2 Generic processor for names that creates abbreviation variants.
4 from collections import defaultdict, namedtuple
8 from icu import Transliterator
11 from nominatim.config import flatten_config_list
12 from nominatim.errors import UsageError
14 ### Configuration section
16 ICUVariant = namedtuple('ICUVariant', ['source', 'replacement'])
18 def configure(rules, normalization_rules):
19 """ Extract and preprocess the configuration for this module.
23 config['replacements'], config['chars'] = _get_variant_config(rules.get('variants'),
25 config['variant_only'] = rules.get('mode', '') == 'variant-only'
30 def _get_variant_config(rules, normalization_rules):
31 """ Convert the variant definition from the configuration into
34 immediate = defaultdict(list)
39 rules = flatten_config_list(rules, 'variants')
41 vmaker = _VariantMaker(normalization_rules)
44 for rule in (section.get('words') or []):
45 vset.update(vmaker.compute(rule))
47 # Intermediate reorder by source. Also compute required character set.
49 if variant.source[-1] == ' ' and variant.replacement[-1] == ' ':
50 replstr = variant.replacement[:-1]
52 replstr = variant.replacement
53 immediate[variant.source].append(replstr)
54 chars.update(variant.source)
56 return list(immediate.items()), ''.join(chars)
60 """ Generater for all necessary ICUVariants from a single variant rule.
62 All text in rules is normalized to make sure the variants match later.
65 def __init__(self, norm_rules):
66 self.norm = Transliterator.createFromRules("rule_loader_normalization",
70 def compute(self, rule):
71 """ Generator for all ICUVariant tuples from a single variant rule.
73 parts = re.split(r'(\|)?([=-])>', rule)
75 raise UsageError("Syntax error in variant rule: " + rule)
77 decompose = parts[1] is None
78 src_terms = [self._parse_variant_word(t) for t in parts[0].split(',')]
79 repl_terms = (self.norm.transliterate(t.strip()) for t in parts[3].split(','))
81 # If the source should be kept, add a 1:1 replacement
85 for froms, tos in _create_variants(*src, src[0], decompose):
86 yield ICUVariant(froms, tos)
88 for src, repl in itertools.product(src_terms, repl_terms):
90 for froms, tos in _create_variants(*src, repl, decompose):
91 yield ICUVariant(froms, tos)
94 def _parse_variant_word(self, name):
96 match = re.fullmatch(r'([~^]?)([^~$^]*)([~$]?)', name)
97 if match is None or (match.group(1) == '~' and match.group(3) == '~'):
98 raise UsageError("Invalid variant word descriptor '{}'".format(name))
99 norm_name = self.norm.transliterate(match.group(2))
103 return norm_name, match.group(1), match.group(3)
106 _FLAG_MATCH = {'^': '^ ',
111 def _create_variants(src, preflag, postflag, repl, decompose):
113 postfix = _FLAG_MATCH[postflag]
114 # suffix decomposition
116 repl = repl + postfix
119 yield ' ' + src, ' ' + repl
122 yield src, ' ' + repl
123 yield ' ' + src, repl
124 elif postflag == '~':
125 # prefix decomposition
126 prefix = _FLAG_MATCH[preflag]
131 yield src + ' ', repl + ' '
134 yield src, repl + ' '
135 yield src + ' ', repl
137 prefix = _FLAG_MATCH[preflag]
138 postfix = _FLAG_MATCH[postflag]
140 yield prefix + src + postfix, prefix + repl + postfix
145 def create(trans_rules, config):
146 """ Create a new token analysis instance for this module.
148 return GenericTokenAnalysis(trans_rules, config)
151 class GenericTokenAnalysis:
152 """ Collects the different transformation rules for normalisation of names
153 and provides the functions to apply the transformations.
156 def __init__(self, to_ascii, config):
157 self.to_ascii = to_ascii
158 self.variant_only = config['variant_only']
161 if config['replacements']:
162 self.replacements = datrie.Trie(config['chars'])
163 for src, repllist in config['replacements']:
164 self.replacements[src] = repllist
166 self.replacements = None
169 def get_variants_ascii(self, norm_name):
170 """ Compute the spelling variants for the given normalized name
171 and transliterate the result.
173 baseform = '^ ' + norm_name + ' ^'
177 if self.replacements is not None:
180 while pos < len(baseform):
181 full, repl = self.replacements.longest_prefix_item(baseform[pos:],
184 done = baseform[startpos:pos]
185 partials = [v + done + r
186 for v, r in itertools.product(partials, repl)
187 if not force_space or r.startswith(' ')]
188 if len(partials) > 128:
189 # If too many variants are produced, they are unlikely
190 # to be helpful. Only use the original term.
193 startpos = pos + len(full)
202 # No variants detected? Fast return.
204 if self.variant_only:
207 trans_name = self.to_ascii.transliterate(norm_name).strip()
208 return [trans_name] if trans_name else []
210 return self._compute_result_set(partials, baseform[startpos:],
211 norm_name if self.variant_only else '')
214 def _compute_result_set(self, partials, prefix, exclude):
217 for variant in partials:
218 vname = (variant + prefix)[1:-1].strip()
220 trans_name = self.to_ascii.transliterate(vname).strip()
222 results.add(trans_name)