parameter is not a parameter, bad error message

[baggepinnen]

The following system gives the cryptic error

ArgumentError: r(t) is not a parameter.

I assume it has something to do with r being an array and the equation using .~, but the error message doesn't really tell me what the problem is, r is certainly a parameter from the user's point of view.

using LinearAlgebra
@component function Temp(; name, r)
    @parameters r(t)[1:3]=r [
        description = "position vector from frame_a to frame_b, resolved in frame_a",
    ]
    @variables fb(t)[1:3]=r [
        description = "force",
    ]
    eqs = cross(r, fb) .~ zeros(3)
    ODESystem(eqs, t; name)
end

Temp(; name=:t, r=zeros(3))

[ven-k]

Error is because the r wasn't passed to ODESystem.

@component function Temp(; name, r)
    :
    ODESystem(eqs, t, [], [r...]; name)
end

Maybe this^1 error message can be changed from throw(ArgumentError("$p is not a parameter.")) to throw(ArgumentError("Either $p is not a parameter or it is not listed as a parameter of the System.")). (It should be possible to say that as check_parameters is used only inside the JumpSystem, ODESystem SDESystem and DiscreteSystem definitions).

[baggepinnen]

If I change the equation to eqs = cross(r, fb) ~ zeros(3) (without the dot in .~) the error goes away without me providing [r...] to the ODESystem constructor, why this inconsistency? Also, the variable fb is not passed to the constructor either, another inconsistency?

[hersle]

I think the reason for the inconsistencies is how check_parameters() are determining parameters based on

1. if they are callable, like r(t)
2. if they are arrays r(t)[1:3]

I guess the "error" here is that r(t)[1:3] represents a (constant?) offset vector, so it should really be defined as r[1:3], in which case the example runs. But a better error should still be thrown. I think the weirdest behavior is that it is possible that isparameters(par) returns false on a parameter defined with @parameters.