Re: [h5md-user] Multiple H5MD root groups per file

[Pierre de Buyl]

On Fri, Sep 20, 2013 at 09:50:47AM +0200, Konrad Hinsen wrote:
> Peter Colberg writes:
> 
>  > The proposal is as follows:
>  > 
>  > -------------------------------------------------------------------
>  > Any group within an HDF5 file may be an H5MD root group, with
>  > the exclusion of groups contained within an H5MD root group.
> 
> Sounds good. The only downside is that it makes reading trajectory
> files a bit more complicated, since the full specification of a
> trajectory is filename plus path to the group. But the benefits
> are worth it in my opinion.

I understand the motivation for the proposal but as Konrad points out there is
more detail needed.

I propose one of these updates:
1. Mandate a list of H5MD paths somewhere at the root of the file (for
autodiscovery of H5MD containers).
2. Allow the extension but mandate that H5MD data be communicated as
file.h5/path/to/some/h5md_root_group.

I prefer 1 but have no immediate proposal for the details (place for the list,
name of the list, etc.).

The hierarchical nature of HDF5 make the idea very natural to allow for H5MD
"file within a file", indeed. I am not 100% that we should do it if we have not
a very (very) straightforward way to open a file and find the H5MD root groups.
We need to consider also the impact on the implementations.

> I have one immediate application for this (in addition to the one you
> mentioned). I have been working on a framework for reproducible
> research for which I just made a first public announcement on the h5py
> mailing list. All the details are on Bitbucket:
> 
>   https://bitbucket.org/khinsen/active_papers/overview
> 
> In this framework, each publishable and shareable unit, containing
> both code and data, is a single HDF5 file. The root group for this
> file contains administrative information, whereas all data has to
> reside in the group /data. With the proposed extension, I could have a
> H5MD trajectory inside an ActivePaper, which is very attractive for my
> plans. In fact, I would probably have done this sooner or later even if
> the H5MD specification did not allow it :-)

This indeed a good motivation! But it is a bit more obvious in your situation as
one would have a link to the trajectories from the administrative part.

Pierre