Can I use throws in constructor?

Question

I have to initialize file objects inside the constructor and for handling the exception, is it efficient using throws or should I go for try/

Answer 1

It's okay to throw an exception in the constructor. I know some of the Java library classes do so (URI for just one example). I think it's better to throw an exception than to return an object in an unknown or invalid state.

Answer 2

You are supposed to take the object to a safe state in the constructor, but let's imagine you want to open a file(for reading) that doesn't exists, throwing the exception is the only way. So it depends on the logic you implement

Answer 3

I'd go with 'throws' if you care about what is initialized in the constructor, which I guess you do.

If you hide the exceptions, then that will probably cause problems later on.

Answer 4

In general it is a good idea to have only simple logic in your constructor (for example setting private fields with arguments values). And set up your objects with other special "assemblers", "preparers" or "factories". The same is about get/set methods- they should be as simple as possible.

Sure it is possible to throw an exception from constuctor. But this is not a good practice.

Answer 5

Consider a constructor, or for that matter any method, to have a contract. The contract for a constructor is very simple - you give me (zero, one or more) parameters and I'll give you back a constructed object. Good practice would suggest that this object should have its internal data structures properly initialised and invariants intact, although there's nothing in the language to enforce this per se.

If, for some reason the constructor cannot hold to this contract, then it should throw an exception. That might be because the parameters passed (if any) were not acceptable (pre-condition failure) or some external problems (file-system full, heap exhaustion, network outage etc.) prevented it.

Answer 6

I think throwing exception in a constructor is an elegant way to indicate an error condition inside the constructor. Otherwise, you would have to create another function which initializes the resources, and call that function after the object is constructed.