convert_darchuk.py

# coding: utf-8
from __future__ import unicode_literals
from unicodecsv import reader, writer


class ConvertException(Exception):
    pass


def guess(tagline, mapping):
    res = []
    keys = mapping.keys()
    keys.sort(key=len, reverse=True)

    while (True):
        for r in keys:
            if tagline.startswith(r):
                res.append(mapping[r])
                tagline = tagline[len(r):].strip(" .")
                break
        else:
            raise ConvertException(tagline)

        if not tagline:
            break

    return list(set(res))

if __name__ == '__main__':
    mapping = {}

    with open("mapping_darchuk.csv", "r") as fp:
        r = reader(fp)

        for l in r:
            mapping[l[0].strip()] = l[1].strip()

    with open("tagset_darchuk_strange.csv", "r") as fp, \
            open("decoded_mapping.csv", "w") as fp_out:
        r = reader(fp)
        w = writer(fp_out)

        for l in r:
            normal = l[1].strip()
            try:
                cnvrtd = guess(normal, mapping)
            except ConvertException as e:
                print("%s: %s" % (normal, e))
                cnvrtd = []

            w.writerow(l + [":".join(cnvrtd)])