nominatim/api/search/db_search_builder.py

   1 # SPDX-License-Identifier: GPL-3.0-or-later
   2 #
   3 # This file is part of Nominatim. (https://nominatim.org)
   4 #
   5 # Copyright (C) 2023 by the Nominatim developer community.
   6 # For a full list of authors see the git log.
   7 """
   8 Convertion from token assignment to an abstract DB search.
   9 """
  10 from typing import Optional, List, Tuple, Iterator
  11 import heapq
  12
  13 from nominatim.api.types import SearchDetails, DataLayer
  14 from nominatim.api.search.query import QueryStruct, Token, TokenType, TokenRange, BreakType
  15 from nominatim.api.search.token_assignment import TokenAssignment
  16 import nominatim.api.search.db_search_fields as dbf
  17 import nominatim.api.search.db_searches as dbs
  18 from nominatim.api.logging import log
  19
  20
  21 def wrap_near_search(categories: List[Tuple[str, str]],
  22                      search: dbs.AbstractSearch) -> dbs.NearSearch:
  23     """ Create a new search that wraps the given search in a search
  24         for near places of the given category.
  25     """
  26     return dbs.NearSearch(penalty=search.penalty,
  27                           categories=dbf.WeightedCategories(categories,
  28                                                             [0.0] * len(categories)),
  29                           search=search)
  30
  31
  32 def build_poi_search(category: List[Tuple[str, str]],
  33                      countries: Optional[List[str]]) -> dbs.PoiSearch:
  34     """ Create a new search for places by the given category, possibly
  35         constraint to the given countries.
  36     """
  37     if countries:
  38         ccs = dbf.WeightedStrings(countries, [0.0] * len(countries))
  39     else:
  40         ccs = dbf.WeightedStrings([], [])
  41
  42     class _PoiData(dbf.SearchData):
  43         penalty = 0.0
  44         qualifiers = dbf.WeightedCategories(category, [0.0] * len(category))
  45         countries=ccs
  46
  47     return dbs.PoiSearch(_PoiData())
  48
  49
  50 class SearchBuilder:
  51     """ Build the abstract search queries from token assignments.
  52     """
  53
  54     def __init__(self, query: QueryStruct, details: SearchDetails) -> None:
  55         self.query = query
  56         self.details = details
  57
  58
  59     @property
  60     def configured_for_country(self) -> bool:
  61         """ Return true if the search details are configured to
  62             allow countries in the result.
  63         """
  64         return self.details.min_rank <= 4 and self.details.max_rank >= 4 \
  65                and self.details.layer_enabled(DataLayer.ADDRESS)
  66
  67
  68     @property
  69     def configured_for_postcode(self) -> bool:
  70         """ Return true if the search details are configured to
  71             allow postcodes in the result.
  72         """
  73         return self.details.min_rank <= 5 and self.details.max_rank >= 11\
  74                and self.details.layer_enabled(DataLayer.ADDRESS)
  75
  76
  77     @property
  78     def configured_for_housenumbers(self) -> bool:
  79         """ Return true if the search details are configured to
  80             allow addresses in the result.
  81         """
  82         return self.details.max_rank >= 30 \
  83                and self.details.layer_enabled(DataLayer.ADDRESS)
  84
  85
  86     def build(self, assignment: TokenAssignment) -> Iterator[dbs.AbstractSearch]:
  87         """ Yield all possible abstract searches for the given token assignment.
  88         """
  89         sdata = self.get_search_data(assignment)
  90         if sdata is None:
  91             return
  92
  93         categories = self.get_search_categories(assignment)
  94
  95         if assignment.name is None:
  96             if categories and not sdata.postcodes:
  97                 sdata.qualifiers = categories
  98                 categories = None
  99                 builder = self.build_poi_search(sdata)
 100             elif assignment.housenumber:
 101                 hnr_tokens = self.query.get_tokens(assignment.housenumber,
 102                                                    TokenType.HOUSENUMBER)
 103                 builder = self.build_housenumber_search(sdata, hnr_tokens, assignment.address)
 104             else:
 105                 builder = self.build_special_search(sdata, assignment.address,
 106                                                     bool(categories))
 107         else:
 108             builder = self.build_name_search(sdata, assignment.name, assignment.address,
 109                                              bool(categories))
 110
 111         if categories:
 112             penalty = min(categories.penalties)
 113             categories.penalties = [p - penalty for p in categories.penalties]
 114             for search in builder:
 115                 yield dbs.NearSearch(penalty, categories, search)
 116         else:
 117             yield from builder
 118
 119
 120     def build_poi_search(self, sdata: dbf.SearchData) -> Iterator[dbs.AbstractSearch]:
 121         """ Build abstract search query for a simple category search.
