ase-extxyz

ASE I/O plugin that wires the extxyz C parser into ASE's read / write machinery.

Install

pip install ase-extxyz

This pulls in extxyz (the parser, with native binary wheels) and ase.

Use

import ase.io
from ase.build import bulk

atoms = bulk('Cu') * 2
ase.io.write('out.xyz', atoms, format='cextxyz')
back = ase.io.read('out.xyz', format='cextxyz')

The plugin registers a format named cextxyz (not extxyz — that name is already taken by ASE's built-in regex-based reader). Pass format='cextxyz' explicitly; auto-detection by extension is intentionally disabled to avoid clashing with the built-in.

Streaming output during optimization / MD

For long runs you don't want to reopen the file on every step. Use ExtXYZTrajectoryWriter — it opens the libc FILE* once and writes each frame through the C writer directly:

from ase.optimize import LBFGS
from ase_extxyz.io import ExtXYZTrajectoryWriter

with ExtXYZTrajectoryWriter('opt.xyz', atoms=atoms) as traj:
    opt = LBFGS(atoms)
    opt.attach(traj, interval=1)   # opt calls traj() per step
    opt.run(fmax=1e-3)

opt.attach(traj) works because the writer is callable; traj() is equivalent to traj.write(atoms) using the atoms captured at construction.

What this is not

ase-extxyz is a thin translation layer between extxyz.Frame (a plain-Python dict + numpy dataclass) and ase.Atoms. The fast C parser lives in the extxyz package; install that alone if you don't need ASE.

License

MIT — same as extxyz.