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 Dataclasses for search results and helper functions to fill them.
10 Data classes are part of the public API while the functions are for
11 internal use only. That's why they are implemented as free-standing functions
12 instead of member functions.
14 from typing import Optional, Tuple, Dict, Sequence, TypeVar, Type, List, cast, Callable
19 import sqlalchemy as sa
21 from .typing import SaSelect, SaRow
22 from .sql.sqlalchemy_types import Geometry
23 from .types import Point, Bbox, LookupDetails
24 from .connection import SearchConnection
25 from .logging import log
26 from .localization import Locales
28 # This file defines complex result data classes.
29 # pylint: disable=too-many-instance-attributes
31 def _mingle_name_tags(names: Optional[Dict[str, str]]) -> Optional[Dict[str, str]]:
32 """ Mix-in names from linked places, so that they show up
33 as standard names where necessary.
39 for k, v in names.items():
40 if k.startswith('_place_'):
42 out[k if outkey in names else outkey] = v
49 class SourceTable(enum.Enum):
50 """ The `SourceTable` type lists the possible sources a result can have.
53 """ The placex table is the main source for result usually containing
57 """ The osmline table contains address interpolations from OSM data.
58 Interpolation addresses are always approximate. The OSM id in the
59 result refers to the OSM way with the interpolation line object.
62 """ TIGER address data contains US addresses imported on the side,
63 see [Installing TIGER data](../customize/Tiger.md).
64 TIGER address are also interpolations. The addresses always refer
65 to a street from OSM data. The OSM id in the result refers to
69 """ The postcode table contains artificial centroids for postcodes,
70 computed from the postcodes available with address points. Results
71 are always approximate.
74 """ The country table provides a fallback, when country data is missing
79 @dataclasses.dataclass
81 """ The `AddressLine` may contain the following fields about a related place
82 and its function as an address object. Most fields are optional.
83 Their presence depends on the kind and function of the address part.
85 category: Tuple[str, str]
86 """ Main category of the place, described by a key-value pair.
89 """ All available names for the place including references, alternative
90 names and translations.
93 """ If true, then the exact area of the place is known. Without area
94 information, Nominatim has to make an educated guess if an address
95 belongs to one place or another.
98 """ If true, this place should be considered for the final address display.
99 Nominatim will sometimes include more than one candidate for
100 the address in the list when it cannot reliably determine where the
101 place belongs. It will consider names of all candidates when searching
102 but when displaying the result, only the most likely candidate should
106 """ [Address rank](../customize/Ranking.md#address-rank) of the place.
109 """ Distance in degrees between the result place and this address part.
111 place_id: Optional[int] = None
112 """ Internal ID of the place.
114 osm_object: Optional[Tuple[str, int]] = None
115 """ OSM type and ID of the place, if such an object exists.
117 extratags: Optional[Dict[str, str]] = None
118 """ Any extra information available about the place. This is a dictionary
119 that usually contains OSM tag key-value pairs.
122 admin_level: Optional[int] = None
123 """ The administrative level of a boundary as tagged in the input data.
124 This field is only meaningful for places of the category
125 (boundary, administrative).
128 local_name: Optional[str] = None
129 """ Place holder for localization of this address part. See
130 [Localization](#localization) below.
134 class AddressLines(List[AddressLine]):
135 """ Sequence of address lines order in descending order by their rank.
138 def localize(self, locales: Locales) -> List[str]:
139 """ Set the local name of address parts according to the chosen
140 locale. Return the list of local names without duplicates.
142 Only address parts that are marked as isaddress are localized
145 label_parts: List[str] = []
148 if line.isaddress and line.names:
149 line.local_name = locales.display_name(line.names)
150 if not label_parts or label_parts[-1] != line.local_name:
151 label_parts.append(line.local_name)
157 @dataclasses.dataclass
159 """ Each entry in the list of search terms contains the
160 following detailed information.
163 """ Internal identifier for the word.
166 """ Normalised and transliterated form of the word.
