Using wrapper Integer class or int primitive in hibernate mapping

Question

In the company that I work for we have this major discussion on whether it should be better to use wrapping classes for primitives (java.lang.Integer, java.lang.Long) or whether to use the primitive types directly in the POJOs that map Entities to Tables in Hibernate.

The idea is that we want these values to not be null in the database.

The arguments in favor of using primitives:

• Handling these values as int means that they can never be null, in this way making it impossible to inadvertently get a null reference on the field.
• int=32/64 bits of memory. Integer = 16 bytes of memory and is also slower

The arguments in favor of using wrapper objects:

• We can add a constraint at the database level to always prevent null values from getting there
• We can end up with misleading data, we can have 0's instead of nulls in the database whenever the user doesn't set a value and buggy data is a tough catch.
• Objects have more expressive power than primitives. We have null values and also integer values, so we can validate them easier using annotations for example (javax.validation.constraints.NotNull).

Answer 1

Use wrappers, make your life simple.

Your data model should dictate this. You should be enforcing nullability in the database anyway.

If they are nullable in the database, then use wrappers. If they are not nullable, and you use wrappers, then you'll get an exception if you try and insert a null into the database.

If your data model doesn't dictate it, then go for a convention, use wrappers all of the time. That way people don't have to think, or decide that a value of 0 means null.

I would also query your assertion that it would less performant. Have you measured this? I mean really measured it? When you're talking to a database, there are a lot more considerations than the difference between 16 bits and 32 bits.

Just use the simple, consistent solution. Use wrappers everywhere, unless somebody gives you a very good reason (with accurate measured statistics) to do otherwise.