Toy language type inference

A toy language has these types:

• primitives: exactly int, float, str, bool, char;
• generics: any other name, like T1, T2, S;
• tuples: an ordered list of types, possibly nested (and possibly empty).

Implement two classes:

• Node(value): value is either a name (str) or a list of Nodes (a tuple). Keep it in an attribute called value.
  • str(node): a name prints as itself; a tuple prints as [ + its children joined by , + ], no spaces.
  • node == other: structural equality (same names, same nesting). Comparing with something that isn't a Node is False.
• Function(parameters, output_type): a list of Nodes and a Node, kept as attributes with those names.
  • str(function): ( + parameters joined by , + ) -> + the output type.

t = Node([Node("int"), Node([Node("T1"), Node("str")])])
str(t)                                    # "[int,[T1,str]]"
str(Function([t, Node("T1")], Node("T1")))  # "([int,[T1,str]],T1) -> T1"
Node([Node("int")]) == Node("int")          # False

No parsing yet: tests build the objects directly.