How can I fill a numpy structured array from a function?

Question

I have created a structured array using numpy. Each structure represents the rgb value of a pixel.

I am trying to work out how to fill the array from a function but I keep getting an 'expected a readable buffer object' error.

I can set individual values from my function but when I try to use 'fromfunction' it fails.

I copied the dtype from the console.

Can anyone point out my mistake?

Do I have to use a 3 dimensional array as opposed to a 2d of structures

import numpy as np

#define structured array
pixel_output = np.zeros((4,2),dtype=('uint8,uint8,uint8'))
#print dtype
print pixel_output.dtype

#function to create structure
def testfunc(x,y):
    return (x,y,x*y)

#I can fill one index of my array from the function.....
pixel_output[(0,0)]=testfunc(2,2)

#But I can't fill the whole array from the function
pixel_output = np.fromfunction(testfunc,(4,2),dtype=[('f0', '|u1'), ('f1', '|u1'), ('f2', '|u1')])

pixel_output = ... does not try to fill the array. It assigns the output of fromfunction to that variable, replacing anything that was there before. — hpaulj
– hpaulj, Commented Oct 19, 2013 at 18:46
The location of your error is relevant. It occurs several levels down in fromfunction, on a line that deals with dtype. What does the documentation for this function say about dtype? — hpaulj
– hpaulj, Commented Oct 19, 2013 at 18:48

hpaulj · Accepted Answer · 2013-10-19 18:43:53Z

X=np.fromfunction(testfunc,(4,2))
pixel_output['f0']=X[0]
pixel_output['f1']=X[1]
pixel_output['f2']=X[2]
print pixel_output

produces

array([[(0, 0, 0), (0, 1, 0)],
       [(1, 0, 0), (1, 1, 1)],
       [(2, 0, 0), (2, 1, 2)],
       [(3, 0, 0), (3, 1, 3)]], 
      dtype=[('f0', 'u1'), ('f1', 'u1'), ('f2', 'u1')])

fromfunction returns a 3 element list of (4,2) arrays. I assign each one, in turn, to the 3 fields of pixel_output. I'll leave the generalization to you.

Another way (assign a tuple to element)

for i in range(4):
    for j in range(2):
        pixel_output[i,j]=testfunc(i,j)

And with the magical functiion http://docs.scipy.org/doc/numpy/reference/generated/numpy.core.records.fromarrays.html#numpy.core.records.fromarrays

pixel_output[:]=np.core.records.fromarrays(X)

When I look at the fromarrays code (with Ipython ??), I see it is doing what I did at first - assign field by field.

for i in range(len(arrayList)):
    _array[_names[i]] = arrayList[i]

Collectives™ on Stack Overflow

How can I fill a numpy structured array from a function?

1 Answer 1

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related