 122             This kind of search requires an additional geographic constraint.
 123         """
 124         if not sdata.housenumbers \
 125            and ((self.details.viewbox and self.details.bounded_viewbox) or self.details.near):
 126             yield dbs.PoiSearch(sdata)
 127
 128
 129     def build_special_search(self, sdata: dbf.SearchData,
 130                              address: List[TokenRange],
 131                              is_category: bool) -> Iterator[dbs.AbstractSearch]:
 132         """ Build abstract search queries for searches that do not involve
 133             a named place.
 134         """
 135         if sdata.qualifiers:
 136             # No special searches over qualifiers supported.
 137             return
 138
 139         if sdata.countries and not address and not sdata.postcodes \
 140            and self.configured_for_country:
 141             yield dbs.CountrySearch(sdata)
 142
 143         if sdata.postcodes and (is_category or self.configured_for_postcode):
 144             penalty = 0.0 if sdata.countries else 0.1
 145             if address:
 146                 sdata.lookups = [dbf.FieldLookup('nameaddress_vector',
 147                                                  [t.token for r in address
 148                                                   for t in self.query.get_partials_list(r)],
 149                                                  'restrict')]
 150                 penalty += 0.2
 151             yield dbs.PostcodeSearch(penalty, sdata)
 152
 153
 154     def build_housenumber_search(self, sdata: dbf.SearchData, hnrs: List[Token],
 155                                  address: List[TokenRange]) -> Iterator[dbs.AbstractSearch]:
 156         """ Build a simple address search for special entries where the
 157             housenumber is the main name token.
 158         """
 159         partial_tokens: List[int] = []
 160         for trange in address:
 161             partial_tokens.extend(t.token for t in self.query.get_partials_list(trange))
 162
 163         sdata.lookups = [dbf.FieldLookup('name_vector', [t.token for t in hnrs], 'lookup_any'),
 164                          dbf.FieldLookup('nameaddress_vector', partial_tokens, 'lookup_all')
 165                         ]
 166         sdata.housenumbers = dbf.WeightedStrings([], [])
 167         yield dbs.PlaceSearch(0.05, sdata, sum(t.count for t in hnrs))
 168
 169
 170     def build_name_search(self, sdata: dbf.SearchData,
 171                           name: TokenRange, address: List[TokenRange],
 172                           is_category: bool) -> Iterator[dbs.AbstractSearch]:
 173         """ Build abstract search queries for simple name or address searches.
 174         """
 175         if is_category or not sdata.housenumbers or self.configured_for_housenumbers:
 176             ranking = self.get_name_ranking(name)
 177             name_penalty = ranking.normalize_penalty()
 178             if ranking.rankings:
 179                 sdata.rankings.append(ranking)
 180             for penalty, count, lookup in self.yield_lookups(name, address):
 181                 sdata.lookups = lookup
 182                 yield dbs.PlaceSearch(penalty + name_penalty, sdata, count)
 183
 184
 185     def yield_lookups(self, name: TokenRange, address: List[TokenRange])\
 186                           -> Iterator[Tuple[float, int, List[dbf.FieldLookup]]]:
 187         """ Yield all variants how the given name and address should best
 188             be searched for. This takes into account how frequent the terms
 189             are and tries to find a lookup that optimizes index use.
 190         """
 191         penalty = 0.0 # extra penalty currently unused
 192
 193         name_partials = self.query.get_partials_list(name)
 194         exp_name_count = min(t.count for t in name_partials)
 195         addr_partials = []
 196         for trange in address:
 197             addr_partials.extend(self.query.get_partials_list(trange))
 198         addr_tokens = [t.token for t in addr_partials]
 199         partials_indexed = all(t.is_indexed for t in name_partials) \
 200                            and all(t.is_indexed for t in addr_partials)
 201
 202         if (len(name_partials) > 3 or exp_name_count < 1000) and partials_indexed:
 203             # Lookup by name partials, use address partials to restrict results.
 204             lookup = [dbf.FieldLookup('name_vector',
 205                                   [t.token for t in name_partials], 'lookup_all')]
 206             if addr_tokens:
 207                 lookup.append(dbf.FieldLookup('nameaddress_vector', addr_tokens, 'restrict'))
 208             yield penalty, exp_name_count, lookup
 209             return
 210
 211         exp_addr_count = min(t.count for t in addr_partials) if addr_partials else exp_name_count
 212         if exp_addr_count < 1000 and partials_indexed:
 213             # Lookup by address partials and restrict results through name terms.
 214             # Give this a small penalty because lookups in the address index are
 215             # more expensive
 216             yield penalty + exp_addr_count/5000, exp_addr_count,\
 217                   [dbf.FieldLookup('name_vector', [t.token for t in name_partials], 'restrict'),
 218                    dbf.FieldLookup('nameaddress_vector', addr_tokens, 'lookup_all')]
 219             return
 220
 221         # Partial term to frequent. Try looking up by rare full names first.
 222         name_fulls = self.query.get_tokens(name, TokenType.WORD)
 223         rare_names = list(filter(lambda t: t.count < 10000, name_fulls))
 224         # At this point drop unindexed partials from the address.
 225         # This might yield wrong results, nothing we can do about that.
 226         if not partials_indexed:
 227             addr_tokens = [t.token for t in addr_partials if t.is_indexed]
 228             penalty += 1.2 * sum(t.penalty for t in addr_partials if not t.is_indexed)
 229         if rare_names:
 230             # Any of the full names applies with all of the partials from the address
 231             lookup = [dbf.FieldLookup('name_vector', [t.token for t in rare_names], 'lookup_any')]
 232             if addr_tokens:
 233                 lookup.append(dbf.FieldLookup('nameaddress_vector', addr_tokens, 'restrict'))
 234             yield penalty, sum(t.count for t in rare_names), lookup
 235
 236         # To catch remaining results, lookup by name and address
 237         # We only do this if there is a reasonable number of results expected.
 238         if min(exp_name_count, exp_addr_count) < 10000:
 239             if all(t.is_indexed for t in name_partials):
 240                 lookup = [dbf.FieldLookup('name_vector',
 241                                           [t.token for t in name_partials], 'lookup_all')]
 242             else:
 243                 # we don't have the partials, try with the non-rare names
 244                 non_rare_names = [t.token for t in name_fulls if t.count >= 1000]
 245                 if not non_rare_names:
 246                     return
 247                 lookup = [dbf.FieldLookup('name_vector', non_rare_names, 'lookup_any')]
 248             if addr_tokens:
 249                 lookup.append(dbf.FieldLookup('nameaddress_vector', addr_tokens, 'lookup_all'))
 250             yield penalty + 0.1 * max(0, 5 - len(name_partials) - len(addr_tokens)),\
 251                   min(exp_name_count, exp_addr_count), lookup
 252
 253
 254     def get_name_ranking(self, trange: TokenRange) -> dbf.FieldRanking:
 255         """ Create a ranking expression for a name term in the given range.
 256         """
 257         name_fulls = self.query.get_tokens(trange, TokenType.WORD)
 258         ranks = [dbf.RankedTokens(t.penalty, [t.token]) for t in name_fulls]
 259         ranks.sort(key=lambda r: r.penalty)
 260         # Fallback, sum of penalty for partials
 261         name_partials = self.query.get_partials_list(trange)
 262         default = sum(t.penalty for t in name_partials) + 0.2
 263         return dbf.FieldRanking('name_vector', default, ranks)
 264
 265
 266     def get_addr_ranking(self, trange: TokenRange) -> dbf.FieldRanking:
 267         """ Create a list of ranking expressions for an address term
 268             for the given ranges.
