88 – Constrained generics

Generics (functions and classes) can be constrained. This means that the variable type must be one of two (or more) specified types.

Consider the function double as a motivating example:

def double[T](value: T) -> T:
    return 2 * value

The fact that the function double is generic explicitly states the relationship between the argument type and the return type, which is preserved in calls like double(2) or double("hello").

However, with the given type hints, it’s not immediately obvious to type hinters what types of values can be passed to the function double. To make that explicit, the type variable T can be annotated with a tuple of two or more types, therefore constraining T to be one of those types:

def double[T: (str, int, float)](value: T) -> T:
    return 2 * value

In this new definition, type checkers will only accept arguments of type str, int, or float, for the function double.

For example, if None is used, mypy will complain:

Value of type variable "T" of "double" cannot be "None"  [type-var]