Scan records in a Fortran unformatted file.

scan_fortran_unformatted.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
from __future__ import division

import itertools
import struct


def scan_fortran_unformatted(fname, marker_size=4):
    """Scan records of a Fortran unformatted file.

    In unformatted IO, Fortran typically writes the length of a record at the
    beginning and at the end of the record.  They are often called as markers.
    Most but not all Fortran runtimes use 4 bytes for the markers.

    Parameters
    ----------
    fname: str
        File name.
    marker_size: int, optional
        Marker size in bytes.  Default is 4.

    Returns
    -------
    record_struct: list of 4-tuple of ints
        Record structure.  Each tuple contains information about a record block:
        the record number, the position of the beginning of the record, the
        number of bytes in the record.
    """
    unpack = struct.unpack

    if not marker_size in (4, 8):
        raise ValueError('marker_size must be either 4 or 8.')
    fmt = 'i' if marker_size == 4 else 'q'

    with open(fname, 'rb') as fh:
        record_struct = []
        for rec_num in itertools.count(1):
            buf = fh.read(marker_size)
            if len(buf) < marker_size:
                break
            # Marker at the beginning of the block.
            marker_beg, = unpack(fmt, buf)
            # Position of the beginning of the record.
            pos_beg = fh.tell()
            # Seek the end of the block.
            fh.seek(marker_beg, 1)
            # Marker at the end of the block.
            marker_end, = unpack(fmt, fh.read(marker_size))
            # Both lengths must be the same.
            assert marker_beg == marker_end

            record_struct.append((rec_num, pos_beg, marker_beg))

    return record_struct