I want to create a list containing indices that would be used to get elements from another list.
A simple case:
A = [5,6,7,8,9,10]
b = 2:4
I want to then do something like
C = A[b]
Which would be like saying C = A[2:4]
I want to later extend this to multidimensional arrays, where e.g b = [2:4, 5:6] and I can simply call A[b] to extract a multidimensional array out of A.
You can define b as a slice object to achieve this:
In[9]:
A = [5,6,7,8,9,10]
b = slice(2,4)
A[b]
Out[9]: [7, 8]
Regarding your other requirement I think if you create a list object containing 2 slice objects then it should achieve what you want:
In[18]:
import numpy as np
a = np.arange(100).reshape(10,10)
b = [slice(1,3), slice(3,4)]
a[b]
Out[18]:
array([[13],
[23]])
You can either use straight python using e.g. slice:
>>> A = [5,6,7,8,9,10]
>>> b = slice(2,4)
>>> A[b]
[7, 8]
But this does not scale very well to nd-arrays. To do this, I'd recommend using numpy's np.s_ function which does exactly what you are looking for, without the need of explicitly constructing a slice for each axis.
>>> b = np.s_[2:4]
>>> A[b]
[7, 8]
This extends nicely to e.g. 2d arrays:
>>> A = np.ones(10, 10)
>>> b = np.s_[2:4, 5:6]
>>> A[b]
array([[ 1.],
[ 1.]])