Strip down a user-provided address to its essentials

geocodeHelper.py

#!python3
"""
@author Philip Kahn
@license MIT
"""


def getBarebonesAddress(dirtyAddress):
    """
    Will take most forms of addresses, and return an address of form

    [Number] [Street] [Zip]

    removing city (encoded in zip), state (encoded in zip), and suite/apt/etc (irrelevant to geocode)

    If there's no street address, returns only a zip code.
    """
    # Strip it and remove any duplications of zips, keeping the last from the canonical field
    dirtyAddress = re.sub(r"^(.*?)\d{5} (\d{5})$", r"\1 \2", dirtyAddress.strip(), 0, re.IGNORECASE | re.MULTILINE)
    # Split addresses of form "2435 Venice Dr E, South Lake Tahoe, CA 96150" and take only the leading part
    workingAddress = dirtyAddress.split(",")[0].strip()
    if dirtyAddress[-5:] != workingAddress[-5:]:
        # If we've truncated the zip code, add it back on
        workingAddress += " " + dirtyAddress[-5:]
    #############################
    # Run through regex filter:
    #
    # 1. Grab a numerical part, and any simple prefixes
    # 2. If it exists, grab a cardinal direction
    # 3. Grab the main part of the name, and any obvious suffixes. While the suffix specification isn't always needed, in some cases it helps
    # 4. Street name cardinal suffixes
    # 5. Check for suites, apts, etc and grab them now so they can be removed
    # 6. Grab trailing character data, for example "Berkeley CA", before a zip
    # 7. Match on a whole zip, return only the first five
    #############################
    bareAddress = re.sub(r"^([0-9\"\'a-z ,\.\/\\\-]+?(?: +(?:n|e|w|s){1,2})?(?: +(?!suite|#|suite#|suite #|ste|ste #|unit|unit #|apt|apt #)(?![a-z]{2}[ ,]+)[0-9\"\'a-z,\.\/\\\-]+?)*(?: +(?:ave|avenue|dr|drive|pkwy|parkway|st|street|rd|road|hwy|highway|pk|park|way|pl|place|ln|lane|cir|circle|ct|court)\.?(?: +(?:n|e|w|s){1,2})?)*)[ ,;]+?(?:(?:;, )*(?:suite|#|suite#|suite #|ste|ste #|unit|unit #|apt|apt #)\.? *[\S]+)*(?: *\d+ )*?(?:[ ,]*[a-z]{3,})*?[ ,]*(?:[ ,]+[a-z]{2}[ ,]+)?([0-9]{5})(?:-[0-9]{4})?$", r"\1 \2", workingAddress, 0, re.IGNORECASE | re.MULTILINE)
    return bareAddress