Which one to use, int or Integer

Integer is a better option, as it can handle null; for int, null would become 0, silently, if resultSet.getInt(..) is used. Otherwise, it might throw some exception, something like, "Unable to set null to a primitive property".

Performance is of little concern here.

• if you choose int, you will end-up adding extra handling code; and that wouldn't benefit you much. Your code will not be clean and straight-forward, lot of boiler-plate code, and you wouldn't even gain performance.
• let me make it clear, for databases, null is not same as zero. Sometimes you end-up entering 0, where null was intended. Imagine the case where user submitted a form, and doesn't supply any value for int. You will end up getting 0 by default. It makes sense, or does that really, when that field is not null in the database.

You should really make your decision based on- what you need your object to do, rather than the performance costs. Deciding based on performance should be done, once a speed issue has been identified with a profiler - the root of all evil and all that.

Look at some of the features of both and use that for your decision, e.g.

• Integer can be null, int cannot. So is the int in the DB a Nullable field?
• Do you need access to the Integer class methods?
• Are you doing arithmetic?

Personally, I always opt for the primitive over the wrapper. But that's just a preference thing, rather than based on any technical merit.

To my mind, the choice between declaring something as int or Integer simply comes down to whether null is a valid value for it or not. Autoboxing (and autounboxing) will take care of any conversion issues where the number simply must be one type or another. Performance (as has been pointed out) is also unlikely to be noticable in almost all cases.

Besides, int should be the natural choice, and is likely to be the most performant should that be an issue anyway. If you need to be able to store nulls, then you have to use Integer (and also ensure that no null references are auto-unboxed for a method that takes simply ints as this will result in a NullPointerException).

Integer is theoretically slower than int, however the performance impact should be minimal unless you are crunching numbers. Also JIT optimizations will reduce the performance loss.

Use the one that better suits your situation in terms of primitive or reference type.

To give you an idea, 2000 Integer would add about 0.5 ms to you query. If you have to serialize this data it could add quite a bit more.

However, correctness should come first. There is no point being very fast but wrong. You have to consider null values and how you handle them. (Unless the column is NOT NULL) You could use Integer.MIN___VALUE or you could use a long field instead of int and use Long.MIN_VALUE for null. Even though it is larger than int, it would still be many times smaller and more efficient than Integer.

I guess it depends among other things on what you are using for accessing the database. With plain old JDBC you could do with ints, while an ORM could silently convert them to Integers anyway. And Integer would allow you to handle nulls.

One scenario to cover would be validation.

Imagine we have the following class:

class Foo{
    @Min(value = 10)
    int val;
}

If the user doesn't provide a value for val in the request, we will get a nasty NumberFormatException.

If int is replaced with Integer, we can use @NotNull and resolve this issue more gracefully.