Re: merging functions from octave-forge

[David Bateman]

John W. Eaton wrote:

> For Octave 3.0 (or perhaps a later 3.x release) I'd like to aim for
> including most of the functions that are currently in Octave forge for
> compatibility with core Matlab functionality (i.e., those in the list
> http://www.mathworks.com/access/helpdesk/help/techdoc/ref/refbookl.html).
>
> Some conditions for including a function in Octave are
>
>  * Copyright status must be clear and distribution terms must be
>    compatible with the GPL.  We must not have any code that is
>    (inadvertantly or not) derived from some proprietary source.
>
>  * Coding style should follow Octave's coding style.
>
>  * Docstrings must be formatted with Texinfo markup so they can be
>    included in the printed manual.
>
>  
I support this fully, and will help a bit with the time. But please can 
others on this list see this as a call for help and make the necessary 
changes to the functions John mentions that they own.

> I did a quick count, and found that the list of functions in Matlab is
> approximately 1066.  Octave includes 390 of those, and Octave forge
> contains 172 of them.  We have the following overlap between Octave
> and Octave forge
>
>  cellstr char chol deal detrend double fieldnames gammaln grid hankel
>  isa isfield isunix lin2mu mu2lin ndims polyder rand randn rmfield
>  struct toeplitz tril triu unix
>  
The hardest to deal with is randn/rand, and some discussion will be 
needed of how to deal with that. The chol function in octave-forge 
relies on a triangular matrix type. I introduced something similar for 
the sparse matrices and would like to do the same for full matrices, but 
in a more generic way (ie. detect PD matrices and triangular matrices in 
the solve code, rather than as a special return type from chol). So chol 
will also be a major effort to drop.

> so that leaves about 150 functions to merge:
>
>  addpath betaln blkdiag brighten builtin calendar clf complex
>  contourf conv2 convhull convhulln cplxpair csvread csvwrite cumtrapz
>  datenum datestr datevec del2 delaunay delaunay3 delaunayn demo
>  dlmread dlmwrite dos drawnow edit ellipj ellipke eomday expm1 factor
>  factorial fill fill3 filter2 fminbnd fplot full funm fzero getfield
>  ginput gradient griddata gtext hadamard ifftshift imread imwrite
>  inputname interp1 interp2 interpft intersect isdir isequal ismember
>  isprime issparse legend legendre magic mat2str meshc mkpp nchoosek
>  nnz nonzeros now nthroot ode23 ode45 optimset orient pareto pascal
>  patch pcg pchip pcolor peaks peaks perms pie plot3 polyarea ppval
>  primes print psi quadl quiver rat rats regexp rmpath rosser rref
>  scatter setdiff setfield setxor sortrows sound soundsc sparse
>  spdiags speye spfun spline spones sprand sprandn spy stem str2double
>  strcmpi strfind strmatch strncmp strncmpi strtok strvcat surf surfc
>  tar text textread trapz tsearch unique unmkpp view voronoi voronoin
>  waitbar wavread wavwrite weekday wilkinson xlsread xmlread xmlwrite
>  zoom
>
>  
The sparse functions like spdiags, speye, spfun, etc are already merged 
into 2.9.1..


> Comments?
>
>  
This should all go together with a major decruftication of octave-forge, 
as there are lots of special casing for different versions of octave, 
and lots of code that is obseleted by other changes in octave (bitand, 
trisolve, SymBand).

Cheers
David

--
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 1 69 35 77 01 (Fax) 
91193 Gif-Sur-Yvette FRANCE

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary