1 # SPDX-License-Identifier: GPL-3.0-or-later
3 # This file is part of Nominatim. (https://nominatim.org)
5 # Copyright (C) 2024 by the Nominatim developer community.
6 # For a full list of authors see the git log.
8 Datastructures for a tokenized query.
10 from typing import List, Tuple, Optional, Iterator
11 from abc import ABC, abstractmethod
16 class BreakType(enum.Enum):
17 """ Type of break between tokens.
20 """ Begin of the query. """
22 """ End of the query. """
24 """ Break between two phrases. """
26 """ Break between words. """
28 """ Break inside a word, for example a hyphen or apostrophe. """
30 """ Break created as a result of tokenization.
31 This may happen in languages without spaces between words.
35 class TokenType(enum.Enum):
39 """ Full name of a place. """
41 """ Word term without breaks, does not necessarily represent a full name. """
42 HOUSENUMBER = enum.auto()
43 """ Housenumber term. """
44 POSTCODE = enum.auto()
45 """ Postal code term. """
47 """ Country name or reference. """
48 QUALIFIER = enum.auto()
49 """ Special term used together with name (e.g. _Hotel_ Bellevue). """
50 NEAR_ITEM = enum.auto()
51 """ Special term used as searchable object(e.g. supermarket in ...). """
54 class PhraseType(enum.Enum):
55 """ Designation of a phrase.
58 """ No specific designation (i.e. source is free-form query). """
60 """ Contains name or type of a POI. """
62 """ Contains a street name optionally with a housenumber. """
64 """ Contains the postal city. """
66 """ Contains the equivalent of a county. """
68 """ Contains a state or province. """
69 POSTCODE = enum.auto()
70 """ Contains a postal code. """
72 """ Contains the country name or code. """
74 def compatible_with(self, ttype: TokenType,
75 is_full_phrase: bool) -> bool:
76 """ Check if the given token type can be used with the phrase type.
78 if self == PhraseType.NONE:
79 return not is_full_phrase or ttype != TokenType.QUALIFIER
80 if self == PhraseType.AMENITY:
81 return ttype in (TokenType.WORD, TokenType.PARTIAL)\
82 or (is_full_phrase and ttype == TokenType.NEAR_ITEM)\
83 or (not is_full_phrase and ttype == TokenType.QUALIFIER)
84 if self == PhraseType.STREET:
85 return ttype in (TokenType.WORD, TokenType.PARTIAL, TokenType.HOUSENUMBER)
86 if self == PhraseType.POSTCODE:
87 return ttype == TokenType.POSTCODE
88 if self == PhraseType.COUNTRY:
89 return ttype == TokenType.COUNTRY
91 return ttype in (TokenType.WORD, TokenType.PARTIAL)
94 @dataclasses.dataclass
96 """ Base type for tokens.
97 Specific query analyzers must implement the concrete token class.
107 def get_category(self) -> Tuple[str, str]:
108 """ Return the category restriction for qualifier terms and
113 @dataclasses.dataclass
115 """ Indexes of query nodes over which a token spans.
120 def __lt__(self, other: 'TokenRange') -> bool:
121 return self.end <= other.start
123 def __le__(self, other: 'TokenRange') -> bool:
124 return NotImplemented
126 def __gt__(self, other: 'TokenRange') -> bool:
127 return self.start >= other.end
129 def __ge__(self, other: 'TokenRange') -> bool:
130 return NotImplemented
132 def replace_start(self, new_start: int) -> 'TokenRange':
133 """ Return a new token range with the new start.
135 return TokenRange(new_start, self.end)
137 def replace_end(self, new_end: int) -> 'TokenRange':
138 """ Return a new token range with the new end.
140 return TokenRange(self.start, new_end)
142 def split(self, index: int) -> Tuple['TokenRange', 'TokenRange']:
143 """ Split the span into two spans at the given index.
144 The index must be within the span.
146 return self.replace_end(index), self.replace_start(index)
149 @dataclasses.dataclass
151 """ List of all tokens of a given type going from one breakpoint to another.
