Using end as a variable to access parts of array

I need to access parts of the array, sometimes from 1:n, and sometimes from 1:end. However, as the program progresses, the length to the end will change. This math is going to be called over and over many many times, and I would like to avoid overhead of finding the length of the array every-time to input that as 'n'. So ideally I would like to be able to use 'end' as a variable. The following does not work

a = 1:10;
b = 'end';
a(5:end) results in Error: The end operator must be used within an array index expression.

Nor can I do something simple like b = end;. Anyone have any ideas how I can do this?

[Cam Salzberger]

Running this code works for me. Is there anything else in your workspace that could be causing this to fail? Or a function called "a" or "end"?

Using "end" will inherently require checking the length of the array under the hood. However, you are correct that it is likely to be faster to do:

a(1:end)

than:

a(1:numel(a))

However, if you are doing 1:end, could you not just do:

a(:)

or just

a

Of course, that wouldn't work with replacing a(5:end).

I'm a little confused about why this is a big concern though. Can you give us more detail on your program logic flow, what your variable sizes are, and how they will change?

-Cam

[Cam Salzberger]

One thing I can suggest is to profile your code as you run it. Try to see if it's really the calls to numel or length that are slowing down your code. Micro-optimization doesn't help if there are larger problems.

Answer 1

[Star Strider]

Using end should work. (I don’t know the MATLAB version when it first indicated the last element in a vector.)

These are equivalent (for a row vector):

Result1 = a(5:end);
Result2 = a(5:numel(a));
Result3 = a(5:length(a));
Result4 = a(5:size(a,2));

For column vectors, or matrices, use the size (link) function. See the documentation for details.

[Shane Sullivan]

The point is, I can not use 'end'. I want my program depending on certain conditions to decide to go to end; but I want to avoid calculating what that end length/size is because I am doing this a crazy large # of times and it is slowing things down. So I am looking for someway possible to use the "end" as a variable essentially.

[Star Strider]

What you want to do and the reason you do not want to use end are not exactly clear. I doubt that using end is slowing your code.

If you want to avoid addressing a vector beyond its length (and throwing an error), something like this will work:

First = 3;
Last = 7;
Out = a(First : min(Last, length(a)));

Answer 2

[KL]

Why do you want to do this? why not just use the length(a),

a(5:length(a))

[Shane Sullivan]

Hi KL, I currently am. However, this is actually part of a many dimension matrix, and I will be iterating through this 100's of thousands of times and trying to improve speed by reducing the number of length or size calls.

[Guillaume]

Note that end is not equivalent to length(a). end is equivalent to numel(a) which will produce vastly different results on anything but vectors.

So:

a(5:numel(a))

[KL]

I agree. I had only considered row vectors but in case of multiple manydimension matrices, why not put them all in a cell array? store their sizes in another cell array beforehand so you could save some time later in your loops/whatever.

C = {rand(2,4);rand(123,56);rand(100,43)};
C_sizes = cellfun(@size,C,'UniformOutput',false);

for k=1:100000
   A = cellfun(@(a,b) a(1:b(1),2:b(2)),C,C_sizes,'UniformOutput',false);
   %or anything else
end

Answer 3

[思诚]

This would work if you really want to call customized end at anywhere (outside subsref) you want.

feval("end",array,k,n) would call end as the end is placed on the kth place with total n subscripts in a subsref.

array: the array to call end on

k: index position. In your case it would be 1

n: total index number. In your case it would be 1

For example:

>>> feval("end",[1,2,3;4,5,6],1,2) % (:,1:end) end=2

ans =

2

>>> feval("end",[1,2,3;4,5,6],1,1) % (1:end) end=6

ans =

6