reformat_iate_col.py 2.74 KB
import argparse
import os


REMOVE_YEAR = True


def main():
    args = parse_arguments()
    if not args.input or not args.output:
        print('Error: Input and output must be selected!')
    reformat_iate_col_by_year(args.input, args.output)


def parse_arguments():
    parser = argparse.ArgumentParser(description='Reformat IATE column.')
    parser.add_argument('-o', '--output', help='output directory')
    required_arguments = parser.add_argument_group('required arguments')
    required_arguments.add_argument('-i', '--input', help='corpora root directory', required=True)
    return parser.parse_args()


def reformat_iate_col_by_year(root_directory, out_corpora_directory):
    for root, dirs, files in os.walk(root_directory):
        for filename in files:
            if filename.endswith('.conllup') or filename.endswith('.conllu'):
                src = os.path.join(root, filename)
                year = get_year(src)
                year_path = os.path.join(out_corpora_directory, year)
                if REMOVE_YEAR:
                    year_path = out_corpora_directory
                os.makedirs(year_path, exist_ok=True)
                dst = os.path.join(year_path, filename)
                reformat_iate(src, dst)


def get_year(filepath):
    with open(filepath, 'r') as conllup_file:
        for line in conllup_file:
            line = line.strip()
            if is_segment(line):
                continue
            elif is_metadata(line):
                name, value = get_metadata(line)
                if name == 'PublicationDate':
                    return value.split('-')[0]
    return '0'


def is_segment(line):
    if line and line[0].isdigit():
        return True
    return False


def is_metadata(line):
    if line.startswith('#'):
        return True
    return False


def get_metadata(line):
    name_value_pair = line.split('=', 1)
    name = name_value_pair[0].lstrip('#').strip()
    value = name_value_pair[1].strip()
    return name, value


def reformat_iate(src, dst):
    cleaned_lines = []

    with open(src, 'r') as conllup_file:
        for line in conllup_file:
            line = line.strip()
            if is_segment(line):
                columns = line.split('\t')
                iate_col = columns[12]
                iate_terms = iate_col.split(';')
                reformatted_terms = []
                for t in iate_terms:
                    reformatted_terms.append(t.split('-')[0])
                new_iate_col = ','.join(reformatted_terms)
                columns[12] = new_iate_col
                cleaned_lines.append('\t'.join(columns))
            else:
                cleaned_lines.append(line)

    with open(dst, 'w') as dst_file:
        dst_file.write('\n'.join(cleaned_lines))


if __name__ == '__main__':
    main()