Dispatching on numbers

Dispatch matches the runtime type of a value, so an overload annotated float accepts only actual float instances, and an int is not a float under this rule. Python's numeric tower, described in PEP 484, treats int as a kind of float, which is in turn a kind of complex, but that promotion is a convention static type checkers follow. Dispatch does not apply it, so annotate a parameter float only when the overload is meant for an actual float value and nothing else:

>>> from bagof.dispatchers import dispatch, NoMethodError
>>> @dispatch
... def halve(x: float) -> float:
...     return x / 2
>>> halve(2.0)
1.0
>>> try:
...     halve(3)          # an int is not a float
... except NoMethodError:
...     print("no overload for int")
no overload for int

To accept any real number, annotate the parameter with numbers.Real instead of float. numbers.Real matches int, float, bool, fractions.Fraction, and any third-party numeric type registered under the numbers abstract base classes, such as NumPy scalars:

>>> from numbers import Real
>>> @dispatch
... def classify(x: Real) -> str:
...     return "a real number"
>>> classify(3)           # int
'a real number'
>>> classify(2.5)         # float
'a real number'

Use numbers.Integral to accept integers, which include int, bool, and NumPy's integer types. Use numbers.Complex to also accept complex, or spell the accepted set out explicitly with a union such as Union[int, float].