Re: [h5md-user] H5MD for proteins

[Pierre de Buyl]

On Tue, Sep 10, 2013 at 08:12:26AM +0200, Konrad Hinsen wrote:
> Pierre de Buyl writes:
> 
>  > H5MD attempts to be rather generic. The module/convention in the other 
> thread
>  > could be a satisfactory solution.
> 
> Indeed. But one question one needs to ask is: what's the goal of
> "bare" H5MD?  What can one do with a trajectory file if all one knows
> is that it's a valid H5MD file?
> 
> For now I see mainly visualization of trajectory contents, plus some
> very specific calculations. For example, one could compute the center
> of mass of a subgroup. Computing the center of mass of two subgroups
> is already impossible without additional assumptions, because the
> positions of two different subgroups could be stored in different
> units that are unknown.

Peter followed your suggestion to use compact datasets and units are back in the
specification, as attributes to the datasets. See
http://nongnu.org/h5md/draft.html#time-dependent-data and
http://nongnu.org/h5md/draft.html#time-independent-data

Anyway, H5MD does not mandate the presence of mass data, nor of species data.
You can have a valid H5MD file with only, say, positions.

>  > Do you mean that Mosaic-HDF5 would sit on top of H5MD?
> 
> Not in general, just for trajectories. A trajectory would be
> an H5MD file with a Mosaic universe (another HDF5 group) sitting
> next to the h5md directory. With a definite name for that universe,
> we'd have our first module.

Seems good to me :-)

>  > > I'll post a few more specific questions about the current H5MD
>  > > definition in separate threads, this message is just to provide
>  > > the context for them.
>  > 
>  > Welcome back then :-)
> 
> Thanks! And thanks for all the work you all did on H5MD. I know
> it's a lot of work which most outsiders don't appreciate as much
> as they should.