]> git.openstreetmap.org Git - nominatim.git/blobdiff - test/python/test_indexing.py
Improved performance of the postcodes query and some code cleaning
[nominatim.git] / test / python / test_indexing.py
index e1a3a4d064433b86394a113933d39b3f89be8e9c..9873e7d72673a8aaeb3e356276a4b558d97be3e8 100644 (file)
@@ -2,28 +2,22 @@
 Tests for running the indexing.
 """
 import itertools
-import psycopg2
 import pytest
 
-from nominatim.indexer.indexer import Indexer
+from nominatim.indexer import indexer
+from nominatim.tokenizer import factory
 
 class IndexerTestDB:
 
-    def __init__(self, name):
-        self.name = name
-        self.conn = None
+    def __init__(self, conn):
         self.placex_id = itertools.count(100000)
         self.osmline_id = itertools.count(500000)
+        self.postcode_id = itertools.count(700000)
 
-    def setup(self):
-        with psycopg2.connect(database='postgres') as conn:
-            conn.set_isolation_level(0)
-            with conn.cursor() as cur:
-                cur.execute('DROP DATABASE IF EXISTS {}'.format(self.name))
-                cur.execute('CREATE DATABASE {}'.format(self.name))
-        self.conn = psycopg2.connect(database=self.name)
+        self.conn = conn
         self.conn.set_isolation_level(0)
         with self.conn.cursor() as cur:
+            cur.execute('CREATE EXTENSION hstore')
             cur.execute("""CREATE TABLE placex (place_id BIGINT,
                                                 class TEXT,
                                                 type TEXT,
@@ -33,12 +27,23 @@ class IndexerTestDB:
                                                 indexed_date TIMESTAMP,
                                                 partition SMALLINT,
                                                 admin_level SMALLINT,
+                                                address HSTORE,
+                                                token_info JSONB,
                                                 geometry_sector INTEGER)""")
             cur.execute("""CREATE TABLE location_property_osmline (
                                place_id BIGINT,
+                               osm_id BIGINT,
+                               address HSTORE,
+                               token_info JSONB,
                                indexed_status SMALLINT,
                                indexed_date TIMESTAMP,
                                geometry_sector INTEGER)""")
+            cur.execute("""CREATE TABLE location_postcode (
+                               place_id BIGINT,
+                               indexed_status SMALLINT,
+                               indexed_date TIMESTAMP,
+                               country_code varchar(2),
+                               postcode TEXT)""")
             cur.execute("""CREATE OR REPLACE FUNCTION date_update() RETURNS TRIGGER
                            AS $$
                            BEGIN
@@ -47,20 +52,30 @@ class IndexerTestDB:
                              END IF;
                              RETURN NEW;
                            END; $$ LANGUAGE plpgsql;""")
-            cur.execute("""CREATE TRIGGER placex_update BEFORE UPDATE ON placex
-                           FOR EACH ROW EXECUTE PROCEDURE date_update()""")
-            cur.execute("""CREATE TRIGGER osmline_update BEFORE UPDATE ON location_property_osmline
-                           FOR EACH ROW EXECUTE PROCEDURE date_update()""")
-
+            cur.execute("""CREATE OR REPLACE FUNCTION placex_prepare_update(p placex,
+                                                      OUT name HSTORE,
+                                                      OUT address HSTORE,
+                                                      OUT country_feature VARCHAR)
+                           AS $$
+                           BEGIN
+                            address := p.address;
+                            name := p.address;
+                           END;
+                           $$ LANGUAGE plpgsql STABLE;
+                        """)
+            cur.execute("""CREATE OR REPLACE FUNCTION
+                             get_interpolation_address(in_address HSTORE, wayid BIGINT)
+                           RETURNS HSTORE AS $$
+                           BEGIN
+                             RETURN in_address;
+                           END;
+                           $$ LANGUAGE plpgsql STABLE;
+                        """)
 
-    def drop(self):
-        if self.conn:
-            self.conn.close()
-            self.conn = None
-        with psycopg2.connect(database='postgres') as conn:
-            conn.set_isolation_level(0)
-            with conn.cursor() as cur:
-                cur.execute('DROP DATABASE IF EXISTS {}'.format(self.name))
+            for table in ('placex', 'location_property_osmline', 'location_postcode'):
+                cur.execute("""CREATE TRIGGER {0}_update BEFORE UPDATE ON {0}
+                               FOR EACH ROW EXECUTE PROCEDURE date_update()
+                            """.format(table))
 
