change indexing order for interpolations

Interpolations are now indexed after rank 30 objects. The housenumber
nodes no longer need information from the interpolations while the
interpolations can make use of precomputed postcodes.

diff --git a/nominatim/indexer/indexer.py b/nominatim/indexer/indexer.py
index 98bb52114bedaa8bc220f6a595d2c1dcf88a2283..555f8704a19c6796da4b97a724cd363d183f7f12 100644 (file)
--- a/nominatim/indexer/indexer.py
+++ b/nominatim/indexer/indexer.py
@@ -160,15 +160,12 @@ class Indexer:
                     minrank, maxrank, self.num_threads)
 
         with self.tokenizer.name_analyzer() as analyzer:
                     minrank, maxrank, self.num_threads)
 
         with self.tokenizer.name_analyzer() as analyzer:

-            for rank in range(max(1, minrank), maxrank):
-                self._index(runners.RankRunner(rank, analyzer))
+            for rank in range(max(1, minrank), maxrank + 1):
+                self._index(runners.RankRunner(rank, analyzer), 20 if rank == 30 else 1)

 
             if maxrank == 30:
                 self._index(runners.RankRunner(0, analyzer))
                 self._index(runners.InterpolationRunner(analyzer), 20)
 
             if maxrank == 30:
                 self._index(runners.RankRunner(0, analyzer))
                 self._index(runners.InterpolationRunner(analyzer), 20)

-                self._index(runners.RankRunner(30, analyzer), 20)
-            else:
-                self._index(runners.RankRunner(maxrank, analyzer))

 
 
     def index_postcodes(self):
 
 
     def index_postcodes(self):

diff --git a/test/python/indexer/test_indexing.py b/test/python/indexer/test_indexing.py
index e303f381e0cea86d3559993942f7f99d3698911d..45c68a33941a2b24141ff1c08e0e82c233efe215 100644 (file)
--- a/test/python/indexer/test_indexing.py
+++ b/test/python/indexer/test_indexing.py
@@ -177,25 +177,16 @@ def test_index_all_by_rank(test_db, threads, test_tokenizer):
         SELECT count(*) FROM placex p WHERE rank_address > 0
           AND indexed_date >= (SELECT min(indexed_date) FROM placex o
                                WHERE p.rank_address < o.rank_address)""") == 0
         SELECT count(*) FROM placex p WHERE rank_address > 0
           AND indexed_date >= (SELECT min(indexed_date) FROM placex o
                                WHERE p.rank_address < o.rank_address)""") == 0

-    # placex rank < 30 objects come before interpolations
+    # placex address ranked objects come before interpolations

     assert test_db.scalar(
     assert test_db.scalar(

-        """SELECT count(*) FROM placex WHERE rank_address < 30
+        """SELECT count(*) FROM placex WHERE rank_address > 0

              AND indexed_date >
                    (SELECT min(indexed_date) FROM location_property_osmline)""") == 0
              AND indexed_date >
                    (SELECT min(indexed_date) FROM location_property_osmline)""") == 0

-    # placex rank = 30 objects come after interpolations
+    # rank 0 comes after all other placex objects

     assert test_db.scalar(
     assert test_db.scalar(

-        """SELECT count(*) FROM placex WHERE rank_address = 30
-             AND indexed_date <
-                   (SELECT max(indexed_date) FROM location_property_osmline)""") == 0
-    # rank 0 comes after rank 29 and before rank 30
-    assert test_db.scalar(
-        """SELECT count(*) FROM placex WHERE rank_address < 30
+        """SELECT count(*) FROM placex WHERE rank_address > 0

              AND indexed_date >
                    (SELECT min(indexed_date) FROM placex WHERE rank_address = 0)""") == 0
              AND indexed_date >
                    (SELECT min(indexed_date) FROM placex WHERE rank_address = 0)""") == 0

-    assert test_db.scalar(
-        """SELECT count(*) FROM placex WHERE rank_address = 30
-             AND indexed_date <
-                   (SELECT max(indexed_date) FROM placex WHERE rank_address = 0)""") == 0

 
 
 @pytest.mark.parametrize("threads", [1, 15])
 
 
 @pytest.mark.parametrize("threads", [1, 15])