Out of Memory for using many global variables

Hi

I have many functions. so instead of use the variables as function arguments to transfer values between functions. I set many variables as global variables. These global variables are not used all the time.

on the top of main function, I did use the following to clear the space.

clc clear all; close all; however, when opening a big excel file(200MB), I ran into out of memory issue for 2014b version. When I change it to use 2017 version. the issue was gone but loading speed for opening excel using xlread() is still very slow. Now I need your help on the following questions.

1. How to reduce memory space with many global variables? How to only assign space when it is used? 2. I have few toolboxes, like statistics, which are only linked to 2014b version. Is it possible to use these toolbox for 2017 version?

Thank you so much for your help. I appreciate it.

[Stephen23]

"I have many functions. so instead of use the variables as function arguments to transfer values between functions. I set many variables as global variables."

I don't see why having many functions requires using global variables. You would likely be much better off passing the data as proper input/output arguments, rather than wasting time using slow and buggy globals.

"on the top of main function, I did use the following to clear the space. clc clear all; close all; "

Why? When a function is first called its workspace is empty (apart from any input arguments), so calling clear all serves no purpose whatsoever and is in fact counter-productive: it will clear all loaded functions and classes, which will then just have to be loaded again when the code is run, thus slowing down your code. It is highly unlikely that clc or close all serve any purpose actually related to the functionality of your functions, so what is the point of calling them? Of course apart from the usual reason why people call those everywhere:

https://en.wikipedia.org/wiki/Cargo_cult_programming

On to your your questions:

"1. How to reduce memory space with many global variables? How to only assign space when it is used?"

Don't use globals.

Pass data properly between workspaces and trust the MATLAB engine to deal with memory and garbage collection.

"2. I have few toolboxes, like statistics, which are only linked to 2014b version. Is it possible to use these toolbox for 2017 version?"

Are these MATLAB toolboxes or third-party toolboxes?

[Matt J]

I have many functions. so instead of use the variables as function arguments to transfer values between functions.

Is this to conserve memory? You will only see a saving when passing variables to functions that actually modify those variables.

[Adam]

Toolboxes can only be used with the version of Matlab for which they are linked. If you upgrade base Matlab you have to upgrade toolboxes equivalently.

As for memory usage, goodness knows what is in memory and when if you have a vast number of global variables.

Functions define scope for variables and when the function ends the variable goes out of scope (putting aside handle-derived objects which you are likely not using) and the garbage collection will release the memory. If you are liberally creating lots of global variables though then they will sit around until you get rid of them.

[roudan]

Thanks guys for your help. I appreciate it.

The functions are created by using uicontrol menu. so I basically choose the values from different dropdown menus each of which will assign values to one or tow variables. Then I move on to data plotting based on the values selected from uicontrol drop down menu.

Now I have 2 questions. 1. If I don't use global variables, how can i save the value when finishing the uicontrol menu and pass on the other functions? 2. do I have to define global variable everwhere even the global variables will not be used in that function? Could I just only define global variables whenever they are used?

Thanks

[Rik]

1. Use the callback of that dropdown and get(hObject,'Value') to call the correct function to plot your data.
2. Don't use globals at all. But if you do use them, you indeed need to only call global for the variables you use in that function.

[Stephen23]

@Yongnuan Liu: you might like to try using nested functions, which are a very intuitive way to transfer data between callbacks and functions:

https://www.mathworks.com/help/matlab/matlab_prog/nested-functions.html

You can easily combine nested functions with local functions and pass output arguments too. Have a look at my FEX submission iregexp for an example of this:

https://www.mathworks.com/matlabcentral/fileexchange/48930-interactive-regular-expression-tool

Don't use globals.

[roudan]

Thanks Stephen. Nice tool. I just downloaded it. Thanks

I have many uicontrol menus , so I cannot put all my function nested under one uicontrol callback function.

The way I am doing is like this:

Main function

Uicontrol1 uicontrol2 Uicontrol3 ...

then all the other functions include uicontrol callback functions are separated. their data are sharing using global variables.

For example, I can usie uicontrol1 to set one variable then use the other uicontrol menu to set the other variables. both variables are global variables. So then I move to different function to plot the data where I need to call the global variables I set using previous Uicontrol menus.

Any comments on how to improve it? Thanks. I appreciate it.

[Rik]

If each function has access to the same figure, you can use guidata.

[roudan]

Thanks Rik, I like to do the other way, I'd like one function have access to all the variables in each uicontrol menus. how to do it? Thanks

[Greg]

doc guidata

[Walter Roberson]

If you have enough data to be running out of memory, you should be using setappdata()/getappdata() instead of guidata(). guidata() is invoked by GUIDE for every callback, and a copy of all the stored data is created each time. With setappdata()/getappdata(), you can pull in copies of just the data you need.

Even then, since the data is large, you should probably instead rewrite your code to use nested functions with shared variables -- or create a handle object containing the array, possibly using https://www.mathworks.com/help/matlab/matlab_oop/handle-compatible-classes-and-heterogeneous-arrays.html