About 8 years ago when the LAMMPS Python package was still very rudimentary and not well documented, I've created the PyLammps interface which wraps the lammps package in a more Pythonic way. It was mostly designed for smaller examples and for interactive use in Jupyter Notebooks. The vision at that time was that it might be useful for tutorials.
Things have significantly improved, especially in 2020 where we did major refactors of the Library/Python interface and added more extensive documentation. I believe it's time to retire PyLammps and direct people to using the lammps Package directly. We might be able to factor out some parts that people may find useful (like a CmdWrapper, IPython-plotting). But overall, the amount of people getting misled into using PyLammps instead of lammps for actual real work has been getting alarming. Specifically for running with mpi4py the interface becomes very fragile due to output parsing.
So the plan is two fold:
deprecate the PyLammps interface, possibly extract useful features in their own helpers
rewrite the PyLammps Howto in terms of the lammps interface using modern facilities.
About 8 years ago when the LAMMPS Python package was still very rudimentary and not well documented, I've created the
PyLammps
interface which wraps thelammps
package in a more Pythonic way. It was mostly designed for smaller examples and for interactive use in Jupyter Notebooks. The vision at that time was that it might be useful for tutorials.Things have significantly improved, especially in 2020 where we did major refactors of the Library/Python interface and added more extensive documentation. I believe it's time to retire PyLammps and direct people to using the
lammps
Package directly. We might be able to factor out some parts that people may find useful (like aCmdWrapper
, IPython-plotting). But overall, the amount of people getting misled into usingPyLammps
instead oflammps
for actual real work has been getting alarming. Specifically for running withmpi4py
the interface becomes very fragile due to output parsing.So the plan is two fold:
PyLammps
interface, possibly extract useful features in their own helperslammps
interface using modern facilities.