Add norm for Vector{T<:Dual}

[dlfivefifty]

julia> norm([dual(1.,2.),dual(3.,2.)])
ERROR: InexactError()
 in float at float.jl:121
 in generic_vecnormInf at linalg/generic.jl:92
 in generic_vecnorm2 at linalg/generic.jl:114
 in vecnorm at linalg/generic.jl:175
 in norm at linalg/generic.jl:191

[jrevels]

see JuliaDiff/ForwardDiff.jl#52 for prior experience with this, the key is to define float(::Dual)

[dlfivefifty]

I’d expect norm(::Vector{T<:Dual}) to return a Dual. Does this change result in that?

On 9 Dec 2015, at 1:57 AM, Jarrett Revels [email protected] wrote:

see JuliaDiff/ForwardDiff.jl#52 JuliaDiff/ForwardDiff.jl#52 for prior experience with this, the key is to define float(::Dual)

—
Reply to this email directly or view it on GitHub #31 (comment).

[jrevels]

The definition I'm thinking of is:

float(d::Dual) = Dual(float(value(d)), float(epsilon(d)))

With that definition, yes, norm(::Vector{T<:Dual}) should return a Dual

[dlfivefifty]

OK, thanks! Are you planning to add this definition to DualNumbers.jl?

On 9 Dec 2015, at 12:10 PM, Jarrett Revels [email protected] wrote:

The definition I'm thinking of is:

float(d::Dual) = Dual(float(value(d)), float(epsilon(d)))
With that definition, yes, norm(::Vector{T<:Dual}) should return a Dual

—
Reply to this email directly or view it on GitHub #31 (comment).

[MikaelSlevinsky]

i thought i added this in #29? Yep, I also added a test.

[dlfivefifty]

This has been implemented apparently.