preallocating an array of random variables, getting varname As used multiple times in model (subsumes As[1])

[SimonEnsemble]

I'd like to create a prior distribution on a function A(x) for x \in [0, 1]. to do so, I represent the function via a list of discrete values of the function evaluated at N points on a grid over [0, 1]. I wish for A(0) to have a uniform prior, but then have the rest of the function values sequentially correlated for a smoothness prior. in a previous version of Turing.jl, I was doing:

As ~ filldist(Normal(), N) # cm²
As[1] ~ Uniform(0.0, 1.0)
for i in 2:N
   As[i] ~ As[i - 1] + Normal(0.0, γ) 
end

but now in v0.33.1, I get the error varname As used multiple times in model (subsumes As[1]).

how can I pre-allocate an array of random variables, then specify their distributions later?

[SimonEnsemble]

ah, this seems to work. the key was to not use ~ twice.

As = Vector{Real}(undef, N)
As[1] ~ Uniform(0.0, 1.0)
for i in 2:N
     As[i] ~ As[i - 1] + Normal(0.0, γ)		
end