167 This form is used for searching.
169 word: Optional[str] = None
170 """ Untransliterated form, if available.
174 WordInfos = Sequence[WordInfo]
177 @dataclasses.dataclass
179 """ Data class collecting information common to all
180 types of search results.
182 source_table: SourceTable
183 category: Tuple[str, str]
186 place_id : Optional[int] = None
187 osm_object: Optional[Tuple[str, int]] = None
188 parent_place_id: Optional[int] = None
189 linked_place_id: Optional[int] = None
190 admin_level: int = 15
192 locale_name: Optional[str] = None
193 display_name: Optional[str] = None
195 names: Optional[Dict[str, str]] = None
196 address: Optional[Dict[str, str]] = None
197 extratags: Optional[Dict[str, str]] = None
199 housenumber: Optional[str] = None
200 postcode: Optional[str] = None
201 wikipedia: Optional[str] = None
203 rank_address: int = 30
204 rank_search: int = 30
205 importance: Optional[float] = None
207 country_code: Optional[str] = None
209 address_rows: Optional[AddressLines] = None
210 linked_rows: Optional[AddressLines] = None
211 parented_rows: Optional[AddressLines] = None
212 name_keywords: Optional[WordInfos] = None
213 address_keywords: Optional[WordInfos] = None
215 geometry: Dict[str, str] = dataclasses.field(default_factory=dict)
218 def lat(self) -> float:
219 """ Get the latitude (or y) of the center point of the place.
221 return self.centroid[1]
225 def lon(self) -> float:
226 """ Get the longitude (or x) of the center point of the place.
228 return self.centroid[0]
231 def calculated_importance(self) -> float:
232 """ Get a valid importance value. This is either the stored importance
233 of the value or an artificial value computed from the place's
236 return self.importance or (0.40001 - (self.rank_search/75.0))
239 def localize(self, locales: Locales) -> None:
240 """ Fill the locale_name and the display_name field for the
241 place and, if available, its address information.
243 self.locale_name = locales.display_name(self.names)
244 if self.address_rows:
245 self.display_name = ', '.join(self.address_rows.localize(locales))
247 self.display_name = self.locale_name
251 BaseResultT = TypeVar('BaseResultT', bound=BaseResult)
253 @dataclasses.dataclass
254 class DetailedResult(BaseResult):
255 """ A search result with more internal information from the database
258 indexed_date: Optional[dt.datetime] = None
261 @dataclasses.dataclass
262 class ReverseResult(BaseResult):
263 """ A search result for reverse geocoding.
265 distance: Optional[float] = None
266 bbox: Optional[Bbox] = None
269 class ReverseResults(List[ReverseResult]):
270 """ Sequence of reverse lookup results ordered by distance.
271 May be empty when no result was found.
275 @dataclasses.dataclass
276 class SearchResult(BaseResult):
277 """ A search result for forward geocoding.
279 bbox: Optional[Bbox] = None
280 accuracy: float = 0.0
284 def ranking(self) -> float:
285 """ Return the ranking, a combined measure of accuracy and importance.
287 return (self.accuracy if self.accuracy is not None else 1) \
288 - self.calculated_importance()
291 class SearchResults(List[SearchResult]):
292 """ Sequence of forward lookup results ordered by relevance.
293 May be empty when no result was found.
297 def _filter_geometries(row: SaRow) -> Dict[str, str]:
298 return {k[9:]: v for k, v in row._mapping.items() # pylint: disable=W0212
299 if k.startswith('geometry_')}
302 def create_from_placex_row(row: Optional[SaRow],
303 class_type: Type[BaseResultT]) -> Optional[BaseResultT]:
304 """ Construct a new result and add the data from the result row
305 from the placex table. 'class_type' defines the type of result
306 to return. Returns None if the row is None.