     def scalar(self, query):
         with self.conn.cursor() as cur:
@@ -87,121 +102,185 @@ class IndexerTestDB:
         next_id = next(self.osmline_id)
         with self.conn.cursor() as cur:
             cur.execute("""INSERT INTO location_property_osmline
-                              (place_id, indexed_status, geometry_sector)
-                              VALUES (%s, 1, %s)""",
-                        (next_id, sector))
+                              (place_id, osm_id, indexed_status, geometry_sector)
+                              VALUES (%s, %s, 1, %s)""",
+                        (next_id, next_id, sector))
+        return next_id
+
+    def add_postcode(self, country, postcode):
+        next_id = next(self.postcode_id)
+        with self.conn.cursor() as cur:
+            cur.execute("""INSERT INTO location_postcode
+                            (place_id, indexed_status, country_code, postcode)
+                            VALUES (%s, 1, %s, %s)""",
+                        (next_id, country, postcode))
         return next_id
 
     def placex_unindexed(self):
         return self.scalar('SELECT count(*) from placex where indexed_status > 0')
 
     def osmline_unindexed(self):
-        return self.scalar('SELECT count(*) from location_property_osmline where indexed_status > 0')
+        return self.scalar("""SELECT count(*) from location_property_osmline
+                              WHERE indexed_status > 0""")
+
+
+@pytest.fixture
+def test_db(temp_db_conn):
+    yield IndexerTestDB(temp_db_conn)
 
 
 @pytest.fixture
-def test_db():
-    db = IndexerTestDB('test_nominatim_python_unittest')
-    db.setup()
-    yield db
-    db.drop()
+def test_tokenizer(tokenizer_mock, def_config, tmp_path):
+    def_config.project_dir = tmp_path
+    return factory.create_tokenizer(def_config)
 
 
 @pytest.mark.parametrize("threads", [1, 15])
-def test_index_full(test_db, threads):
+def test_index_all_by_rank(test_db, threads, test_tokenizer):
     for rank in range(31):
         test_db.add_place(rank_address=rank, rank_search=rank)
     test_db.add_osmline()
 
-    assert 31 == test_db.placex_unindexed()
-    assert 1 == test_db.osmline_unindexed()
+    assert test_db.placex_unindexed() == 31
+    assert test_db.osmline_unindexed() == 1
 
-    idx = Indexer('dbname=test_nominatim_python_unittest', threads)
+    idx = indexer.Indexer('dbname=test_nominatim_python_unittest', test_tokenizer, threads)
     idx.index_by_rank(0, 30)
 
-    assert 0 == test_db.placex_unindexed()
-    assert 0 == test_db.osmline_unindexed()
+    assert test_db.placex_unindexed() == 0
+    assert test_db.osmline_unindexed() == 0
 
-    assert 0 == test_db.scalar("""SELECT count(*) from placex
-                               WHERE indexed_status = 0 and indexed_date is null""")
+    assert test_db.scalar("""SELECT count(*) from placex
+                             WHERE indexed_status = 0 and indexed_date is null""") == 0
     # ranks come in order of rank address
-    assert 0 == test_db.scalar("""
+    assert test_db.scalar("""
         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)""")
+                               WHERE p.rank_address < o.rank_address)""") == 0
     # placex rank < 30 objects come before interpolations
-    assert 0 == test_db.scalar(
+    assert test_db.scalar(
         """SELECT count(*) FROM placex WHERE rank_address < 30
-             AND indexed_date > (SELECT min(indexed_date) FROM location_property_osmline)""")
+             AND indexed_date >
+                   (SELECT min(indexed_date) FROM location_property_osmline)""") == 0
     # placex rank = 30 objects come after interpolations
-    assert 0 == 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)""")
+             AND indexed_date <
+                   (SELECT max(indexed_date) FROM location_property_osmline)""") == 0
     # rank 0 comes after rank 29 and before rank 30
-    assert 0 == test_db.scalar(
+    assert test_db.scalar(
         """SELECT count(*) FROM placex WHERE rank_address < 30
-             AND indexed_date > (SELECT min(indexed_date) FROM placex WHERE rank_address = 0)""")
-    assert 0 == test_db.scalar(
+             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)""")
+             AND indexed_date <
+                   (SELECT max(indexed_date) FROM placex WHERE rank_address = 0)""") == 0
 
 
 @pytest.mark.parametrize("threads", [1, 15])
-def test_index_partial_without_30(test_db, threads):
+def test_index_partial_without_30(test_db, threads, test_tokenizer):
     for rank in range(31):
         test_db.add_place(rank_address=rank, rank_search=rank)
     test_db.add_osmline()
 
-    assert 31 == test_db.placex_unindexed()
-    assert 1 == test_db.osmline_unindexed()
+    assert test_db.placex_unindexed() == 31
+    assert test_db.osmline_unindexed() == 1
 
-    idx = Indexer('dbname=test_nominatim_python_unittest', threads)
+    idx = indexer.Indexer('dbname=test_nominatim_python_unittest',
+                          test_tokenizer, threads)
     idx.index_by_rank(4, 15)
 
-    assert 19 == test_db.placex_unindexed()
-    assert 1 == test_db.osmline_unindexed()
+    assert test_db.placex_unindexed() == 19
+    assert test_db.osmline_unindexed() == 1
 
