Delete a column from an array of uncertain size?

Hi everyone.. I know there are answers out there regarding removing a column from an array with x dimensions--this much I have no problem doing.

I'm trying to turn some code into a function that I can use to solve the same type of problem with anywhere from 2 to 7 dimensions (or infinite, if I can write a general enough code).

What I have to do, in a certain part of the code, is take an array of grid points and drop the first point on the second dimension.

E.g. if it was a matrix I would write:

newarray = oldarray(:,2:end);

if it was a 4D array I would write:

newarray = oldarray(:,2:end,:,:);

What do you think the most efficient/general way to code this would be?

Thanks!

[Image Analyst]

You know in advance what "x" is , right?

[Amanda]

Yes, but it is determined by the person using the function.

I can, for example, write

if num_dimensions == 2
newarray = oldarray(:,2:end);
elseif num_dimensions == 3
newarray = oldarray(:,2:end,:);
end

but that's not very efficient, I don't think.

My current thought is that I simply create the grid arrays within the function, and drop the first entry of the problematic dimensions before doing so... but this creates some other issues and so I'm curious to know if there is another option.

Answer 1

[David Young]

The trick needed is to use the fact that cell arrays can be expanded into comma-separated lists, so can represent any number of subscripts. It works like this.

Test data:

x = 1 + randi(9);   % random no. dimensions from 2 to 10
oldarray = rand(repmat(3, 1, x)); % 3 x 3 x 3 ... array

Computation:

% get cell array of subscript arguments representing whole of oldarray
subs = arrayfun(@(s) {1:s}, size(oldarray));
% change second subscript to start from 2
subs{2} = 2:size(oldarray,2);
% create new array by indexing old array
newarray = oldarray(subs{:});

Check newarray is correct size, and look at one column (noting that trailing ones are always OK regardless of the number of dimensions):

disp(size(oldarray));
disp(size(newarray));
disp(oldarray(1, :, 1, 1, 1, 1, 1, 1, 1, 1, 1));
disp(newarray(1, :, 1, 1, 1, 1, 1, 1, 1, 1, 1));

[Amanda]

Thanks! This answer and the answer by Walter below both work and both make me so happy. I chose this one at the 'accepted' answer since a test ran in 0.002623 seconds versus 0.014156 seconds.

Answer 2

[Walter Roberson]

s = size(oldarray);
[r, c, p] = size(oldarray);  %deliberate that 3 outputs are given for array that might be more dimensions
newarray = reshape(oldarray,r,c,p);
newarray = reshape(newarray(:,2:end,:), [r, c-1, s(3:end)]);

That is, the 3 dimensional case covers the rest.