csv_to_arff.py

import csv
import argparse
import os
from configparser import ConfigParser
import helpers

def csv_to_arff(csv_filepath, attribute_filepath, output_arff_filepath, relation_name="dataset"):
    # Read attribute definitions from the attribute file
    with open(attribute_filepath, 'r') as attr_file:
        attribute_definitions = attr_file.read()

    # Open the CSV file and ARFF output file
    with open(csv_filepath, 'r') as csv_file, open(output_arff_filepath, 'w') as arff_file:
        csv_reader = csv.reader(csv_file)

        # Write the relation name and attribute definitions to the ARFF file
        arff_file.write(f"@relation {relation_name}\n\n")
        arff_file.write(attribute_definitions + "\n")

        # Extract the header (column names) from the CSV file
        header = next(csv_reader)

        # Write the data section header in the ARFF file
        arff_file.write("\n@data\n")

        # Write each row of the CSV file to the ARFF file
        for row in csv_reader:
            processed_row = []
            for value in row:
                if ' ' in value:  # If a value contains a space, enclose it in single quotes
                    processed_row.append(f"'{value}'")
                else:
                    processed_row.append(value)
            # Convert each row to a comma-separated string and write to ARFF
            arff_file.write(",".join(processed_row) + "\n")

    print(f"Conversion completed: {output_arff_filepath} created.")

def main():
    parser = argparse.ArgumentParser()
    parser.add_argument('indir', help="""
        A directory containing a config.ini file. All classifier results will be
        output to this file.
    """)
    args = parser.parse_args()

    config_file = helpers.assert_dir_contains_config(args.indir)

    # Load the configuration file
    config = ConfigParser()
    config.optionxform = str  # preserve case in config keys
    config.read(config_file)

    # Loop over each section in the config file and process the conversions
    for section in config.sections():
        if section.startswith('meta'):
            # Read the paths and details from the config file for this section
            csv_filepath = config[section].get('csv_filepath')
            attribute_filepath = config[section].get('attribute_filepath')
            output_arff_filepath = config[section].get('output_arff_filepath')
            relation_name = config[section].get('relation_name', fallback='dataset')

            # Check if paths exist
            if not os.path.exists(csv_filepath):
                raise FileNotFoundError(f"CSV file not found: {csv_filepath}")
            if not os.path.exists(attribute_filepath):
                raise FileNotFoundError(f"Attribute file not found: {attribute_filepath}")

            # Call the csv_to_arff function for this section
            csv_to_arff(csv_filepath, attribute_filepath, output_arff_filepath, relation_name)

if __name__ == "__main__":
    main()