Error when computing jacobian for broadcasted function

[mattiasvillani]

Simple broadcasting errors for me. See below.
I am on Julia 1.10.0-beta2+0.x64.linux.gnu

within the following environment:

Status ~/.julia/environments/v1.10/Project.toml
[c29ec348] AbstractDifferentiation v0.5.2
[9f5e2b26] Diffractor v0.2.1
[31c24e10] Distributions v0.25.100

This shows the issue (verified that I get the same problem with gradient from AbstractDifferentiation.jl):

using Diffractor: DiffractorForwardBackend
using AbstractDifferentiation: derivative, jacobian
jacobian(DiffractorForwardBackend(), θ -> sin(θ[1]) + sin(θ[2]), [0.3, 0.8])  #  works
jacobian(DiffractorForwardBackend(), θ -> sum(sin, θ), [0.3, 0.8])                # works
jacobian(DiffractorForwardBackend(), θ -> sum(sin.(θ)), [0.3, 0.8])               # errors

last command errors with (truncated error):

ERROR: MethodError: no method matching Diffractor.TangentBundle{1, Float64, Diffractor.TaylorTangent{Tuple{Float64}}}(::Float64, ::Tuple{Float64})
Stacktrace:
[1] createinstance(::Type{Diffractor.TangentBundle{1, Float64, Diffractor.TaylorTangent{Tuple{Float64}}}}, ::Float64, ::Vararg{Float64})
@ Diffractor ~/.julia/packages/Diffractor/vzpvE/src/tangent.jl:351
[2] _getindex
@ Diffractor ~/.julia/packages/StructArrays/dNQpc/src/structarray.jl:353 [inlined]
[3] getindex
@ Diffractor ~/.julia/packages/StructArrays/dNQpc/src/structarray.jl:348 [inlined]
[4] iterate
@ Diffractor ./abstractarray.jl:1212 [inlined]
[5] iterate
@ Diffractor ./abstractarray.jl:1210 [inlined]
[6] iterate
@ Diffractor ./generator.jl:44 [inlined]
[7] _collect
@ Diffractor ./array.jl:852 [inlined]
[8] collect_similar
@ Diffractor ./array.jl:761 [inlined]
[9] map
@ Diffractor ./abstractarray.jl:3273 [inlined]
[10] rebundle(A::StructArrays.StructVector{Diffractor.TangentBundle{1, Float64, Diffractor.TaylorTangent{Tuple{Float64}}}, Tuple{Vector{Float64}, Vector{Float64}}, Int64})
@ Diffractor ~/.julia/packages/Diffractor/vzpvE/src/tangent.jl:375
[11] (::Diffractor.∂☆{1})(zc::ZeroBundle{1, typeof(copy)}, bc::Diffractor.CompositeBundle{1, Base.Broadcast.Broadcasted{…}, Tuple{…}})
@ Diffractor ~/.julia/packages/Diffractor/vzpvE/src/stage1/broadcast.jl:27
[12] materialize
@ Diffractor ./broadcast.jl:903 [inlined]
[13] (::Diffractor.∂☆internal{1})(::ZeroBundle{1, Type{Base.Broadcast.Broadcasted}}, ::Vararg{Diffractor.AbstractTangentBundle{1}})
@ Diffractor ~/.julia/packages/Diffractor/vzpvE/src/stage1/forward.jl:121 [inlined]
[14] ∂☆
@ Diffractor ~/.julia/packages/Diffractor/vzpvE/src/stage1/forward.jl:160 [inlined]
[15] #61
@ Diffractor ./REPL[45]:1 [inlined]
[16] (::Diffractor.∂☆recurse{1})(::ZeroBundle{1, var"#61#62"}, ::Diffractor.TangentBundle{1, Vector{Float64}, Diffractor.TaylorTangent{Tuple{Vector{Float64}}}})
@ Diffractor ~/.julia/packages/Diffractor/vzpvE/src/stage1/recurse_fwd.jl:0
[17] (::Diffractor.∂☆internal{1})(::ZeroBundle{1, var"#61#62"}, ::Vararg{Diffractor.AbstractTangentBundle{1}})
@ Diffractor ~/.julia/packages/Diffractor/vzpvE/src/stage1/forward.jl:121
[18] (::Diffractor.∂☆{1})(::ZeroBundle{1, var"#61#62"}, ::Vararg{Diffractor.AbstractTangentBundle{1}})
@ Diffractor ~/.julia/packages/Diffractor/vzpvE/src/stage1/forward.jl:160
[19] (::Diffractor.var"#pushforward#359"{var"#61#62", Tuple{Vector{Float64}}})(vs::Tuple{Vector{Float64}})
@ Diffractor ~/.julia/packages/Diffractor/vzpvE/src/AbstractDifferentiation.jl:13
[20] (::Diffractor.var"#358#362"{Diffractor.var"#pushforward#359"{var"#61#62", Tuple{Vector{Float64}}}})(cols::Vector{Float64})
@ Diffractor ~/.julia/packages/AbstractDifferentiation/eEkWP/src/AbstractDifferentiation.jl:526
[21] mapslices(f::Diffractor.var"#358#362"{Diffractor.var"#pushforward#359"{var"#61#62", Tuple{Vector{Float64}}}}, A::Matrix{Float64}; dims::Int64)
@ Base ./abstractarray.jl:3179
[22] jacobian(b::DiffractorForwardBackend, f::Function, args::Vector{Float64})
@ Diffractor ~/.julia/packages/AbstractDifferentiation/eEkWP/src/AbstractDifferentiation.jl:524
[23] top-level scope
@ REPL[45]:1
Some type information was truncated. Use show(err) to see complete types.

[mattiasvillani]

Just noted that the following works

using Diffractor
Diffractor.gradient(θ -> sum(sin.(θ)), [0.3, 0.8])

but

using Diffractor
Diffractor.jacobian(θ -> sum(sin.(θ)), [0.3, 0.8])

errors with message

julia> Diffractor.jacobian(θ -> sum(sin.(θ)), [0.3, 0.8])
ERROR: UndefVarError: jacobian not defined
Stacktrace:
[1] getproperty(x::Module, f::Symbol)
@ Base ./Base.jl:31

[oxinabox]

There is no function Diffractor.jacobian only AbstractDifferentiation.jacobian.
Diffractor.gradient is not part of the public API, we make zero stability promises for it.

[mattiasvillani]

Ok, thanks. I take it as the original issue is still a (potential) bug.

[oxinabox]

This may or may not be fixed in #224