chunkify

Splits a vector or cell array into evenly-sized chunks.

Description

C = CHUNKIFY(DATA, CHUNKSIZE)

Turns a matrix DATA into chunks of data with CHUNKSIZE elements in each cell entry of C. If DATA's size is not evenly divisible by CHUNKSIZE, the last entry in C will contain fewer items.

Examples:

chunkify([1 2 3 4], 2) --> {[1 2] [3 4]}
chunkify([1 2 3 4 5], 2) --> {[1 2] [3 4] [5]}

chunkify({'1' '2' '3' '4'}, 2) --> {{'1' '2'} {'3' '4'}}

Motivating scenario:

I wrote this function because I needed to process a lot of data in separate chunks, sometimes to have different machines run each chunk, sometimes to just load in chunks of the data files so I could balance I/O and memory constraints. In both these cases, I found it handy to have this simple function to break up my index vector. Because the number of data items was not always an exact multiple of my desired chunk size, I used this function instead of Matlab's built-in RESHAPE (putting one chunk per row, for example).

TODO:
If DATA is multi-dimensional, the dimension on which to chunkify is given by DIM.
   chunkify([1 2 3 4], 2, 2) --> {[1 2] [3 4]}
   chunkify([1 2 3 4], 2, 1) --> {[1 2 3 4]}
   chunkify([1 2; 3 4], 2, 1) --> {[1 2] [3 4]}

• MATLAB release: MATLAB 7.0.4 (R14SP2)

[S. Bart]

Another piece of junk, and guess what this time: this junky chunky code has a Copyright!

[Duane Hanselman]

Why would someone want to do this so often that one needs copyrighted code to do it? The simplicity and limited functionality of this code gives it dubious utility. If the copyright notice and permission notice take up more bytes that the code, one has to question either the copyright or the code or both. I'm not seeing something important and useful enough here to justify sharing it with others.

[Gerald Dalley]

I decided to remove the boilerplate copyright and add a motivating scenario. Use this function if you think it's useful. If not, don't. The new version should be up by 22 Feb or earlier.

[John D'Errico]

Good help, explaining clearly when to use this tool. No boilerplate anymore. I'll be honest, before you added the section explaining its use, I had absolutely no idea where one might find a use for this tool. That helped a lot. And, since its a task that I do fairly often, even though its not hard to do, I like it. Thanks for fixing it up.

[Micah Richert]

You can do this in one line using accumarray, and it easily solves the multi-dimension issue:
accumarray([floor((0:numel(DATA)-1)/CHUNKSIZE)'+1 ones(5,1)],DATA,[],@(x) {x})