Action items post-release

[johnmyleswhite]

• [ ] Implement AllResults API from ForwardDiff
• [ ] Implement box-constrained dynamic direction choice for constrained optimization
• [ ] Implement method to calculate Jacobians
• [x] Write some basic benchmarks
• [ ] Port all tests from Calculus.jl

[pkofod]

Implement AllResults API from ForwardDiff

seems like this is now gone, is it still the way you want to do it, or do you want to follow the DiffBase-style?

ref: docs http://www.juliadiff.org/ForwardDiff.jl/upgrade.html

# old v0.1 style
answer, results = ForwardDiff.hessian(f, x, AllResults)
v = ForwardDiff.value(results)
g = ForwardDiff.gradient(results)
h = ForwardDiff.hessian(results) # == answer

# old v0.2 style
out = HessianResult(x)
ForwardDiff.hessian!(out, f, x)
v = ForwardDiff.value(out)
g = ForwardDiff.gradient(out)
h = ForwardDiff.hessian(out)

# current v0.3 style
using DiffBase
out = DiffBase.HessianResult(x)
ForwardDiff.hessian!(out, f, x)
v = DiffBase.value(out)
g = DiffBase.gradient(out)
h = DiffBase.hessian(out)