nominatim/api/search/token_assignment.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 Create query interpretations where each vertice in the query is assigned
   9 a specific function (expressed as a token type).
  10 """
  11 from typing import Optional, List, Iterator
  12 import dataclasses
  13
  14 import nominatim.api.search.query as qmod
  15 from nominatim.api.logging import log
  16
  17 # pylint: disable=too-many-return-statements,too-many-branches
  18
  19 @dataclasses.dataclass
  20 class TypedRange:
  21     """ A token range for a specific type of tokens.
  22     """
  23     ttype: qmod.TokenType
  24     trange: qmod.TokenRange
  25
  26
  27 PENALTY_TOKENCHANGE = {
  28     qmod.BreakType.START: 0.0,
  29     qmod.BreakType.END: 0.0,
  30     qmod.BreakType.PHRASE: 0.0,
  31     qmod.BreakType.WORD: 0.1,
  32     qmod.BreakType.PART: 0.2,
  33     qmod.BreakType.TOKEN: 0.4
  34 }
  35
  36 TypedRangeSeq = List[TypedRange]
  37
  38 @dataclasses.dataclass
  39 class TokenAssignment: # pylint: disable=too-many-instance-attributes
  40     """ Representation of a possible assignment of token types
  41         to the tokens in a tokenized query.
  42     """
  43     penalty: float = 0.0
  44     name: Optional[qmod.TokenRange] = None
  45     address: List[qmod.TokenRange] = dataclasses.field(default_factory=list)
  46     housenumber: Optional[qmod.TokenRange] = None
  47     postcode: Optional[qmod.TokenRange] = None
  48     country: Optional[qmod.TokenRange] = None
  49     category: Optional[qmod.TokenRange] = None
  50     qualifier: Optional[qmod.TokenRange] = None
  51
  52
  53     @staticmethod
  54     def from_ranges(ranges: TypedRangeSeq) -> 'TokenAssignment':
  55         """ Create a new token assignment from a sequence of typed spans.
  56         """
  57         out = TokenAssignment()
  58         for token in ranges:
  59             if token.ttype == qmod.TokenType.PARTIAL:
  60                 out.address.append(token.trange)
  61             elif token.ttype == qmod.TokenType.HOUSENUMBER:
  62                 out.housenumber = token.trange
  63             elif token.ttype == qmod.TokenType.POSTCODE:
  64                 out.postcode = token.trange
  65             elif token.ttype == qmod.TokenType.COUNTRY:
  66                 out.country = token.trange
  67             elif token.ttype == qmod.TokenType.CATEGORY:
  68                 out.category = token.trange
  69             elif token.ttype == qmod.TokenType.QUALIFIER:
  70                 out.qualifier = token.trange
  71         return out
  72
  73
  74 class _TokenSequence:
  75     """ Working state used to put together the token assignements.
  76
  77         Represents an intermediate state while traversing the tokenized
  78         query.
  79     """
  80     def __init__(self, seq: TypedRangeSeq,
  81                  direction: int = 0, penalty: float = 0.0) -> None:
  82         self.seq = seq
  83         self.direction = direction
  84         self.penalty = penalty
  85
  86
  87     def __str__(self) -> str:
  88         seq = ''.join(f'[{r.trange.start} - {r.trange.end}: {r.ttype.name}]' for r in self.seq)
  89         return f'{seq} (dir: {self.direction}, penalty: {self.penalty})'
  90
  91
  92     @property
  93     def end_pos(self) -> int:
  94         """ Return the index of the global end of the current sequence.
  95         """
  96         return self.seq[-1].trange.end if self.seq else 0
  97
  98
  99     def has_types(self, *ttypes: qmod.TokenType) -> bool:
 100         """ Check if the current sequence contains any typed ranges of
 101             the given types.
 102         """
 103         return any(s.ttype in ttypes for s in self.seq)
 104
 105
 106     def is_final(self) -> bool:
 107         """ Return true when the sequence cannot be extended by any
 108             form of token anymore.
