UndefVarError: SOI not defined

[Esnilg]

Hello, I have been trying to run the example model but I have this error when running

const SOI = StructDualDynProg.SOI

what I need to do

[blegat]

Which version of StructDualDynProg are you using ?

[Esnilg]

Hi

I Have

Pkg.status("StructDualDynProg")

• StructDualDynProg 0.0.1

[blegat]

You need to use at lease version 0.2 to have the SOI constant defined. I would recommend updating the package

[Esnilg]

Thanks, today update and now works, but as I do to know the status, objective value, solution x?

[blegat]

You need to do

sol = SOI.optimize!(...)
SOI.last_result(sol).lowerbound
SOI.last_result(sol).upperbound
...

You can see the other fields of the result here.

[Esnilg]

Thank you very much for all your help, I found what I want but I wanted one last question, how do I better understand this data structure of SOI.last_result (sun) .paths? for example I want to know the solution of the master node, x1, after x2, ...., xT

[blegat]

SOI.last_result(sol).paths gives this structure:

StructDualDynProg.jl/src/SDDP/algorithm.jl

Lines 61 to 63 in 349c9bc

	struct Paths{NodeT, TT, SolT} <: SOI.AbstractPaths
	paths::Vector{Vector{Tuple{NodeT, Vector{SDDPPath{TT, SolT}}}}}
	end

As explained here:

StructDualDynProg.jl/src/SDDP/algorithm.jl

Lines 65 to 70 in 349c9bc

	# At the master node, all paths are still concentrated in the same one
	node_paths = paths.paths[1]
	@assert length(node_paths) == 1
	paths_vec = node_paths[1][2]
	@assert length(paths_vec) == 1
	Ks = paths_vec[1].K

To get the path for the master node you need to do

SOI.last_result(sol).paths.paths[1][1][2][1]

Then you will have this structure

StructDualDynProg.jl/src/SDDP/path.jl

Lines 36 to 53 in 349c9bc

	mutable struct SDDPPath{TT<:SOI.AbstractTransition, SolT<:SOI.AbstractSolution}
	# Solution pool for the last node of the path
	pool::SolutionPool{TT, SolT}
	# The three following arguments are vectors since when we merge paths,
	# we need to keep track of the difference of `z` that they have accumulated
	# while diverging for computing the variance `σ^2` in [`meanstdpaths`](@ref) at the end.
	# Sum of the objective solution (without θ) at each node of the path
	z::Vector{Float64}
	# Probability of the path
	proba::Vector{Float64}
	# Number of walks in the path
	K::Vector{Int}
	function SDDPPath{TT}(sol::SolT, z, proba, K) where {TT, SolT}
	new{TT, SolT}(SolutionPool{TT}(sol), z, proba, K)
	end
	end

and the solution is in the pool field.
I agree that we should add a more user friendly way to access it

[Esnilg]

Hello, I wanted to thank you, I was able to better understand the data structure with typeof and fieldnames. But I have a question, how do I use model2lattice

[blegat]

It doesn't exist anymore. I assume you read it from the README, I just fixed it:
ac3b1d0