Re: Disaggregating input arguments for sub2ind

[Alexander Barth]

On Wed, Feb 19, 2014 at 3:39 PM, Juan Pablo Carbajal <address@hidden> wrote:

On Wed, Feb 19, 2014 at 2:38 PM, Nir Krakauer <address@hidden> wrote:
> Dear all,
>
> I have an n-D grid whose size is given by grid_size, and an array of
> subscripts with size npoints*ndims, for example:
>
> grid_size = [5 4];
> grid_subs = [1 4; 5 1; 3 2];
>
> What's a general way to convert the subscripts to single indices, achieving
> the same result as, for example,
> grid_inds = sub2ind (grid_size, grid_subs(:, 1), grid_subs(:, 2)); #2-D case
> grid_inds = sub2ind (grid_size, grid_subs(:, 1), grid_subs(:, 2),
> grid_subs(:, 3)); #3-D case
> if the number of dimensions in the grid may vary between calls?
>
> Nir
>

> _______________________________________________
> Help-octave mailing list
> address@hidden
> https://mailman.cae.wisc.edu/listinfo/help-octave
>

Nir,

Besides having  eval statement with the parametrized string, I
wouldn't know any other way, since it seems sub2ind doesn't accept
zipped arguments... something like the **kargs in python.
I hope Jordi corrects me on this.

Dear Nir,

You can, but you have to use cell arrays for the indices. For example:
>> A = magic(5)
A =

   17   24    1    8   15
   23    5    7   14   16
    4    6   13   20   22
   10   12   19   21    3
   11   18   25    2    9


>> ind = {2,3}
ind =
{
  [1,1] =  2
  [1,2] =  3
}

Elements of ind can also be vectors (e.g. ind = {[ 2 3],[3 3]} as in your question)

>> lindex = sub2ind(size(A),ind{:})
lindex =  12
>> A(lindex)
ans =  7
>> A(2,3)
ans =  7

If you want to access the corresponding element in A you can use directly this:

>> A(ind{:})
ans =  7

As indexing calls the method subsref, one could also use this subsref(A,idx) where idx = substruct ('()',ind), but I found the the previous method is much faster.

To use ind2sub, you can use this:

ind = cell(ndims(A),1);
[ind{:}] = ind2sub(size(A),lindex)

In my package divand, I had to use such constructs to make the code work in an arbitrary number of dimensions.
 

Regards,

Alexander

 

_______________________________________________
Help-octave mailing list
address@hidden
https://mailman.cae.wisc.edu/listinfo/help-octave