Issue: Cannot init variables via Vector{Vector{}} indexing.

[ofSingularMind]

I had a data variable ω for which I needed to process the components of in separate nodes. I formatted ω as a Vector{Vector{}} and passed it to the model in data = () field. I defined the portion of my model specific to ω as below.

for i in 1:size(ω, 1)
  ω[i][1] ~ UniSGP_AMF_Grad(x[i], v, w, θ, 1) 
  ω[i][2] ~ UniSGP_AMF_Grad(x[i], v, w, θ, 2) 
  ω[i][3] ~ UniSGP_AMF_Grad(x[i], v, w, θ, 3) 
end

Unfortunately, this did not work and lead to node variables not being updated. The solution was to format data object ω as a matrix and to utilize row, column indexing as below.

for i in 1:size(ω, 1)
  ω[i, 1] ~ GenericNode(x[i], v, w, θ, 1) 
  ω[i, 2] ~ GenericNode(x[i], v, w, θ, 2) 
  ω[i, 3] ~ GenericNode(x[i], v, w, θ, 3) 
end

This works fine, but maybe we should allow for Vector{Vector{}} formatting of data variables that need to be split up.

Thanks! Alex

[wouterwln]

@ofSingularMind this first model specification should have thrown an error when running the model macro, since, as far as I know, we don't allow double indexing on the left hand side. On my PC, this model macro does not even parse and already throws. Can you double check?

[ofSingularMind]

Ah, yes, that's right. I "solved" that issue with the following, which does compile ^_^

for i in 1:size(ω, 1)
  ω_i = ω[i][1]
  ω_j = ω[i][2]
  ω_k = ω[i][3]
  
  ω_i ~ UniSGP_AMF_Grad(x[i], v, w, θ, 1) 
  ω_j ~ UniSGP_AMF_Grad(x[i], v, w, θ, 2) 
  ω_k ~ UniSGP_AMF_Grad(x[i], v, w, θ, 3) 
end