+-- SPDX-License-Identifier: GPL-2.0-only
+--
+-- This file is part of Nominatim. (https://nominatim.org)
+--
+-- Copyright (C) 2022 by the Nominatim developer community.
+-- For a full list of authors see the git log.
+
+-- Get tokens used for searching the given place.
+--
+-- These are the tokens that will be saved in the search_name table.
+CREATE OR REPLACE FUNCTION token_get_name_search_tokens(info JSONB)
+ RETURNS INTEGER[]
+AS $$
+ SELECT (info->>'names')::INTEGER[]
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+-- Get tokens for matching the place name against others.
+--
+-- This should usually be restricted to full name tokens.
+CREATE OR REPLACE FUNCTION token_get_name_match_tokens(info JSONB)
+ RETURNS INTEGER[]
+AS $$
+ SELECT (info->>'names')::INTEGER[]
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+-- Return the housenumber tokens applicable for the place.
+CREATE OR REPLACE FUNCTION token_get_housenumber_search_tokens(info JSONB)
+ RETURNS INTEGER[]
+AS $$
+ SELECT (info->>'hnr_tokens')::INTEGER[]
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+-- Return the housenumber in the form that it can be matched during search.
+CREATE OR REPLACE FUNCTION token_normalized_housenumber(info JSONB)
+ RETURNS TEXT
+AS $$
+ SELECT info->>'hnr';
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+CREATE OR REPLACE FUNCTION token_has_addr_street(info JSONB)
+ RETURNS BOOLEAN
+AS $$
+ SELECT info->>'street' is not null;
+$$ LANGUAGE SQL IMMUTABLE;
+
+
+CREATE OR REPLACE FUNCTION token_has_addr_place(info JSONB)
+ RETURNS BOOLEAN
+AS $$
+ SELECT info->>'place_match' is not null;
+$$ LANGUAGE SQL IMMUTABLE;
+
+
+CREATE OR REPLACE FUNCTION token_matches_street(info JSONB, street_tokens INTEGER[])
+ RETURNS BOOLEAN
+AS $$
+ SELECT (info->>'street')::INTEGER[] && street_tokens
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+CREATE OR REPLACE FUNCTION token_matches_place(info JSONB, place_tokens INTEGER[])
+ RETURNS BOOLEAN
+AS $$
+ SELECT (info->>'place_match')::INTEGER[] && place_tokens
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+CREATE OR REPLACE FUNCTION token_addr_place_search_tokens(info JSONB)
+ RETURNS INTEGER[]
+AS $$
+ SELECT (info->>'place_search')::INTEGER[]
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+CREATE OR REPLACE FUNCTION token_get_address_keys(info JSONB)
+ RETURNS SETOF TEXT
+AS $$
+ SELECT * FROM jsonb_object_keys(info->'addr');
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+CREATE OR REPLACE FUNCTION token_get_address_search_tokens(info JSONB, key TEXT)
+ RETURNS INTEGER[]
+AS $$
+ SELECT (info->'addr'->key->>0)::INTEGER[];
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+CREATE OR REPLACE FUNCTION token_matches_address(info JSONB, key TEXT, tokens INTEGER[])
+ RETURNS BOOLEAN
+AS $$
+ SELECT (info->'addr'->key->>1)::INTEGER[] && tokens;
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+CREATE OR REPLACE FUNCTION token_get_postcode(info JSONB)
+ RETURNS TEXT
+AS $$
+ SELECT info->>'postcode';
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+-- Return token info that should be saved permanently in the database.
+CREATE OR REPLACE FUNCTION token_strip_info(info JSONB)
+ RETURNS JSONB
+AS $$
+ SELECT NULL::JSONB;
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+--------------- private functions ----------------------------------------------
+
-- Functions for term normalisation and access to the 'word' table.
CREATE OR REPLACE FUNCTION transliteration(text) RETURNS text
$$
LANGUAGE plpgsql;
+
-- Create housenumber tokens from an OSM addr:housenumber.
-- The housnumber is split at comma and semicolon as necessary.
-- The function returns the normalized form of the housenumber suitable
-- for comparison.
