8

Suppose I have a 2D numPy array such as:

a = [ [1, 2, 3], [4, 5, 6], [7, 8, 9] ]

How to I find the index of the row for which I know multiple values? For example, if it is known that the 0th column is 2 and the 1st column is 5, I would like to know the row index where this condition is met (row 1 in this case).

In my application, the first two columns are (x,y) coordinates, and the third column is information about that coordinate. I am trying to find particular coordinates in a list so I can change the value in the third column.

EDIT: To clarify, here is a non-square example:

a = [ [1, 2, 3, 4, 5, 6], [7, 8, 9, 10, 11, 12], [13, 14, 15, 16, 17, 18] ]

Suppose I know the row I am looking for has 13 in the 0th column, and 14 in the 1st column. I would like to return the index of that row. In this case, I would like to return the index 2 (2nd row).

Or better yet, I would like to edit the 4th column of the row that has 13 in the 0th column and 14 in the 1st column. Here is a solution I found to the case I have described (changing the value to 999):

a[(a[:,0]==13) & (a[:,1]==14), 3] = 999

gives:

a = [ [1, 2, 3, 4, 5, 6], [7, 8, 9, 10, 11, 12], [13, 14, 15, 999, 17, 18] ]

I'm sorry if this was unclear. Could someone point out in my original post (above the edit) how this could be interpreted differently, because I am having trouble seeing it.

Thanks.

EDIT 2: Fixed mistake in first edit (shown in bold)

I can now see how I made this whole thing confusing for everyone. The solution to my problem is well described in condition b) of eat's solution. Thank you.

Improve this question
5
  • I'm buffled. After your edit, what's your result for: sum(a[0]== 13)? Thanks Commented Feb 6, 2011 at 23:31
  • Unfortunately I screwed up when editing my post the first time, which I have now amended. Thank you for you patience. Commented Feb 7, 2011 at 2:11
  • glad to here that you have been able to solve your problem. If you feel that my answer was valuable to you, you may consider to upvote it ;-). Thanks Commented Feb 7, 2011 at 11:10
  • @eat: I can't upvote until I have 15 reputation, but I did mark it as my accepted answer. Thanks. Commented Feb 7, 2011 at 12:24
  • @lookitsmark: but you seem to be quite close to 15 allready ;-). Thanks Commented Feb 7, 2011 at 12:41

3 Answers 3

Reset to default
9 
In [80]: a = np.array([ [1, 2, 3], [4, 5, 6], [7, 8, 9] ])
In [81]: a
Out[81]: 
array([[1, 2, 3],
       [4, 5, 6],
       [7, 8, 9]])

a==2 returns a boolean numpy array, showing where the condition is True:

In [82]: a==2
Out[82]: 
array([[False,  True, False],
       [False, False, False],
       [False, False, False]], dtype=bool)

You can find any columns where this is True by using np.any(...,axis=0):

In [83]: np.any(a==2,axis=0)
Out[83]: array([False,  True, False], dtype=bool)

In [84]: np.any(a==5,axis=0)
Out[84]: array([False,  True, False], dtype=bool)

You can find where both conditions are simultaneously true by using &:

In [85]: np.any(a==2,axis=0) & np.any(a==5,axis=0)
Out[85]: array([False,  True, False], dtype=bool)

Finally, you can find the index of the columns where the conditions are simultaneously True using np.where:

In [86]: np.where(np.any(a==2,axis=0) & np.any(a==5,axis=0))
Out[86]: (array([1]),)
Improve this answer
Sign up to request clarification or add additional context in comments.

Comments

4

Here are ways to handle conditions on columns or rows, inspired by the Zen of Python.

In []: import this
The Zen of Python, by Tim Peters

Beautiful is better than ugly.
Explicit is better than implicit.
...

So following the second advice:
a) conditions on column(s), applied to row(s):

In []: a= arange(12).reshape(3, 4)
In []: a
Out[]:
array([[ 0,  1,  2,  3],
       [ 4,  5,  6,  7],
       [ 8,  9, 10, 11]])
In []: a[2, logical_and(1== a[0, :], 5== a[1, :])]+= 12
In []: a
Out[]:
array([[ 0,  1,  2,  3],
       [ 4,  5,  6,  7],
       [ 8, 21, 10, 11]])

b) conditions on row(s), applied to column(s):

In []: a= a.T
In []: a
Out[]:
array([[ 0,  4,  8],
       [ 1,  5, 21],
       [ 2,  6, 10],
       [ 3,  7, 11]])
In []: a[logical_and(1== a[:, 0], 5== a[:, 1]), 2]+= 12
In []: a
Out[]:
array([[ 0,  4,  8],
       [ 1,  5, 33],
       [ 2,  6, 10],
       [ 3,  7, 11]])

So I hope this really makes sense to allways be explicit when accessing columns and rows. Code is typically read by people with various backgrounds.

Improve this answer

Comments

2

Doing 

np.where(np.any(a==2,axis=0) & np.any(a==5,axis=0))

as unutbu suggested will not use the information that 2 is in the 0th column, and 5 is in the 1st. So, for a = np.array([[5, 2, 3], [2, 5, 6], [7, 8, 9]]), it will mistakenly return (array([0, 1]),)

Instead, you can use 

np.where((a[0]==2) & (a[1]==5))

to get the correct result (array([1]),).

Furthermore, if you want to edit the 2nd column of that particular row, you can skip the np.where and just reference it with: a[2][(a[0]==2) & (a[1]==5)]. This will work also for assignments, for example a[2][(a[0]==2) & (a[1]==5)] = 11.

Improve this answer

3 Comments

Your assignment example didn't quite give the results I was hoping for. For a = np.array([[5, 2, 3], [2, 5, 6], [7, 8, 9]]) using a[2][(a[0]==2) & (a[1]==5)] = 11 returns array([ [5, 2, 3], [2, 5, 6], [7, 11, 9] ]) when I wanted it to return array([ [5, 2, 3], [2, 5, 11], [7, 8, 9] ]). However, I have figured out the solution that produces this result: a[(a[0]==2) & (a[1]==5), 2] = 11
@lookitsmarc: based on your comment to J0ker5, I'll like to ask you to calriify your question. Would you care to give an example of a nonsquare a, which fullfills your requirements? IMHO your current question somehow mixes rows and cols with each other. Thanks
I have edited my original post with some more clarification. I hope that is sufficient.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.