 269         """
 270         todo: List[Tuple[int, int, dbf.RankedTokens]] = []
 271         heapq.heappush(todo, (0, trange.start, dbf.RankedTokens(0.0, [])))
 272         ranks: List[dbf.RankedTokens] = []
 273
 274         while todo: # pylint: disable=too-many-nested-blocks
 275             neglen, pos, rank = heapq.heappop(todo)
 276             for tlist in self.query.nodes[pos].starting:
 277                 if tlist.ttype in (TokenType.PARTIAL, TokenType.WORD):
 278                     if tlist.end < trange.end:
 279                         chgpenalty = PENALTY_WORDCHANGE[self.query.nodes[tlist.end].btype]
 280                         if tlist.ttype == TokenType.PARTIAL:
 281                             penalty = rank.penalty + chgpenalty \
 282                                       + max(t.penalty for t in tlist.tokens)
 283                             heapq.heappush(todo, (neglen - 1, tlist.end,
 284                                                   dbf.RankedTokens(penalty, rank.tokens)))
 285                         else:
 286                             for t in tlist.tokens:
 287                                 heapq.heappush(todo, (neglen - 1, tlist.end,
 288                                                       rank.with_token(t, chgpenalty)))
 289                     elif tlist.end == trange.end:
 290                         if tlist.ttype == TokenType.PARTIAL:
 291                             ranks.append(dbf.RankedTokens(rank.penalty
 292                                                           + max(t.penalty for t in tlist.tokens),
 293                                                           rank.tokens))
 294                         else:
 295                             ranks.extend(rank.with_token(t, 0.0) for t in tlist.tokens)
 296                         if len(ranks) >= 10:
 297                             # Too many variants, bail out and only add
 298                             # Worst-case Fallback: sum of penalty of partials
 299                             name_partials = self.query.get_partials_list(trange)
 300                             default = sum(t.penalty for t in name_partials) + 0.2
 301                             ranks.append(dbf.RankedTokens(rank.penalty + default, []))
 302                             # Bail out of outer loop
 303                             todo.clear()
 304                             break
 305
 306         ranks.sort(key=lambda r: len(r.tokens))
 307         default = ranks[0].penalty + 0.3
 308         del ranks[0]
 309         ranks.sort(key=lambda r: r.penalty)
 310
 311         return dbf.FieldRanking('nameaddress_vector', default, ranks)
 312
 313
 314     def get_search_data(self, assignment: TokenAssignment) -> Optional[dbf.SearchData]:
 315         """ Collect the tokens for the non-name search fields in the
 316             assignment.
 317         """
 318         sdata = dbf.SearchData()
 319         sdata.penalty = assignment.penalty
 320         if assignment.country:
 321             tokens = self.query.get_tokens(assignment.country, TokenType.COUNTRY)
 322             if self.details.countries:
 323                 tokens = [t for t in tokens if t.lookup_word in self.details.countries]
 324                 if not tokens:
 325                     return None
 326             sdata.set_strings('countries', tokens)
 327         elif self.details.countries:
 328             sdata.countries = dbf.WeightedStrings(self.details.countries,
 329                                                   [0.0] * len(self.details.countries))
 330         if assignment.housenumber:
 331             sdata.set_strings('housenumbers',
 332                               self.query.get_tokens(assignment.housenumber,
 333                                                     TokenType.HOUSENUMBER))
 334         if assignment.postcode:
 335             sdata.set_strings('postcodes',
 336                               self.query.get_tokens(assignment.postcode,
 337                                                     TokenType.POSTCODE))
 338         if assignment.qualifier:
 339             sdata.set_qualifiers(self.query.get_tokens(assignment.qualifier,
 340                                                        TokenType.QUALIFIER))
 341
 342         if assignment.address:
 343             sdata.set_ranking([self.get_addr_ranking(r) for r in assignment.address])
 344         else:
 345             sdata.rankings = []
 346
 347         return sdata
 348
 349
 350     def get_search_categories(self,
 351                               assignment: TokenAssignment) -> Optional[dbf.WeightedCategories]:
 352         """ Collect tokens for category search or use the categories
 353             requested per parameter.
 354             Returns None if no category search is requested.
 355         """
 356         if assignment.category:
 357             tokens = [t for t in self.query.get_tokens(assignment.category,
 358                                                        TokenType.CATEGORY)
 359                       if not self.details.categories
 360                          or t.get_category() in self.details.categories]
 361             return dbf.WeightedCategories([t.get_category() for t in tokens],
 362                                           [t.penalty for t in tokens])
 363
 364         if self.details.categories:
 365             return dbf.WeightedCategories(self.details.categories,
 366                                           [0.0] * len(self.details.categories))
 367
 368         return None
 369
 370
 371 PENALTY_WORDCHANGE = {
 372     BreakType.START: 0.0,
 373     BreakType.END: 0.0,
 374     BreakType.PHRASE: 0.0,
 375     BreakType.WORD: 0.1,
 376     BreakType.PART: 0.2,
 377     BreakType.TOKEN: 0.4
 378 }