Merge pull request #3957 from jayaddison/issue-2714/linked-places-default-language

Indexing: add default-language placename from linked places

lib-sql/functions/placex_triggers.sql

@@ -29,6 +29,7 @@ DECLARE
   location RECORD;
   result prepare_update_info;
   extra_names HSTORE;
+  default_language VARCHAR(10);
 BEGIN
   IF not p.address ? '_inherited' THEN
     result.address := p.address;
@@ -85,6 +86,13 @@ BEGIN
 
     IF location.name is not NULL THEN
       {% if debug %}RAISE WARNING 'Names original: %, location: %', result.name, location.name;{% endif %}
+
+      -- Add the linked-place (e.g. city) name as a searchable placename in the default language (if any)
+      default_language := get_country_language_code(location.country_code);
+      IF default_language is not NULL AND NOT location.name ? ('name:' || default_language) THEN
+        location.name := location.name || hstore('name:' || default_language, location.name->'name');
+      END IF;
+
       -- Add all names from the place nodes that deviate from the name
       -- in the relation with the prefix '_place_'. Deviation means that
       -- either the value is different or a given key is missing completely

src/nominatim_db/indexer/indexer.py

@@ -56,7 +56,7 @@ class Indexer:
                         cur.execute('ANALYZE')
 
             while True:
-                if await self.index_by_rank(0, 4) > 0:
+                if await self.index_by_rank(1, 4) > 0:
                     _analyze()
 
                 if await self.index_boundaries() > 100:
@@ -68,6 +68,9 @@ class Indexer:
                 if await self.index_by_rank(26, 30) > 1000:
                     _analyze()
 
+                # Special case: rank zero depends on the previously-indexed [1..30] ranks
+                await self.index_by_rank(0, 0)
+
                 if await self.index_postcodes() > 100:
                     _analyze()
 
@@ -147,8 +150,11 @@ class Indexer:
                 total += await self._index(runners.RankRunner(rank, analyzer),
                                            batch=batch, total_tuples=total_tuples.get(rank, 0))
 
-            if maxrank == 30:
+            # Special case: rank zero depends on ranks [1..30]
+            if minrank == 0:
                 total += await self._index(runners.RankRunner(0, analyzer))
+
+            if maxrank == 30:
                 total += await self._index(runners.InterpolationRunner(analyzer), batch=20)
 
         return total

test/bdd/features/db/import/linking.feature

@@ -297,9 +297,8 @@ Feature: Linking of places
          | R1         | LabelPlace |
 
 
-    @skip
     Scenario: Linked places expand default language names
-        Given the grid
+        Given the grid with origin CO
             | 1 |   | 2 |
             |   | 9 |   |
             | 4 |   | 3 |

test/python/indexer/test_indexing.py

@@ -226,12 +226,12 @@ async def test_index_partial_with_30(test_db, threads, test_tokenizer):
     idx = indexer.Indexer('dbname=test_nominatim_python_unittest', test_tokenizer, threads)
     await idx.index_by_rank(28, 30)
 
-    assert test_db.placex_unindexed() == 27
+    assert test_db.placex_unindexed() == 28
     assert test_db.osmline_unindexed() == 0
 
     assert test_db.scalar("""
                     SELECT count(*) FROM placex
-                      WHERE indexed_status = 0 AND rank_address between 1 and 27""") == 0
+                      WHERE indexed_status = 0 AND rank_address between 0 and 27""") == 0
 
 
 @pytest.mark.parametrize("threads", [1, 15])