Source code for ase.io.trajectory

"""Trajectory"""
import contextlib
import io
import warnings
from typing import Tuple

import numpy as np

from ase import __version__
from ase.atoms import Atoms
from ase.calculators.calculator import PropertyNotImplementedError
from ase.calculators.singlepoint import SinglePointCalculator, all_properties
from ase.io.formats import is_compressed
from ase.io.jsonio import decode, encode
from ase.io.pickletrajectory import PickleTrajectory
from ase.parallel import world
from ase.utils import tokenize_version

__all__ = ['Trajectory', 'PickleTrajectory']


[docs]def Trajectory(filename, mode='r', atoms=None, properties=None, master=None, comm=world): """A Trajectory can be created in read, write or append mode. Parameters: filename: str The name of the file. Traditionally ends in .traj. mode: str The mode. 'r' is read mode, the file should already exist, and no atoms argument should be specified. 'w' is write mode. The atoms argument specifies the Atoms object to be written to the file, if not given it must instead be given as an argument to the write() method. 'a' is append mode. It acts as write mode, except that data is appended to a preexisting file. atoms: Atoms object The Atoms object to be written in write or append mode. properties: list of str If specified, these calculator properties are saved in the trajectory. If not specified, all supported quantities are saved. Possible values: energy, forces, stress, dipole, charges, magmom and magmoms. master: bool Controls which process does the actual writing. The default is that process number 0 does this. If this argument is given, processes where it is True will write. The atoms, properties and master arguments are ignores in read mode. """ if mode == 'r': return TrajectoryReader(filename) return TrajectoryWriter(filename, mode, atoms, properties, master=master, comm=comm)
[docs]class TrajectoryWriter: """Writes Atoms objects to a .traj file.""" def __init__(self, filename, mode='w', atoms=None, properties=None, master=None, comm=world): """A Trajectory writer, in write or append mode. Parameters: filename: str The name of the file. Traditionally ends in .traj. mode: str The mode. 'r' is read mode, the file should already exist, and no atoms argument should be specified. 'w' is write mode. The atoms argument specifies the Atoms object to be written to the file, if not given it must instead be given as an argument to the write() method. 'a' is append mode. It acts as write mode, except that data is appended to a preexisting file. atoms: Atoms object The Atoms object to be written in write or append mode. properties: list of str If specified, these calculator properties are saved in the trajectory. If not specified, all supported quantities are saved. Possible values: energy, forces, stress, dipole, charges, magmom and magmoms. master: bool Controls which process does the actual writing. The default is that process number 0 does this. If this argument is given, processes where it is True will write. comm: MPI communicator MPI communicator for this trajectory writer, by default world. Passing a different communicator facilitates writing of different trajectories on different MPI ranks. """ if master is None: master = comm.rank == 0 self.filename = filename self.mode = mode self.atoms = atoms self.properties = properties self.master = master self.comm = comm self.description = {} self.header_data = None self.multiple_headers = False self._open(filename, mode) def __enter__(self): return self def __exit__(self, exc_type, exc_value, tb): self.close() def set_description(self, description): self.description.update(description) def _open(self, filename, mode): import ase.io.ulm as ulm if mode not in 'aw': raise ValueError('mode must be "w" or "a".') if self.master: self.backend = ulm.open(filename, mode, tag='ASE-Trajectory') if len(self.backend) > 0 and mode == 'a': with Trajectory(filename) as traj: atoms = traj[0] self.header_data = get_header_data(atoms) else: self.backend = ulm.DummyWriter()
[docs] def write(self, atoms=None, **kwargs): """Write the atoms to the file. If the atoms argument is not given, the atoms object specified when creating the trajectory object is used. Use keyword arguments to add extra properties:: writer.write(atoms, energy=117, dipole=[0, 0, 1.0]) """ if atoms is None: atoms = self.atoms for image in atoms.iterimages(): self._write_atoms(image, **kwargs)
