[CP2K:1576] Re: UNIT in &CELL and UNIT for output

[Teodoro Laino]

Hi Toon,

> with (real) atomic units and nothing else, but this is probably not acceptable 
> for everyone. netCDF or hdf5 are good starting points, but a good layer on top 
> of that is hard to design. What would you think of a questionnaire about output 
> usage patterns? That could give us a few good clues.
>   
That sounds like a good idea (if you will find responsive users, which 
are kind of rare beasts to find ;-) ).
Unfortunately, standardization/design of formats is a quite difficult 
topic, and asymptotically, only the most
user-friendly (and not the best devised) will survived and will be used.
Anyway.. having people interested in this topic is pretty important.. 
and if you'll manage to find time to
come out with a kind of library which CP2K can sit on the top of, this 
would be simply perfect.

I don't know the standard you guys use for MD-Tracks.. maybe that one is 
already a good starting point.

A presto,
Teo
> I have a bit of experience with the topic, although this is all still low 
> profile. Our MD analysis code (MD-Tracks) converts output files from several MD 
> codes into a sort of standard binary database format. After the conversion, we 
> do a lot of post-processing without having to care about unit conversions or 
> file formats. In fact the results are only converted to a non-atomic-unit when 
> writing values to a spreadsheet or when plotting them. This is all pretty 
> satisfactory, mainly because that standard format allowed us to implement a 
> whole bunch of analysis features in very little time. There are still some 
> obvious improvements that should be implemented in the database format. I plan 
> to do this next summer. It would be nice to have a first serious draft of a 
> standard by then.
>
> cheers,
> Toon
>
>
>
>
> >
>