Cython fortran-file

November 9, 2021 ยท View on GitHub

Build and Test PyPI Conda-forge

This package is a fast and easy reader for record-based binary format, as written by Fortran.

Installation

This packages requires a compiler to be installed on you system. If you're using Linux or MacOSX you should be fine. If you're using Windows, please get in touch with me or open an issue.

pip install cython_fortran_file

That's all!

Using it

Here is a simple demonstration:

from cython_fortran_file import FortranFile as FF

f = FF("/path/to/my/fortran/file.dat", mode="r")

# Skip 5 records
f.skip(5)

# Read one record (an array of `double`)
data = f.read_vector("d")
assert data.dtype == np.float64

# Read one record (an array of signed `int`)
data = f.read_vector("i")
assert data.dtype == np.int32

# Read one *single* int. This will fail if there is more to read!
v = f.read_int()

# f _looks_ like a regular file descriptor
f.tell()
f.seek(0)
f.close()

It also supports with statement so that you can't forget to close the file handler

with FF("/path/to/my/fortran/file.dat") as f:
    f.skip(5)

    data = f.read_vector("d")  # Read a vector of doubles (as usual)

Data types

The argument of read_vector follows the convention of the python struct package (C convention). See https://docs.python.org/3.5/library/struct.html#format-characters for the full list. Here is a shorter version:

ArgumentFortran typeC typeNumpy dtype
dreal(8)doublenp.float64
freal(4)floatnp.float32
iintegerintnp.int32
linteger(8)longnp.int64