How to get a Slice/View rom an arbirary vec/iterable of integers(ie indixes)?

[AnatolyBuga]

Hello team, and thank you for this great library!

I wonder if anyone came across/has done something like this before:

I need to take a slice from an array based om an arbitrary, dynamic vec of integers(indexes). After thoroughly digging docs I did not find any way to do that.

Firstly, Slice and slice info can be created from a single int or a range, but not from iterable or vec.
Secondly, multi_slice method doesn't allow dynamic input because it take a tuple as argument.

Here is

let ar = arr2(&[[1.,2.,3., 4.,5.,6.],[0., 0., 0., 0., 0., 0.,], [7., 8., 9., 10., 11., 12.]]);

Say, I want to keep only indexes 0, 3 and 4 and rows 0 and 2. So essentially I want something like this:

dbg!(ar.slice(s![&[0,2], &[0, 3, 4] ));

and get:

[[1.0, 4.0, 5.0],
 [7.0, 10.0, 11.0]]

Many thanks for any suggestions

[bluss]

Views and thus slicing can only represent consistent strides. I.e you make a lattice that you overlay on top of the original array, skipping as many elements as the stride for that axis says. Thus we can't do this, since we don't have a regular grid here, we are skipping different numbers of steps in different places.

However, .select(), the method, can copy out particular indexes along an axis.

[bluss]

The illustrations here can be useful. ndarray can't do all of that of course, but the stride concept is the same https://towardsdatascience.com/advanced-numpy-master-stride-tricks-with-25-illustrated-exercises-923a9393ab20

[AnatolyBuga]

Understood, thanks for prompt responce. Will look into utilizing .select() and this looks like a great set of exercises to get better understanding of how strides work. Thank you.