]> git.openstreetmap.org Git - nominatim.git/blobdiff - nominatim/tools/migration.py
Merge pull request #2856 from t-tomek/patch-1
[nominatim.git] / nominatim / tools / migration.py
index 76726e8c59d7ab6d58ca310d77ec41963a04a90f..10bca15c117c78e0210f82aaf68aab0ad96b6920 100644 (file)
@@ -7,20 +7,26 @@
 """
 Functions for database migration to newer software versions.
 """
+from typing import List, Tuple, Callable, Any
 import logging
 
+from psycopg2 import sql as pysql
+
+from nominatim.config import Configuration
 from nominatim.db import properties
-from nominatim.db.connection import connect
-from nominatim.version import NOMINATIM_VERSION
+from nominatim.db.connection import connect, Connection
+from nominatim.version import NOMINATIM_VERSION, version_str
 from nominatim.tools import refresh
 from nominatim.tokenizer import factory as tokenizer_factory
 from nominatim.errors import UsageError
 
 LOG = logging.getLogger()
 
-_MIGRATION_FUNCTIONS = []
+VersionTuple = Tuple[int, int, int, int]
+
+_MIGRATION_FUNCTIONS : List[Tuple[VersionTuple, Callable[..., None]]] = []
 
-def migrate(config, paths):
+def migrate(config: Configuration, paths: Any) -> int:
     """ Check for the current database version and execute migrations,
         if necesssary.
     """
@@ -46,8 +52,9 @@ def migrate(config, paths):
         has_run_migration = False
         for version, func in _MIGRATION_FUNCTIONS:
             if db_version <= version:
-                LOG.warning("Runnning: %s (%s)", func.__doc__.split('\n', 1)[0],
-                            '{0[0]}.{0[1]}.{0[2]}-{0[3]}'.format(version))
+                title = func.__doc__ or ''
+                LOG.warning("Running: %s (%s)", title.split('\n', 1)[0],
+                            version_str(version))
                 kwargs = dict(conn=conn, config=config, paths=paths)
                 func(**kwargs)
                 conn.commit()
@@ -59,15 +66,14 @@ def migrate(config, paths):
             tokenizer = tokenizer_factory.get_tokenizer_for_db(config)
             tokenizer.update_sql_functions(config)
 
-        properties.set_property(conn, 'database_version',
-                                '{0[0]}.{0[1]}.{0[2]}-{0[3]}'.format(NOMINATIM_VERSION))
+        properties.set_property(conn, 'database_version', version_str())
 
         conn.commit()
 
     return 0
 
 
-def _guess_version(conn):
+def _guess_version(conn: Connection) -> VersionTuple:
     """ Guess a database version when there is no property table yet.
         Only migrations for 3.6 and later are supported, so bail out
         when the version seems older.
@@ -87,7 +93,8 @@ def _guess_version(conn):
 
 
 
-def _migration(major, minor, patch=0, dbpatch=0):
+def _migration(major: int, minor: int, patch: int = 0,
+               dbpatch: int = 0) -> Callable[[Callable[..., None]], Callable[..., None]]:
     """ Decorator for a single migration step. The parameters describe the
         version after which the migration is applicable, i.e before changing
         from the given version to the next, the migration is required.
@@ -100,7 +107,7 @@ def _migration(major, minor, patch=0, dbpatch=0):
         process, so the migration functions may leave a temporary state behind
         there.
     """
-    def decorator(func):
+    def decorator(func: Callable[..., None]) -> Callable[..., None]:
         _MIGRATION_FUNCTIONS.append(((major, minor, patch, dbpatch), func))
         return func
 
@@ -108,7 +115,7 @@ def _migration(major, minor, patch=0, dbpatch=0):
 
 
 @_migration(3, 5, 0, 99)
-def import_status_timestamp_change(conn, **_):
+def import_status_timestamp_change(conn: Connection, **_: Any) -> None:
     """ Add timezone to timestamp in status table.
 
         The import_status table has been changed to include timezone information
@@ -120,19 +127,19 @@ def import_status_timestamp_change(conn, **_):
 
 
 @_migration(3, 5, 0, 99)
-def add_nominatim_property_table(conn, config, **_):
+def add_nominatim_property_table(conn: Connection, config: Configuration, **_: Any) -> None:
     """ Add nominatim_property table.
     """
     if not conn.table_exists('nominatim_properties'):
         with conn.cursor() as cur:
-            cur.execute("""CREATE TABLE nominatim_properties (
-                               property TEXT,
-                               value TEXT);
-                           GRANT SELECT ON TABLE nominatim_properties TO "{}";
-                        """.format(config.DATABASE_WEBUSER))
+            cur.execute(pysql.SQL("""CREATE TABLE nominatim_properties (
+                                        property TEXT,
+                                        value TEXT);
+                                     GRANT SELECT ON TABLE nominatim_properties TO {};
+                                  """).format(pysql.Identifier(config.DATABASE_WEBUSER)))
 
 @_migration(3, 6, 0, 0)
-def change_housenumber_transliteration(conn, **_):
+def change_housenumber_transliteration(conn: Connection, **_: Any) -> None:
     """ Transliterate housenumbers.
 