311 return class_type(source_table=SourceTable.PLACEX,
312 place_id=row.place_id,
313 osm_object=(row.osm_type, row.osm_id),
314 category=(row.class_, row.type),
315 parent_place_id = row.parent_place_id,
316 linked_place_id = getattr(row, 'linked_place_id', None),
317 admin_level = getattr(row, 'admin_level', 15),
318 names=_mingle_name_tags(row.name),
320 extratags=row.extratags,
321 housenumber=row.housenumber,
322 postcode=row.postcode,
323 wikipedia=row.wikipedia,
324 rank_address=row.rank_address,
325 rank_search=row.rank_search,
326 importance=row.importance,
327 country_code=row.country_code,
328 centroid=Point.from_wkb(row.centroid),
329 geometry=_filter_geometries(row))
332 def create_from_osmline_row(row: Optional[SaRow],
333 class_type: Type[BaseResultT]) -> Optional[BaseResultT]:
334 """ Construct a new result and add the data from the result row
335 from the address interpolation table osmline. 'class_type' defines
336 the type of result to return. Returns None if the row is None.
338 If the row contains a housenumber, then the housenumber is filled out.
339 Otherwise the result contains the interpolation information in extratags.
344 hnr = getattr(row, 'housenumber', None)
346 res = class_type(source_table=SourceTable.OSMLINE,
347 place_id=row.place_id,
348 parent_place_id = row.parent_place_id,
349 osm_object=('W', row.osm_id),
350 category=('place', 'houses' if hnr is None else 'house'),
352 postcode=row.postcode,
353 country_code=row.country_code,
354 centroid=Point.from_wkb(row.centroid),
355 geometry=_filter_geometries(row))
358 res.extratags = {'startnumber': str(row.startnumber),
359 'endnumber': str(row.endnumber),
360 'step': str(row.step)}
362 res.housenumber = str(hnr)
367 def create_from_tiger_row(row: Optional[SaRow],
368 class_type: Type[BaseResultT],
369 osm_type: Optional[str] = None,
370 osm_id: Optional[int] = None) -> Optional[BaseResultT]:
371 """ Construct a new result and add the data from the result row
372 from the Tiger data interpolation table. 'class_type' defines
373 the type of result to return. Returns None if the row is None.
375 If the row contains a housenumber, then the housenumber is filled out.
376 Otherwise the result contains the interpolation information in extratags.
381 hnr = getattr(row, 'housenumber', None)
383 res = class_type(source_table=SourceTable.TIGER,
384 place_id=row.place_id,
385 parent_place_id = row.parent_place_id,
386 osm_object=(osm_type or row.osm_type, osm_id or row.osm_id),
387 category=('place', 'houses' if hnr is None else 'house'),
388 postcode=row.postcode,
390 centroid=Point.from_wkb(row.centroid),
391 geometry=_filter_geometries(row))
394 res.extratags = {'startnumber': str(row.startnumber),
395 'endnumber': str(row.endnumber),
396 'step': str(row.step)}
398 res.housenumber = str(hnr)
403 def create_from_postcode_row(row: Optional[SaRow],
404 class_type: Type[BaseResultT]) -> Optional[BaseResultT]:
405 """ Construct a new result and add the data from the result row
406 from the postcode table. 'class_type' defines
407 the type of result to return. Returns None if the row is None.
412 return class_type(source_table=SourceTable.POSTCODE,
413 place_id=row.place_id,
414 parent_place_id = row.parent_place_id,
415 category=('place', 'postcode'),
416 names={'ref': row.postcode},
417 rank_search=row.rank_search,
418 rank_address=row.rank_address,
419 country_code=row.country_code,
420 centroid=Point.from_wkb(row.centroid),
421 geometry=_filter_geometries(row))
424 def create_from_country_row(row: Optional[SaRow],
425 class_type: Type[BaseResultT]) -> Optional[BaseResultT]:
426 """ Construct a new result and add the data from the result row
427 from the fallback country tables. 'class_type' defines
428 the type of result to return. Returns None if the row is None.
