integrate returns undefined .. Bug in zero denominator?

The fundamental theorem of integral caculus can be extented to this slightly 
stronger statement under the right conditions (which apply in this case).

integrate(f(x),x,a,b) = limit(F(x),x,b) - limit(F(x),x,a)

diff(F(X),x) = f(x)

Then solving it this way could be a solution to the integrate bug.

Rich




----- Original Message ----- 
From: "John Lapeyre" <pdl at johnlapeyre.com>
To: <maxima at math.utexas.edu>
Sent: Wednesday, September 10, 2008 4:53 PM
Subject: Re: [Maxima] integrate returns undefined .. Bug in zero 
denominator?


> This is not too pretty, but it seems to remove some bad
> solutions. For simplicity, it assumes the eqn is of the form
> someexpr = 0.  It checks some solutions of univariate
> equations.  Multivariate cases and some complicated cases
> are supposed to fall through. It also assumes that if
> limit() returns a number, it is more reliable than when
> solve() returns a solution!
>
> mysolve(sin(x-y)/(x-y),x);
>
> `solve' is using arc-trig functions to get a solution.
> Some solutions will be lost.
> (%o5) []
>
> ------------
> removebadsolns(expr,solns,var) := block( [s,var1,lim],
>    for s in solns do (
>    var1:inpart(s,1),
>    if mapatom(var1) and var1 = var then (
>      lim:limit(expr,var,inpart(s,2)),
>      if numberp(lim) and lim # 0 then
>        solns:delete(s,solns))),
>     solns);
>
> mysolve(expr,var) := block([solns],
>  solns:solve(expr,var),
>  if not listp(var) then
>     solns:removebadsolns(expr,solns,var) else solns);
> _______________________________________________
> Maxima mailing list
> Maxima at math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima