CODEWITHSUNDEEP
matlab
Oriol
Oriol

Reputation: 288010

Can I use a modified variable without creating a new one in Matlab?

I want to use a modified version of a variable (without changing the original one) only once, so I don't want to save it in a new variable.

For example:

a = [1 -2 3];
copy = a;
copy(a < 0) = 0;
b = [4 5 6] .* copy;

Can I simplify that to something like the following?

a = [1 -2 3];
b = [4 5 6] .* a(<0 = 0);

Edit:

The example above is just an example. The general problem is how to get the copy produced by the code below, without creating it.

% `a` is a vector
% `cond` is a logical vector such as `size(a) == size(cond)`
% `num` is a number
copy = a;
copy(cond) = num;

Upvotes: 1

Views: 53

Answers (4)

Oriol
Oriol

Reputation: 288010

For the general problem

% `a` is a vector
% `cond` is a logical vector such as `size(a) == size(cond)`
% `num` is a number
copy = a;
copy(cond) = num;

Instead of creating copy, one of the following can be used:

  • a + (-a+num) .* cond

  • a .* ~cond + num * cond

Warning: doesn't work if num is one of the following: nan, inf, -inf (and maybe more)

Performance

It seems that a + (-a+num) .* cond is faster:

a = rand(1, 10000);  cond = a < 0.5;  num = 10;

tic; for i=1:100000 a + (-a+num) .* cond; end; toc;
% Elapsed time is 14.764796 seconds.

tic; for i=1:100000 a .* ~cond + num * cond; end; toc;
% Elapsed time is 29.842845 seconds.

Upvotes: 1

Luis Mendo
Luis Mendo

Reputation: 112659

You can do it with an anonymous function which applies the desired modification to your variable:

f = @(x) max(x,0); % this will apply the desired operation to a
a = [1 -2 3];
b = [4 5 6] .* f(a);

The modification required on a needs to be simple enough, so that it can be defined using an anonymous function. Specifically, according to the above link,

Anonymous functions can contain only a single executable statement.

Upvotes: 0

Bas Swinckels
Bas Swinckels

Reputation: 18488

In this particular case, you can do

b = [4 5 6] .* max(0, a)

But why do you ask this? Do you want to save memory when you do something similar with big matrices? Or do you want to write more compact code?

Upvotes: 2

Prashant Kumar
Prashant Kumar

Reputation: 22519

I don't think the syntax you are looking for exists in Matlab.

Obviously, your real use case might be more complicated than this, but I would change your code to this to remove the extraneous copy variable:

a = [1 -2 3];
b = [4 5 6] .* a;
b(a < 0) = 0;

See: http://www.mathworks.com/matlabcentral/answers/25235

Upvotes: 0

Related Questions