nominatim/tokenizer/sanitizers/clean_postcodes.py

   1 # SPDX-License-Identifier: GPL-2.0-only
   2 #
   3 # This file is part of Nominatim. (https://nominatim.org)
   4 #
   5 # Copyright (C) 2022 by the Nominatim developer community.
   6 # For a full list of authors see the git log.
   7 """
   8 Sanitizer that filters postcodes by their officially allowed pattern.
   9
  10 Arguments:
  11     convert-to-address: If set to 'yes' (the default), then postcodes that do
  12                         not conform with their country-specific pattern are
  13                         converted to an address component. That means that
  14                         the postcode does not take part when computing the
  15                         postcode centroids of a country but is still searchable.
  16                         When set to 'no', non-conforming postcodes are not
  17                         searchable either.
  18 """
  19 import re
  20
  21 from nominatim.errors import UsageError
  22 from nominatim.tools import country_info
  23
  24 class _PostcodeMatcher:
  25     """ Matches and formats a postcode according to the format definition.
  26     """
  27     def __init__(self, country_code, config):
  28         if 'pattern' not in config:
  29             raise UsageError("Field 'pattern' required for 'postcode' "
  30                              f"for country '{country_code}'")
  31
  32         pc_pattern = config['pattern'].replace('d', '[0-9]').replace('l', '[A-Z]')
  33
  34         self.norm_pattern = re.compile(f'\\s*(?:{country_code.upper()}[ -]?)?(.*)\\s*')
  35         self.pattern = re.compile(pc_pattern)
  36
  37         self.output = config.get('output', r'\g<0>')
  38
  39
  40     def match(self, postcode):
  41         """ Match the given postcode against the postcode pattern for this
  42             matcher. Returns a `re.Match` object if the match was successful
  43             and None otherwise.
  44         """
  45         # Upper-case, strip spaces and leading country code.
  46         normalized = self.norm_pattern.fullmatch(postcode.upper())
  47
  48         if normalized:
  49             return self.pattern.fullmatch(normalized.group(1))
  50
  51         return None
  52
  53
  54     def normalize(self, match):
  55         """ Return the default format of the postcode for the given match.
  56             `match` must be a `re.Match` object previously returned by
  57             `match()`
  58         """
  59         return match.expand(self.output)
  60
  61
  62 class _PostcodeSanitizer:
  63
  64     def __init__(self, config):
  65         self.convert_to_address = config.get_bool('convert-to-address', True)
  66         # Objects without a country code can't have a postcode per definition.
  67         self.country_without_postcode = {None}
  68         self.country_matcher = {}
  69
  70         for ccode, prop in country_info.iterate('postcode'):
  71             if prop is False:
  72                 self.country_without_postcode.add(ccode)
  73             elif isinstance(prop, dict):
  74                 self.country_matcher[ccode] = _PostcodeMatcher(ccode, prop)
  75             else:
  76                 raise UsageError(f"Invalid entry 'postcode' for country '{ccode}'")
  77
  78
  79     def __call__(self, obj):
  80         if not obj.address:
  81             return
  82
  83         postcodes = ((i, o) for i, o in enumerate(obj.address) if o.kind == 'postcode')
  84
  85         for pos, postcode in postcodes:
  86             formatted = self.scan(postcode.name, obj.place.country_code)
  87
  88             if formatted is None:
  89                 if self.convert_to_address:
  90                     postcode.kind = 'unofficial_postcode'
  91                 else:
  92                     obj.address.pop(pos)
  93             else:
  94                 postcode.name = formatted[0]
  95                 postcode.set_attr('lookup', formatted[1])
  96
  97
  98     def scan(self, postcode, country):
  99         """ Check the postcode for correct formatting and return the
 100             normalized version. Returns None if the postcode does not
 101             correspond to the oficial format of the given country.
 102         """
 103         if country in self.country_without_postcode:
 104             return None
 105
 106         matcher = self.country_matcher.get(country)
 107         if matcher is not None:
 108             match = matcher.match(postcode)
 109             if match is None:
 110                 return None
 111             return matcher.normalize(match), ' '.join(match.groups())
 112
 113         return postcode.upper(), ''
 114
 115
 116
 117 def create(config):
 118     """ Create a housenumber processing function.
 119     """
 120
 121     return _PostcodeSanitizer(config)