-CREATE OR REPLACE FUNCTION create_housenumber_id(housenumber TEXT)
- RETURNS TEXT
+CREATE OR REPLACE FUNCTION create_housenumbers(housenumbers TEXT[],
+ OUT tokens TEXT,
+ OUT normtext TEXT)
AS $$
-DECLARE
- normtext TEXT;
BEGIN
- SELECT array_to_string(array_agg(trans), ';')
- INTO normtext
- FROM (SELECT lookup_word as trans, getorcreate_housenumber_id(lookup_word)
+ SELECT array_to_string(array_agg(trans), ';'), array_agg(tid)::TEXT
+ INTO normtext, tokens
+ FROM (SELECT lookup_word as trans, getorcreate_housenumber_id(lookup_word) as tid
FROM (SELECT make_standard_name(h) as lookup_word
- FROM regexp_split_to_table(housenumber, '[,;]') h) x) y;
-
- return normtext;
+ FROM unnest(housenumbers) h) x) y;
END;
$$ LANGUAGE plpgsql STABLE STRICT;
+
CREATE OR REPLACE FUNCTION getorcreate_housenumber_id(lookup_word TEXT)
RETURNS INTEGER
AS $$
LANGUAGE plpgsql;
-CREATE OR REPLACE FUNCTION getorcreate_postcode_id(postcode TEXT)
- RETURNS INTEGER
+CREATE OR REPLACE FUNCTION create_postcode_id(postcode TEXT)
+ RETURNS BOOLEAN
AS $$
DECLARE
+ r RECORD;
lookup_token TEXT;
- lookup_word TEXT;
return_word_id INTEGER;
BEGIN
- lookup_word := upper(trim(postcode));
- lookup_token := ' ' || make_standard_name(lookup_word);
- SELECT min(word_id) FROM word
- WHERE word_token = lookup_token and word = lookup_word
+ lookup_token := ' ' || make_standard_name(postcode);
+ FOR r IN
+ SELECT word_id FROM word
+ WHERE word_token = lookup_token and word = postcode
and class='place' and type='postcode'
- INTO return_word_id;
- IF return_word_id IS NULL THEN
- return_word_id := nextval('seq_word');
- INSERT INTO word VALUES (return_word_id, lookup_token, lookup_word,
- 'place', 'postcode', null, 0);
- END IF;
- RETURN return_word_id;
-END;
-$$
-LANGUAGE plpgsql;
-
-
-CREATE OR REPLACE FUNCTION getorcreate_country(lookup_word TEXT,
- lookup_country_code varchar(2))
- RETURNS INTEGER
- AS $$
-DECLARE
- lookup_token TEXT;
- return_word_id INTEGER;
-BEGIN
- lookup_token := ' '||trim(lookup_word);
- SELECT min(word_id) FROM word
- WHERE word_token = lookup_token and country_code=lookup_country_code
- INTO return_word_id;
- IF return_word_id IS NULL THEN
- return_word_id := nextval('seq_word');
- INSERT INTO word VALUES (return_word_id, lookup_token, null,
- null, null, lookup_country_code, 0);
- END IF;
- RETURN return_word_id;
-END;
-$$
-LANGUAGE plpgsql;
-
-
-CREATE OR REPLACE FUNCTION getorcreate_amenity(lookup_word TEXT, normalized_word TEXT,
- lookup_class text, lookup_type text)
- RETURNS INTEGER
- AS $$
-DECLARE
- lookup_token TEXT;
- return_word_id INTEGER;
-BEGIN
- lookup_token := ' '||trim(lookup_word);
- SELECT min(word_id) FROM word
- WHERE word_token = lookup_token and word = normalized_word
- and class = lookup_class and type = lookup_type
- INTO return_word_id;
- IF return_word_id IS NULL THEN
- return_word_id := nextval('seq_word');
- INSERT INTO word VALUES (return_word_id, lookup_token, normalized_word,
- lookup_class, lookup_type, null, 0);
- END IF;
- RETURN return_word_id;
-END;
-$$
-LANGUAGE plpgsql;
-
+ LOOP
+ RETURN false;
+ END LOOP;
-CREATE OR REPLACE FUNCTION getorcreate_amenityoperator(lookup_word TEXT,
- normalized_word TEXT,
- lookup_class text,
- lookup_type text,
- op text)
- RETURNS INTEGER
- AS $$
-DECLARE
- lookup_token TEXT;
- return_word_id INTEGER;
-BEGIN
- lookup_token := ' '||trim(lookup_word);
- SELECT min(word_id) FROM word
- WHERE word_token = lookup_token and word = normalized_word
- and class = lookup_class and type = lookup_type and operator = op
- INTO return_word_id;
- IF return_word_id IS NULL THEN
- return_word_id := nextval('seq_word');
- INSERT INTO word VALUES (return_word_id, lookup_token, normalized_word,
- lookup_class, lookup_type, null, 0, op);
- END IF;
- RETURN return_word_id;
+ INSERT INTO word VALUES (nextval('seq_word'), lookup_token, postcode,
+ 'place', 'postcode', null, 0);
+ RETURN true;
END;
$$
LANGUAGE plpgsql;
LANGUAGE plpgsql;
-CREATE OR REPLACE FUNCTION getorcreate_name_id(lookup_word TEXT)
- RETURNS INTEGER
- AS $$
-DECLARE
-BEGIN
- RETURN getorcreate_name_id(lookup_word, '');
-END;
-$$
-LANGUAGE plpgsql;
-
-- Normalize a string and lookup its word ids (partial words).
