Source code for ase.io.rmc6f

import re
import time
import numpy as np

from ase.atoms import Atoms
from ase.utils import reader, writer
from ase.cell import Cell

__all__ = ['read_rmc6f', 'write_rmc6f']

ncols2style = {9: 'no_labels',
               10: 'labels',
               11: 'magnetic'}


def _read_construct_regex(lines):
    """
    Utility for constructing  regular expressions used by reader.
    """
    lines = [l.strip() for l in lines]
    lines_re = '|'.join(lines)
    lines_re = lines_re.replace(' ', r'\s+')
    lines_re = lines_re.replace('(', r'\(')
    lines_re = lines_re.replace(')', r'\)')
    return '({})'.format(lines_re)


def _read_line_of_atoms_section(fields):
    """
    Process `fields` line of Atoms section in rmc6f file and output extracted
    info as atom id (key) and list of properties for Atoms object (values).

    Parameters
    ----------
    fields: list[str]
        List of columns from line in rmc6f file.


    Returns
    ------
    atom_id: int
        Atom ID
    properties: list[str|float]
        List of Atoms properties based on rmc6f style.
        Basically, have 1) element and fractional coordinates for 'labels'
        or 'no_labels' style and 2) same for 'magnetic' style except adds
        the spin.
        Examples for 1) 'labels' or 'no_labels' styles or 2) 'magnetic' style:
            1) [element, xf, yf, zf]
            2) [element, xf, yf, zf, spin]
    """
    # Atom id
    atom_id = int(fields[0])

    # Get style used for rmc6f from file based on number of columns
    ncols = len(fields)
    style = ncols2style[ncols]

    # Create the position dictionary
    properties = list()
    element = str(fields[1])
    if style == 'no_labels':
        # id element xf yf zf ref_num ucell_x ucell_y ucell_z
        xf = float(fields[2])
        yf = float(fields[3])
        zf = float(fields[4])
        properties = [element, xf, yf, zf]

    elif style == 'labels':
        # id element label xf yf zf ref_num ucell_x ucell_y ucell_z
        xf = float(fields[3])
        yf = float(fields[4])
        zf = float(fields[5])
        properties = [element, xf, yf, zf]

    elif style == 'magnetic':
        # id element xf yf zf ref_num ucell_x ucell_y ucell_z M: spin
        xf = float(fields[2])
        yf = float(fields[3])
        zf = float(fields[4])
        spin = float(fields[10].strip("M:"))
        properties = [element, xf, yf, zf, spin]
    else:
        raise Exception("Unsupported style for parsing rmc6f file format.")

    return atom_id, properties


def _read_process_rmc6f_lines_to_pos_and_cell(lines):
    """
    Processes the lines of rmc6f file to atom position dictionary and cell

    Parameters
    ----------
    lines: list[str]
        List of lines from rmc6f file.

    Returns
    ------
    pos : dict{int:list[str|float]}
        Dict for each atom id and Atoms properties based on rmc6f style.
        Basically, have 1) element and fractional coordinates for 'labels'
        or 'no_labels' style and 2) same for 'magnetic' style except adds
        the spin.
        Examples for 1) 'labels' or 'no_labels' styles or 2) 'magnetic' style:
            1) pos[aid] = [element, xf, yf, zf]
            2) pos[aid] = [element, xf, yf, zf, spin]
    cell: Cell object
        The ASE Cell object created from cell parameters read from the 'Cell'
        section of rmc6f file.
    """

    # Inititalize output pos dictionary
    pos = {}

    # Defined the header an section lines we process
    header_lines = [
        "Number of atoms:",
        "Supercell dimensions:",
        "Cell (Ang/deg):",
        "Lattice vectors (Ang):"]
    sections = ["Atoms"]

    # Construct header and sections regex
    header_lines_re = _read_construct_regex(header_lines)
    sections_re = _read_construct_regex(sections)

    section = None
    header = True

    # Remove any lines that are blank
    lines = [line for line in lines if line != '']

    # Process each line of rmc6f file
    pos = {}
    for line in lines:

        # check if in a section
        m = re.match(sections_re, line)
        if m is not None:
            section = m.group(0).strip()
            header = False
            continue

        # header section
        if header:
            field = None
            val = None

            # Regex that matches whitespace-separated floats
            float_list_re = r'\s+(\d[\d|\s\.]+[\d|\.])'
            m = re.search(header_lines_re + float_list_re, line)
            if m is not None:
                field = m.group(1)
                val = m.group(2)

            if field is not None and val is not None:

                if field == "Number of atoms:":
                    pass
                    """
                    NOTE: Can just capture via number of atoms ingested.
                          Maybe use in future for a check.
                    code: natoms = int(val)
                    """

                if field.startswith('Supercell'):
                    pass
                    """
                    NOTE: wrapping back down to unit cell is not
                          necessarily needed for ASE object.

