LAMMPS DumpReader does not return the correct coordinates for unwrapped trajectories.

[hmacdope]

Expected behavior

Unscaled coordinates written with LAMMPS in a lammpstrj dump format should be parsed correctly.

Actual behavior

The DumpReader Reader does not check what convention coordinates are written in. (xs,ys,zs) vs (xu,yu,zu). The transform distances.transform_StoR(ts.positions, ts.dimensions) is also called in _read_next_timestep regardless of whether the coordinates are scaled or not, resulting in insanely large coordinates if unscaled (xu,yu,zu) coordinates are input.

Code to reproduce the behavior

import MDAnalysis as mda
import numpy as np
from numpy import testing

n_atoms = 1536
# do it by reading the file by hand
def parse_lammpstrj(file, n_atoms):
    with open(file, "r") as f:
        lines = f.readlines()
        atomsegs = []
        for i, line in enumerate(lines):
            if "ITEM: ATOMS" in line:
                atomseg = lines[i+1:i+n_atoms+1]
                atomsegs.append(atomseg)
        coords = np.empty([len(atomsegs),n_atoms,3])
        for frame in range(len(atomsegs)):
            atom_frame = atomsegs[frame]
            for atom in range(len(atom_frame)):
                crd_line = atom_frame[atom]
                idx, _, xu, yu, zu = crd_line.split()
                xu = float(xu)
                yu = float(yu)
                zu = float(zu)
                coords[frame, atom,:] = [xu,yu,zu]
        return coords

coords = parse_lammpstrj("wat_unwraped.lammpsdump",n_atoms)

# do it with MDA
u = mda.Universe("wat_unwraped.lammpsdump")
coords_mda = np.empty([len(u.trajectory),n_atoms,3])
for i,ts in enumerate(u.trajectory):
    coords_mda[i,:,:] = ts.positions

np.testing.assert_allclose(coords_mda,coords)
....

Current version of MDAnalysis

• Which version are you using: 2.0.0-dev0
• Which version of Python: 3.7.5
• Which operating system: Ubuntu Eoan Ermine

[hmacdope]

I should note that this is associated with the need to unhardcode various paths in in the LAMMPS reader.
See #2333 for example.