 109         """
 110         # Country and category must be the final term for left-to-right
 111         return len(self.seq) > 1 and \
 112                self.seq[-1].ttype in (qmod.TokenType.COUNTRY, qmod.TokenType.CATEGORY)
 113
 114
 115     def appendable(self, ttype: qmod.TokenType) -> Optional[int]:
 116         """ Check if the give token type is appendable to the existing sequence.
 117
 118             Returns None if the token type is not appendable, otherwise the
 119             new direction of the sequence after adding such a type. The
 120             token is not added.
 121         """
 122         if ttype == qmod.TokenType.WORD:
 123             return None
 124
 125         if not self.seq:
 126             # Append unconditionally to the empty list
 127             if ttype == qmod.TokenType.COUNTRY:
 128                 return -1
 129             if ttype in (qmod.TokenType.HOUSENUMBER, qmod.TokenType.QUALIFIER):
 130                 return 1
 131             return self.direction
 132
 133         # Name tokens are always acceptable and don't change direction
 134         if ttype == qmod.TokenType.PARTIAL:
 135             return self.direction
 136
 137         # Other tokens may only appear once
 138         if self.has_types(ttype):
 139             return None
 140
 141         if ttype == qmod.TokenType.HOUSENUMBER:
 142             if self.direction == 1:
 143                 if len(self.seq) == 1 and self.seq[0].ttype == qmod.TokenType.QUALIFIER:
 144                     return None
 145                 if len(self.seq) > 2 \
 146                    or self.has_types(qmod.TokenType.POSTCODE, qmod.TokenType.COUNTRY):
 147                     return None # direction left-to-right: housenumber must come before anything
 148             elif self.direction == -1 \
 149                  or self.has_types(qmod.TokenType.POSTCODE, qmod.TokenType.COUNTRY):
 150                 return -1 # force direction right-to-left if after other terms
 151
 152             return self.direction
 153
 154         if ttype == qmod.TokenType.POSTCODE:
 155             if self.direction == -1:
 156                 if self.has_types(qmod.TokenType.HOUSENUMBER, qmod.TokenType.QUALIFIER):
 157                     return None
 158                 return -1
 159             if self.direction == 1:
 160                 return None if self.has_types(qmod.TokenType.COUNTRY) else 1
 161             if self.has_types(qmod.TokenType.HOUSENUMBER, qmod.TokenType.QUALIFIER):
 162                 return 1
 163             return self.direction
 164
 165         if ttype == qmod.TokenType.COUNTRY:
 166             return None if self.direction == -1 else 1
 167
 168         if ttype == qmod.TokenType.CATEGORY:
 169             return self.direction
 170
 171         if ttype == qmod.TokenType.QUALIFIER:
 172             if self.direction == 1:
 173                 if (len(self.seq) == 1
 174                     and self.seq[0].ttype in (qmod.TokenType.PARTIAL, qmod.TokenType.CATEGORY)) \
 175                    or (len(self.seq) == 2
 176                        and self.seq[0].ttype == qmod.TokenType.CATEGORY
 177                        and self.seq[1].ttype == qmod.TokenType.PARTIAL):
 178                     return 1
 179                 return None
 180             if self.direction == -1:
 181                 return -1
 182
 183             tempseq = self.seq[1:] if self.seq[0].ttype == qmod.TokenType.CATEGORY else self.seq
 184             if len(tempseq) == 0:
 185                 return 1
 186             if len(tempseq) == 1 and self.seq[0].ttype == qmod.TokenType.HOUSENUMBER:
 187                 return None
 188             if len(tempseq) > 1 or self.has_types(qmod.TokenType.POSTCODE, qmod.TokenType.COUNTRY):
 189                 return -1
 190             return 0
 191
 192         return None
 193
 194
 195     def advance(self, ttype: qmod.TokenType, end_pos: int,
 196                 btype: qmod.BreakType) -> Optional['_TokenSequence']:
 197         """ Return a new token sequence state with the given token type
 198             extended.