                    code: supercell = [int(x) for x in val.split()]
                    """

                if field.startswith('Cell'):
                    cellpar = [float(x) for x in val.split()]
                    cell = Cell.fromcellpar(cellpar)

                if field.startswith('Lattice'):
                    pass
                    """
                    NOTE: Have questions about RMC fractionalization matrix for
                          conversion in data2config vs. the ASE matrix.
                          Currently, just support the Cell section.
                    """

        # main body section
        if section is not None:
            if section == 'Atoms':
                atom_id, atom_props = _read_line_of_atoms_section(line.split())
                pos[atom_id] = atom_props

    return pos, cell


def _write_output_column_format(columns, arrays):
    """
    Helper function to build output for data columns in rmc6f format

    Parameters
    ----------
    columns: list[str]
        List of keys in arrays. Will be columns in the output file.
    arrays: dict{str:np.array}
        Dict with arrays for each column of rmc6f file that are
        property of Atoms object.

    Returns
    ------
    property_ncols : list[int]
        Number of columns for each property.
    dtype_obj: np.dtype
        Data type object for the columns.
    formats_as_str: str
        The format for printing the columns.

    """
    fmt_map = {'d': ('R', '%14.6f '),
               'f': ('R', '%14.6f '),
               'i': ('I', '%8d '),
               'O': ('S', '%s'),
               'S': ('S', '%s'),
               'U': ('S', '%s'),
               'b': ('L', ' %.1s ')}

    property_types = []
    property_ncols = []
    dtypes = []
    formats = []

    # Take each column and setup formatting vectors
    for column in columns:
        array = arrays[column]
        dtype = array.dtype

        property_type, fmt = fmt_map[dtype.kind]
        property_types.append(property_type)

        # Flags for 1d vectors
        is_1d = len(array.shape) == 1
        is_1d_as_2d = len(array.shape) == 2 and array.shape[1] == 1

        # Construct the list of key pairs of column with dtype
        if (is_1d or is_1d_as_2d):
            ncol = 1
            dtypes.append((column, dtype))
        else:
            ncol = array.shape[1]
            for c in range(ncol):
                dtypes.append((column + str(c), dtype))

        # Add format and number of columns for this property to output array
        formats.extend([fmt] * ncol)
        property_ncols.append(ncol)

    # Prepare outputs to correct data structure
    dtype_obj = np.dtype(dtypes)
    formats_as_str = ''.join(formats) + '\n'

    return property_ncols, dtype_obj, formats_as_str


def _write_output(filename, header_lines, data, fmt, order=None):
    """
    Helper function to write information to the filename

    Parameters
    ----------
    filename : file|str
        A file like object or filename
    header_lines : list[str]
        Header section of output rmc6f file
    data: np.array[len(atoms)]
        Array for the Atoms section to write to file. Has
        the columns that need to be written on each row
    fmt: str
        The format string to use for writing each column in
        the rows of data.
    order : list[str]
        If not None, gives a list of atom types for the order
        to write out each.
    """
    fd = filename

    # Write header section
    for line in header_lines:
        fd.write("%s \n" % line)

    # If specifying the order, fix the atom id and write to file
    natoms = data.shape[0]
    if order is not None:
        new_id = 0
        for atype in order:
            for i in range(natoms):
                if atype == data[i][1]:
                    new_id += 1
                    data[i][0] = new_id
                    fd.write(fmt % tuple(data[i]))
    # ...just write rows to file
    else:
        for i in range(natoms):
            fd.write(fmt % tuple(data[i]))


