infinity correct maxima



Barton Willis wrote:
> After committing, I did change a few things that seemed to be bugs. With my
> version
> (not yet on CVS), I get
> 
> (%i1) load("altsimp.lisp")$
> 
> (%i2) nusum(n*n!,n,0,n);
> Dependent equations eliminated:  (1)
> (%o2) (n+1)!-1
> 
> (%i3) nusum(n^4*4^n/binomial(2*n,n),n,0,n);
> (%o3) (2*(n+1)*(63*n^4+112*n^3+18*n^2-22*n+3)*4^n)/(693*binomial
> (2*n,n))-2/(33*7)
> 
> (%i4) unsum(%,n);
> (%o4) (n^4*4^n)/binomial(2*n,n)
> 
> (%i5) unsum(product(i^2,i,1,n),n);
> (%o5) (product(i^2,i,1,n-1))*(n-1)*(n+1)
> 
> I'll do some more tests; if everything checks out, I'll make these changes.
> 
> I don't want altsimp to be a distraction from fixing bugs & extending
> Maxima
> is useful ways. Altsimp  is pretty much a 100% re-write. Over the years, a
> great
> deal of effort has gone into simp.lisp and simplus. A from scratch re-write
> might
> be a bad idea. A more conservative method for making Maxima infinity
> correct might
> be better.

I find simplus and friends pretty much undocumented and unreadable.  A 
100% rewrite doesn't sound like a bad idea to me.

A while back, I made a branch to try to make inf behave better.  You can 
look there if you want to see it, but it's pretty gross.  I never 
finished it because it causes the test suite to fail in many places 
because lots of the code either purposely or accidentally expect inf-inf 
to be 0, and similar stuff.

Ray