Closed BeastyBlacksmith closed 6 years ago
Yeah, I was thinking about this when doing the change. It seems that BVP boundaries are now the only thing that don't follow the u,p,t
pattern. Wrapping the u
in a tuple before sending is essentially free, so that's an option. Basically it would (res,u,p,t)
where u
is the u[1]
and u[end]
tuple, while t[1]
and t[end]
correspond to that. We can put t
on the end for the other BVP as well for consistency.
I like that choice
Great. I'll implement it as the last changes of the v4.0.0 upgrade. This will make it easy to wrap other multi-point BVP solvers like those in ODEInterface.jl too, so I'll get on that.
I find it very restrictive to pass only
u[1]
andu[end]
to the residual function ofTwoPointBVProblem
therefore one cannot access the tspan.One possibility is to pass solution types that only hold
u[1]
andu[end]
asu
. The other possibility I see is to explicitly passtspan
.