Re: [h5md-user] box and observables

[Pierre de Buyl]

Hi,

Just replying a bit on the box issue, as I cannot keep track of all the topics
currently discussed on the list.

On Mon, Sep 23, 2013 at 11:48:27AM +0200, Konrad Hinsen wrote:
> Felix Höfling writes:
> 
>  > The reason that there are several box groups is primarily that different  
>  > subgroups have different sampling intervals. For trajectory subgroups, the 
>  
> 
> That was my understanding as well (there is even a sentence saying
> so), but apparently Peter disagrees.

Within a particles group, data can be sampled at different intervals. To make
the point clear, else we would have unique step/time datasets for a particles
group. One may wish to sample different coordinates at different time intervals.

>  > For this reason (and since the /particle root group may be absent),
>  > the box is also stored in /observables where it is considered a
>  > physical observable (NPT simulations), not just an appendage to
>  > some real data.
> 
> Having the box as an observable is indeed quite reasonable.

We had a realllllly long discussion on whether to put the box in /particles or
in /observables. The current solution is the result of that.

>  > Assuming that there is only one simulation box, the
>  > box is not stored inside each particle subgroup but at the main
>  > level. Actually, this latter point deviates from the structure in
>  > /particles and may be debated.
> 
> Now that I understand the reasoning behind the current layout, I'd say
> the fundamental problem is elsewhere: it's that the "observables" group
> can contain two very distinct kinds of subgroups: those that define
> systems-wide observables, and those that group together observables for
> different subsystems.
> 
> There is no such distinction in the "particles" group, where all data
> is stored by subsystem, implying that the whole system is just a special
> case of a subsystem.
> 
> Why not take the same point of view for "observables"? We'd then have
> only subsystem-naming groups at the top level, and all observables, 
> including "box", inside them.

I cannot comment on this yet but I have the impression that starting to give a
rigid structure to /observables might lead to premature complexification [1].

Cheers,

Pierre

[1] Can I say "Premature complexification is the root of all evil" ?