|
From: | Felix Höfling |
Subject: | Re: [h5md-user] New commits, "id" and rc |
Date: | Thu, 02 Aug 2012 11:20:55 +0200 |
User-agent: | Opera Mail/12.00 (Linux) |
Am 31.07.2012, 18:08 Uhr, schrieb Peter Colberg <address@hidden>:
On Tue, Jul 31, 2012 at 05:45:23PM +0200, Felix Höfling wrote:I have some minor suggestions for "observables", see the attached commit. Mainly, I think we should make it possible to have subgroups for storing partial pressures or partial energy contributions. Global quantities shall be stored at the top level still. If you agree I will upload the commit.This is a good point, and I would like to extend it a bit. When I read through the specification, I stumbled upon the explicit mentioning of observables ("temperature", "total_energy", …), and also noticed that "pressure" was missing (which Felix has fixed in his patch). The selection of observables seems a bit arbitrary to me, and I wonder what exactly we should standardise for the observables group. I regard as important the specification of time-dependent data sets. But besides that, I do not see the point of restricting the group structure or dataset names of the observables group, unless we restrict the H5MD file format to applications that define these quantities. What do you think about the specification of observables? Should H5MD specify only the use of time-dependent datasets in observables, or should it also specify exact names and physical interpretation? Peter
I would like to read the specification as follows: if you intend to store the system's potential energy in the H5MD file then do it in the "observables" group and use the name "potential_energy". Of course, it is not mandatory to have this information and every user is free to invent new names for whatever type of data. But _suggesting_ common names for the most common variables eases the analysis of H5MD files from different sources.The same applies for the trajectory group, where, e.g., "forces" and "species" are optional datasets. Actually, reading the current version of the draft, it is not clear whether force is optional or mandatory. Actually, I would favour having almost all datasets optional, the file format should not enforce storing information that is not needed within a specific application/project. Only few things should be mandatory, e.g. the "h5md" group. Other are conditionally required, e.g., the "box" datasets is only needed if particle positions are given (although it may be present without a trajectory group).
Minor remark: The name of the unit attribute is not specified: "The “position”, “velocity” and “force” datasets possess an optional attribute that is the unit of their respective data (“nm” for the position, for instance)."
Regards, Felix -- Dr Felix Höfling Max Planck Institute for Intelligent Systems (formerly Max Planck Institute for Metals Research) Heisenbergstr. 3 70569 Stuttgart Germany Phone: +49 711 689 1938 Fax: +49 711 689 1922
[Prev in Thread] | Current Thread | [Next in Thread] |