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 Datastructures for a tokenized query.
10 from typing import List, Tuple, Optional, Iterator
11 from abc import ABC, abstractmethod
15 class BreakType(enum.Enum):
16 """ Type of break between tokens.
19 """ Begin of the query. """
21 """ End of the query. """
23 """ Break between two phrases. """
25 """ Break between words. """
27 """ Break inside a word, for example a hyphen or apostrophe. """
29 """ Break created as a result of tokenization.
30 This may happen in languages without spaces between words.
34 class TokenType(enum.Enum):
38 """ Full name of a place. """
40 """ Word term without breaks, does not necessarily represent a full name. """
41 HOUSENUMBER = enum.auto()
42 """ Housenumber term. """
43 POSTCODE = enum.auto()
44 """ Postal code term. """
46 """ Country name or reference. """
47 QUALIFIER = enum.auto()
48 """ Special term used together with name (e.g. _Hotel_ Bellevue). """
49 CATEGORY = enum.auto()
50 """ Special term used as searchable object(e.g. supermarket in ...). """
53 class PhraseType(enum.Enum):
54 """ Designation of a phrase.
57 """ No specific designation (i.e. source is free-form query). """
59 """ Contains name or type of a POI. """
61 """ Contains a street name optionally with a housenumber. """
63 """ Contains the postal city. """
65 """ Contains the equivalent of a county. """
67 """ Contains a state or province. """
68 POSTCODE = enum.auto()
69 """ Contains a postal code. """
71 """ Contains the country name or code. """
73 def compatible_with(self, ttype: TokenType) -> bool:
74 """ Check if the given token type can be used with the phrase type.
76 if self == PhraseType.NONE:
78 if self == PhraseType.AMENITY:
79 return ttype in (TokenType.WORD, TokenType.PARTIAL,
80 TokenType.QUALIFIER, TokenType.CATEGORY)
81 if self == PhraseType.STREET:
82 return ttype in (TokenType.WORD, TokenType.PARTIAL, TokenType.HOUSENUMBER)
83 if self == PhraseType.POSTCODE:
84 return ttype == TokenType.POSTCODE
85 if self == PhraseType.COUNTRY:
86 return ttype == TokenType.COUNTRY
88 return ttype in (TokenType.WORD, TokenType.PARTIAL)
91 @dataclasses.dataclass
93 """ Base type for tokens.
94 Specific query analyzers must implement the concrete token class.
105 def get_category(self) -> Tuple[str, str]:
106 """ Return the category restriction for qualifier terms and
110 @dataclasses.dataclass
112 """ Indexes of query nodes over which a token spans.
117 def __lt__(self, other: 'TokenRange') -> bool:
118 return self.end <= other.start
121 def __le__(self, other: 'TokenRange') -> bool:
122 return NotImplemented
125 def __gt__(self, other: 'TokenRange') -> bool:
126 return self.start >= other.end
129 def __ge__(self, other: 'TokenRange') -> bool:
130 return NotImplemented
133 def replace_start(self, new_start: int) -> 'TokenRange':
134 """ Return a new token range with the new start.
136 return TokenRange(new_start, self.end)
139 def replace_end(self, new_end: int) -> 'TokenRange':
140 """ Return a new token range with the new end.
142 return TokenRange(self.start, new_end)
145 def split(self, index: int) -> Tuple['TokenRange', 'TokenRange']:
146 """ Split the span into two spans at the given index.
147 The index must be within the span.
149 return self.replace_end(index), self.replace_start(index)
152 @dataclasses.dataclass
154 """ List of all tokens of a given type going from one breakpoint to another.
161 def add_penalty(self, penalty: float) -> None:
162 """ Add the given penalty to all tokens in the list.
164 for token in self.tokens:
165 token.penalty += penalty
168 @dataclasses.dataclass
170 """ A node of the querry representing a break between terms.
174 starting: List[TokenList] = dataclasses.field(default_factory=list)
176 def has_tokens(self, end: int, *ttypes: TokenType) -> bool:
177 """ Check if there are tokens of the given types ending at the
180 return any(tl.end == end and tl.ttype in ttypes for tl in self.starting)
183 def get_tokens(self, end: int, ttype: TokenType) -> Optional[List[Token]]:
184 """ Get the list of tokens of the given type starting at this node
185 and ending at the node 'end'. Returns 'None' if no such
188 for tlist in self.starting:
189 if tlist.end == end and tlist.ttype == ttype:
194 @dataclasses.dataclass
196 """ A normalized query part. Phrases may be typed which means that
197 they then represent a specific part of the address.
204 """ A tokenized search query together with the normalized source
205 from which the tokens have been parsed.
207 The query contains a list of nodes that represent the breaks
208 between words. Tokens span between nodes, which don't necessarily
209 need to be direct neighbours. Thus the query is represented as a
210 directed acyclic graph.
212 When created, a query contains a single node: the start of the
213 query. Further nodes can be added by appending to 'nodes'.
216 def __init__(self, source: List[Phrase]) -> None:
218 self.nodes: List[QueryNode] = \
219 [QueryNode(BreakType.START, source[0].ptype if source else PhraseType.NONE)]
222 def num_token_slots(self) -> int:
223 """ Return the length of the query in vertice steps.
225 return len(self.nodes) - 1
228 def add_node(self, btype: BreakType, ptype: PhraseType) -> None:
229 """ Append a new break node with the given break type.
230 The phrase type denotes the type for any tokens starting
233 self.nodes.append(QueryNode(btype, ptype))
236 def add_token(self, trange: TokenRange, ttype: TokenType, token: Token) -> None:
237 """ Add a token to the query. 'start' and 'end' are the indexes of the
238 nodes from which to which the token spans. The indexes must exist
239 and are expected to be in the same phrase.
240 'ttype' denotes the type of the token and 'token' the token to
243 If the token type is not compatible with the phrase it should
244 be added to, then the token is silently dropped.
246 snode = self.nodes[trange.start]
247 if snode.ptype.compatible_with(ttype):
248 tlist = snode.get_tokens(trange.end, ttype)
250 snode.starting.append(TokenList(trange.end, ttype, [token]))
255 def get_tokens(self, trange: TokenRange, ttype: TokenType) -> List[Token]:
256 """ Get the list of tokens of a given type, spanning the given
257 nodes. The nodes must exist. If no tokens exist, an
258 empty list is returned.
260 return self.nodes[trange.start].get_tokens(trange.end, ttype) or []
263 def get_partials_list(self, trange: TokenRange) -> List[Token]:
264 """ Create a list of partial tokens between the given nodes.
265 The list is composed of the first token of type PARTIAL
266 going to the subsequent node. Such PARTIAL tokens are
269 return [next(iter(self.get_tokens(TokenRange(i, i+1), TokenType.PARTIAL)))
270 for i in range(trange.start, trange.end)]
273 def iter_token_lists(self) -> Iterator[Tuple[int, QueryNode, TokenList]]:
274 """ Iterator over all token lists in the query.
276 for i, node in enumerate(self.nodes):
277 for tlist in node.starting:
281 def find_lookup_word_by_id(self, token: int) -> str:
282 """ Find the first token with the given token ID and return
283 its lookup word. Returns 'None' if no such token exists.
284 The function is very slow and must only be used for
287 for node in self.nodes:
288 for tlist in node.starting:
289 for t in tlist.tokens:
291 return f"[{tlist.ttype.name[0]}]{t.lookup_word}"