Dearests,
using Symbolics
@variables p, α, N
D = Differential(α)
D((p^α + 1 - p)^N) / log(p) |> expand_derivatives |> x -> substitute(x, α => 1) |> simplify
correctly returns N*p*(1^(N - 1)).
I was wondering if a substitution rule should be implemented such that 1^x -> 1 no matter what x is. BTW, Mathematica does it automatically.
f[ α_, p_, N_] := (p^α_ + 1 - p)^N
D[f[a, p, N], a]/(Log[p]) /. a -> 1 # returns N * p
It might be a good idea. Other simplifications area already done by default (you can check them here).
I’ll add a PR for this.
What if x=∞ or -∞?
1 is still a reasonable result. (math people think of it as an indeterminate form, but that tends to be less useful than defining it to equal 1.)
Open an issue. Note that Symbolics is a typed symbolic system, so @variables p is @variables p::Real which is the “standard” type and can make these tricks. I forget the name, but it’s like @variables p::SafeReal has different rule sets, for example omitting a/a == 1, which is safe for the case of a=0. The default is what users generally want, but the safe mode gives people a way to opt in to a slightly slower system that will work for all edge cases, giving the best of both worlds.
Made pull request PR #469 at SymbolicUtils.jl