 199         """
 200         newdir = self.appendable(ttype)
 201         if newdir is None:
 202             return None
 203
 204         if not self.seq:
 205             newseq = [TypedRange(ttype, qmod.TokenRange(0, end_pos))]
 206             new_penalty = 0.0
 207         else:
 208             last = self.seq[-1]
 209             if btype != qmod.BreakType.PHRASE and last.ttype == ttype:
 210                 # extend the existing range
 211                 newseq = self.seq[:-1] + [TypedRange(ttype, last.trange.replace_end(end_pos))]
 212                 new_penalty = 0.0
 213             else:
 214                 # start a new range
 215                 newseq = list(self.seq) + [TypedRange(ttype,
 216                                                       qmod.TokenRange(last.trange.end, end_pos))]
 217                 new_penalty = PENALTY_TOKENCHANGE[btype]
 218
 219         return _TokenSequence(newseq, newdir, self.penalty + new_penalty)
 220
 221
 222     def _adapt_penalty_from_priors(self, priors: int, new_dir: int) -> bool:
 223         if priors == 2:
 224             self.penalty += 1.0
 225         elif priors > 2:
 226             if self.direction == 0:
 227                 self.direction = new_dir
 228             else:
 229                 return False
 230
 231         return True
 232
 233
 234     def recheck_sequence(self) -> bool:
 235         """ Check that the sequence is a fully valid token assignment
 236             and addapt direction and penalties further if necessary.
 237
 238             This function catches some impossible assignments that need
 239             forward context and can therefore not be exluded when building
 240             the assignment.
 241         """
 242         # housenumbers may not be further than 2 words from the beginning.
 243         # If there are two words in front, give it a penalty.
 244         hnrpos = next((i for i, tr in enumerate(self.seq)
 245                        if tr.ttype == qmod.TokenType.HOUSENUMBER),
 246                       None)
 247         if hnrpos is not None:
 248             if self.direction != -1:
 249                 priors = sum(1 for t in self.seq[:hnrpos] if t.ttype == qmod.TokenType.PARTIAL)
 250                 if not self._adapt_penalty_from_priors(priors, -1):
 251                     return False
 252             if self.direction != 1:
 253                 priors = sum(1 for t in self.seq[hnrpos+1:] if t.ttype == qmod.TokenType.PARTIAL)
 254                 if not self._adapt_penalty_from_priors(priors, 1):
 255                     return False
 256
 257         return True
 258
 259
 260     def _get_assignments_postcode(self, base: TokenAssignment,
 261                                   query_len: int)  -> Iterator[TokenAssignment]:
 262         """ Yield possible assignments of Postcode searches with an
 263             address component.
 264         """
 265         assert base.postcode is not None
 266
 267         if (base.postcode.start == 0 and self.direction != -1)\
 268            or (base.postcode.end == query_len and self.direction != 1):
 269             log().comment('postcode search')
 270             # <address>,<postcode> should give preference to address search
 271             if base.postcode.start == 0:
 272                 penalty = self.penalty
 273                 self.direction = -1 # name searches are only possbile backwards
 274             else:
 275                 penalty = self.penalty + 0.1
 276                 self.direction = 1 # name searches are only possbile forewards
 277             yield dataclasses.replace(base, penalty=penalty)
 278
 279
 280     def get_assignments(self, query: qmod.QueryStruct) -> Iterator[TokenAssignment]:
 281         """ Yield possible assignments for the current sequence.
 282
 283             This function splits up general name assignments into name
 284             and address and yields all possible variants of that.
