CODEWITHSUNDEEP
pythonarraysnumpy
Gauthier
Gauthier

Reputation: 41945

Numpy complex array multiplication gives unsupported operand type(s)

This is working:

In [49]: type([1, 2, 3])
Out[49]: list

In [50]: type(array([1, 2, 3]))
Out[50]: numpy.ndarray

In [52]: 1j*array([1, 2, 3])
Out[52]: array([ 0.+1.j,  0.+2.j,  0.+3.j])

When I try to do this on my own list asdf:

In [46]: type(asdf)
Out[46]: list

In [47]: type(array(asdf))
Out[47]: numpy.ndarray

In [48]: 1j*array(asdf)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-48-dd0d8b4948e8> in <module>()
----> 1 1j*array(asdf)

TypeError: unsupported operand type(s) for *: 'complex' and 'numpy.ndarray'

The multiplication seemed happy with the short list, what could possibly go wrong for asdf? It is rather large, 20347 elements, but that shouldn't matter?

Upvotes: 1

Views: 559

Answers (1)

hpaulj
hpaulj

Reputation: 231385

The Python interpreter translates * to a call to __mul__ (or one of its variants). And in the case of 3 * obj, it will use the method of the obj.

Building on your comment that asdf contains strings:

In [205]: A=np.array(['A','B','C'])
In [206]: B=np.array([1,2,3])
In [207]: B.__mul__(3)
Out[207]: array([3, 6, 9])
In [208]: A.__mul__(3)
Out[208]: NotImplemented

For the array of strings the method returns 'NotImplemented', without saying anything about why. That probably is why the error message says numpy.array, the type of A, rather than the dtype of A. It's an interpreter error message, not a numpy one.

There is a version of multiply that works with strings:

In [213]: np.char.multiply(A,3)
Out[213]: 
array(['AAA', 'BBB', 'CCC'],

But not with complex multipliers.

Upvotes: 2

Related Questions