Cannot update vector parameters with

[TorkelE]

I am not fully sure if this should work, but saw some stuff in the code that sugegsts that it should.

MWE. The two p terms should become 0 at time 1.0, causing decay of X and Y, this does not happen.

using ModelingToolkit, OrdinaryDiffEqDefault, Plots
using ModelingToolkit: t_nounits as t, D_nounits as D
import ModelingToolkit: SymbolicDiscreteCallback

# Creates a standard MTK system.
@parameters p(t)[1:2] d
@variables X(t) Y(t)
eqs = [
    D(X) ~ p[1] - d*X,
    D(Y) ~ p[2] - d*Y
]
devent = SymbolicDiscreteCallback([1.0] => [p ~ [0.0, 0.0]]; discrete_parameters = [p])
@mtkbuild sys = ODESystem(eqs, t; discrete_events = devent)

sim_cond = [X => 1.0, Y => 2.0, p => [1.0, 2.0], d => 1.0]
oprob = ODEProblem(sys, sim_cond, 2.0)
plot(solve(oprob))

[sablonl]

Facing the issue in my code, where using a matrix parameter with callbacks greatly improve the performance time of mtkcompile.

An additional information: sometimes scalarizing helps, but here it leads to another error during the call to mtkcompile:

devent = ModelingToolkit.SymbolicDiscreteCallback([1.0] => [Symbolics.scalarize(p ~ [0.0, 0.0])...]; discrete_parameters = [Symbolics.scalarize(p)...])

@named sys = ODESystem(eqs, t; discrete_events = devent)
csys = mtkcompile(sys) # ERROR: Expected discrete variable (p(t))[2] in callback to be a parameter

Just scalarizeing the equations doesn't help either.

A possible workaround is to directly use DiffEqCallbacks.DiscreteCallback, but this results in inconsistencies when retrieving observed variables; i.e, assuming the parameter was constant with its last known value.