Re: matlab compatibility of od45 with adaptive stepping

[Carlo De Falco]

On 11 Oct 2015, at 12:57, Sebastian Schoeps <address@hidden> wrote:

> That's what I meant ;) But Carlo indicated that we might have to rewrite a 
> lot. I think we just need to call dense output before returning the results 
> if the input time vector has more than two entries. 
> 
> Maybe we could additionally add an (Matlab incompatible) switch in odeset 
> that forces ode45 to use the input vector for the actual time stepping. I 
> would like that....
> 
> Sebastian

What I meant is, as Jacopo said, that matlab only stores solution at required 
timesteps, while Octave currently also saves timesteps created by adaptivity.
This is quite simple to implement but is completely different from the current 
very convoluted looping strategy in integrate_adaptive.
I am working out a patch right now.

c.