ICU: better letter identification in normalization

[nominatim.git] / lib-sql / tokenizer / icu_tokenizer.sql

diff --git a/lib-sql/tokenizer/icu_tokenizer.sql b/lib-sql/tokenizer/icu_tokenizer.sql
index ffe6648c38e959c6279efb2d1898d835514f32a7..a3dac8ddcbe82eb5fd6057bd81bb9b823befa159 100644 (file)
--- a/lib-sql/tokenizer/icu_tokenizer.sql
+++ b/lib-sql/tokenizer/icu_tokenizer.sql
@@ -1,3 +1,10 @@
+-- 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.
@@ -34,40 +41,59 @@ AS $$
 $$ LANGUAGE SQL IMMUTABLE STRICT;
 
 
-CREATE OR REPLACE FUNCTION token_addr_street_match_tokens(info JSONB)
-  RETURNS INTEGER[]
+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 $$
-  SELECT (info->>'street')::INTEGER[]
+  SELECT info->>'place' is not null;
+$$ 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_addr_place_match_tokens(info JSONB)
-  RETURNS INTEGER[]
+CREATE OR REPLACE FUNCTION token_matches_place(info JSONB, place_tokens INTEGER[])
+  RETURNS BOOLEAN
 AS $$
-  SELECT (info->>'place_match')::INTEGER[]
+  SELECT (info->>'place')::INTEGER[] <@ place_tokens
 $$ LANGUAGE SQL IMMUTABLE STRICT;
 
 
 CREATE OR REPLACE FUNCTION token_addr_place_search_tokens(info JSONB)
   RETURNS INTEGER[]
 AS $$
-  SELECT (info->>'place_search')::INTEGER[]
+  SELECT (info->>'place')::INTEGER[]
 $$ LANGUAGE SQL IMMUTABLE STRICT;
 
 
-DROP TYPE IF EXISTS token_addresstoken CASCADE;
-CREATE TYPE token_addresstoken AS (
-  key TEXT,
-  match_tokens INT[],
-  search_tokens INT[]
-);
+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_tokens(info JSONB)
-  RETURNS SETOF token_addresstoken
+
+CREATE OR REPLACE FUNCTION token_get_address_search_tokens(info JSONB, key TEXT)
+  RETURNS INTEGER[]
+AS $$
+  SELECT (info->'addr'->>key)::INTEGER[];
+$$ LANGUAGE SQL IMMUTABLE STRICT;
+
+
+CREATE OR REPLACE FUNCTION token_matches_address(info JSONB, key TEXT, tokens INTEGER[])
+  RETURNS BOOLEAN
 AS $$
-  SELECT key, (value->>1)::int[] as match_tokens,
-         (value->>0)::int[] as search_tokens
-  FROM jsonb_each(info->'addr');
+  SELECT (info->'addr'->>key)::INTEGER[] <@ tokens;
 $$ LANGUAGE SQL IMMUTABLE STRICT;
 
 
@@ -127,15 +153,34 @@ BEGIN
         VALUES (term_id, term, 'w', json_build_object('count', term_count));
     END IF;
 
-    IF term_count < {{ max_word_freq }} THEN
-      partial_tokens := array_merge(partial_tokens, ARRAY[term_id]);
-    END IF;
+    partial_tokens := array_merge(partial_tokens, ARRAY[term_id]);
   END LOOP;
 END;
 $$
 LANGUAGE plpgsql;
 
 
+CREATE OR REPLACE FUNCTION getorcreate_partial_word(partial TEXT)
+  RETURNS INTEGER
+  AS $$
+DECLARE
+  token INTEGER;
+BEGIN
+  SELECT min(word_id) INTO token
+    FROM word WHERE word_token = partial and type = 'w';
+
+  IF token IS NULL THEN
+    token := nextval('seq_word');
+    INSERT INTO word (word_id, word_token, type, info)
+        VALUES (token, partial, 'w', json_build_object('count', 0));
+  END IF;
+
+  RETURN token;
+END;
+$$
+LANGUAGE plpgsql;
+
+
 CREATE OR REPLACE FUNCTION getorcreate_hnr_id(lookup_term TEXT)
   RETURNS INTEGER
   AS $$
@@ -155,3 +200,26 @@ BEGIN
 END;
 $$
 LANGUAGE plpgsql;
+
+
+CREATE OR REPLACE FUNCTION create_analyzed_hnr_id(norm_term TEXT, lookup_terms TEXT[])
+  RETURNS INTEGER
+  AS $$
+DECLARE
+  return_id INTEGER;
+BEGIN
+  SELECT min(word_id) INTO return_id
+    FROM word WHERE word = norm_term and type = 'H';
+
+  IF return_id IS NULL THEN
+    return_id := nextval('seq_word');
+    INSERT INTO word (word_id, word_token, type, word, info)
+      SELECT return_id, lookup_term, 'H', norm_term,
+             json_build_object('lookup', lookup_terms[1])
+        FROM unnest(lookup_terms) as lookup_term;
+  END IF;
+
+  RETURN return_id;
+END;
+$$
+LANGUAGE plpgsql;