implicit-known_traj

[Peter230655]

I may have an idea how to handle something like surface(x, y).
What I have done is this:
I introduce
surface_x = sm.Function('surface_x')(x)
surface_y = sm.Function('surface_y')(y)

They represent the same function, but in surface_x, I treat y as parameters, and x as variables and vice versa.
I made a simple simulation, which looks like it may work, but early in the game.
If I get it, I will push it, so you can see it.

[moorepants]

You don't have to find a hack for everything, I can simply support the function z(x,y ).

[Peter230655]

What a shame! I thought my hack was really useful. :-) :-)

Jest aside: Playing around, I feel that supporting z(x_1, ..., x_n) is probably useful.

[moorepants]

I agree, I think I will add it. But I could merge what I have and then add the support later. I don't think it would change the API, you'd just do:

Problem(
    known_trajectory_map={
        z(x, y): calc_z_xy
        z.diff(x): calc_dzdx,
        z.diff(y): calc_dxdy,
    }
)

[Peter230655]

Would you allow z to be dependent on all state variables or just two of them?

I agree, this can wait. opty has so many new things, particularly than you can bound the eoms.
I think opty-dev is now substantially different from the released version.

NB:
It looks like my hack is working :-)
I will push it once it is beautified, so you can look at it if you want to.

[moorepants]

If I support more than one variable it would be anything in the free vector.

[Peter230655]

This would be perfect!
While my hack seems to work, it does add lots of lines to the code.