Re: Translating ODE from Matlab to Octave

[c.]

On 26 Nov 2013, at 17:18, Bård Skaflestad <address@hidden> wrote:

>> no, you can't call 'lsode' like that. 
>> 
>> 'lsode' does not return the time vector 't',
> 
> Thanks a lot for alerting me to that fact.  Now I just have to ask
> myself why I didn't derive that conclusion from the documentation...

The documentation [1] gives two possibilities for calling lsode:

 — Loadable Function: [x, istate, msg] = lsode (fcn, x_0, t)
 — Loadable Function: [x, istate, msg] = lsode (fcn, x_0, t, t_crit)

where 'istate' and æmsg' are flags describing the succes / failure of the 
integration. a

As you see there is no 't' in the list of outputs, maybe you just assumed it 
would work that way because matlab ode solvers do ...
'lsode' does not have a matlab compatible interface because there is no 'lsode' 
function in matlab, you can find other solvers with 
more matlab-like interface in the odepkg forge package [2].

a matlab compatible ode15s is not implemented yet and it won't be in 3.8 
either, but that is a planned feature that might be in 4.0.

> Are you saying that LSODE does not support automatic step size selection
> or does the solver "simply" not expose the selected time steps to the
> caller?

lsode does use internally adaptive time-stepping but only returns the solution 
evaluated at a set of requested time steps.

c.


[1] 
http://www.gnu.org/software/octave/doc/interpreter/Ordinary-Differential-Equations.html#Ordinary-Differential-Equations
[2] http://octave.sourceforge.net/odepkg/