Looking further at Matlab I see that their is an interesting disconnect between the form [t y] = ode45(...) and sol = ode45(...). The structure version, using sol, is far more flexible. With dense output being generated using deval(sol, tspan, [varIdx...]).
In a sense what I am currently during is a bit of a hybrid where both versions (the grided, and dense output versions) having the diagnostics etc. Now I am less clear as I think about this if this is the correct way of doing this, should I only support the dense version where the solver grid points are saved and then a mechanism like deval is used to get gridded solutions?
Looking further at Matlab I see that their is an interesting disconnect between the form
[t y] = ode45(...)andsol = ode45(...). The structure version, usingsol, is far more flexible. With dense output being generated usingdeval(sol, tspan, [varIdx...]).In a sense what I am currently during is a bit of a hybrid where both versions (the grided, and dense output versions) having the diagnostics etc. Now I am less clear as I think about this if this is the correct way of doing this, should I only support the dense version where the solver grid points are saved and then a mechanism like
devalis used to get gridded solutions?