In the context of inheritance, when B is subclass of A , an instance of B can be used anywhere where an instance of A is valid. This is the characteristic of polymorphism. However I have a question.
I know it's correct:
A oa = new A();
B ob = new B();
A oab = new B();
However, is this also correct?
B oba = new A();
And if it is, I'd like to know why.
Think that B is Poodle and A is Cachorro . All Poodle is Cachorro , but the opposite is not always valid.
For example:
// Ok. Em uma variável do tipo Poodle, dá para colocar um Poodle sem problemas.
Poodle w = new Poodle();
// Ok. Poodle é um Cachorro, então dá para colocar em uma variável do tipo Cachorro.
Cachorro x = new Poodle();
// Ok. Cria algum vira-latas genérico e coloca em uma variável do tipo Cachorro.
Cachorro y = new Cachorro();
// Não! Não dá para se afirmar que um cachorro genérico é um poodle.
Poodle z = new Cachorro();
Not correct because B is more specific than A . This would compile if you did a cast :
B oab = (B) new A ();
This is because although this is the new operator, an object coming from a variable of type A could have been created from a subclass C that was not compatible with B :
class C extends A{...}
A myA = new C ();
B myB = myA; // aqui precisa falhar a computação
But if you put cast it's basically telling the compiler to trust you.