-----maxima-bounces at math.utexas.edu wrote: -----
>But when did solve start returning numeric answers?
The option variable algexact controls this. The test is
(%i85) eqs : [4*x^2-y^2 = 12,x*y-x = 2]$
Solve with algexact : true (non-default)
(%i86) solve(eqs,[x,y]), algexact : true$
The solution is a mess. Convert to 235 digit big floats
(%i87) bsol : bfloat(rectform(%)), fpprec : 235$
Convert to doubles and sort
(%i88) bsol : sort(float(bsol))$
Solve using algexact : false (default)
(%i89) sol : sort(float(solve(eqs,[x,y]))), algexact : false;
(%o89) [[x=-1.733751846381093,y=-0.15356757100197], ...]
Compare to bsol
(%i90) bsol-sol;
(%o90) [[0=-8.2147165736401462*10^-8, ...
Not so good -- this is with 5.14.0. Is there another option variable
that controls the accuracy of float solutions when algexact : false?
Barton