Most Pythonic (efficient) method to create array to fill it with values

Question

My question is simple: what is the most efficient (preferred) method in Python to create an array and fill it with values column by column?

First thing that comes into my mind is just to create array empty array (or array filled with zeros) and step by step fill it based on column number.

import random
import numpy as np

values = np.zeros([10, 20])
# or values = np.empty([10,20])
for i in range(len(values.T)):
    if i == 0:
        values[:,i] = 1
    elif i < 10:
        values[:,i] = random.sample(range(1, 100), 10)
    else:
        values[:,i] =  values[:,i] - values[:,i - 1] * i

EDIT: What I am filling the columns with is not as important here as the most efficent way to do it.

Well that obviously depends how you want to fill the matrix. Here you can do it with less numpy calls. — willeM_ Van Onsem
– willeM_ Van Onsem, Commented Jul 19, 2017 at 12:37
I want to fill it column by column since some calculations, as in the example, depend on values from previous column. — An economist
– An economist, Commented Jul 19, 2017 at 12:39
But here these do not really depend: the first row consists out of ones. The next 9 out of 10s and finally it will start alternating between -10 and 10. — willeM_ Van Onsem
– willeM_ Van Onsem, Commented Jul 19, 2017 at 12:40
About the performance: I guess what you are doing is quite fine. Preallocate the correct size and fill it. Everything else like creating the array from multi-dimensional lists or appending rows to an existing numpy array is slower. — Joe
– Joe, Commented Jul 19, 2017 at 12:42
Ok, that is what I was looking for, since I am concerned more about proper way to do it rather than the calculations itself. I admit my example is stupid and do not fully reflect the complexity of calculations, since they will be based on other array. — An economist
– An economist, Commented Jul 19, 2017 at 12:45

Jonas Adler · Accepted Answer · 2017-07-19 12:53:56Z

1

You could use resize:

values = np.zeros([20])

for i in range(len(values)):
    if i == 0:
        values[i] = 1
    elif i < 10:
        values[i] = random.sample(range(1, 100), 1)
    else:
        values[i] = values[i] - values[i - 1] * i 

values = np.resize(values, [10, 20])

This is efficient since it mostly works on small arrays, defining each column, and then performs only one operation in order to create a larger array.

edited Jul 19, 2017 at 12:53

answered Jul 19, 2017 at 12:46

Jonas Adler

10.9k5 gold badges50 silver badges76 bronze badges

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

1 Comment

Jonas Adler Over a year ago

You said the values vary column by column, not row by row?

Collectives™ on Stack Overflow

Most Pythonic (efficient) method to create array to fill it with values

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