是否可以将不同的子类传递给同一PCollection？

Question

Is it possible to have a PCollection in Beam that "holds" different objects of the same superclass such that

PCollection<T extends SuperClass> result =  input.apply(ParDo.of(new DoFnThatReturnsSubClasses()));

instead of

PCollection<SubClassA> A = input.apply(ParDo.of(new DoFnSubClassesA()));
PCollection<SubClassB> B = input.apply(ParDo.of(new DoFnSubClassesB()));

My superclass holds some basic fields as follows

public class SuperClass {

String key; 
Long timestamp;

}

But my subclasses will contain an additional field which have the same variable name of different data types. Ie

public class SubClassA extends SuperClass {   
    List<HashMap<String, String>> value = new ArrayList<>();       
}

Another subclass,

public class SubClassB extends SuperClass {   
    String value;       
}

What I really want to achieve is to allow the field value to be of different data types. The only thing I could think was dividing this up into different subclasses. I realise this might be more of a Java programming question and poor understanding of the language.

Answer 1

Yes, this is possible. You can have a PCollection that contains a mix of SubclassX objects. You'll need to make sure that the Coder that you are using for the PCollection can encode and decode all the possible subclasses.

Your DoFns downstream will receive Superclass objects, and you'll need to downcast to the appropriate subclass.