[Python] How can I control conversion from custom object to numpy.longdouble?

I have a custom class that (for historical reasons) stores high-precision numbers as two integers representing the integer part and the nano-precision non-integer part, e.g.:

class Custom(object):
def __init__(self, int_, nano_):
self._int = int_
self._nano = nano_

def __float__(self):
return self._int + self._nano * 1e-9

def __int__(self):
return self._int


However, when I try to convert this to numpy.longdouble it's going through the __float__ method:

>>> a = Custom(1234567890, 123456789)
>>> print(numpy.longdouble(a))
1234567890.1234567165
>>> print(numpy.longdouble(float(a)))
1234567890.1234567165
>>> print(numpy.longdouble(a._int) + numpy.longdouble(a._nano * 1e-9))
1234567890.123456789


The final example there is the result I want. So the question is, how can I override the default behaviour of casting to numpy.longdouble?

Continue reading...