Merge pull request #253 from JuliaDiffEq/myb/one

Add diff rule for `one`

Author: ChrisRackauckas

Project.toml

@@ -1,7 +1,7 @@
 name = "ModelingToolkit"
 uuid = "961ee093-0014-501f-94e3-6117800e7a78"
 authors = ["Chris Rackauckas <accounts@chrisrackauckas.com>"]
-version = "1.2.9"
+version = "1.2.10"
 
 [deps]
 DiffEqBase = "2b5f629d-d688-5b77-993f-72d75c75574e"

src/differentials.jl

@@ -113,6 +113,7 @@ end
 
 derivative(::typeof(+), args::NTuple{N,Any}, ::Val) where {N} = 1
 derivative(::typeof(*), args::NTuple{N,Any}, ::Val{i}) where {N,i} = Operation(*, deleteat!(collect(args), i))
+derivative(::typeof(one), args::Tuple{<:Any}, ::Val) = 0
 
 function count_order(x)
     @assert !(x isa Symbol) "The variable $x must have an order of differentiation that is greater or equal to 1!"

test/derivatives.jl

@@ -72,3 +72,13 @@ isequal(ModelingToolkit.derivative(Operation(+, [x*y, y, z]), 1), 1)
               Expression[ModelingToolkit.Constant(1)  ModelingToolkit.Constant(0)
                          Differential(t)(x)           ModelingToolkit.Constant(1)
                          ModelingToolkit.Constant(0)  ModelingToolkit.Constant(0)])
+
+# issue 252
+@variables beta, alpha, delta
+@variables x1, x2, x3
+
+# expression
+tmp = beta * (alpha * exp(x1) * x2 ^ (alpha - 1) + 1 - delta) / x3
+# derivative w.r.t. x1 and x2
+t1 = ModelingToolkit.gradient(tmp, [x1, x2])
+@test_nowarn ModelingToolkit.gradient(t1[1], [beta])