Adding priorities to UserDemand/Time causes initialization to fail

[Huite]

Running a model with all UserDemand at 1 runs fine, but assigning (random) priorities 1 to 3 results in this error below.

I haven't changed anything else in the model, so allocation isn't used.
In that case, the whole grouping per priority should maybe not happen at all?

❯ c:\bin\ribasim\bin\ribasim.exe ribasim.toml
┌ Info: Starting a Ribasim simulation.
│   cli.ribasim_version = "2024.4.0"
│   starttime = 2017-01-01T00:00:00
└   endtime = 2019-01-01T00:00:00
ERROR: BoundsError: attempt to access 0-element Vector{Float64} at index [1]
Stacktrace:
  [1] getindex
    @ .\essentials.jl:13 [inlined]
  [2] get_scalar_interpolation(starttime::Dates.DateTime, t_end::Float64, time::StructArrays.StructVector{Ribasim.UserDemandTimeV1, @NamedTuple{node_id::Vector{Int64}, time::Vector{Dates.DateTime}, demand::Vector{Float64}, return_fa
ctor::Vector{Float64}, min_level::Vector{Float64}, priority::Vector{Int64}}, Int64}, node_id::Ribasim.NodeID, param::Symbol; default_value::Float64)
    @ Ribasim D:\buildAgent\work\ecd2b8f9b25b1609\ribasim\core\src\util.jl:215
  [3] Ribasim.UserDemand(db::SQLite.DB, config::Ribasim.config.Config)
    @ Ribasim D:\buildAgent\work\ecd2b8f9b25b1609\ribasim\core\src\read.jl:703
  [4] Ribasim.Parameters(db::SQLite.DB, config::Ribasim.config.Config)
    @ Ribasim D:\buildAgent\work\ecd2b8f9b25b1609\ribasim\core\src\read.jl:893
  [5] Ribasim.Model(config::Ribasim.config.Config)
    @ Ribasim D:\buildAgent\work\ecd2b8f9b25b1609\ribasim\core\src\model.jl:51
  [6] run(config::Ribasim.config.Config)
    @ Ribasim D:\buildAgent\work\ecd2b8f9b25b1609\ribasim\core\src\main.jl:11
  [7] (::Ribasim.var"#206#208"{Ribasim.config.Config})()
    @ Ribasim D:\buildAgent\work\ecd2b8f9b25b1609\ribasim\core\src\main.jl:66
  [8] with_logstate(f::Function, logstate::Any)
    @ Base.CoreLogging .\logging.jl:515
  [9] with_logger
    @ Base.CoreLogging .\logging.jl:627 [inlined]
 [10] #205
    @ Ribasim D:\buildAgent\work\ecd2b8f9b25b1609\ribasim\core\src\main.jl:59 [inlined]
 [11] open(::Ribasim.var"#205#207"{Ribasim.config.Config}, ::String, ::Vararg{String}; kwargs::@Kwargs{})
    @ Base .\io.jl:396
 [12] open
    @ Base .\io.jl:393 [inlined]
 [13] main(ARGS::Vector{String})
    @ Ribasim D:\buildAgent\work\ecd2b8f9b25b1609\ribasim\core\src\main.jl:56
 [14] main()
    @ Ribasim D:\buildAgent\work\ecd2b8f9b25b1609\ribasim\core\src\main.jl:24
 [15] top-level scope
    @ none:1

[Huite]

I reckon the problem is here:

        elseif node_id in time_node_ids
            push!(demand_from_timeseries, true)
            for p in priorities
                push!(demand, 0.0)
                if p in keys(time_priority_dict)
                    demand_p_itp, is_valid = get_scalar_interpolation(
                        config.starttime,
                        t_end,
                        time_priority_dict[p],
                        node_id,
                        :demand;
                        default_value = 0.0,
                    )
                    if is_valid
                        push!(demand_itp_node_id, demand_p_itp)
                    else
                        @error "The demand(t) relationship for UserDemand #$node_id of priority $p from the time table has repeated timestamps, this can not be interpolated."
                        errors = true
                    end
                else
                    demand_p_itp = LinearInterpolation([0.0, 0.0], trivial_timespan)
                    push!(demand_itp_node_id, demand_p_itp)
                end
            end
            push!(demand_itp, demand_itp_node_id)

In my case, I have a node 118, with priority 3.

We first make the time_priority_dict. In this case, priority 1 => (data for node 131, 132).
We iterate over the node_ids. 118 is first. It fetches the data for priority 1 first, with time_priority_dict[p]. This obviously doesn't contain the data for node 118. This then results in zero-sized vector inside of get_scalar_interpolation, which causes the index error.

I think line if p in keys(time_priority_dict) is a thinko? This should check for the presence of the node_id in the group instead.

In general, my feeling is the entire function UserDemand is in need of some restructuring. It's over 110 lines.
Static data should probably be dealt with in a separate function, and the transient data in another function.
Also not clear (haven't read super carefully to be fair): why are the data processed in order of priority?
If this really matters, it would be good to have some comments.