review: refactor: SignaturePrinter prints type parameter bounds, no name

[pvojtechovsky]

Signature shouldn't contain name of type parameter reference, because these three methods declarations:

<T extends String> void m(T a);
<S extends String> void m(S b);
void m(String c)

Should have the same signature: m(String)

WDYT?

[surli]

What if the type parameter is defined at the class level? Such as:

class Foo<T> {
 void m(T a);
}

or

class Bar<T extends String> {
 void m(T a);
}

[pvojtechovsky]

What if the type parameter is defined at the class level?

It should be same - the super class of the type parameter should be used.

[surli]

It should be same - the super class of the type parameter should be used.

But in case of Foo you don't know that T extends String... So it would be:
m(Object) or m(T)?

[pvojtechovsky]

in case of

class Foo<T> {
 void m(T a);
}

m has signature m(Object)

[monperrus]

OK for me. Thanks Pavel.

[pvojtechovsky]

I see now, more cases, which are probably not handled well too.

<T extends List<String>> void m(T a);
<T extends List<Integer>> void m(T b);
<T extends List> void m(T c);
void m(List<String> d);
void m(List<?> e);
void m(List f);

all of them should have same signature: m(List)

WDYT?

If you agree than we should use CtTypeReference#getTypeErasure() during signature printing too. May be it is task for another small PR.

[surli]

I'm ok with the other changes you propose. For me they could be implemented in the same PR or in another one, as you prefer.