 285         """
 286         base = TokenAssignment.from_ranges(self.seq)
 287
 288         num_addr_tokens = sum(t.end - t.start for t in base.address)
 289         if num_addr_tokens > 50:
 290             return
 291
 292         # Postcode search (postcode-only search is covered in next case)
 293         if base.postcode is not None and base.address:
 294             yield from self._get_assignments_postcode(base, query.num_token_slots())
 295
 296         # Postcode or country-only search
 297         if not base.address:
 298             if not base.housenumber and (base.postcode or base.country or base.category):
 299                 log().comment('postcode/country search')
 300                 yield dataclasses.replace(base, penalty=self.penalty)
 301         else:
 302             # <postcode>,<address> should give preference to postcode search
 303             if base.postcode and base.postcode.start == 0:
 304                 self.penalty += 0.1
 305             # Use entire first word as name
 306             if self.direction != -1:
 307                 log().comment('first word = name')
 308                 yield dataclasses.replace(base, name=base.address[0],
 309                                           penalty=self.penalty,
 310                                           address=base.address[1:])
 311
 312             # Use entire last word as name
 313             if self.direction == -1 or (self.direction == 0 and len(base.address) > 1):
 314                 log().comment('last word = name')
 315                 yield dataclasses.replace(base, name=base.address[-1],
 316                                           penalty=self.penalty,
 317                                           address=base.address[:-1])
 318
 319             # variant for special housenumber searches
 320             if base.housenumber:
 321                 yield dataclasses.replace(base, penalty=self.penalty)
 322
 323             # Use beginning of first word as name
 324             if self.direction != -1:
 325                 first = base.address[0]
 326                 if (not base.housenumber or first.end >= base.housenumber.start)\
 327                    and (not base.qualifier or first.start >= base.qualifier.end):
 328                     base_penalty = self.penalty
 329                     if (base.housenumber and base.housenumber.start > first.start) \
 330                        or len(query.source) > 1:
 331                         base_penalty += 0.25
 332                     for i in range(first.start + 1, first.end):
 333                         name, addr = first.split(i)
 334                         penalty = base_penalty + PENALTY_TOKENCHANGE[query.nodes[i].btype]
 335                         log().comment(f'split first word = name ({i - first.start})')
 336                         yield dataclasses.replace(base, name=name, penalty=penalty,
 337                                                   address=[addr] + base.address[1:])
 338
 339             # Use end of last word as name
 340             if self.direction != 1:
 341                 last = base.address[-1]
 342                 if (not base.housenumber or last.start <= base.housenumber.end)\
 343                    and (not base.qualifier or last.end <= base.qualifier.start):
 344                     base_penalty = self.penalty
 345                     if base.housenumber and base.housenumber.start < last.start:
 346                         base_penalty += 0.4
 347                     if len(query.source) > 1:
 348                         base_penalty += 0.25
 349                     for i in range(last.start + 1, last.end):
 350                         addr, name = last.split(i)
 351                         penalty = base_penalty + PENALTY_TOKENCHANGE[query.nodes[i].btype]
 352                         log().comment(f'split last word = name ({i - last.start})')
 353                         yield dataclasses.replace(base, name=name, penalty=penalty,
 354                                                   address=base.address[:-1] + [addr])
 355
 356
 357
 358 def yield_token_assignments(query: qmod.QueryStruct) -> Iterator[TokenAssignment]:
 359     """ Return possible word type assignments to word positions.
 360
 361         The assignments are computed from the concrete tokens listed
 362         in the tokenized query.
 363
 364         The result includes the penalty for transitions from one word type to
 365         another. It does not include penalties for transitions within a
 366         type.
 367     """
 368     todo = [_TokenSequence([], direction=0 if query.source[0].ptype == qmod.PhraseType.NONE else 1)]
 369
 370     while todo:
 371         state = todo.pop()
 372         node = query.nodes[state.end_pos]
 373
 374         for tlist in node.starting:
 375             newstate = state.advance(tlist.ttype, tlist.end, node.btype)
 376             if newstate is not None:
 377                 if newstate.end_pos == query.num_token_slots():
 378                     if newstate.recheck_sequence():
 379                         log().var_dump('Assignment', newstate)
 380                         yield from newstate.get_assignments(query)
 381                 elif not newstate.is_final():
 382                     todo.append(newstate)