image thumbnail

Memory-efficient anonymous functions

version 1.1.0.0 (2.44 KB) by Matt J
Obtain anonymous function clean of non-essential workspace variables

365 Downloads

Updated 06 Nov 2015

View License

It may or may not be widely known, but as of the time of this submission (R2015a), anonymous function handles can store large, unused chunks of memory from the workspace where they are created. Often, anonymous functions are used transiently and the effect of this is never felt. If, however, you return an anonymous function handle as the output of a function, or save it to a .mat file, or broadcast it to workers using the Parallel Toolbox's PARFOR command, you may see lots of overhead from this hidden data. For examples and further discussion, see these threads
http://www.mathworks.com/matlabcentral/answers/115569-what-extra-data-is-stored-by-an-anonymous-function

http://www.mathworks.com/matlabcentral/answers/115254-parfor-behavior-sensitive-to-comments

This submission contains the function AFSLIM, which will let you create an anonymous function that stores only the extra parameter variables that you specify. (Naturally, of course, you must specify all the variables that it needs.)

EXAMPLE: The following code (important - must be run as a function, not from the command line!!) generates two files containing an anonymous function with the same functional behavior. However, tst1.mat consumes 259 MB whereas tst2.mat consumes only 1 KB.

function test

b=2;

fun1=@(x)x+b; %anonymous function with parameter 'b'

fun2=afslim(fun1,b); %slim version

b=rand(6000);

save tst1 fun1
save tst2 fun2

end

If you are given an existing anonymous function which already contains lots of unwanted extra data, this submission also contains AFCLEAN, which tries to strip the extra data away. However, AFCLEAN relies on a certain amount of undocumented MATLAB and breaks anonymous functions that refer to nested or subfunctions. When it works, it seems to work well, but use at your own risk!

Cite As

Matt J (2021). Memory-efficient anonymous functions (https://www.mathworks.com/matlabcentral/fileexchange/45749-memory-efficient-anonymous-functions), MATLAB Central File Exchange. Retrieved .

MATLAB Release Compatibility
Created with R2013b
Compatible with any release
Platform Compatibility
Windows macOS Linux
Acknowledgements

Inspired: Sparse Grid Interpolation

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!