Change_independent_variable and connections

[aml5600]

Question❓

It looks like the change_independent_variable transformation needs to have the connections expanded. This fails:

    @mtkmodel ConnectSys begin
        @components begin
            in = RealInput()
            out = RealOutput()
        end
        @variables begin
            x(t)
            y(t)
        end
        @equations begin
            connect(in, out)
            in.u ~ x
            D(x) ~ out.u
            D(y) ~ 1
        end
    end
    @named sys = ConnectSys()
    sys = complete(sys; flatten = false)
    new_sys = change_independent_variable(sys, sys.y)
    ss = structural_simplify(new_sys)

failing on structural_simplify stuff. The systems from the connection equation are still wrt t.

@hersle is it possible to work through the connection systems similar to how you did everything else? should we enforce that all connections are expanded before calling the transformation? Thanks!

[hersle]

I haven't tested it on systems with connections. But a connection is equivalent to one or more equations that it expands to, right? If change_independent_variable can be adapted to do the substitution on the components inside the connect(...) statements, I think it should work. Can one do that without expanding the connection equations?

Otherwise, like you said, a workaround is to expand connections first.

[aml5600]

So I tried expanding connections before changing variables, but this necessarily flattens the system before going into change_independent_variable. Which then the resulting system is flat, and that breaks all my downstream stuff.

I tried overloading the transform function to specialize on Equation, and another to specialize on Connection, and an inbetween to glue the two. The Connection overload transforms each system in connection.systems.

The function would succeed, but then something was failing when trying to show/display the system in the REPL, so looks like I didn't reassemble something correctly =/ will keep trying, though.

[aml5600]

FYI I asked another question on slack about array variables, which are also a requirement for me unfortunately. Happy to discuss if you have any ideas on that front! Thanks again by the way!

[hersle]

I tried overloading the transform function to specialize on Equation, and another to specialize on Connection, and an inbetween to glue the two. The Connection overload transforms each system in connection.systems.

That would also be my first thought, to just recursively apply change_independent_variable to the systems in the connection.

[aml5600]

@hersle I went ahead and made #3625 . Let me know what you think.

[aml5600]

Fixed with #3625