-    assert 0 == test_db.scalar("""
+    assert test_db.scalar("""
                     SELECT count(*) FROM placex
-                      WHERE indexed_status = 0 AND not rank_address between 4 and 15""")
+                      WHERE indexed_status = 0 AND not rank_address between 4 and 15""") == 0
 
 
 @pytest.mark.parametrize("threads", [1, 15])
-def test_index_partial_with_30(test_db, threads):
+def test_index_partial_with_30(test_db, threads, test_tokenizer):
     for rank in range(31):
         test_db.add_place(rank_address=rank, rank_search=rank)
     test_db.add_osmline()
 
-    assert 31 == test_db.placex_unindexed()
-    assert 1 == test_db.osmline_unindexed()
+    assert test_db.placex_unindexed() == 31
+    assert test_db.osmline_unindexed() == 1
 
-    idx = Indexer('dbname=test_nominatim_python_unittest', threads)
+    idx = indexer.Indexer('dbname=test_nominatim_python_unittest', test_tokenizer, threads)
     idx.index_by_rank(28, 30)
 
-    assert 27 == test_db.placex_unindexed()
-    assert 0 == test_db.osmline_unindexed()
+    assert test_db.placex_unindexed() == 27
+    assert test_db.osmline_unindexed() == 0
 
-    assert 0 == test_db.scalar("""
+    assert test_db.scalar("""
                     SELECT count(*) FROM placex
-                      WHERE indexed_status = 0 AND rank_address between 1 and 27""")
+                      WHERE indexed_status = 0 AND rank_address between 1 and 27""") == 0
 
 @pytest.mark.parametrize("threads", [1, 15])
-def test_index_boundaries(test_db, threads):
+def test_index_boundaries(test_db, threads, test_tokenizer):
     for rank in range(4, 10):
         test_db.add_admin(rank_address=rank, rank_search=rank)
     for rank in range(31):
         test_db.add_place(rank_address=rank, rank_search=rank)
     test_db.add_osmline()
 
-    assert 37 == test_db.placex_unindexed()
-    assert 1 == test_db.osmline_unindexed()
+    assert test_db.placex_unindexed() == 37
+    assert test_db.osmline_unindexed() == 1
 
-    idx = Indexer('dbname=test_nominatim_python_unittest', threads)
+    idx = indexer.Indexer('dbname=test_nominatim_python_unittest', test_tokenizer, threads)
     idx.index_boundaries(0, 30)
 
-    assert 31 == test_db.placex_unindexed()
-    assert 1 == test_db.osmline_unindexed()
+    assert test_db.placex_unindexed() == 31
+    assert test_db.osmline_unindexed() == 1
 
-    assert 0 == test_db.scalar("""
+    assert test_db.scalar("""
                     SELECT count(*) FROM placex
-                      WHERE indexed_status = 0 AND class != 'boundary'""")
+                      WHERE indexed_status = 0 AND class != 'boundary'""") == 0
+
+
+@pytest.mark.parametrize("threads", [1, 15])
+def test_index_postcodes(test_db, threads, test_tokenizer):
+    for postcode in range(1000):
+        test_db.add_postcode('de', postcode)
+    for postcode in range(32000, 33000):
+        test_db.add_postcode('us', postcode)
+
+    idx = indexer.Indexer('dbname=test_nominatim_python_unittest', test_tokenizer, threads)
+    idx.index_postcodes()
+
+    assert test_db.scalar("""SELECT count(*) FROM location_postcode
+                                  WHERE indexed_status != 0""") == 0
+
+
+@pytest.mark.parametrize("analyse", [True, False])
+def test_index_full(test_db, analyse, test_tokenizer):
+    for rank in range(4, 10):
+        test_db.add_admin(rank_address=rank, rank_search=rank)
+    for rank in range(31):
+        test_db.add_place(rank_address=rank, rank_search=rank)
+    test_db.add_osmline()
+    for postcode in range(1000):
+        test_db.add_postcode('de', postcode)
+
+    idx = indexer.Indexer('dbname=test_nominatim_python_unittest', test_tokenizer, 4)
+    idx.index_full(analyse=analyse)
+
+    assert test_db.placex_unindexed() == 0
+    assert test_db.osmline_unindexed() == 0
+    assert test_db.scalar("""SELECT count(*) FROM location_postcode
+                             WHERE indexed_status != 0""") == 0
+
+
+@pytest.mark.parametrize("threads", [1, 15])
+def test_index_reopen_connection(test_db, threads, monkeypatch, test_tokenizer):
+    monkeypatch.setattr(indexer.WorkerPool, "REOPEN_CONNECTIONS_AFTER", 15)
+
+    for _ in range(1000):
+        test_db.add_place(rank_address=30, rank_search=30)
+
+    idx = indexer.Indexer('dbname=test_nominatim_python_unittest', test_tokenizer, threads)
+    idx.index_by_rank(28, 30)
+
+    assert test_db.placex_unindexed() == 0