Index is out of bounds when using for loop

Question

I want to create a new array with specific elements of the original array. I created a minimal example and there it works what I want to achieve, but on the actual data it doesn't and I can't figure out the essential difference.

First the minimal example: I want to get the numbers of every row that stand under the letters A-C and save it in the array B_feat.

 import numpy as np
 years = 5 #A-E
 yearsf = 3 #A-C
 B_new =(['0','A','B','C','D','E','A','B','C','D','E'],
 ['X','2','3','3','3','4','6','5','4','3','4'],
 ['Y','3','4','6','7','3','2','4','7','9','8'],
 ['Z','3','4','6','3','4','6','9','1','4','7']) 
 B_feat = np.zeros((3,2*yearsf)) 
 i=0
 for row in B_feat:
 j=0 
 k=0
 for element in row:
 B_feat[i][j:int(j+yearsf)]=B_new[i+1][k+1:int(k+yearsf+1)]
 j+=yearsf
 k+=years
 i+=1 
 print B_feat

and I receive

[[ 2. 3. 3. 6. 5. 4.]
 [ 3. 4. 6. 2. 4. 7.]
 [ 3. 4. 6. 6. 9. 1.]]

Now with the actual data I have:

 years = 9
 yearsf = 4
 np.shape(B_new) = (244, 181)
 np.shape(B_feat) = (243, 76)

I want to have a new array B_feat that ignores the first row and column of B_new, then skips 9 columns and starts "extracting" always the first 4 out of the next 9 row elements.

 import numpy as np
 i=0
 for row in B_feat:
 j=0 
 k=0
 for element in row:
 B_feat[i][j:int(j+yearsf)]=B_new[i+1][int(k+1+years):int(k+years+yearsf+1)]
 j+=yearsf
 k+=years
 i+=1 

When running the code, I receive the following error:

 IndexError: index 80 is out of bounds for axis 0 with size 76

I don't really understand this error since I thought axis 0 is down the rows (where I have 243 for B_feat) and I could't figure out where the index goes up to 80.

As I'm new to python and this forum, please let me know if I can improve my question or anything is unclearly stated.

Answer 1

Short answer, you are getting out of range as the error suggest, and this is also true for the minimal example that seems to work. I do not know why it works for the minimal example in the first place.

Detailled answer: In the minimal example, B_newhas 6 columns indexed from 0 to 5. Your internal loop iterate over the number of columns of B_new. And for each iteration, it increments j by yearsf which is 3. At the third iteration, j is 6 wich is larger than the max index 5. You have the exact same issue with the index k.

Suggestion: The number of iteration in your internal loop must be the number of group of columns to process, 2 in your minimal example. By group of column, I simply mean a set of columns below 'A-C'. Your loops can be turned into something like this:

i=0
for row in B_feat:
 for j,k in zip( range(0,B_feat.shape[1], yearsf),
 range(1,B_new.shape[1], years)):
 print 'i = ', i, ', j = ', j, ', k = ', k
 B_feat[i][j:int(j+yearsf)]=B_new[i+1][k:int(k+yearsf)]
 i+=1

Notice that I remove the +1 in the slice of B_new and set k to start at 1 And I will advise you to turn the external loop into something like for i in range(B_feat.shape[0]) and remove the statement i+=1

• Dave: Great, I adjusted the code in the following way and it worked out! Thanks a lot for your help! Unfortunately I can't post the code here in a nice format, but I also added the external loop;)
• innoSPG: @Dave, there is absolutely no need to post your final code here. You solved your problem and accepted an answer to help those whose have similar problem in the future.