping with assigning values

[krabat]

In my usecase the automatic data migration did not work. Here, my "User" List was changed to a Map, where the key has to be set to the User.id:

public class DataRoot1 {//OLD
	public List<User> users = new ArrayList<>();
}
public class DataRoot2 {//CURRENT
	public Map<Long, User> users = new HashMap<>();
}
public class User {
	public Long id;
	public String name;
}

How can I migrate my EclipseStore data where some new Data has to be set/calculated?

[hg-ms]

Hi,
There is no simple build in solution to convert between lists and maps. The simplest solution may be to do that conversion in the application. To do so:

1. Add the new map to your class, don’t replace the original list
2. Add some code to your application that initialized the new map with the values form the original list, don’t forget the store
3. Start the app, execute the migration code.
4. Optional: remove the old list from the class

Another option would be to implement a custom LegacyTypeHandler that does the required mapping, but this is a quite complex task.

[krabat]

Thank you for your answer.
I have to admit I hoped for another solution. Something like this:

public interface MigrationExecutor<O, C> {
	void migrate(O oldInstance, C currentInstance);
}
public class MigrationExecutorImpl<DataRoot1, DataRoot2> implements MigrationExecutor {
	@Override
	public void migrate(DataRoot1 dataRoot1, DataRoot2 dataRoot2) {
		dataRoot1.users.forEach(user -> dataRoot2.users.put(user.id, user));
	}
}

I think data migration between modifications of DataRoot classes is very important. Ideally it would feature defined data migration in some way as Flyway does it for DBs.

In your example you execute the migration code manually at application start. That's ok for data migration of one alteration. But consider the use-case where the persisted data lacks several iterations behind the modified DataRoot definition in the application code. Then you would benefit from the Flyway approach as the data state is checked and then only the exactly required data migrations are done in the right order.

(edit: minor clarification)