[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [h5md-user] Species Data Type

From: Konrad Hinsen
Subject: Re: [h5md-user] Species Data Type
Date: Mon, 9 Sep 2013 12:11:10 +0200

Pierre de Buyl writes:

 > Extensions of "species" could be discussed later on, however (after
 > we settle on a base format that will receive the version number
 > 1.0).

I understand that not everything can be settled for version 1.0, but
the current definition for species information is close to empty.  All
it says that if a program wants to store species information, it
should do so using integer arrays arranged in a specific way.  A set
of random integers would be perfectly compatible with the definition,
and yet completely useless.

What I'd expect at a minimum from a species definition is a list of
promises that a program reading a trajectory may safely assume.
For example, that particles with identical properties have the same
number, meaning that particles with a different number differ in at
least one relevant property.

Beyond such basic rules, I suspect there is little one can define in a
meaningful way for the whole range of applications that H5MD is good
for. As someone who mostly simulates molecular systems whose particles
are "real" atoms, I'd be tempted to use the chemical element number
for the species information, but of course this doesn't carry over to
different models. And even for atomic-scale simulations, someone else
may want to make a distinction between different isotopes.

I faced similar problems in my Mosaic project, and came up wit the
concept of "conventions". The idea is that different domains of
application can come up with their own conventions for naming and
numbering things. They can then write down their conventions and give
them a name, which is recorded in a Mosaic data item. That approach
keeps domain-specific details out of a general format specification,
and provides openness for future extensions.

I think the same approach could work for H5MD: provide a "convention"
field in the H5MD metadata, and let people work out conventions for
their respective domains. For the species information, this could range
from a simple convention such as "use chemical element numbers" to
more elaborate ones such as "the species is an index into an array
named xxx that stores additional information".

Konrad Hinsen
Centre de Biophysique Moléculaire, CNRS Orléans
Synchrotron Soleil - Division Expériences
Saint Aubin - BP 48
91192 Gif sur Yvette Cedex, France
Tel. +33-1 69 35 97 15
E-Mail: research AT khinsen DOT fastmail DOT net
ORCID: http://orcid.org/0000-0003-0330-9428
Twitter: @khinsen

reply via email to

[Prev in Thread] Current Thread [Next in Thread]