[Ktechlab-devel] simulator

[Celelibi]

2008/11/10 Glen <gcanaday at gmail.com>

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
> Perhaps I should term what I mean an "expression solver," then? It's not
> meant to solve differential equations nor anything that complex.
>

So, I think x^2 + log(x) = 0 is too complex.

Though anything that can be solved using simple steps shouldn't be too
> hard as long as the expression is built with simple components. The
> parser itself will be more difficult!
>
> i^2 + log(i) is very much in the range of what I mean. Log functions
> have been built into many math libraries for a long time. Every 3D
> modeler out there also solves differentials, but that's going a bit far
> unless someone else wants to put it in.


I'm not sure to understand. (I'm just a poor frog, my english is not
perfect.)
So you think a "generic solver" could solve i^2 + log(i) = 0 ?
I think (but not sure) it has been proved that such equations can't be
solved without using numerical approximation.

What argument have you against just saying in the component :
- which solver the component uses
- what parameter we give to the solver
?

This would be very flexible. We could have a solver that uses linear
algebra, another that just take a "simple equation" and try to solve it,
another that try to solve numerically using Newton method etc.

And since there is a so wide range of possible solver, these should be
pluggable to ktechlab. That in order to add solvers without needing a new
release of the whole "ktech system".
Ok numerical solving works pretty well for almost every differetiable
function. But it may not be quick and/or accruate enough for everything. So
when adding a new component, we may need to add a new solver.


Celelibi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/ktechlab-devel/attachments/20081110/91caaf4d/attachment.html>