[Numpy-discussion] Automatic matrices

[Keith Goodman]

On 12/21/06, Sven Schreiber <svetosch at gmx.net> wrote:
> Keith Goodman schrieb:
>
> > How about diag?
> >
>
> There was a thread about this (in which you participated, I believe);
> for matrices you should now use m.diagonal() I think. So diag doesn't
> qualify.

I think the different results returned by x.diagonal and M.diagonal(x)
is confusing:

>> x

matrix([[-0.87175207,  1.57394765],
        [-1.7135918 , -1.5183181 ]])
>> x.diagonal()
matrix([[-0.87175207, -1.5183181 ]])   <-----matrix
>> M.diagonal(x)
array([-0.87175207, -1.5183181 ])       <-----array

321 	def diagonal(a, offset=0, axis1=0, axis2=1):
322 	    """diagonal(a, offset=0, axis1=0, axis2=1) returns the given diagonals
323 	    defined by the last two dimensions of the array.
324 	    """
325 	    return asarray(a).diagonal(offset, axis1, axis2)

Maybe this asarray could be changed to asanyarray?