Extending 'gradient' to handle function handles

[John W. Eaton]

On  1-Feb-2008, Søren Hauberg wrote:

| Hi,
|   The 'gradient' function currently only allows you to estimate the
| gradient of discrete data (i.e. data in a matrix). I think it would make
| sense if the 'gradient' function was also defined for function handles,
| such that you could do something like this:
| 
|   f = @sin;
|   df_dx = gradient(f, 0); # calculates the gradient at x = 0
| 
| Is this something there's interest in? The attached patch implements
| this using a simple central difference scheme. For multi-dimensional
| functions the API is like this:
| 
|   f = @(x,y) sin(x).*cos(x);
|   [dx, dy] = gradient(f, rand(7,2)); # calculate the gradient in 7
| random points
| 
| Thoughts?

This seems like a reasonable extension.  Would you like to turn your
previous patch into an hg changeset with ChangeLog entry?  It could
also use a few style changes for consistency with the rest of Octave.

jwe