Niall Jackson n.jackson12 at imperial.ac.uk
Wed Jul 9 16:02:49 UTC 2014

Dear Konrad,

Thank you for your reply. I suspected that this might be the case - I'll 
have a look into writing a Python script to interpret LAMMPS output 
files and feed them into MMTK.

On 09/07/14 17:01, Konrad Hinsen wrote:
> Niall Jackson writes:
>   > We are trying to use nMOLDYN with LAMMPS trajectories. Ideally, we would
>   > like to produce a new LAMMPS dump style that generates MMTK style dump
>   > files directly. Is there a document or manual section which outlines the
>   > structure of the MMTK style NetCDF file?
> Unfortunately not, and for a reason: it is probably very difficult to
> generate such netCDF files without using MMTK.
> Most of the contents of a trajectory file are plain netCDF arrays that
> hardly require much documentation. But there is also the variable
> "description", which contains a string that defines the complete
> universe for which the simulation was run. That string is a Python
> expression that gets evaluated in the right context to reconstruct the
> system. This is a nice trick, and quite efficient, but not at all
> portable. At the very least, you need Python and a class hierarchy
> with a similar structure to MMTK's.
> For this reason, all trajectory converters to or from MMTK's format
> that I know of are written in Python and use MMTK.
> My long-term plan is to transition to a new format that is fully
> documented, more flexible, and has better performance:
>    https://mosaic-data-model.github.io/mosaic-specification/h5md_mosaic_module.html
> It won't happen rapidly though, because it is very difficult to get
> funding for such work.
> Konrad.

More information about the mmtk mailing list