Re: pcg.m replacement for OCTAVE

[Jordi Gutiérrez Hermoso]

On 6 September 2011 23:04, Andrew Knyazev <address@hidden> wrote:
> Attach please find a proposed update of OCTAVE's function pcg.m It
> is backward compatible and tested. The changes are described in the
> comments and demos of the code. Could you please replace the current
> pcg.m with the new revision attached in OCTAVE?

Thank you for your suggested patches. The reference you added will
come in handy, and more demos and tests are also welcome. I will let
Piotr comment on the flexible CG idea itself, but at a first glance,
it looks like a useful addition. For my part, I only want to comment
on a few minor stylistic things:

    - We prefer to use ## comments to Matlab-style % comments.

    - Indentation and whitespace should be consistent.

    - Code duplication is undesirable. The last two tests have a lot
      of this. It would be better to have them together as a single
      test.

    - Octave is not an acronym, so there is no need to SHOUT IT. ;-)

These are all minor things that I or Piotr can edit in order to
extract your thoughtful contributions.

> P.S. I also want to add a few new functions to OCTAVE, e.g.,:
>
> http://www.mathworks.com/matlabcentral/fileexchange/32425-best-polynomial-approximation-in-uniform-norm
>
> http://www.mathworks.com/matlabcentral/fileexchange/27279-laplacian-in-1d-2d-or-3d
>
> http://www.mathworks.com/matlabcentral/fileexchange/26962-majorization-check
>
> How do I do it, if you could give me an advice, please?

At first glance, these functions do not cover core Matlab
compatibility (please correct me if I'm wrong), so they generally
would not go into Octave core itself, but into our sister project
Octave-Forge (http://octave.sf.net). However, they seem generally
useful. I think the polynomial function may go in core with the other
polynomial functions. The majorisation function perhaps could go in
with general functions, although I wish I could think of something
more specific. The Laplacian function might go with the special matrix
functions (hilb, toeplitz and such).

These functions would also have to be revised for style for inclusion
in Octave or Octave-Forge, but this is a minor point.

I think the best place for these functions would be Agora, if I ever
got around to finishing that. :-/ Do you have any grad students who
can do web development that you could send my way? ;-)

Thanks for your work and contributions,
- Jordi G. H.