Rolling concatenation array of numpy arrays

Question

I want to implement a rolling concatenation function for numpy array of arrays. For example, if my numpy array is the following:

 [[1.0]
  [1.5]
  [1.6]
  [1.8]
  ...
  ...
  [1.2]
  [1.3]
  [1.5]]

then, for a window size of 3, my function should return:

 [[1.0]
  [1.0 1.5]
  [1.0 1.5 1.6]
  [1.5 1.6 1.8]
  ...
  ...
  [1.2 1.3 1.5]]

The input array could have elements of different shapes as well. For example, if input is:

[[1.0]
 [1.5]
 [1.6 1.7]
 [1.8]
 ...
 ...
 [1.2]
 [1.3]
 [1.5]]

then output should be:

 [[1.0]
  [1.0 1.5]
  [1.0 1.5 1.6 1.7]
  [1.5 1.6 1.7 1.8]
  ...
  ...
  [1.2 1.3 1.5]]

If you are willing to pad with NaNs/some-other-invalid-specifier to keep a 2D shaped array, take a look here - stackoverflow.com/questions/40683601/… — Divakar
– Divakar, Commented Apr 11, 2017 at 10:33
This isn't really a numpy question, it looks more like a list concatenation. You're starting and finishing with a ragged list, which means there's nothing in numpy that will make your life easier. — Daniel F
– Daniel F, Commented Apr 11, 2017 at 10:52
@DanielForsman all those individual elements are actually numpy arrays and not lists. the overall structure is an array of arrays — Vaibhav Agarwal
– Vaibhav Agarwal, Commented Apr 11, 2017 at 11:11

Daniel F · Accepted Answer · 2017-04-11 12:57:22Z

1

First, make your array into a list. There's no purpose in having an array of arrays in numpy.

l = arr.tolist()           #l is a list of arrays

Now use list comprehension to get your elements, and concatenate them with np.r_

l2 = [np.r_[tuple(l[max(i - n, 0):i])] for i in range(1, len(l)+1)]

answered Apr 11, 2017 at 11:44

Daniel F

14.5k2 gold badges34 silver badges59 bronze badges

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

1 Answer 1