numpy sum of unsigned and signed number becomes floating point

Question

Why the answer is floating point? I thought it should be an integer.

>>> import numpy as np
>>> np.uint64(12)
12
>>> np.uint64(12) + int(4)
16.0
>>> 

Answer 1

Based on a recent reddit post.

This is the way Numpy functions are designed to return consistent values for the whole range of input data.

Both 16-bit unsigned and signed integers would fit into int32, therefore

>>> np.uint16(4) + np.int16(5)
9

However for uint64 and int64 the smallest datatype that would fit all possible values is float64.