Good programming practice...

Asked by Raviteja on 22 Sep 2011

Latest activity Edited by Jitin Beri on 12 Nov 2018

54 views (last 30 days)

Hello all,

Please explain good MATLAB programming practice methods. It will help to the guys who are new to programming like me.

Previously I used

for i=1:10

after following some suggestions from this answers pages I learnt to use

for i1=1:100

This is the good way to write programs.

Like this, as a professional programmer, please mention some good programming practice techniques.

It will useful to all!

2 Comments

Michael Leung on 22 Sep 2011

You are right! i is undesirable for the loop variable as matlab treats i as the complex number, sqrt(-1).

I personalty use ii and jj for my loops, but as long as its not the lonesome "i" it will speed up your code.

Steven on 19 Oct 2011

However, the difference seems to be indistinguishable considering a certain precision.

tic;

for i1 = 1:100000

x = sin(i1);

end

toc

Elapsed time is 0.003597 seconds.

tic;

for i = 1:100000

x = sin(i);

end

toc

Elapsed time is 0.003569 seconds.

16 Answers

Link

Answer by Daniel Shub on 23 Sep 2011

I suggest developing a coding "style." Some good places to start are given in the FAQ:

http://matlab.wikia.com/wiki/FAQ#Where_can_I_find_MATLAB_Style_Guides_or_Coding_Standards.3F

I also would consider this book reviewed by Loren

http://blogs.mathworks.com/loren/2011/02/10/book-review-the-elements-of-matlab-style/

although I should say I have not looked at it yet. It is on my list of things to do.

1 Comment

K E on 26 Jan 2012

Elements of Matlab Style is great

Link

Answer by Jake Bolanski on 22 Sep 2011

It's always good to comment in front of your code so that you won't have trouble remembering why you did what you did. It also helps others recognize your code and what you're doing.

1 Comment

Jan on 22 Sep 2011

I agree. If you want to share your code, or do this ever in the future, it is a good idea to use the standard comment style of Matlab:

function [Output, ...] = Fcn(Input, ...)

% H1 line

% Help text

% [Output, ...] = Fcn(Input, ...)

% ...

% Date, author

Link

Answer by Jan on 23 Sep 2011

Read the FAQ

It is very likely, that the questions, which concern other users frequently, do concern you also. It is very efficient to profit from the mistakes of others, instead to implement them by your own.

2 Comments

Daniel Shub on 23 Sep 2011

I wish I could vote many many times.

Bjorn Gustavsson on 23 Sep 2011

"It is very efficient to profit from the mistakes of others, instead to implement them by your own." was one of the funniest (sadly fun, funily said) statements I've come across this week!

Link

Answer by Fangjun Jiang on 22 Sep 2011

Where do you want to start? I'll add as I come up with some.

pre-allocate large size variables using zeros(), cell()

0 Comments

Link

Answer by Jan on 22 Sep 2011

Avoid overloading builtin functions. It happens very often, that a user creates a variable called "max" and is surprised that the MAX function does not work anymore:

max = 19;
... 100 lines later:
B   = max(1:10)
??? Index exceeds matrix dimensions.

A confusing but correct error message!

1 Comment

Jan on 26 Jan 2012

@Steven: As you've found out, longer symbols do not need more time (as long as the lookup table is not poluted by EVALs). While runtime is not a problem, debug time is. "B=max(1:10)" is confusing.

Link

Answer by Jan on 22 Sep 2011

Do not start a script or function by:

clear all

This removes all loaded functions from the memory. Reloading them will need a lot of time such that a program can be 100 times slower in a not so unlikely worst case.

I admit that a clear or the equivalent clear variables can help to detect crude typos in names of a variable, if this variable is existing before the script. But this is a really rare case and if a function is used, MLint detects this much better.

But there is a frightening large number of codes starting with clear all. I assume this is recommended in some tutorials which have been written without deeper insight into MATLAB.

