1 # SPDX-License-Identifier: GPL-3.0-or-later
3 # This file is part of Nominatim. (https://nominatim.org)
5 # Copyright (C) 2023 by the Nominatim developer community.
6 # For a full list of authors see the git log.
8 Data structures for more complex fields in abstract search descriptions.
10 from typing import List, Tuple, Iterator, cast
13 import sqlalchemy as sa
14 from sqlalchemy.dialects.postgresql import ARRAY
16 from nominatim.typing import SaFromClause, SaColumn, SaExpression
17 from nominatim.api.search.query import Token
19 @dataclasses.dataclass
20 class WeightedStrings:
21 """ A list of strings together with a penalty.
24 penalties: List[float]
26 def __bool__(self) -> bool:
27 return bool(self.values)
30 def __iter__(self) -> Iterator[Tuple[str, float]]:
31 return iter(zip(self.values, self.penalties))
34 def get_penalty(self, value: str, default: float = 1000.0) -> float:
35 """ Get the penalty for the given value. Returns the given default
36 if the value does not exist.
39 return self.penalties[self.values.index(value)]
45 @dataclasses.dataclass
46 class WeightedCategories:
47 """ A list of class/type tuples together with a penalty.
49 values: List[Tuple[str, str]]
50 penalties: List[float]
52 def __bool__(self) -> bool:
53 return bool(self.values)
56 def __iter__(self) -> Iterator[Tuple[Tuple[str, str], float]]:
57 return iter(zip(self.values, self.penalties))
60 def get_penalty(self, value: Tuple[str, str], default: float = 1000.0) -> float:
61 """ Get the penalty for the given value. Returns the given default
62 if the value does not exist.
65 return self.penalties[self.values.index(value)]
71 def sql_restrict(self, table: SaFromClause) -> SaExpression:
72 """ Return an SQLAlcheny expression that restricts the
73 class and type columns of the given table to the values
75 Must not be used with an empty list.
78 if len(self.values) == 1:
79 return sa.and_(table.c.class_ == self.values[0][0],
80 table.c.type == self.values[0][1])
82 return sa.or_(*(sa.and_(table.c.class_ == c, table.c.type == t)
83 for c, t in self.values))
86 @dataclasses.dataclass(order=True)
88 """ List of tokens together with the penalty of using it.
93 def with_token(self, t: Token, transition_penalty: float) -> 'RankedTokens':
94 """ Create a new RankedTokens list with the given token appended.
95 The tokens penalty as well as the given transision penalty
96 are added to the overall penalty.
98 return RankedTokens(self.penalty + t.penalty + transition_penalty,
99 self.tokens + [t.token])
102 @dataclasses.dataclass
104 """ A list of rankings to be applied sequentially until one matches.
105 The matched ranking determines the penalty. If none matches a
106 default penalty is applied.
110 rankings: List[RankedTokens]
112 def normalize_penalty(self) -> float:
113 """ Reduce the default and ranking penalties, such that the minimum
114 penalty is 0. Return the penalty that was subtracted.
117 min_penalty = min(self.default, min(r.penalty for r in self.rankings))
119 min_penalty = self.default
120 if min_penalty > 0.0:
121 self.default -= min_penalty
122 for ranking in self.rankings:
123 ranking.penalty -= min_penalty
127 def sql_penalty(self, table: SaFromClause) -> SaColumn:
128 """ Create an SQL expression for the rankings.
132 return sa.func.weigh_search(table.c[self.column],
133 [f"{{{','.join((str(s) for s in r.tokens))}}}"
134 for r in self.rankings],
135 [r.penalty for r in self.rankings],
139 @dataclasses.dataclass
141 """ A list of tokens to be searched for. The column names the database
142 column to search in and the lookup_type the operator that is applied.
143 'lookup_all' requires all tokens to match. 'lookup_any' requires
144 one of the tokens to match. 'restrict' requires to match all tokens
145 but avoids the use of indexes.
151 def sql_condition(self, table: SaFromClause) -> SaColumn:
152 """ Create an SQL expression for the given match condition.
154 col = table.c[self.column]
155 if self.lookup_type == 'lookup_all':
156 return col.contains(self.tokens)
157 if self.lookup_type == 'lookup_any':
158 return cast(SaColumn, col.overlap(self.tokens))
160 return sa.func.array_cat(col, sa.text('ARRAY[]::integer[]'),
161 type_=ARRAY(sa.Integer())).contains(self.tokens)
165 """ Search fields derived from query and token assignment
166 to be used with the SQL queries.
170 lookups: List[FieldLookup] = []
171 rankings: List[FieldRanking]
173 housenumbers: WeightedStrings = WeightedStrings([], [])
174 postcodes: WeightedStrings = WeightedStrings([], [])
175 countries: WeightedStrings = WeightedStrings([], [])
177 qualifiers: WeightedCategories = WeightedCategories([], [])
180 def set_strings(self, field: str, tokens: List[Token]) -> None:
181 """ Set on of the WeightedStrings properties from the given
182 token list. Adapt the global penalty, so that the
183 minimum penalty is 0.
186 min_penalty = min(t.penalty for t in tokens)
187 self.penalty += min_penalty
188 wstrs = WeightedStrings([t.lookup_word for t in tokens],
189 [t.penalty - min_penalty for t in tokens])
191 setattr(self, field, wstrs)
194 def set_qualifiers(self, tokens: List[Token]) -> None:
195 """ Set the qulaifier field from the given tokens.
198 min_penalty = min(t.penalty for t in tokens)
199 self.penalty += min_penalty
200 self.qualifiers = WeightedCategories([t.get_category() for t in tokens],
201 [t.penalty - min_penalty for t in tokens])
204 def set_ranking(self, rankings: List[FieldRanking]) -> None:
205 """ Set the list of rankings and normalize the ranking.
208 for ranking in rankings:
210 self.penalty += ranking.normalize_penalty()
211 self.rankings.append(ranking)
213 self.penalty += ranking.default