ICU: add an index over word_ids

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

diff --git a/lib-sql/tokenizer/legacy_tokenizer.sql b/lib-sql/tokenizer/legacy_tokenizer.sql
index d78c800c7ee2a5332b1695d8b2ca6fa5233c95b3..2b734e6f2a95a5cfe97243e81e3b4c47485a3d92 100644 (file)
--- a/lib-sql/tokenizer/legacy_tokenizer.sql
+++ b/lib-sql/tokenizer/legacy_tokenizer.sql
@@ -7,6 +7,7 @@ AS $$
   SELECT (info->>'names')::INTEGER[]
 $$ LANGUAGE SQL IMMUTABLE STRICT;
 
   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.
 -- Get tokens for matching the place name against others.
 --
 -- This should usually be restricted to full name tokens.


@@ -17,6 +18,85 @@ AS $$
 $$ LANGUAGE SQL IMMUTABLE STRICT;
 
 
 $$ 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_normalized_postcode(postcode TEXT)
+  RETURNS TEXT
+AS $$
+  SELECT CASE WHEN postcode SIMILAR TO '%(,|;)%' THEN NULL ELSE upper(trim(postcode))END;
+$$ 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
 -- Return token info that should be saved permanently in the database.
 CREATE OR REPLACE FUNCTION token_strip_info(info JSONB)
   RETURNS JSONB


@@ -75,26 +155,25 @@ END;
 $$
 LANGUAGE plpgsql;
 
 $$
 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 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 $$
   AS $$

-DECLARE
-  normtext TEXT;

 BEGIN
 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 (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;
 
 END;
 $$ LANGUAGE plpgsql STABLE STRICT;
 

+

 CREATE OR REPLACE FUNCTION getorcreate_housenumber_id(lookup_word TEXT)
   RETURNS INTEGER
   AS $$
 CREATE OR REPLACE FUNCTION getorcreate_housenumber_id(lookup_word TEXT)
   RETURNS INTEGER
   AS $$


@@ -117,100 +196,26 @@ $$
 LANGUAGE plpgsql;
 
 
 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 $$
 DECLARE
   AS $$
 DECLARE

+  r RECORD;

   lookup_token TEXT;
   lookup_token TEXT;

-  lookup_word TEXT;

   return_word_id INTEGER;
 BEGIN
   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'
           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;
 $$
 LANGUAGE plpgsql;
 END;
 $$
 LANGUAGE plpgsql;


@@ -239,16 +244,6 @@ $$
 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[]
 -- Normalize a string and lookup its word ids (partial words).
 CREATE OR REPLACE FUNCTION addr_ids_from_name(lookup_word TEXT)
   RETURNS INTEGER[]


@@ -303,36 +298,6 @@ $$
 LANGUAGE plpgsql STABLE STRICT;
 
 
 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 $$
 CREATE OR REPLACE FUNCTION make_keywords(src HSTORE)
   RETURNS INTEGER[]
   AS $$


@@ -341,26 +306,21 @@ DECLARE
   s TEXT;
   w INTEGER;
   words TEXT[];
   s TEXT;
   w INTEGER;
   words TEXT[];

-  item RECORD;
+  value TEXT;

   j INTEGER;
 BEGIN
   result := '{}'::INTEGER[];
 
   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;
 
 
     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
     words := string_to_array(s, ' ');
     IF array_upper(words, 1) IS NOT NULL THEN
       FOR j IN 1..array_upper(words, 1) LOOP


@@ -373,24 +333,23 @@ BEGIN
       END LOOP;
     END IF;
 
       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 IF;

-      END LOOP;
+      END IF;

     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
       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;
         IF NOT (ARRAY[w] <@ result) THEN
           result := result || w;
         END IF;