ignore frequent partial search terms

[nominatim.git] / sql / functions.sql

diff --git a/sql/functions.sql b/sql/functions.sql
index 548362ad21230775dc9d0bd889cfa416d39649ec..7faadc167c159a7b82af887dd1ba13a94090ca76 100644 (file)
--- a/sql/functions.sql
+++ b/sql/functions.sql
@@ -83,18 +83,24 @@ END;
 $$
 LANGUAGE 'plpgsql' IMMUTABLE;
 
 $$
 LANGUAGE 'plpgsql' IMMUTABLE;
 

+-- returns NULL if the word is too common

 CREATE OR REPLACE FUNCTION getorcreate_word_id(lookup_word TEXT) 
   RETURNS INTEGER
   AS $$
 DECLARE
   lookup_token TEXT;
   return_word_id INTEGER;
 CREATE OR REPLACE FUNCTION getorcreate_word_id(lookup_word TEXT) 
   RETURNS INTEGER
   AS $$
 DECLARE
   lookup_token TEXT;
   return_word_id INTEGER;

+  count INTEGER;

 BEGIN
   lookup_token := trim(lookup_word);
 BEGIN
   lookup_token := trim(lookup_word);

-  SELECT min(word_id) FROM word WHERE word_token = lookup_token and class is null and type is null into return_word_id;
+  SELECT min(word_id), max(search_name_count) FROM word WHERE word_token = lookup_token and class is null and type is null into return_word_id, count;

   IF return_word_id IS NULL THEN
     return_word_id := nextval('seq_word');
     INSERT INTO word VALUES (return_word_id, lookup_token, regexp_replace(lookup_token,E'([^0-9])\\1+',E'\\1','g'), null, null, null, null, 0, null);
   IF return_word_id IS NULL THEN
     return_word_id := nextval('seq_word');
     INSERT INTO word VALUES (return_word_id, lookup_token, regexp_replace(lookup_token,E'([^0-9])\\1+',E'\\1','g'), null, null, null, null, 0, null);

+  ELSE
+    IF count > get_maxwordfreq() THEN
+      return_word_id := NULL;
+    END IF;

   END IF;
   RETURN return_word_id;
 END;
   END IF;
   RETURN return_word_id;
 END;


@@ -317,7 +323,7 @@ BEGIN
       FOR j IN 1..array_upper(words, 1) LOOP
         IF (words[j] != '') THEN
           w = getorcreate_word_id(words[j]);
       FOR j IN 1..array_upper(words, 1) LOOP
         IF (words[j] != '') THEN
           w = getorcreate_word_id(words[j]);

-          IF NOT (ARRAY[w] <@ result) THEN
+          IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN

             result := result || w;
           END IF;
         END IF;
             result := result || w;
           END IF;
         END IF;


@@ -330,7 +336,7 @@ BEGIN
         s := make_standard_name(words[j]);
         IF s != '' THEN
           w := getorcreate_word_id(s);
         s := make_standard_name(words[j]);
         IF s != '' THEN
           w := getorcreate_word_id(s);

-          IF NOT (ARRAY[w] <@ result) THEN
+          IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN

             result := result || w;
           END IF;
         END IF;
             result := result || w;
           END IF;
         END IF;


@@ -379,7 +385,7 @@ BEGIN
     FOR j IN 1..array_upper(words, 1) LOOP
       IF (words[j] != '') THEN
         w = getorcreate_word_id(words[j]);
     FOR j IN 1..array_upper(words, 1) LOOP
       IF (words[j] != '') THEN
         w = getorcreate_word_id(words[j]);

-        IF NOT (ARRAY[w] <@ result) THEN
+        IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN

           result := result || w;
         END IF;
       END IF;
           result := result || w;
         END IF;
       END IF;


@@ -392,7 +398,7 @@ BEGIN
       s := make_standard_name(words[j]);
       IF s != '' THEN
         w := getorcreate_word_id(s);
       s := make_standard_name(words[j]);
       IF s != '' THEN
         w := getorcreate_word_id(s);

-        IF NOT (ARRAY[w] <@ result) THEN
+        IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN

           result := result || w;
         END IF;
       END IF;
           result := result || w;
         END IF;
       END IF;


@@ -1234,6 +1240,13 @@ DECLARE
   result BOOLEAN;
 BEGIN
 
   result BOOLEAN;
 BEGIN
 

+  -- deferred delete
+  IF OLD.indexed_status = 100 THEN
+    --DEBUG: RAISE WARNING 'placex_update_delete % %',NEW.osm_type,NEW.osm_id;
+    delete from placex where place_id = OLD.place_id;
+    RETURN NULL;
+  END IF;
+

   IF NEW.indexed_status != 0 OR OLD.indexed_status = 0 OR NEW.linked_place_id is not null THEN
     RETURN NEW;
   END IF;
   IF NEW.indexed_status != 0 OR OLD.indexed_status = 0 OR NEW.linked_place_id is not null THEN
     RETURN NEW;
   END IF;