433 return class_type(source_table=SourceTable.COUNTRY,
434 category=('place', 'country'),
435 centroid=Point.from_wkb(row.centroid),
437 rank_address=4, rank_search=4,
438 country_code=row.country_code,
439 geometry=_filter_geometries(row))
442 async def add_result_details(conn: SearchConnection, results: List[BaseResultT],
443 details: LookupDetails) -> None:
444 """ Retrieve more details from the database according to the
445 parameters specified in 'details'.
448 log().section('Query details for result')
449 if details.address_details:
450 log().comment('Query address details')
451 await complete_address_details(conn, results)
452 if details.linked_places:
453 log().comment('Query linked places')
454 for result in results:
455 await complete_linked_places(conn, result)
456 if details.parented_places:
457 log().comment('Query parent places')
458 for result in results:
459 await complete_parented_places(conn, result)
461 log().comment('Query keywords')
462 for result in results:
463 await complete_keywords(conn, result)
464 for result in results:
465 result.localize(details.locales)
468 def _result_row_to_address_row(row: SaRow, isaddress: Optional[bool] = None) -> AddressLine:
469 """ Create a new AddressLine from the results of a database query.
471 extratags: Dict[str, str] = getattr(row, 'extratags', {}) or {}
472 if 'linked_place' in extratags:
473 extratags['place'] = extratags['linked_place']
475 names = _mingle_name_tags(row.name) or {}
476 if getattr(row, 'housenumber', None) is not None:
477 names['housenumber'] = row.housenumber
479 if isaddress is None:
480 isaddress = getattr(row, 'isaddress', True)
482 return AddressLine(place_id=row.place_id,
483 osm_object=None if row.osm_type is None else (row.osm_type, row.osm_id),
484 category=(getattr(row, 'class'), row.type),
487 admin_level=row.admin_level,
488 fromarea=row.fromarea,
490 rank_address=row.rank_address,
491 distance=row.distance)
494 def _get_address_lookup_id(result: BaseResultT) -> int:
495 assert result.place_id
496 if result.source_table != SourceTable.PLACEX or result.rank_search > 27:
497 return result.parent_place_id or result.place_id
499 return result.linked_place_id or result.place_id
502 async def _finalize_entry(conn: SearchConnection, result: BaseResultT) -> None:
503 assert result.address_rows is not None
504 if result.category[0] not in ('boundary', 'place')\
505 or result.category[1] not in ('postal_code', 'postcode'):
506 postcode = result.postcode
507 if not postcode and result.address:
508 postcode = result.address.get('postcode')
509 if postcode and ',' not in postcode and ';' not in postcode:
510 result.address_rows.append(AddressLine(
511 category=('place', 'postcode'),
512 names={'ref': postcode},
513 fromarea=False, isaddress=True, rank_address=5,
515 if result.country_code:
516 async def _get_country_names() -> Optional[Dict[str, str]]:
517 t = conn.t.country_name
518 sql = sa.select(t.c.name, t.c.derived_name)\
519 .where(t.c.country_code == result.country_code)
520 for cres in await conn.execute(sql):
521 names = cast(Dict[str, str], cres[0])
523 names.update(cast(Dict[str, str], cres[1]))
527 country_names = await conn.get_cached_value('COUNTRY_NAME',
531 result.address_rows.append(AddressLine(
532 category=('place', 'country'),
534 fromarea=False, isaddress=True, rank_address=4,
536 result.address_rows.append(AddressLine(
537 category=('place', 'country_code'),
538 names={'ref': result.country_code}, extratags = {},
539 fromarea=True, isaddress=False, rank_address=4,
543 def _setup_address_details(result: BaseResultT) -> None:
544 """ Retrieve information about places that make up the address of the result.
