How to modify an anonymous function with another function

I have the following function

function [y1sig,y2sig] = lab1sim(A,B,L,sig)
    Cs = 333.333;
    fs = 1e6;
    t = -5:(1/fs):5;
    d1 = pdist([0,2*A;B,L],'euclidean')/Cs;
    d2 = pdist([0,A;B,L],'euclidean')/Cs;
    y1sig = sig(t - d1);
    y2sig = sig(t - d2);
end

which is supposed to take a function of t (sig) and convert it to be t minus some delay. So for example (t)^2 should become (t-delay)^2. My function as it stands just returns sig exactly as it was.

sig = @(t) 1000*cos(880*pi*t).*heaviside(t);

[Jan]

What does not work? Which problem do you try to solve?

[Max Rothstein]

y1sig and y2sig are returned unmodified, they should have t replaced with (t-d1) or (t-d2)

Answer 1

[Steven Lord]

As written, y1sig and y2sig aren't function handles. You could make them function handles like this:

f1 = @(d) sind(d);
f2 = @(x) f1(x+90); % f2 is a shifted version of f1

theta = 0:360;
axis([0 360 -1 1])
hold on
plot(theta, f1(theta), 'r-', theta, f2(theta), 'k--')

To anticipate your next question no, there is no way to have f2 displayed as @(x) sind(x+90) or @(d) sind(d+90). At least not without converting f1 into text with func2str, manipulating the text, and converting to a function handle with str2func. If you were to go that route, be careful: @(d+90) sind+90(d+90) is not a valid function handle. Don't make the clbuttic mistake.

[Max Rothstein]

I changed my function to this:

function [y1sig, y2sig] = lab1sim(A,B,L,sig)
    Cs = 333.333;
    fs = 1e6;
    t = -5:(1/fs):5;
    d1 = pdist([0,2*A;B,L],'euclidean')/Cs;
    d2 = pdist([0,A;B,L],'euclidean')/Cs;
    y1sig = @(n) sig(n-d1);
    y2sig = @(n) sig(n-d2);
end

I cant plot y2sig or y2sig because they appear to have no link back to the original sig function handle

[Steven Lord]

You didn't show us A, B, L, or t. But when I tried running with arbitrary values of those four variables it seemed to work.

sig = @(t) 1000*cos(880*pi*t).*heaviside(t);
A = 2; B = 4; L = 8;

Cs = 333.333;
fs = 1e6;
t = -5:(1/fs):5;
d1 = pdist([0,2*A;B,L],'euclidean')/Cs;
d2 = pdist([0,A;B,L],'euclidean')/Cs;
y1sig = @(n) sig(n-d1);
y2sig = @(n) sig(n-d2);

Q = y1sig(1:10);

Can you show us the specific values of A, B, L, and t that you're using? If you can't show the values, at least showing us the class and size of each could be useful.

whos A B L t

[Walter Roberson]

At least not without converting f1 into text with func2str, manipulating the text, and converting to a function handle with str2func.

Or by using the Symbolic Toolbox.

The route of converting to string has some risk. When an anonymous function mentions a variable that is not a parameter, the function handle generated has a copy of the variable as of the time the handle was defined, and it is that copy of the value that is used every time the function executes, no matter what changes you might make to the variable that was mentioned. But when you convert the function handle to text, then there is no way for the text to say "use my local copy of the variable phase", so when you convert to text and back to function handle, it would be the current value of that variable that would get captured, not the value that was associated with the anonymous function.

This is not an issue if you went the symbolic toolbox route: when you created the symbolic evaluation of the function handle, the value of the saved variable would be copied in (and the name of the variable would drop out.)