SV: Re: [MMTK] Temporary implementation of integrator in python

[Rune Larsen]

--- khinsen at cea.fr skrev:

> On 27.09.2005, at 01:33, Rune Larsen wrote:
> 
> > and with the MMTK package. I've hoped to find a
> > template integrator written intirely in python but
> > I've only come across C implementations. Now I'am
> a
> 
> I have had a velocity-verlet implementation in
> Python lying around  
> for a while, but I didn't think anyone would be
> interested... From  
> the next release, it will be in the example
> collection. I'll send you  
> a copy already, of course.
> 
> > Now for the questions :
> >
> > Does Universe.configration() return a real copi of
> the
> > atom positions ?
> 
> No. If you want a copy, make one: 
> copy(universe.configuration())
> 
> > Does Universe.energy() evaluate the energy before
> > returning ?
> 
> Yes.
> 
> > (risk of sounding stupid) Do I have to negate the
> > energy gradient of
> > Universe.energyAndGradient()
> > to get the force or is it implied ? And are they
> both
> > evaluated before they are returned ?
> 
> What you get is the gradient, so you do have to
> negate it if you want  
> the forces. And yes, both are evaluated before they
> are returned.
> 
> > My integration is made within a for-loop, and I
> write
> > a trajectory file using snapshot ... is there a
> better
> 
> That's just fine.
> 
> > way ? TrajectoryOutput takes "step numbers" as
> > arguments but I dont know where to define the
> "step"
> 
> The snapshot generator generates th.em
> automatically, every snapshot  
> counts as one step.
> 
> Konrad.
> --

Thank you for the fast reply. The python verlet
implementation helped a lot. 

I have just one more question. I would like to use the
TrajectoryAction objects: TranslationRemover and
RotationRemover. How can I call these two methods from
within my integrator ?. My Integrator is derived from
Dynamics.Integrator.