Speaking for myself (and myself only) from now on, I didn't bother replying because it seems to me that you are simply misguided. As someone said, it is technically feasible. Almost anything is. But that doesn't mean it should be done (esp. bearing in mind the cost of developing a good solution to your problem).
I personally signaled my interest for something much more general than only derivatives. But as Dongfei's goal seems to be focused in that direction, I didn't want to disturb him further from it. If I want more, I can always do it by myself. About the possibility of using Maple or any other tool: of course, that's what any programmer would do. Like I said, the interest of a compile-time simplifier appears when you have some compile-time integral values that you don't know in advance in your expression. Of course you can specialize for N=1, N=2, etc... until N is too big for the expression to be simplifiable. It might also be "periodically simplifiable" (for example for N being a factor of some value), in which case specialization can still do the trick, I agree. But a more general tool can be handy as well. It's just horribly hard to do, never denied that. Regards Bruno