The relevant piece of Python code from fractions.py looks like this:

    if type(numerator) is int is type(denominator):
        # *very* normal case
        g = math.gcd(numerator, denominator)
        if denominator < 0:
            g = -g
    else:
        g = _gcd(numerator, denominator)

I wonder whether we should simply drop the type check: math.gcd should accept anything that supports __index__ (and that includes NumPy integer types and anything that implements numbers.Integral).

OTOH, math.gcd will return a plain int, and if we have some subclass myint of int, we might want all the arithmetic to stay in the domain of myints (though I'm not really sure _why_ we'd want that). In that case we'll need to restore the pure Python _gcd implementation.