OSM-XML duplicate address detector

dupeaddresses.py

#!/usr/bin/env python3
#
# Detects duplicate address specifications (addr:street, addr:housenumber
# and possibly addr:unit) in an OSM XML file.
#
# Released into the public domain.

from xml.dom.minidom import parseString as domParse
import xml.dom as dom

def _fishOutAddressTags(nd):
	ret = {}
	for subnd in nd.childNodes:
		if subnd.nodeType != subnd.ELEMENT_NODE:
			continue
		if subnd.tagName == "tag":
			k = subnd.getAttribute("k")
			if k.startswith("addr:"):
				ret[k] = subnd.getAttribute("v")
	return ret

def fishOutAddressData(xmlstring):
	ret = {}
	d = domParse(xmlstring)

	for nd in d.documentElement.childNodes:
		if nd.nodeType != nd.ELEMENT_NODE:
			continue
		if nd.tagName == "node":
			myid = "n" + nd.getAttribute("id")
		elif nd.tagName == "way":
			myid = "w" + nd.getAttribute("id")
		elif nd.tagName == "relation":
			myid = "r" + nd.getAttribute("id")
		else:
			continue

		if nd.getAttribute("action") == "delete":
			# ne'er mind
			continue

		myaddrs = _fishOutAddressTags(nd)
		if len(myaddrs) > 0:
			if myid in ret:
				raise ValueError("duplicate ID " + myid)
			ret[myid] = myaddrs

	return ret

def pairwise(iterable, f):
	l = list(iterable)
	for na in range(len(l)):
		for nb in range(na+1, len(l)):
			f(l[na], l[nb])

def prettyAddr(addr):
	ret = "{0} {1}".format(addr["addr:street"], addr["addr:housenumber"])
	if "addr:unit" in addr:
		ret += "/" + addr["addr:unit"]
	return ret

def validateAddr(eid, addr):
	if "addr:street" not in addr:
		raise ValueError(eid + " has no street")
	if "addr:housenumber" not in addr:
		raise ValueError(eid + " has no house number")

def isAddrEq(a, b):
	(aid, aaddr) = a
	(bid, baddr) = b

	validateAddr(*a)
	validateAddr(*b)

	if aaddr["addr:street"] != baddr["addr:street"]:
		return False
	if aaddr["addr:housenumber"] != baddr["addr:housenumber"]:
		return False
	if "addr:unit" not in aaddr:
		if "addr:unit" in baddr:
			raise ValueError("{0} mostly equals to {1} but {0} has no unit while {1} has".format(aid, bid))
		return True
	elif "addr:unit" not in baddr:
		if "addr:unit" in aaddr:
			raise ValueError("{0} mostly equals to {1} but {0} has a unit while {1} has not".format(aid, bid))
		return True
	else:
		return (aaddr["addr:unit"] == baddr["addr:unit"])

def raiseIfAddrEq(a, b):
	if isAddrEq(a, b):
		raise ValueError("{0}'s address ({2}) is equal to {1}'s address ({3})".format(a[0], b[0], prettyAddr(a[1]), prettyAddr(b[1])))

if __name__ == '__main__':
	from sys import argv
	ad = {}
	with open(argv[1], "r") as f:
		ad = fishOutAddressData(f.read())
		pairwise(ad.items(), raiseIfAddrEq)