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 normalize(self, postcode):
  41         """ Return the normalized version of the postcode. If the given postcode
  42             does not correspond to the usage-pattern, return null.
  43         """
  44         # Upper-case, strip spaces and leading country code.
  45         normalized = self.norm_pattern.fullmatch(postcode.upper())
  46
  47         if normalized:
  48             match = self.pattern.fullmatch(normalized.group(1))
  49             return match.expand(self.output) if match else None
  50
  51         return None
  52
  53
  54 class _PostcodeSanitizer:
  55
  56     def __init__(self, config):
  57         self.convert_to_address = config.get_bool('convert-to-address', True)
  58         # Objects without a country code can't have a postcode per definition.
  59         self.country_without_postcode = {None}
  60         self.country_matcher = {}
  61
  62         for ccode, prop in country_info.iterate('postcode'):
  63             if prop is False:
  64                 self.country_without_postcode.add(ccode)
  65             elif isinstance(prop, dict):
  66                 self.country_matcher[ccode] = _PostcodeMatcher(ccode, prop)
  67             else:
  68                 raise UsageError(f"Invalid entry 'postcode' for country '{ccode}'")
  69
  70
  71     def __call__(self, obj):
  72         if not obj.address:
  73             return
  74
  75         postcodes = ((i, o) for i, o in enumerate(obj.address) if o.kind == 'postcode')
  76
  77         for pos, postcode in postcodes:
  78             formatted = self.scan(postcode.name, obj.place.country_code)
  79
  80             if formatted is None:
  81                 if self.convert_to_address:
  82                     postcode.kind = 'unofficial_postcode'
  83                 else:
  84                     obj.address.pop(pos)
  85             else:
  86                 postcode.name = formatted
  87
  88
  89     def scan(self, postcode, country):
  90         """ Check the postcode for correct formatting and return the
  91             normalized version. Returns None if the postcode does not
  92             correspond to the oficial format of the given country.
  93         """
  94         if country in self.country_without_postcode:
  95             return None
  96
  97         if country in self.country_matcher:
  98             return self.country_matcher[country].normalize(postcode)
  99
 100         return postcode.upper()
 101
 102
 103
 104 def create(config):
 105     """ Create a housenumber processing function.
 106     """
 107
 108     return _PostcodeSanitizer(config)