Derivative of a multivariate function handle

I have a function handle here with symbolic array variable 'y'. I would like to take derivative of the function for which I am using the eval function. But after this the symbolic array variable 'y' disappears and we have y1 and y2. So I cannot carry out the substitution as in the original function. Is there a way of preserving the symbolic array variable after differentiation? I need this as I will be using it later in fsolve. I actually have many such functions for which I would like to determine the variable array using fsolve later. The code below

syms y [1 2]
g = @(y) [y(1)*cos(y(2))+y(2)*sin(y(1))-0.5]

g = function_handle with value:
    @(y)[y(1)*cos(y(2))+y(2)*sin(y(1))-0.5]

vpa(g([0.5 0.7]))
g1 = eval(['@(y)' char(-diff(g(y),y(1),1))])

g1 = function_handle with value:
    @(y)-cos(y2)-y2*cos(y1)

g1([0.5 0.7])

Answer 1

[Matt J]

Is this what you want?

syms y [1 2]

g = [y(1)*cos(y(2))+y(2)*sin(y(1))-0.5];

Jfun=matlabFunction(jacobian(g,y))

Jfun = function_handle with value:
    @(y1,y2)[cos(y2)+y2.*cos(y1),sin(y1)-y1.*sin(y2)]

Jfun(0.5,0.7)

ans = 1×2

1.3791 0.1573

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Swami]

Yes, exactly! Thank you!

[Matt J]

If so, please Accept-click the answer.

Answer 2

[nick]

Hi Swami,

I understand that you want to create a function handle for the differentiated function in which values can be substituted. You can use 'matlabFunction' function to convert the symboic function into function handle, as shown:

  syms y [1 2]
  g = @(y) [y(1)*cos(y(2))+y(2)*sin(y(1))-0.5];
  vpa(g([0.5 0.7]))
  g_diff = diff(g(y),y(1));
  g1 = matlabFunction(g_diff, 'Vars', {y});
  g1([0.5 0.7])

ans = 1.3791

You may refer to the following documentation to know more about 'matlabFunction' :

https://mathworks.com/help/symbolic/sym.matlabfunction.html

[Swami]

Yes, this is what I wanted! Thank you :)