how to get the gradient wrt to some inputs but not others? scoping?

gradient_scope_test.jl

using ReverseDiff
using Base.Test


mutable struct data
    X::Array{Float64, 2}
end

const D = data(zeros(Float64, 2, 2))

function f1(params)
    X = [1 2;
         3 4]
    sum(params[1]' * X[:, 1] - (params[1] .* params[2])' * X[:, 2].^2)
end

"query database to get new X, want gradient wrt params"
function f2(params, d)
    sum(params[1]' * d.X[:, 1] - (params[1] .* params[2])' * d.X[:, 2].^2)
end

f2(params) = f2(params, D)

function scope_test()
    D.X =  float.([6 7; 1 3])
    f2_tape = ReverseDiff.GradientTape(f2, [1, 2])
    D.X = float.([1 2; 3 4])
    grad = ReverseDiff.gradient!(f2_tape, [3,4])
    return grad
end

function scope_test2()
    D.X = float.([1 2; 3 4])
    f2_tape = ReverseDiff.GradientTape(f2, [1, 2])
    D.X = float.([1 2; 3 4])
    grad = ReverseDiff.gradient!(f2_tape, [3,4])
    return grad
end

D.X = float.([1 2; 3 4])

@test f1([3,4]) == f2([3, 4], D)

f1_tape = ReverseDiff.GradientTape(f1, [3,4])

@test ReverseDiff.gradient!(f1_tape, [3,4]) == scope_test() # fails! uses original values
@test ReverseDiff.gradient!(f1_tape, [3,4]) == scope_test2()