This happens in numpy as well, and was confusing the first time I saw it:

    b = np.ones(1)
    c = np.ones(1)
    c = c/(2**73)
    b += c

    >>>  ufunc 'add' output (typecode 'O') could not be coerced to provided output parameter (typecode 'd') according to the casting rule ''same_kind''

but `b = b + c` does work.

[1] https://stackoverflow.com/questions/12588986/typeerror-gener...