Assigning 2D-array function return values for numpy array

Question

I have a function that works perfectly. It takes a numpy 2D array and do something by the array then returns it back. I try to use the returned value to fill an array by a condition. look at below code:

>>> import numpy as np
>>> x=np.random.randint(20, size=(4,5))
>>> y=np.zeros_like(x) * np.nan
>>> x
array([[19,  0,  6, 17,  5],
       [18, 18, 10, 19,  9],
       [ 2,  5, 10,  5, 15],
       [ 9,  3,  0,  6,  9]])
>>> y
array([[ nan,  nan,  nan,  nan,  nan],
       [ nan,  nan,  nan,  nan,  nan],
       [ nan,  nan,  nan,  nan,  nan],
       [ nan,  nan,  nan,  nan,  nan]])
>>> y[ x>15 ] = 1000
>>> y
array([[ 1000.,    nan,    nan,  1000.,    nan],
       [ 1000.,  1000.,    nan,  1000.,    nan],
       [   nan,    nan,    nan,    nan,    nan],
       [   nan,    nan,    nan,    nan,    nan]])

problem is when add a function like.

>>> def foo(x):
    return x*2

>>> y[ x>15 ] = foo(x)

Warning (from warnings module):
  File "__main__", line 1
FutureWarning: assignment exception type will change in the future
Traceback (most recent call last):
  File "<pyshell#59>", line 1, in <module>
    y[ x>15 ] = foo(x)
ValueError: boolean index array should have 1 dimension

or something like:

>>> _=foo(x)
>>> y[ x>15 ]=_
Traceback (most recent call last):
  File "<pyshell#64>", line 1, in <module>
    y[ x>15 ]=_
ValueError: boolean index array should have 1 dimension

why it does not work any more!?

In your first try you use a scalar. In the second one you have a 2D array. Their sizes don't match. You can use np.where(x>15, foo(x), y). — user2285236
– user2285236, Commented Aug 2, 2017 at 9:32
@IgnacioVergaraKausel I have some conditions. in one of them y must be 0 and in other y must be 1 else y must be nan. I need nan's. — pd shah
– pd shah, Commented Aug 2, 2017 at 9:37
@ayhan yes thats right, or reshape the result ? instead using of where — pd shah
– pd shah, Commented Aug 2, 2017 at 9:38

Daniel F · Accepted Answer · 2017-08-02 09:49:18Z

1

Your major problem is that a boolean-indexed array returns a 1d array, so:

y[x > 15]
[Out]: array([nan, nan, nan, nan, nan])

So if you want to assign to it you need a 1-d array of the same size (or something that broadcasts to 1d, like a 0d scalar as you did in your first example).

So either boolean slice your input

y[x > 15] = foo(x[x > 15])  # or
y[x > 15] = foo(x)[x > 15]

Or use np.where, which preserves the shape.

y = np.where(x > 15, foo(x), y)

The first option is faster, but np.where is generally more clear and extensible.

answered Aug 2, 2017 at 9:49

Daniel F

14.5k2 gold badges34 silver badges59 bronze badges

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

Collectives™ on Stack Overflow

Assigning 2D-array function return values for numpy array

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related