Re: matconcat

[Karim Belabas <Karim.Belabas@math.u-bordeaux1.fr>]

* Andreas Enge [2012-02-13 20:12]:
> On Mon, Feb 13, 2012 at 08:01:55PM +0100, Karim Belabas wrote:
> > Please provide "easy" building blocks of the required dimensions. :-)
> > The "best" I can come up with is
> > 
> > {
> >   matconcat([A,          vector(2),   vector(3);
> >              vectorv(2), B,           matrix(2,3);
> >              vectorv(3), matrix(3,2), C]);
> > }
> 
> Sort of easy, but I agree it will become annoying with more and more diagonal
> blocks.
> 
> One could argue that this should already be handled by matdiagonal,
> instead of creating a 3x3 matrix with matrix entries.
> 
> Maybe two functions? matconcat for the cases where mathematicians draw
> straight lines on a blackboard, and en extended matdiagonal (with a new
> name, if need be) when they draw little boxes on a diagonal?

I ended up modifying a little the original implementation, when
dimensions don't match, choosing the "x = x*Identity" semantic
for scalars 'x'. The algorithm uses 2 passes: we first determine block
sizes, then extend submatrices by 0s until they match the block size.

This covers all my original use cases, is reasonably coherent, not too
hard to document; and straightforward to implement.

The rigid alternative [ raise exception when anything doesn't match ]
made matconcat() essentially useless: my examples could all be
implemented using existing GP functions, with comparable readability.
What makes matconcat() really useful in my scripts are the expanding
zeroes.

Please test and comment ! :-)

    K.B.
-- 
Karim Belabas, IMB (UMR 5251)  Tel: (+33) (0)5 40 00 26 17
Universite Bordeaux 1          Fax: (+33) (0)5 40 00 69 50
351, cours de la Liberation    http://www.math.u-bordeaux1.fr/~belabas/
F-33405 Talence (France)       http://pari.math.u-bordeaux1.fr/  [PARI/GP]
`