546 result.address_rows = AddressLines()
548 result.address_rows.append(AddressLine(
549 place_id=result.place_id,
550 osm_object=result.osm_object,
551 category=result.category,
553 extratags=result.extratags or {},
554 admin_level=result.admin_level,
555 fromarea=True, isaddress=True,
556 rank_address=result.rank_address, distance=0.0))
557 if result.source_table == SourceTable.PLACEX and result.address:
558 housenumber = result.address.get('housenumber')\
559 or result.address.get('streetnumber')\
560 or result.address.get('conscriptionnumber')
561 elif result.housenumber:
562 housenumber = result.housenumber
566 result.address_rows.append(AddressLine(
567 category=('place', 'house_number'),
568 names={'ref': housenumber},
569 fromarea=True, isaddress=True, rank_address=28, distance=0))
570 if result.address and '_unlisted_place' in result.address:
571 result.address_rows.append(AddressLine(
572 category=('place', 'locality'),
573 names={'name': result.address['_unlisted_place']},
574 fromarea=False, isaddress=True, rank_address=25, distance=0))
577 async def complete_address_details(conn: SearchConnection, results: List[BaseResultT]) -> None:
578 """ Retrieve information about places that make up the address of the result.
580 for result in results:
581 _setup_address_details(result)
583 ### Lookup entries from place_address line
585 lookup_ids = [{'pid': r.place_id,
586 'lid': _get_address_lookup_id(r),
587 'names': list(r.address.values()) if r.address else [],
588 'c': ('SRID=4326;' + r.centroid.to_wkt()) if r.centroid else '' }
589 for r in results if r.place_id]
594 ltab = sa.func.JsonArrayEach(sa.type_coerce(lookup_ids, sa.JSON))\
595 .table_valued(sa.column('value', type_=sa.JSON))
598 taddr = conn.t.addressline
600 sql = sa.select(ltab.c.value['pid'].as_integer().label('src_place_id'),
601 t.c.place_id, t.c.osm_type, t.c.osm_id, t.c.name,
602 t.c.class_, t.c.type, t.c.extratags,
603 t.c.admin_level, taddr.c.fromarea,
604 sa.case((t.c.rank_address == 11, 5),
605 else_=t.c.rank_address).label('rank_address'),
606 taddr.c.distance, t.c.country_code, t.c.postcode)\
607 .join(taddr, sa.or_(taddr.c.place_id == ltab.c.value['pid'].as_integer(),
608 taddr.c.place_id == ltab.c.value['lid'].as_integer()))\
609 .join(t, taddr.c.address_place_id == t.c.place_id)\
610 .order_by('src_place_id')\
611 .order_by(sa.column('rank_address').desc())\
612 .order_by((taddr.c.place_id == ltab.c.value['pid'].as_integer()).desc())\
613 .order_by(sa.case((sa.func.CrosscheckNames(t.c.name, ltab.c.value['names']), 2),
614 (taddr.c.isaddress, 0),
615 (sa.and_(taddr.c.fromarea,
616 t.c.geometry.ST_Contains(
617 sa.func.ST_GeomFromEWKT(
618 ltab.c.value['c'].as_string()))), 1),
620 .order_by(taddr.c.fromarea.desc())\
621 .order_by(taddr.c.distance.desc())\
622 .order_by(t.c.rank_search.desc())
625 current_result = None
626 current_rank_address = -1
627 for row in await conn.execute(sql):
628 if current_result is None or row.src_place_id != current_result.place_id:
629 current_result = next((r for r in results if r.place_id == row.src_place_id), None)
630 assert current_result is not None
631 current_rank_address = -1
633 location_isaddress = row.rank_address != current_rank_address
635 if current_result.country_code is None and row.country_code:
636 current_result.country_code = row.country_code
638 if row.type in ('postcode', 'postal_code') and location_isaddress:
639 if not row.fromarea or \
640 (current_result.address and 'postcode' in current_result.address):
641 location_isaddress = False
643 current_result.postcode = None
645 assert current_result.address_rows is not None
646 current_result.address_rows.append(_result_row_to_address_row(row, location_isaddress))
647 current_rank_address = row.rank_address
649 for result in results:
650 await _finalize_entry(conn, result)
653 ### Finally add the record for the parent entry where necessary.