CREATE OR REPLACE FUNCTION addr_ids_from_name(lookup_word TEXT)
RETURNS INTEGER[]
LANGUAGE plpgsql STABLE STRICT;
-CREATE OR REPLACE FUNCTION create_country(src HSTORE, country_code varchar(2))
- RETURNS VOID
- AS $$
-DECLARE
- s TEXT;
- w INTEGER;
- words TEXT[];
- item RECORD;
- j INTEGER;
-BEGIN
- FOR item IN SELECT (each(src)).* LOOP
-
- s := make_standard_name(item.value);
- w := getorcreate_country(s, country_code);
-
- words := regexp_split_to_array(item.value, E'[,;()]');
- IF array_upper(words, 1) != 1 THEN
- FOR j IN 1..array_upper(words, 1) LOOP
- s := make_standard_name(words[j]);
- IF s != '' THEN
- w := getorcreate_country(s, country_code);
- END IF;
- END LOOP;
- END IF;
- END LOOP;
-END;
-$$
-LANGUAGE plpgsql;
-
-
CREATE OR REPLACE FUNCTION make_keywords(src HSTORE)
RETURNS INTEGER[]
AS $$
s TEXT;
w INTEGER;
words TEXT[];
- item RECORD;
+ value TEXT;
j INTEGER;
BEGIN
result := '{}'::INTEGER[];
- FOR item IN SELECT (each(src)).* LOOP
-
- s := make_standard_name(item.value);
- w := getorcreate_name_id(s, item.value);
+ FOR value IN SELECT unnest(regexp_split_to_array(svals(src), E'[,;]')) LOOP
+ -- full name
+ s := make_standard_name(value);
+ w := getorcreate_name_id(s, value);
IF not(ARRAY[w] <@ result) THEN
result := result || w;
END IF;
- w := getorcreate_word_id(s);
-
- IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN
- result := result || w;
- END IF;
-
+ -- partial single-word terms
words := string_to_array(s, ' ');
IF array_upper(words, 1) IS NOT NULL THEN
FOR j IN 1..array_upper(words, 1) LOOP
END LOOP;
END IF;
- words := regexp_split_to_array(item.value, E'[,;()]');
- IF array_upper(words, 1) != 1 THEN
- FOR j IN 1..array_upper(words, 1) LOOP
- s := make_standard_name(words[j]);
- IF s != '' THEN
- w := getorcreate_word_id(s);
- IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN
- result := result || w;
- END IF;
+ -- consider parts before an opening braket a full word as well
+ words := regexp_split_to_array(value, E'[(]');
+ IF array_upper(words, 1) > 1 THEN
+ s := make_standard_name(words[1]);
+ IF s != '' THEN
+ w := getorcreate_name_id(s, words[1]);
+ IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN
+ result := result || w;
END IF;
- END LOOP;
+ END IF;
END IF;
- s := regexp_replace(item.value, '市$', '');
- IF s != item.value THEN
+ s := regexp_replace(value, '市$', '');
+ IF s != value THEN
s := make_standard_name(s);
IF s != '' THEN
- w := getorcreate_name_id(s, item.value);
+ w := getorcreate_name_id(s, value);
IF NOT (ARRAY[w] <@ result) THEN
result := result || w;
END IF;
END;
$$
LANGUAGE plpgsql;
-
-
-CREATE OR REPLACE FUNCTION create_poi_search_terms(obj_place_id BIGINT,
- in_partition SMALLINT,
- parent_place_id BIGINT,
- address HSTORE,
- country TEXT,
- housenumber TEXT,
- initial_name_vector INTEGER[],
- geometry GEOMETRY,
- OUT name_vector INTEGER[],
- OUT nameaddress_vector INTEGER[])
- AS $$
-DECLARE
- parent_name_vector INTEGER[];
- parent_address_vector INTEGER[];
- addr_place_ids INTEGER[];
-
- addr_item RECORD;
- parent_address_place_ids BIGINT[];
- filtered_address HSTORE;
-BEGIN
- nameaddress_vector := '{}'::INTEGER[];
-
- SELECT s.name_vector, s.nameaddress_vector
- INTO parent_name_vector, parent_address_vector
- FROM search_name s
- WHERE s.place_id = parent_place_id;
-
- -- Find all address tags that don't appear in the parent search names.
