Marc A. Murison wrote: > > Leon Granowitz wrote: > > > James Phillips wrote: > > > My simple example of mathcad's weak symbolics is this bug, which has > > > been present and repeatable since about v3: the conversion of -b+c > > > into b+c when you select it from a-b+c ... > > > > > What's the problem with the a-b+c example? The minus in front of the b > > is not the sign of b but the subtraction operator between a and b. Or > > am I missing something? > > -- > > Syntactical elegance is in the eye of the parser, I suppose, but most people > would probably consider a-b+c to be equivalent to a+(-b)+c. That is, the > leaves of the expression tree contain a, -b, and c, and the two nodes both > contain the operation "+". In Mathcad, selecting the subexpression -b+c in > fact gives you b+c when you paste it in elsewhere, requiring you to manually > reinsert the minus on b. This is *very* annoying, since it needlessly > introduces nasty opportunities for user error. We already have too much to > occupy our minds when in the midst of grunging through a tough problem -- the > software should take care of silly bookkeeping such as this. > > But, for whatever (possibly valid) reasons, MathSoft refuses to get a clue. > Over the years, it has been demonstrated over and over again that the > programmers at MathSoft use at best toy algebra to test their interface, and > that they don't understand at all the kinds of things that would make a > symbolic manipulator's life easier. Silly problems like this get reported and > are seemingly routinely ignored, as evidenced by their continued appearance in > version after version. They really should consult with and *listen* to people > who grind through real problems for a living, but they never seem to. A big > part of the problem seems to be that, right or wrong, symbolics has a low > priority at MathSoft. They have to allocate their finite resources as they see > fit, but I sure do wish they'd spend some quality time on the symbolic > interface. > > -- > Marc A. Murison > Astronomical Applications Dept. > U.S. Naval Observatory, Washington, D.C. > http://riemann.usno.navy.mil/murison/ > mailto:email@example.com > Utinam logica falsa tuam philosophiam totam suffodiant!
O.K. already. I still think it's trivial. At least you see what it is doing. By the way, I can't select b + c in version 6+, but I can in version 7+.
Here's a more serious problem with Mathcad. It gives the answer 0 for the indeterminate form 0/0 (even for a symbolic form such as f(x)/g(x) at x = a where f(a) = g(a) = 0 and limit as x --> a of f(x)/g(x) is non-zero). Because you may not realize that both the numerator and denominator are zero you are not aware of the wrong answer. This problem should be given a much higer priority than the one above. Oddly, the solution is more than halfway there in Mathcad. Mathcad gives a divide by zero error if the denominator is zero and the numerator is non-zero. All they need to do is provide a indeterminate form message instead of 0 when the numerator is also 0 (for numeric problems) and make use of their limit function for the symbolic case.
-- Leon Granowitz | firstname.lastname@example.org Principal Engineer | email@example.com Raytheon | (617) 275-2665