def _write_atoms(self, atoms, **kwargs): b = self.backend if self.header_data is None: b.write(version=1, ase_version=__version__) if self.description: b.write(description=self.description) # Atomic numbers and periodic boundary conditions are written # in the header in the beginning. # # If an image later on has other numbers/pbc, we write a new # header. All subsequent images will then have their own header # whether or not their numbers/pbc change. self.header_data = get_header_data(atoms) write_header = True else: if not self.multiple_headers: header_data = get_header_data(atoms) self.multiple_headers = not headers_equal(self.header_data, header_data) write_header = self.multiple_headers write_atoms(b, atoms, write_header=write_header) calc = atoms.calc if calc is None and len(kwargs) > 0: calc = SinglePointCalculator(atoms) if calc is not None: if not hasattr(calc, 'get_property'): calc = OldCalculatorWrapper(calc) c = b.child('calculator') c.write(name=calc.name) if hasattr(calc, 'todict'): c.write(parameters=calc.todict()) for prop in all_properties: if prop in kwargs: x = kwargs[prop] else: if self.properties is not None: if prop in self.properties: x = calc.get_property(prop, atoms) else: x = None else: try: x = calc.get_property(prop, atoms, allow_calculation=False) except (PropertyNotImplementedError, KeyError): # KeyError is needed for Jacapo. # XXX We can perhaps remove this. x = None if x is not None: if prop in ['stress', 'dipole']: x = x.tolist() c.write(prop, x) info = {} for key, value in atoms.info.items(): try: encode(value) except TypeError: warnings.warn(f'Skipping "{key}" info.') else: info[key] = value if info: b.write(info=info) b.sync()
[docs] def close(self): """Close the trajectory file.""" self.backend.close()
def __len__(self): return self.comm.sum_scalar(len(self.backend))
[docs]class TrajectoryReader: """Reads Atoms objects from a .traj file.""" def __init__(self, filename): """A Trajectory in read mode. The filename traditionally ends in .traj. """ self.filename = filename self.numbers = None self.pbc = None self.masses = None self._open(filename) def __enter__(self): return self def __exit__(self, exc_type, exc_value, tb): self.close() def _open(self, filename): import ase.io.ulm as ulm self.backend = ulm.open(filename, 'r') self._read_header() def _read_header(self): b = self.backend if b.get_tag() != 'ASE-Trajectory': raise OSError('This is not a trajectory file!') if len(b) > 0: self.pbc = b.pbc self.numbers = b.numbers self.masses = b.get('masses') self.constraints = b.get('constraints', '[]') self.description = b.get('description') self.version = b.version self.ase_version = b.get('ase_version')
[docs] def close(self): """Close the trajectory file.""" self.backend.close()
def __getitem__(self, i=-1): if isinstance(i, slice): return SlicedTrajectory(self, i) b = self.backend[i] if 'numbers' in b: # numbers and other header info was written alongside the image: atoms = read_atoms(b, traj=self) else: # header info was not written because they are the same: atoms = read_atoms(b, header=[self.pbc, self.numbers, self.masses, self.constraints], traj=self) if 'calculator' in b: results = {} implemented_properties = [] c = b.calculator for prop in all_properties: if prop in c: results[prop] = c.get(prop) implemented_properties.append(prop) calc = SinglePointCalculator(atoms, **results) calc.name = b.calculator.name calc.implemented_properties = implemented_properties if 'parameters' in c: calc.parameters.update(c.parameters) atoms.calc = calc return atoms def __len__(self): return len(self.backend) def __iter__(self): for i in range(len(self)): yield self[i]
