Populate a numpy array from index, value pairs without iterating

Question

I want to associate values into an array by index, without iterating. Picture the following, but for hundreds of thousands of values:

to_populate = [
    [[0, 0], 1],
    [[1, 1], 100],
    [[2, 3], 29],
    [[3, 2], 33],
]

mat = np.empty((4, 4))
mat[:] = np.nan

for idx, val in to_populate:
    x, y = idx
    mat[x, y] = val

# array([[  1.,  nan,  nan,  nan],
#        [ nan, 100.,  nan,  nan],
#        [ nan,  nan,  nan,  29.],
#        [ nan,  nan,  33.,  nan]])

I'm thinking of it as conceptually like the reverse of np.argwhere, but with specific values rather than a predicate condition.

mat = np.random.randn(3, 3)

# array([[ 0.89298522,  0.41059024,  0.32770948],
#        [-0.91956498, -0.11774805, -1.42625182],
#        [ 1.28644586, -0.06951971, -0.88742959]])

np.argwhere(mat < 0)

# array([[1, 0],
#        [1, 1],
#        [1, 2],
#        [2, 1],
#        [2, 2]])

FBruzzesi · Accepted Answer · 2020-05-20 13:20:00Z

2

You can "unpack" to_populate into row indexes, column indexes and values and then use slicing:

import numpy as np
to_populate = [
    [[0, 0], 1],
    [[1, 1], 100],
    [[2, 3], 29],
    [[3, 2], 33],
]
# flatten the list
to_populate = np.array([[i[0][0], i[0][1], i[1]] for i in to_populate])

idx = to_populate[:,0]
idy = to_populate[:,1]
values = to_populate[:,2]

mat = np.full((4,4),np.nan)
mat[idx,idy]=values

mat
array([[  1.,  nan,  nan,  nan],
       [ nan, 100.,  nan,  nan],
       [ nan,  nan,  nan,  29.],
       [ nan,  nan,  33.,  nan]])

edited May 20, 2020 at 13:20

answered May 20, 2020 at 13:11

FBruzzesi

6,6143 gold badges19 silver badges42 bronze badges

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

1 Comment

Sooheon Kim Over a year ago

Thank you! to_populate was being created from dataframe columns, so I can save a step and just index by the columns directly.

Collectives™ on Stack Overflow

Populate a numpy array from index, value pairs without iterating

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related