CODEWITHSUNDEEP
performancenumpyindexing
Louis Thibault
Louis Thibault

Reputation: 21390

Is there a better way of getting elements with x,y coordinates from a numpy array?

It is possible to index a numpy array with a tuple of sequences such that tpl[0] is a a sequence of x coordinates and tple[1] is a sequence of y coordinates. One simply needs to index the array with the tuple, thus: other_array[tpl].

I currently have coordinates stored in a 2D array such that the vector ar[0] corresponds to my x values and ar[1] corresponds to my y values.

Right now, I'm indexing other_array by creating a tuple: other_array((ar[0], ar[1])). Unfortunately, this operation is running in a tight loop, so any amount of performance I can squeeze out would be highly beneficial. Creating a tuple can add a bit of overhead if performed 10^8 times! Is there a faster, numpythonic way of indexing with such a matrix of xy coordinates?

Thank you very much!

Upvotes: 2

Views: 1222

Answers (1)

Warren Weckesser
Warren Weckesser

Reputation: 114781

You can index a numpy array with another array, so you don't have to create a tuple. For example:

In [199]: other_array
Out[199]: 
array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14]])

In [200]: ar
Out[200]: 
array([[0, 2, 1],
       [1, 3, 0]])

In [201]: other_array[ar[0], ar[1]]
Out[201]: array([ 1, 13,  5])

If that doesn't answer your question, could you include a simple working example in your question that shows what you are currently doing?

Upvotes: 1

Related Questions