ndarrays2Carray.py - Exporting Numpy ndarrays to C arrays declared inside C files

ndarrays2Carray.py

"""
npy2Carray.py - A np.ndarray to C array converter

This script converts a numpy array to a C array and saves it in a file.

Author: Daniel Enériz
Version: 1.0

License: GPL-3.0

Example of use:

    If you declare the following array in Python:

        >>> import numpy as np
        >>> A = np.arange(24).reshape((2, 3, 4))
        >>> A
        array([[[ 0,  1,  2,  3],
                [ 4,  5,  6,  7],
                [ 8,  9, 10, 11]],

            [[12, 13, 14, 15],
                [16, 17, 18, 19],
                [20, 21, 22, 23]]])

    You can save it as a C array in a file with the following code:

        >>> npy2Carray(A, 'A', mode='w')

    There will be a file named 'arrays.c' in the current directory with the following content:

        const int A[2][3][4] = {
            {
                {0, 1, 2, 3},
                {4, 5, 6, 7},
                {8, 9, 10, 11}
            },
            {
                {12, 13, 14, 15},
                {16, 17, 18, 19},
                {20, 21, 22, 23}
            }
        };
"""

import numpy as np

def _write_array_values(array : np.ndarray, f, fmt : str = '{}', ident : int = 0):
    """Writes the values of a numpy array to a file. This auxiliary function is
    used by npy2Carray. Recursively calls itself to write the
    values of multidimensional arrays.
    
    Args:
        array (np.ndarray): Numpy array to write.
        f (file): File to write to.
        fmt (str, optional): Format to write the values. Defaults to '{}'.
        ident (int, optional): Number of identation levels. Defaults to 0.
    """

    # If shape is 1D, write the values in a single line
    if len(array.shape) == 1:

        # Indent the line
        for i in range(ident):
            f.write(' ')

        # Write the opening brace
        f.write('{')

        for i in range(array.shape[0]):
            f.write(fmt.format(array[i]))
            if i < array.shape[0] - 1:
                f.write(', ')
            else:
                f.write('}')
        
    # If shape has more than 1 dimension, call this function a number of times
    # equal to the first dimension
    else:
        for i in range(ident):
            f.write(' ')

        f.write('{\n')
        
        for i in range(array.shape[0]):
            _write_array_values(array[i], f, fmt, ident + 4)
            if i < array.shape[0] - 1:
                f.write(',\n')
            else:
                f.write('\n')
                for i in range(ident):
                    f.write(' ')
                f.write('}')

default_header = "// This file was automatically generated by npy2Carrays.py,\n" \
               + "// a script by Daniel Enériz that creates C arrays from numpy\n" \
               + "// arrays and saves them on a file." \

# Save data as a C header file defining a float array
def npy2Carray(array : np.ndarray, name : 'str',
    file_path : str = 'arrays.c', mode : str = 'a', fmt : str = '{}',
    ifndef : bool = True, header : str = default_header):
    """Saves a numpy array as a C array in a file
    
    Args:
        array (np.ndarray): Numpy array to save
        name (str): Name of the array. Must be a valid C identifier.
        file_path (str, optional): Path to the file. Defaults to 'arrays.c'.
        mode (str, optional): Mode to open the file, must be 'w' or 'a'.
            Defaults to 'a'.
        fmt (str, optional): Format to write the values. Defaults to '{}'.
            Must be a valid format for the type of the array.
        ifndef (bool, optional): If True, the file will be wrapped in an #ifndef
            statement to avoid multiple definitions. Defaults to True.
        header (str, optional): Header to add to the file. Defaults to
            `default_header`.
    """

    # Check if array is a np.ndarray
    if not isinstance(array, np.ndarray):
        raise TypeError('array must be a numpy array')

    # Check if name is a string and is a valid C identifier
    if not isinstance(name, str):
        raise TypeError('name must be a string')
    
    # Check if file_path is a string and is a valid file
    if not isinstance(file_path, str):
        raise TypeError('file_path must be a string')
    
    # Check if mode is a string and is valid
    if not isinstance(mode, str):
        raise TypeError('mode must be a string')
    if mode not in ['w', 'a']:
        raise ValueError('mode must be \'w\' or \'a\'')

    # Open the file
    f = open(file_path, mode)

    # Determine the type of the array from the dtype
    if 'float' in array.dtype.name:
        c_type = 'float'
    elif 'int' in array.dtype.name:
        c_type = 'int'
    else:
        raise TypeError('array must be an array of floats or integers')

    # Add the header
    f.write(header)

    # Add the #ifndef statement if required
    if ifndef:
        f.write('\n\n#ifndef {}\n#define {}\n'.format(name.upper(), name.upper()))
    
    # Write the array declaration
    f.write('\nconst {} {}'.format(c_type, name))
    
    # Iterate over the dimensions of the array to add the shape
    for dim in array.shape:
        f.write('[{}]'.format(dim))
    
    # Write the equal sign and the opening brace
    f.write(' = ')

    # Iterate over the dimensions of the array and start writing the values
    _write_array_values(array, f, '{}')

    # Add a semicolon and a new line
    f.write(';\n\n')

    # Write the number of dimensions of the array
    f.write('const unsigned int {}_ndim = {};\n'.format(name, array.ndim))
    
    # Write the shape of the array in ndim unsigned ints
    for i, dim in enumerate(array.shape):
        f.write('const unsigned int {}_shape_{} = {};\n'.format(name, i, dim))

    # Add the #endif statement if required
    if ifndef:
        f.write('\n#endif')