[Maxima-discuss] varieties of invert, was: ratsimp not working as expected

[Robert D. <rob...@gm...>]

On 2014-03-07, andre maute <and...@gm...> wrote:

> --- ratsimp-bug3.max ---------------
> display2d : false;
> d : 4;
> A : genmatrix(lambda([k,l],v[[k,l]]-v[[0,l]]),d,d);
> Ainv : invert(A);
> --- ratsimp-bug3.max ---------------

Pretty sure you can avoid this bug by using a different implementation
of 'invert'. Try load(invert); and then the code above -- works for me.
That loads an implementation of the adjoint method -- the built-in
default is based on LU factorization if I remember correctly.

The problem is that there are various ways to compute the inverse and
they work better or worse in different cases. Recently (maybe a year
ago) yours truly changed the default implementation (after discussion
on the mailing list). We've bumped into various problems with that so
I've been meaning to rework it again. Sorry for the bother.

I don't know whether this particular problem is at the heart of the
ratsimp trouble you originally encountered.

best

Robert Dodier