[docs]@reader def read_rmc6f(filename, atom_type_map=None): """ Parse a RMCProfile rmc6f file into ASE Atoms object Parameters ---------- filename : file|str A file like object or filename. atom_type_map: dict{str:str} Map of atom types for conversions. Mainly used if there is an atom type in the file that is not supported by ASE but want to map to a supported atom type instead. Example to map deuterium to hydrogen: atom_type_map = { 'D': 'H' } Returns ------ structure : Atoms The Atoms object read in from the rmc6f file. """ fd = filename lines = fd.readlines() # Process the rmc6f file to extract positions and cell pos, cell = _read_process_rmc6f_lines_to_pos_and_cell(lines) # create an atom type map if one does not exist from unique atomic symbols if atom_type_map is None: symbols = [atom[0] for atom in pos.values()] atom_type_map = {atype: atype for atype in symbols} # Use map of tmp symbol to actual symbol for atom in pos.values(): atom[0] = atom_type_map[atom[0]] # create Atoms from read-in data symbols = [] scaled_positions = [] spin = None magmoms = [] for atom in pos.values(): if len(atom) == 4: element, x, y, z = atom else: element, x, y, z, spin = atom element = atom_type_map[element] symbols.append(element) scaled_positions.append([x, y, z]) if spin is not None: magmoms.append(spin) atoms = Atoms(scaled_positions=scaled_positions, symbols=symbols, cell=cell, magmoms=magmoms, pbc=[True, True, True]) return atoms
[docs]@writer def write_rmc6f(filename, atoms, order=None, atom_type_map=None): """ Write output in rmc6f format - RMCProfile v6 fractional coordinates Parameters ---------- filename : file|str A file like object or filename. atoms: Atoms object The Atoms object to be written. order : list[str] If not None, gives a list of atom types for the order to write out each. atom_type_map: dict{str:str} Map of atom types for conversions. Mainly used if there is an atom type in the Atoms object that is a placeholder for a different atom type. This is used when the atom type is not supported by ASE but is in RMCProfile. Example to map hydrogen to deuterium: atom_type_map = { 'H': 'D' } """ # get atom types and how many of each (and set to order if passed) atom_types = set(atoms.symbols) if order is not None: if set(atom_types) != set(order): raise Exception("The order is not a set of the atom types.") atom_types = order atom_count_dict = atoms.symbols.formula.count() natom_types = [str(atom_count_dict[atom_type]) for atom_type in atom_types] # create an atom type map if one does not exist from unique atomic symbols if atom_type_map is None: symbols = set(np.array(atoms.symbols)) atom_type_map = {atype: atype for atype in symbols} # HEADER SECTION # get type and number of each atom type atom_types_list = [atom_type_map[atype] for atype in atom_types] atom_types_present = ' '.join(atom_types_list) natom_types_present = ' '.join(natom_types) header_lines = [ "(Version 6f format configuration file)", "(Generated by ASE - Atomic Simulation Environment https://wiki.fysik.dtu.dk/ase/ )", # noqa: E501 "Metadata date: " + time.strftime('%d-%m-%Y'), "Number of types of atoms: {} ".format(len(atom_types)), "Atom types present: {}".format(atom_types_present), "Number of each atom type: {}".format(natom_types_present), "Number of moves generated: 0", "Number of moves tried: 0", "Number of moves accepted: 0", "Number of prior configuration saves: 0", "Number of atoms: {}".format(len(atoms)), "Supercell dimensions: 1 1 1"] # magnetic moments if atoms.has('magmoms'): spin_str = "Number of spins: {}" spin_line = spin_str.format(len(atoms.get_initial_magnetic_moments())) header_lines.extend([spin_line]) density_str = "Number density (Ang^-3): {}" density_line = density_str.format(len(atoms) / atoms.get_volume()) cell_angles = [str(x) for x in atoms.cell.cellpar()] cell_line = "Cell (Ang/deg): " + ' '.join(cell_angles) header_lines.extend([density_line, cell_line]) # get lattice vectors from cell lengths and angles # NOTE: RMCProfile uses a different convention for the fractionalization # matrix cell_parameters = atoms.cell.cellpar() cell = Cell.fromcellpar(cell_parameters).T x_line = ' '.join(['{:12.6f}'.format(i) for i in cell[0]]) y_line = ' '.join(['{:12.6f}'.format(i) for i in cell[1]]) z_line = ' '.join(['{:12.6f}'.format(i) for i in cell[2]]) lat_lines = ["Lattice vectors (Ang):", x_line, y_line, z_line] header_lines.extend(lat_lines) header_lines.extend(['Atoms:']) # ATOMS SECTION # create columns of data for atoms (fr_cols) fr_cols = ['id', 'symbols', 'scaled_positions', 'ref_num', 'ref_cell'] if atoms.has('magmoms'): fr_cols.extend('magmom') # Collect data to be written out natoms = len(atoms) arrays = {} arrays['id'] = np.array(range(1, natoms + 1, 1), int) arrays['symbols'] = np.array(atoms.symbols) arrays['ref_num'] = np.zeros(natoms, int) arrays['ref_cell'] = np.zeros((natoms, 3), int) arrays['scaled_positions'] = np.array(atoms.get_scaled_positions()) # get formatting for writing output based on atom arrays ncols, dtype_obj, fmt = _write_output_column_format(fr_cols, arrays) # Pack fr_cols into record array data = np.zeros(natoms, dtype_obj) for column, ncol in zip(fr_cols, ncols): value = arrays[column] if ncol == 1: data[column] = np.squeeze(value) else: for c in range(ncol): data[column + str(c)] = value[:, c] # Use map of tmp symbol to actual symbol for i in range(natoms): data[i][1] = atom_type_map[data[i][1]] # Write the output _write_output(filename, header_lines, data, fmt, order=order)