taguqoxora.py

#!/usr/bin/python

"""
Turn a common_crawl identifier into a normal-looking URL
"""

import sys
import re

def is_port(s):
    if re.findall(r"\A\d+\Z", s):
        if int(s) <= 65535:
            return True
    return False

for line in sys.stdin:
    url = line.rstrip()
    rest, schema =  url.rsplit(":", 1)

    try:
        domain, path = rest.split('/', 1)
    except ValueError:
        domain = rest
        path = ''

    try:
        maybe_path, maybe_port = path.rsplit(":", 1)
    except ValueError:
        port = None
    else:
        if is_port(maybe_port):
            path = maybe_path
            port = maybe_port
        else:
            port = None

    sys.stdout.write(schema + '://' + '.'.join(domain.split('.')[::-1]) + (':' + port if port is not None else '') + '/' + path + "\n")