VoidTypeName should not exist about javapoet

[JakeWharton]

I like the idea here. We can't totally eliminate VoidTypeName because of TypeNames.forClass, but we can make it non-public for that case.

As to the suggestions, I'm not sure which approach I'd prefer yet. Need to think about it a bit.

[gk5885]

I understand the principle here, but both of the standard APIs for working with types in Java represent void as a type:

• https://docs.oracle.com/javase/7/docs/api/java/lang/Void.html#TYPE
• https://docs.oracle.com/javase/7/docs/api/javax/lang/model/type/NoType.html

IMO, the decision was made for us and IllegalArgumentException just isn't that bad.

[eamonnmcmanus]

Neither of those is part of an API for defining the type of something, though. You can never get void.class as the result of Field.getType() or Method.getParameterTypes()[0], and there's no way with reflection or annotation-processing to create a field, say, where you could accidentally specify a type of void.class or NoType.

I think having the Java type system help here would be beneficial to users, and I don't really see much of a downside. I do think that TypeNames.forClass and TypeNames.forTypeMirror should throw if given void.class or NoType respectively, and that there should be TypeNames.forClassReturnType and TypeNames.forTypeMirrorReturnType that do the right thing for those before forwarding to the existing methods.

[swankjesse]

No action to take here with the current codebase.