157 def add_penalty(self, penalty: float) -> None:
158 """ Add the given penalty to all tokens in the list.
160 for token in self.tokens:
161 token.penalty += penalty
164 @dataclasses.dataclass
166 """ A node of the query representing a break between terms.
170 starting: List[TokenList] = dataclasses.field(default_factory=list)
172 def has_tokens(self, end: int, *ttypes: TokenType) -> bool:
173 """ Check if there are tokens of the given types ending at the
176 return any(tl.end == end and tl.ttype in ttypes for tl in self.starting)
178 def get_tokens(self, end: int, ttype: TokenType) -> Optional[List[Token]]:
179 """ Get the list of tokens of the given type starting at this node
180 and ending at the node 'end'. Returns 'None' if no such
183 for tlist in self.starting:
184 if tlist.end == end and tlist.ttype == ttype:
189 @dataclasses.dataclass
191 """ A normalized query part. Phrases may be typed which means that
192 they then represent a specific part of the address.
199 """ A tokenized search query together with the normalized source
200 from which the tokens have been parsed.
202 The query contains a list of nodes that represent the breaks
203 between words. Tokens span between nodes, which don't necessarily
204 need to be direct neighbours. Thus the query is represented as a
205 directed acyclic graph.
207 When created, a query contains a single node: the start of the
208 query. Further nodes can be added by appending to 'nodes'.
211 def __init__(self, source: List[Phrase]) -> None:
213 self.nodes: List[QueryNode] = \
214 [QueryNode(BreakType.START, source[0].ptype if source else PhraseType.NONE)]
216 def num_token_slots(self) -> int:
217 """ Return the length of the query in vertice steps.
219 return len(self.nodes) - 1
221 def add_node(self, btype: BreakType, ptype: PhraseType) -> None:
222 """ Append a new break node with the given break type.
223 The phrase type denotes the type for any tokens starting
226 self.nodes.append(QueryNode(btype, ptype))
228 def add_token(self, trange: TokenRange, ttype: TokenType, token: Token) -> None:
229 """ Add a token to the query. 'start' and 'end' are the indexes of the
230 nodes from which to which the token spans. The indexes must exist
231 and are expected to be in the same phrase.
232 'ttype' denotes the type of the token and 'token' the token to
235 If the token type is not compatible with the phrase it should
236 be added to, then the token is silently dropped.
238 snode = self.nodes[trange.start]
239 full_phrase = snode.btype in (BreakType.START, BreakType.PHRASE)\
240 and self.nodes[trange.end].btype in (BreakType.PHRASE, BreakType.END)
241 if snode.ptype.compatible_with(ttype, full_phrase):
242 tlist = snode.get_tokens(trange.end, ttype)
244 snode.starting.append(TokenList(trange.end, ttype, [token]))
248 def get_tokens(self, trange: TokenRange, ttype: TokenType) -> List[Token]:
249 """ Get the list of tokens of a given type, spanning the given
250 nodes. The nodes must exist. If no tokens exist, an
251 empty list is returned.
253 return self.nodes[trange.start].get_tokens(trange.end, ttype) or []
255 def get_partials_list(self, trange: TokenRange) -> List[Token]:
256 """ Create a list of partial tokens between the given nodes.
257 The list is composed of the first token of type PARTIAL
258 going to the subsequent node. Such PARTIAL tokens are
261 return [next(iter(self.get_tokens(TokenRange(i, i+1), TokenType.PARTIAL)))
262 for i in range(trange.start, trange.end)]
264 def iter_token_lists(self) -> Iterator[Tuple[int, QueryNode, TokenList]]:
265 """ Iterator over all token lists in the query.
267 for i, node in enumerate(self.nodes):
268 for tlist in node.starting:
271 def find_lookup_word_by_id(self, token: int) -> str:
272 """ Find the first token with the given token ID and return
273 its lookup word. Returns 'None' if no such token exists.
274 The function is very slow and must only be used for
277 for node in self.nodes:
278 for tlist in node.starting:
279 for t in tlist.tokens:
281 return f"[{tlist.ttype.name[0]}]{t.lookup_word}"