> There is no resimplification after subst with an evaluating function:
>
> subst(2,x,zeta(x)) => zeta(2) (really 'zeta(2) )
> ev( 'zeta(2) , nouns) => %pi^2/6 -- invoking evaluator
> zeta(2) => %pi^2/6
>
> subst(2,x,psi[0](x)) => 1-%gamma
> psi[0](2) => 1-%gamma
>
> All mathematical functions like zeta should become simplifying functions.
But from the user perspective this is not a real difference. It's just
that zeta is a little more annoying to use.
So if I understand correctly when you say nintegrate should not be a
simplifying function you want that it produces an error when it can't
get a numerical result?
--
Andrej