`MultiPhaseTPFlash` throws ArgumentError: reducing over an empty collection is not allowed

[abcdvvvv]

Summary

A call to MultiPhaseTPFlash aborts with the error above. It appears during the TP flash when assembling/aggregating phase results (likely a reduce/maximum/sum over an empty collection). Please help confirm the root cause and the correct fix.

Minimal Reproducible Example

# The mixture model is as reported in #454.
p = 2.99e6
t = 450.74
n = [3.68, 7.069, 2.3402, 0.8063]
res = Clapeyron.tp_flash2(mix, p, t, n, MultiPhaseTPFlash(full_tpd=true))

Output:

ERROR: ArgumentError: reducing over an empty collection is not allowed; consider supplying `init` to the reducer
Stacktrace:
  [1] _empty_reduce_error()
    @ Base .\reduce.jl:311
  [2] mapreduce_empty(f::Function, op::Function, T::Type)
    @ Base .\reduce.jl:313
  [3] reduce_empty(op::Base.MappingRF{typeof(eltype), typeof(promote_type)}, ::Type{Vector{Float64}})
    @ Base .\reduce.jl:350
  [4] reduce_empty_iter
    @ .\reduce.jl:373 [inlined]
  [5] mapreduce_empty_iter(f::Function, op::Function, itr::Vector{Vector{Float64}}, ItrEltype::Base.HasEltype)
    @ Base .\reduce.jl:369
  [6] _mapreduce
    @ .\reduce.jl:421 [inlined]
  [7] _mapreduce_dim
    @ .\reducedim.jl:334 [inlined]
  [8] mapreduce
    @ .\reducedim.jl:326 [inlined]
  [9] reduce
    @ .\abstractarray.jl:1775 [inlined]
 [10] initial_beta!(comps::Vector{Vector{Float64}}, z::Vector{Float64})    
    @ Clapeyron Clapeyron.jl\src\methods\property_solvers\multicomponent\tp_flash\multiphase.jl:1163
 [11] tp_flash_multi(model::PR{…}, p::Float64, T::Float64, nn::Vector{…}, options::MultiPhaseTPFlash{…})
    @ Clapeyron Clapeyron.jl\src\methods\property_solvers\multicomponent\tp_flash\multiphase.jl:224
 [12] tp_flash_impl
    @ Clapeyron.jl\src\methods\property_solvers\multicomponent\tp_flash\multiphase.jl:178 [inlined]
 [13] tp_flash2(model::PR{…}, p::Float64, T::Float64, n::Vector{…}, method::MultiPhaseTPFlash{…})
    @ Clapeyron Clapeyron.jl\src\methods\property_solvers\multicomponent\flash\PT.jl:86

Thank you for your time!

[longemen3000]

hmm, i will check what i can do, i need to update that whole flash code to use FlashResult in the iterations.

in the meantime, i got this little code for vlle, if you are interested:

function vlle_tp_flash(model,p,T,z)
   #TODO: create a TPFlashMethod with initial guesses for the LL and LV split, verbose, etc...
    res1 = tp_flash2(model,p,T,z,MichelsenTPFlash(equilibrium = :vle))
    if iszero(res1.fractions[1])
        return res1
    end

    res2 = tp_flash2(model,p,T,res1.compositions[1],MichelsenTPFlash(equilibrium = :lle))
    if iszero(res2.fractions[1]) || iszero(res2.fractions[2])
        return res1
    end

    spec = FlashSpecifications(p = p,T = T)

    comps = [res2.compositions[1],res2.compositions[2],res1.compositions[2]]
    βv,βl = res1.fractions[1],res1.fractions[2]
    βl1,βl2 = βl*res2.fractions[1],βl*res2.fractions[2]
    fracs = [βl1,βl2,βv]
    vols = [res2.volumes[1],res2.volumes[2],res1.volumes[2]]
    res0 = FlashResult(comps,fracs,vols,FlashData(p,T))
    return xy_flash(model,spec,z,res0)
end

[abcdvvvv]

in the meantime, i got this little code for vlle, if you are interested:

Nice, I can understand the code. But I'm not currently working on VLLE-related tasks, so I don't have any test cases on hand.

You can set a short path for pure components. And, I have performance concerns about the numerous small arrays allocated within the function.

[abcdvvvv]

For example, could you store comps, fracs in FlashResult as Tuple?