Repository navigation
float128(0).item() returns numpy.float128 #8547
Description
float128 types cannot be represented in native Python. The conversion method .item() returns a numpy.float128, contrary to documentation and to the behaviour for all other dtypes:
In [307]: type(float128(0).item())
Out[307]: numpy.float128
Not sure what the right approach here is; for the other dtypes that cannot be represented natively, I don't like the approach either.
Activity
seberg commentedon May 20, 2020
I think this is the best approach since there is no safe converesion, and you could use
float(...)otherwise. Yes, its a bit strange, but I think raising an error is worse...(We discussed this briefly on the triage meeting)