@@ -1248,13 +1261,6 @@ BEGIN
     RETURN NEW;
   END IF;
 
     RETURN NEW;
   END IF;
 

-  -- deferred delete
-  IF OLD.indexed_status = 100 THEN
-    --DEBUG: RAISE WARNING 'placex_update_delete % %',NEW.osm_type,NEW.osm_id;
-    delete from placex where place_id = OLD.place_id;
-    RETURN NULL;
-  END IF;
-

   IF OLD.indexed_status != 0 THEN
     --DEBUG: RAISE WARNING 'placex_update_0 % %',NEW.osm_type,NEW.osm_id;
 
   IF OLD.indexed_status != 0 THEN
     --DEBUG: RAISE WARNING 'placex_update_0 % %',NEW.osm_type,NEW.osm_id;
 


@@ -1504,23 +1510,29 @@ BEGIN
 -- RAISE WARNING 'get_osm_rel_members, label';
       FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['label']) as member LOOP
 
 -- RAISE WARNING 'get_osm_rel_members, label';
       FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['label']) as member LOOP
 

-        select * from placex where osm_type = upper(substring(relMember.member,1,1))::char(1) 
-          and osm_id = substring(relMember.member,2,10000)::bigint order by rank_search desc limit 1 into linkedPlacex;
+        FOR linkedPlacex IN select * from placex where osm_type = upper(substring(relMember.member,1,1))::char(1) 
+          and osm_id = substring(relMember.member,2,10000)::bigint order by rank_search desc limit 1 LOOP

 
 

-        -- If we don't already have one use this as the centre point of the geometry
-        IF NEW.centroid IS NULL THEN
-          NEW.centroid := coalesce(linkedPlacex.centroid,st_centroid(linkedPlacex.geometry));
-        END IF;
+          -- If we don't already have one use this as the centre point of the geometry
+          IF NEW.centroid IS NULL THEN
+            NEW.centroid := coalesce(linkedPlacex.centroid,st_centroid(linkedPlacex.geometry));
+          END IF;

 
 

-        -- merge in the label name, re-init word vector
-        NEW.name := linkedPlacex.name || NEW.name;
-        name_vector := make_keywords(NEW.name);
+          -- merge in the label name, re-init word vector
+          IF NOT linkedPlacex.name IS NULL THEN
+            NEW.name := linkedPlacex.name || NEW.name;
+            name_vector := make_keywords(NEW.name);
+          END IF;
+
+          -- merge in extra tags
+          IF NOT linkedPlacex.extratags IS NULL THEN
+            NEW.extratags := linkedPlacex.extratags || NEW.extratags;
+          END IF;

 
 

-        -- merge in extra tags
-        NEW.extratags := linkedPlacex.extratags || NEW.extratags;
+          -- mark the linked place (excludes from search results)
+          UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id;

 
 

-        -- mark the linked place (excludes from search results)
-        UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id;
+        END LOOP;

 
       END LOOP;
 
 
       END LOOP;
 


@@ -1528,33 +1540,39 @@ BEGIN
 
         FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['admin_center','admin_centre']) as member LOOP
 
 
         FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['admin_center','admin_centre']) as member LOOP
 

-          select * from placex where osm_type = upper(substring(relMember.member,1,1))::char(1) 
-            and osm_id = substring(relMember.member,2,10000)::bigint order by rank_search desc limit 1 into linkedPlacex;
+          FOR linkedPlacex IN select * from placex where osm_type = upper(substring(relMember.member,1,1))::char(1) 
+            and osm_id = substring(relMember.member,2,10000)::bigint order by rank_search desc limit 1 LOOP

 
 

-          -- For an admin centre we also want a name match - still not perfect, for example 'new york, new york'
-          -- But that can be fixed by explicitly setting the label in the data
-          IF make_standard_name(NEW.name->'name') = make_standard_name(linkedPlacex.name->'name') 
-            AND NEW.rank_search = linkedPlacex.rank_search THEN
+            -- For an admin centre we also want a name match - still not perfect, for example 'new york, new york'
+            -- But that can be fixed by explicitly setting the label in the data
+            IF make_standard_name(NEW.name->'name') = make_standard_name(linkedPlacex.name->'name') 
+              AND NEW.rank_address = linkedPlacex.rank_address THEN

 
 
 
 

-            -- If we don't already have one use this as the centre point of the geometry
-            IF NEW.centroid IS NULL THEN
-              NEW.centroid := coalesce(linkedPlacex.centroid,st_centroid(linkedPlacex.geometry));
-            END IF;
+              -- If we don't already have one use this as the centre point of the geometry
+              IF NEW.centroid IS NULL THEN
+                NEW.centroid := coalesce(linkedPlacex.centroid,st_centroid(linkedPlacex.geometry));
+              END IF;

 
 

