[SciPy-Dev] ARPACK fixes before 0.9

[Aric Hagberg]

On Mon, Nov 29, 2010 at 4:38 AM, Pauli Virtanen <pav at iki.fi> wrote:
> Sun, 28 Nov 2010 20:45:25 -0700, Aric Hagberg wrote:

>> (Similarly should eigs_symmetric be eigsh  to match numpy.linalg.eigh?.
>> It would need to be extended to handle the complex case).
>
> There doesn't seem to be any special support for Hermitian matrices in
> Arpack, so I'm not sure which is better, (i) to raise an error on complex-
> valued input, or (ii) to fall back to eigs().

The simplest is to raise an error though it would be nice, and fitting
of the name eigsh, to fall back to eigs().

If you chose the fall-back plan then the option which='BE' isn't
implemented so you'll have to raise an error on that.   The others can
be mapped - LA'->'LR', 'SA'->'SR'.  Also it might be good to cast the
returned eigenvalues as real (the eigenvectors will in general be
complex).

Aric