PyLammps: Add access to thermo output and dump information about lammps HOT 5 CLOSED

@rbberger is still under consideration? or should we close is issue?

from lammps.

@giacomofiorin can you have a look at this avg/chunk reader? Still need to write some example and/or docs. But it should be usable as follows:

from lammps import AvgChunkFile

c = AvgChunkFile(filename)
c.timesteps # list of timesteps in file
c.chunks    # list of chunks in file

# c.chunks[chunk_id][property_name]
c.chunks[0]['id']
c.chunks[0]['coord']
c.chunks[0]['density/count']  # list of density/count values over time of chunk 1

from lammps.

Will do ASAP, thanks for implementing it!

from lammps.

@rbberger: sorry for the delay. I misunderstood that the new class AvgChunkFile was going to make use of LAMMPS API functions (it can actually be used in the stand-alone interpreter).

Overall the class does the intended job. One thing that I recommend is replacing:

assert(num_chunks == int(parts[1]))

with an exception like:

if not (num_chunks == int(parts[1])):
  raise Exception("Currently, changing numbers of chunks are not supported.")

because the case of changing number of chunks over time is not an uncommon one (e.g. chunks along a Cartesian axis when running NPT, or chunks computed for a sub-set of the system).

I think it's possible to amend your parser, but ultimately this is an issue of the way the fix ave/chunk file is written to begin with. I think it'd be useful to define additional variants of the styles bin/1d through bin/3d where not only the grid spacings but also the boundaries can be pre-allocated (similarly to bin/cylinder and bin/sphere). This would make parsing of the ave/chunk output files much easier: empty chunks can be easily handled by having Ncount written as zero.

from lammps.

please don't close issues until the PR is merged.

from lammps.