-            -- merge in the name, re-init word vector
-            NEW.name := linkedPlacex.name || NEW.name;
-            name_vector := make_keywords(NEW.name);
+              -- merge in the name, re-init word vector
+              IF NOT linkedPlacex.name IS NULL THEN
+                NEW.name := linkedPlacex.name || NEW.name;
+                name_vector := make_keywords(NEW.name);
+              END IF;

 
 

-            -- merge in extra tags
-            NEW.extratags := linkedPlacex.extratags || NEW.extratags;
+              -- merge in extra tags
+              IF NOT linkedPlacex.extratags IS NULL THEN
+                NEW.extratags := linkedPlacex.extratags || NEW.extratags;
+              END IF;

 
 

-            -- mark the linked place (excludes from search results)
-            UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id;
+              -- mark the linked place (excludes from search results)
+              UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id;

 
 

-            -- keep a note of the node id in case we need it for wikipedia in a bit
-            linked_node_id := linkedPlacex.osm_id;
-          END IF;
+              -- keep a note of the node id in case we need it for wikipedia in a bit
+              linked_node_id := linkedPlacex.osm_id;
+            END IF;
+
+          END LOOP;

 
         END LOOP;
 
 
         END LOOP;
 


@@ -1565,7 +1583,7 @@ BEGIN
 
         FOR linkedPlacex IN select placex.* from placex WHERE
           make_standard_name(name->'name') = make_standard_name(NEW.name->'name')
 
         FOR linkedPlacex IN select placex.* from placex WHERE
           make_standard_name(name->'name') = make_standard_name(NEW.name->'name')

-          AND placex.rank_search = NEW.rank_search
+          AND placex.rank_address = NEW.rank_address

           AND placex.place_id != NEW.place_id
           AND placex.osm_type = 'N'::char(1) AND placex.rank_search < 26
           AND st_covers(NEW.geometry, placex.geometry)
           AND placex.place_id != NEW.place_id
           AND placex.osm_type = 'N'::char(1) AND placex.rank_search < 26
           AND st_covers(NEW.geometry, placex.geometry)


@@ -1639,6 +1657,8 @@ BEGIN
     location_rank_search := 0;
     location_distance := 0;
     location_parent := NULL;
     location_rank_search := 0;
     location_distance := 0;
     location_parent := NULL;

+    -- added ourself as address already
+    address_havelevel[NEW.rank_address] := true;

     -- RAISE WARNING '  getNearFeatures(%,''%'',%,''%'')',NEW.partition, place_centroid, search_maxrank, isin_tokens;
     FOR location IN SELECT * from getNearFeatures(NEW.partition, place_centroid, search_maxrank, isin_tokens) LOOP
 
     -- RAISE WARNING '  getNearFeatures(%,''%'',%,''%'')',NEW.partition, place_centroid, search_maxrank, isin_tokens;
     FOR location IN SELECT * from getNearFeatures(NEW.partition, place_centroid, search_maxrank, isin_tokens) LOOP
 


@@ -1763,9 +1783,9 @@ DECLARE
 BEGIN
   -- RAISE WARNING 'placex_delete % %',OLD.osm_type,OLD.osm_id;
 
 BEGIN
   -- RAISE WARNING 'placex_delete % %',OLD.osm_type,OLD.osm_id;
 

-  update placex set linked_place_id = null where linked_place_id = OLD.place_id;
+  update placex set linked_place_id = null, indexed_status = 2 where linked_place_id = OLD.place_id and indexed_status = 0;

   --DEBUG: RAISE WARNING 'placex_delete:01 % %',OLD.osm_type,OLD.osm_id;
   --DEBUG: RAISE WARNING 'placex_delete:01 % %',OLD.osm_type,OLD.osm_id;

-  update placex set indexed_status = 2 where linked_place_id = OLD.place_id and indexed_status = 0;
+  update placex set linked_place_id = null where linked_place_id = OLD.place_id;

   --DEBUG: RAISE WARNING 'placex_delete:02 % %',OLD.osm_type,OLD.osm_id;
 
   IF OLD.rank_address < 30 THEN
   --DEBUG: RAISE WARNING 'placex_delete:02 % %',OLD.osm_type,OLD.osm_id;
 
   IF OLD.rank_address < 30 THEN


@@ -2178,7 +2198,7 @@ BEGIN
 
   FOR location IN select * from get_addressdata(for_place_id) where isaddress order by rank_address desc LOOP
     currresult := trim(get_name_by_language(location.name, languagepref));
 
   FOR location IN select * from get_addressdata(for_place_id) where isaddress order by rank_address desc LOOP
     currresult := trim(get_name_by_language(location.name, languagepref));

-    IF currresult != prevresult AND currresult IS NOT NULL THEN
+    IF currresult != prevresult AND currresult IS NOT NULL AND result[(100 - location.rank_address)] IS NULL THEN

       result[(100 - location.rank_address)] := trim(get_name_by_language(location.name, languagepref));
       prevresult := currresult;
     END IF;
       result[(100 - location.rank_address)] := trim(get_name_by_language(location.name, languagepref));
       prevresult := currresult;
     END IF;