python+numpy: why does numpy.log throw an attribute error if its operand is too big?

Question

Running

np.log(math.factorial(21))

throws an AttributeError: log. Why is that? I could imagine a ValueError, or some sort of UseYourHighSchoolMathsError, but why the attribute error?

I suppose NumPy first tries to convert the argument to one of its scalar types and then tries to access some NumPy-specific attribute. math.factorial(21) excedes the size of numpy.uint64, so it can't be converted to a NumPy scalar. Of course NumPy should throw a ValueError! — Sven Marnach
– Sven Marnach, Commented May 17, 2011 at 15:05
See also AttributeError in python/numpy when constructing function for certain values — tripleee
– tripleee, Commented Jul 2, 2016 at 8:30

Robert Kern · Accepted Answer · 2011-05-18 00:11:51Z

33

The result of math.factorial(21) is a Python long. numpy cannot convert it to one of its numeric types, so it leaves it as dtype=object. The way that unary ufuncs work for object arrays is that they simply try to call a method of the same name on the object. E.g.

np.log(np.array([x], dtype=object)) <-> np.array([x.log()], dtype=object)

Since there is no .log() method on a Python long, you get the AttributeError.

answered May 18, 2011 at 0:11

Robert Kern

13.5k3 gold badges37 silver badges32 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Thomas Ahle Over a year ago

Wow, this explains a lot of things

Rafal R · Accepted Answer · 2012-03-05 17:59:30Z

5

Prefer the math.log() function, that does the job even on long integers.

edited Mar 5, 2012 at 17:59

Rafal R

22.7k5 gold badges64 silver badges79 bronze badges

answered Mar 5, 2012 at 14:16

florent

511 silver badge1 bronze badge

Collectives™ on Stack Overflow

python+numpy: why does numpy.log throw an attribute error if its operand is too big?

2 Answers 2

1 Comment

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related