- SELECT hstore(array_agg(ARRAY[k, v])) INTO filtered_address
- FROM (SELECT skeys(address) as k, svals(address) as v) a
- WHERE not addr_ids_from_name(v) && parent_address_vector
- AND k not in ('country', 'street', 'place', 'postcode',
- 'housenumber', 'streetnumber', 'conscriptionnumber');
-
- -- Compute all search terms from the addr: tags.
- IF filtered_address IS NOT NULL THEN
- FOR addr_item IN
- SELECT * FROM
- get_places_for_addr_tags(in_partition, geometry, filtered_address, country)
- LOOP
- IF addr_item.place_id is null THEN
- nameaddress_vector := array_merge(nameaddress_vector,
- addr_item.keywords);
- CONTINUE;
- END IF;
-
- IF parent_address_place_ids is null THEN
- SELECT array_agg(parent_place_id) INTO parent_address_place_ids
- FROM place_addressline
- WHERE place_id = parent_place_id;
- END IF;
-
- IF not parent_address_place_ids @> ARRAY[addr_item.place_id] THEN
- nameaddress_vector := array_merge(nameaddress_vector,
- addr_item.keywords);
-
- INSERT INTO place_addressline (place_id, address_place_id, fromarea,
- isaddress, distance, cached_rank_address)
- VALUES (obj_place_id, addr_item.place_id, not addr_item.isguess,
- true, addr_item.distance, addr_item.rank_address);
- END IF;
- END LOOP;
- END IF;
-
- name_vector := initial_name_vector;
-
- -- Check if the parent covers all address terms.
- -- If not, create a search name entry with the house number as the name.
- -- This is unusual for the search_name table but prevents that the place
- -- is returned when we only search for the street/place.
-
- IF housenumber is not null and not nameaddress_vector <@ parent_address_vector THEN
- name_vector := array_merge(name_vector,
- ARRAY[getorcreate_housenumber_id(make_standard_name(housenumber))]);
- END IF;
-
- IF not address ? 'street' and address ? 'place' THEN
- addr_place_ids := addr_ids_from_name(address->'place');
- IF not addr_place_ids <@ parent_name_vector THEN
- -- make sure addr:place terms are always searchable
- nameaddress_vector := array_merge(nameaddress_vector, addr_place_ids);
- -- If there is a housenumber, also add the place name as a name,
- -- so we can search it by the usual housenumber+place algorithms.
- IF housenumber is not null THEN
- name_vector := array_merge(name_vector,
- ARRAY[getorcreate_name_id(make_standard_name(address->'place'))]);
- END IF;
- END IF;
- END IF;
-
- -- Cheating here by not recomputing all terms but simply using the ones
- -- from the parent object.
- nameaddress_vector := array_merge(nameaddress_vector, parent_name_vector);
- nameaddress_vector := array_merge(nameaddress_vector, parent_address_vector);
-
-END;
-$$
-LANGUAGE plpgsql;
projects / nominatim.git / blobdiff