class SlicedTrajectory: """Wrapper to return a slice from a trajectory without loading from disk. Initialize with a trajectory (in read mode) and the desired slice object.""" def __init__(self, trajectory, sliced): self.trajectory = trajectory self.map = range(len(self.trajectory))[sliced] def __getitem__(self, i): if isinstance(i, slice): # Map directly to the original traj, not recursively. traj = SlicedTrajectory(self.trajectory, slice(0, None)) traj.map = self.map[i] return traj return self.trajectory[self.map[i]] def __len__(self): return len(self.map) def get_header_data(atoms): return {'pbc': atoms.pbc.copy(), 'numbers': atoms.get_atomic_numbers(), 'masses': atoms.get_masses() if atoms.has('masses') else None, 'constraints': list(atoms.constraints)} def headers_equal(headers1, headers2): assert len(headers1) == len(headers2) eq = True for key in headers1: eq &= np.array_equal(headers1[key], headers2[key]) return eq class VersionTooOldError(Exception): pass def read_atoms(backend, header: Tuple = None, traj: TrajectoryReader = None, _try_except: bool = True) -> Atoms: from ase.constraints import dict2constraint if _try_except: try: return read_atoms(backend, header, traj, False) except Exception as ex: if (traj is not None and tokenize_version(__version__) < tokenize_version(traj.ase_version)): msg = ('You are trying to read a trajectory file written ' f'by ASE-{traj.ase_version} from ASE-{__version__}. ' 'It might help to update your ASE') raise VersionTooOldError(msg) from ex else: raise b = backend if header: pbc, numbers, masses, constraints = header else: pbc = b.pbc numbers = b.numbers masses = b.get('masses') constraints = b.get('constraints', '[]') atoms = Atoms(positions=b.positions, numbers=numbers, cell=b.cell, masses=masses, pbc=pbc, info=b.get('info'), constraint=[dict2constraint(d) for d in decode(constraints)], momenta=b.get('momenta'), magmoms=b.get('magmoms'), charges=b.get('charges'), tags=b.get('tags')) return atoms def write_atoms(backend, atoms, write_header=True): b = backend if write_header: b.write(pbc=atoms.pbc.tolist(), numbers=atoms.numbers) if atoms.constraints: if all(hasattr(c, 'todict') for c in atoms.constraints): b.write(constraints=encode(atoms.constraints)) if atoms.has('masses'): b.write(masses=atoms.get_masses()) b.write(positions=atoms.get_positions(), cell=atoms.get_cell().tolist()) if atoms.has('tags'): b.write(tags=atoms.get_tags()) if atoms.has('momenta'): b.write(momenta=atoms.get_momenta()) if atoms.has('initial_magmoms'): b.write(magmoms=atoms.get_initial_magnetic_moments()) if atoms.has('initial_charges'): b.write(charges=atoms.get_initial_charges())
[docs]def read_traj(fd, index): trj = TrajectoryReader(fd) for i in range(*index.indices(len(trj))): yield trj[i]
@contextlib.contextmanager def defer_compression(fd): """Defer the file compression until all the configurations are read.""" # We do this because the trajectory and compressed-file # internals do not play well together. # Be advised not to defer compression of very long trajectories # as they use a lot of memory. if is_compressed(fd): with io.BytesIO() as bytes_io: try: # write the uncompressed data to the buffer yield bytes_io finally: # write the buffered data to the compressed file bytes_io.seek(0) fd.write(bytes_io.read()) else: yield fd
[docs]def write_traj(fd, images): """Write image(s) to trajectory.""" if isinstance(images, Atoms): images = [images] with defer_compression(fd) as fd_uncompressed: trj = TrajectoryWriter(fd_uncompressed) for atoms in images: trj.write(atoms)
class OldCalculatorWrapper: def __init__(self, calc): self.calc = calc try: self.name = calc.name except AttributeError: self.name = calc.__class__.__name__.lower() def get_property(self, prop, atoms, allow_calculation=True): try: if (not allow_calculation and self.calc.calculation_required(atoms, [prop])): return None except AttributeError: pass method = 'get_' + {'energy': 'potential_energy', 'magmom': 'magnetic_moment', 'magmoms': 'magnetic_moments', 'dipole': 'dipole_moment'}.get(prop, prop) try: result = getattr(self.calc, method)(atoms) except AttributeError: raise PropertyNotImplementedError return result def convert(name): import os t = TrajectoryWriter(name + '.new') for atoms in PickleTrajectory(name, _warn=False): t.write(atoms) t.close() os.rename(name, name + '.old') os.rename(name + '.new', name) def main(): import optparse parser = optparse.OptionParser(usage='python -m ase.io.trajectory ' 'a1.traj [a2.traj ...]', description='Convert old trajectory ' 'file(s) to new format. ' 'The old file is kept as a1.traj.old.') opts, args = parser.parse_args() for name in args: convert(name) if __name__ == '__main__': main()