1 Comment

Daniel Shub on 20 Oct 2011

Even the best sometimes do it:

http://blogs.mathworks.com/loren/2009/06/16/rooting-around-in-matlab-part-2/

Link

Answer by Daniel Shub on 23 Sep 2011

Edited by Jitin Beri on 12 Nov 2018

Come to Answers before using global or eval. There is almost always a better way to do it.

http://matlab.wikia.com/wiki/FAQ#Why_is_it_advised_to_avoid_using_the_.22eval.22_function.3F

http://www.mathworks.co.uk/support/tech-notes/1100/1103.html

1 Comment

Jan on 23 Sep 2011

I wish, that you do *not* come to Answers, but proceed to the above two links immediately... But I vote the helpful links +1.

Link

Answer by Daniel Shub on 23 Sep 2011

Use the functional form of load and save, and almost every other function except maybe help and doc

instead of

load filename.mat

use

data = load('filename.mat')

2 Comments

Jan on 23 Sep 2011

Same for SAVE.

And catch the output: Data = load(FileName), otherwise you could find unexpected variables in your workspace like "max", which will shadow existing functions to your surprise. See http://www.mathworks.com/matlabcentral/answers/16484-good-programming-practice#answer_22299

Daniel Shub on 23 Sep 2011

@jan you are correct, catching the output of load is an important piece of good practice.

Link

Answer by Jan on 22 Sep 2011

Use English for comments and the help text. Although you might assume, that you will never share the code, even a discussion in this forum is much harder, if all variables and descriptions are in Italian.

1 Comment

Walter Roberson on 23 Sep 2011

This one I cannot really agree with. It trades the possibility that one might want to share a section of the code in Answers or CSSM, against the near certainty that the code will need to be easily understandable by other people in one's workplace.

There are also locations were writing the comments and help text in English would violate various language preservation laws.

Link

Answer by Jan on 22 Sep 2011

Include a check of number and type of inputs if all functions.

This will cost several seconds runtime, but will save hours of debugging.

0 Comments

Link

Answer by Jan on 23 Sep 2011

Read the tips for program development in the dokumentation:

http://www.mathworks.de/help/techdoc/matlab_prog/f10-60481.html

And if you are on the way, read the rest of the documentation also, at least the "Getting Strarted" chapters.

0 Comments

Link

Answer by Jan on 23 Sep 2011

This is a list of helpful links: http://www.ee.columbia.edu/~marios/matlab/matlab_tricks.html

E.g. this is exhaustive: http://www.ee.columbia.edu/~marios/matlab/MatlabStyle1p5.pdf

0 Comments

Link

Answer by Daniel Shub on 20 Oct 2011

Surprised Doug didn't chime in with his blog post: http://blogs.mathworks.com/videos/2010/03/08/top-10-matlab-code-practices-that-make-me-cry/

1 Comment

Fangjun Jiang on 20 Oct 2011

+1. I remember reading Doug's post. That's a good one.

Link

Answer by Daniel Shub on 27 Jan 2012

It is only fair that if I plug Doug's blog, I should plug Loren's blog also:

http://blogs.mathworks.com/loren/2012/01/13/best-practices-for-programming-matlab/

0 Comments

Link

Answer by Raviteja on 23 Sep 2011

I got good help from matlab datasheets for matlab programming tips

http://www.mathworks.in/help/pdf_doc/matlab/programming_tips.pdf

10 Comments

Show 7 older comments

Walter Roberson on 23 Sep 2011

@Mohammad: http://www.flickr.com/photos/gothamoddisee/216748915/

(Be sure to read the caption)

mohammad on 23 Sep 2011

Nice, thanks Walter

mohammad on 24 Sep 2011

I use this: www.google.com/bookmarks

but it's to be nice if here has itself Bookmarks

Link

Answer by Steven on 19 Oct 2011

However, the difference seems to be indistinguishable considering a certain precision whatever the number of repetitions.

tic;
for i1 = 1:100000
    x = sin(i1);
end
toc
Elapsed time is 0.003597 seconds.
tic;
for i = 1:100000
    x = sin(i);
end
toc
Elapsed time is 0.003569 seconds.

1 Comment

Jan on 27 Jan 2012

@Steven: You have posted this detail 3 times in this thread. Please consider, that the runtime is not affected by the length of the name of the variable, because internally the variable is accessed by a memory pointer taken from a lookup-table. This lookup-table is created when the function is loaded the first time only.

You are now following this question