655 parent_lookup_ids = list(filter(lambda e: e['pid'] != e['lid'], lookup_ids))
656 if parent_lookup_ids:
657 ltab = sa.func.JsonArrayEach(sa.type_coerce(parent_lookup_ids, sa.JSON))\
658 .table_valued(sa.column('value', type_=sa.JSON))
659 sql = sa.select(ltab.c.value['pid'].as_integer().label('src_place_id'),
660 t.c.place_id, t.c.osm_type, t.c.osm_id, t.c.name,
661 t.c.class_, t.c.type, t.c.extratags,
664 .where(t.c.place_id == ltab.c.value['lid'].as_integer())
666 for row in await conn.execute(sql):
667 current_result = next((r for r in results if r.place_id == row.src_place_id), None)
668 assert current_result is not None
669 assert current_result.address_rows is not None
671 current_result.address_rows.append(AddressLine(
672 place_id=row.place_id,
673 osm_object=(row.osm_type, row.osm_id),
674 category=(row.class_, row.type),
675 names=row.name, extratags=row.extratags or {},
676 admin_level=row.admin_level,
677 fromarea=True, isaddress=True,
678 rank_address=row.rank_address, distance=0.0))
680 ### Now sort everything
681 def mk_sort_key(place_id: Optional[int]) -> Callable[[AddressLine], Tuple[bool, int, bool]]:
682 return lambda a: (a.place_id != place_id, -a.rank_address, a.isaddress)
684 for result in results:
685 assert result.address_rows is not None
686 result.address_rows.sort(key=mk_sort_key(result.place_id))
689 def _placex_select_address_row(conn: SearchConnection,
690 centroid: Point) -> SaSelect:
692 return sa.select(t.c.place_id, t.c.osm_type, t.c.osm_id, t.c.name,
693 t.c.class_.label('class'), t.c.type,
694 t.c.admin_level, t.c.housenumber,
695 t.c.geometry.is_area().label('fromarea'),
697 t.c.geometry.distance_spheroid(
698 sa.bindparam('centroid', value=centroid, type_=Geometry)).label('distance'))
701 async def complete_linked_places(conn: SearchConnection, result: BaseResult) -> None:
702 """ Retrieve information about places that link to the result.
704 result.linked_rows = AddressLines()
705 if result.source_table != SourceTable.PLACEX:
708 sql = _placex_select_address_row(conn, result.centroid)\
709 .where(conn.t.placex.c.linked_place_id == result.place_id)
711 for row in await conn.execute(sql):
712 result.linked_rows.append(_result_row_to_address_row(row))
715 async def complete_keywords(conn: SearchConnection, result: BaseResult) -> None:
716 """ Retrieve information about the search terms used for this place.
718 Requires that the query analyzer was initialised to get access to
721 t = conn.t.search_name
722 sql = sa.select(t.c.name_vector, t.c.nameaddress_vector)\
723 .where(t.c.place_id == result.place_id)
725 result.name_keywords = []
726 result.address_keywords = []
728 t = conn.t.meta.tables['word']
729 sel = sa.select(t.c.word_id, t.c.word_token, t.c.word)
731 for name_tokens, address_tokens in await conn.execute(sql):
732 for row in await conn.execute(sel.where(t.c.word_id.in_(name_tokens))):
733 result.name_keywords.append(WordInfo(*row))
735 for row in await conn.execute(sel.where(t.c.word_id.in_(address_tokens))):
736 result.address_keywords.append(WordInfo(*row))
739 async def complete_parented_places(conn: SearchConnection, result: BaseResult) -> None:
740 """ Retrieve information about places that the result provides the
743 result.parented_rows = AddressLines()
744 if result.source_table != SourceTable.PLACEX:
747 sql = _placex_select_address_row(conn, result.centroid)\
748 .where(conn.t.placex.c.parent_place_id == result.place_id)\
749 .where(conn.t.placex.c.rank_search == 30)
751 for row in await conn.execute(sql):
752 result.parented_rows.append(_result_row_to_address_row(row))