When Should You Flatten Attributes?

Flattening attributes is just another way to conceptually “add” an attribute from one entity to another. A generally better approach is to traverse the object graph directly through relationships. Enterprise Objects makes this easy by supporting the notion of key paths.

The difference between flattening attributes and traversing the object graph (either programmatically or by using key paths) is that the values of flattened attributes are tied to the database rather than the object graph. If an enterprise object in the object graph changes, a flattened attribute can quickly get out of sync.

For example, suppose you flatten a departmentName attribute into an Employee object. If a user then changes an employee’s department reference to a different department or changes the name of the department itself, the flattened attribute won’t reflect the change until the changes in the object graph are committed to the database and the data is refetched (this is because flattened attributes are derived attributes—see “Definition (Derived Attributes)” for more details). However, if you’re using key paths in this scenario, users see changes to data as soon as they happen in the object graph. This ensures that your application’s view of the data remains internally consistent.

Therefore, you should use flattened attributes only in the following cases:

• If you want to combine multiple tables joined by a one-to-one relationship to form a logical unit. For example, you might have employee data that’s spread across multiple tables such as ADDRESS, BENEFITS, and so on. If you have no need to access these tables individually (that is, if you’d never need to create an Address object since the address data is always subsumed in the Employee object), then it makes sense to flatten attributes from those entities into the Employee entity.
• If your application is read-only.
• If you’re using vertical inheritance mapping. See “Vertical Mapping”.