         The database schema switched from saving raw housenumbers in
@@ -163,7 +170,7 @@ def change_housenumber_transliteration(conn, **_):
 
 
 @_migration(3, 7, 0, 0)
-def switch_placenode_geometry_index(conn, **_):
+def switch_placenode_geometry_index(conn: Connection, **_: Any) -> None:
     """ Replace idx_placex_geometry_reverse_placeNode index.
 
         Make the index slightly more permissive, so that it can also be used
@@ -180,7 +187,7 @@ def switch_placenode_geometry_index(conn, **_):
 
 
 @_migration(3, 7, 0, 1)
-def install_legacy_tokenizer(conn, config, **_):
+def install_legacy_tokenizer(conn: Connection, config: Configuration, **_: Any) -> None:
     """ Setup legacy tokenizer.
 
         If no other tokenizer has been configured yet, then create the
@@ -194,15 +201,16 @@ def install_legacy_tokenizer(conn, config, **_):
                                            and column_name = 'token_info'""",
                                         (table, ))
                 if has_column == 0:
-                    cur.execute('ALTER TABLE {} ADD COLUMN token_info JSONB'.format(table))
+                    cur.execute(pysql.SQL('ALTER TABLE {} ADD COLUMN token_info JSONB')
+                                .format(pysql.Identifier(table)))
         tokenizer = tokenizer_factory.create_tokenizer(config, init_db=False,
                                                        module_name='legacy')
 
-        tokenizer.migrate_database(config)
+        tokenizer.migrate_database(config) # type: ignore[attr-defined]
 
 
 @_migration(4, 0, 99, 0)
-def create_tiger_housenumber_index(conn, **_):
+def create_tiger_housenumber_index(conn: Connection, **_: Any) -> None:
     """ Create idx_location_property_tiger_parent_place_id with included
         house number.
 
@@ -219,7 +227,7 @@ def create_tiger_housenumber_index(conn, **_):
 
 
 @_migration(4, 0, 99, 1)
-def create_interpolation_index_on_place(conn, **_):
+def create_interpolation_index_on_place(conn: Connection, **_: Any) -> None:
     """ Create idx_place_interpolations for lookup of interpolation lines
         on updates.
     """
@@ -230,10 +238,10 @@ def create_interpolation_index_on_place(conn, **_):
 
 
 @_migration(4, 0, 99, 2)
-def add_step_column_for_interpolation(conn, **_):
+def add_step_column_for_interpolation(conn: Connection, **_: Any) -> None:
     """ Add a new column 'step' to the interpolations table.
 
-        Also convers the data into the stricter format which requires that
+        Also converts the data into the stricter format which requires that
         startnumbers comply with the odd/even requirements.
     """
     if conn.table_has_column('location_property_osmline', 'step'):
@@ -265,7 +273,7 @@ def add_step_column_for_interpolation(conn, **_):
 
 
 @_migration(4, 0, 99, 3)
-def add_step_column_for_tiger(conn, **_):
+def add_step_column_for_tiger(conn: Connection, **_: Any) -> None:
     """ Add a new column 'step' to the tiger data table.
     """
     if conn.table_has_column('location_property_tiger', 'step'):
@@ -280,7 +288,7 @@ def add_step_column_for_tiger(conn, **_):
 
 
 @_migration(4, 0, 99, 4)
-def add_derived_name_column_for_country_names(conn, **_):
+def add_derived_name_column_for_country_names(conn: Connection, **_: Any) -> None:
     """ Add a new column 'derived_name' which in the future takes the
         country names as imported from OSM data.
     """
@@ -290,7 +298,7 @@ def add_derived_name_column_for_country_names(conn, **_):
 
 
 @_migration(4, 0, 99, 5)
-def mark_internal_country_names(conn, config, **_):
+def mark_internal_country_names(conn: Connection, config: Configuration, **_: Any) -> None:
     """ Names from the country table should be marked as internal to prevent
         them from being deleted. Only necessary for ICU tokenizer.
     """
@@ -307,3 +315,62 @@ def mark_internal_country_names(conn, config, **_):
                     names = {}
                 names['countrycode'] = country_code
                 analyzer.add_country_names(country_code, names)
+
+
+@_migration(4, 1, 99, 0)
+def add_place_deletion_todo_table(conn: Connection, **_: Any) -> None:
+    """ Add helper table for deleting data on updates.
+
+        The table is only necessary when updates are possible, i.e.
+        the database is not in freeze mode.
+    """
+    if conn.table_exists('place'):
+        with conn.cursor() as cur:
+            cur.execute("""CREATE TABLE IF NOT EXISTS place_to_be_deleted (
+                             osm_type CHAR(1),
+                             osm_id BIGINT,
+                             class TEXT,
+                             type TEXT,
+                             deferred BOOLEAN)""")
+
+
+@_migration(4, 1, 99, 1)
+def split_pending_index(conn: Connection, **_: Any) -> None:
+    """ Reorganise indexes for pending updates.
+    """
+    if conn.table_exists('place'):
+        with conn.cursor() as cur:
+            cur.execute("""CREATE INDEX IF NOT EXISTS idx_placex_rank_address_sector
+                           ON placex USING BTREE (rank_address, geometry_sector)
+                           WHERE indexed_status > 0""")
+            cur.execute("""CREATE INDEX IF NOT EXISTS idx_placex_rank_boundaries_sector
+                           ON placex USING BTREE (rank_search, geometry_sector)
+                           WHERE class = 'boundary' and type = 'administrative'
+                                 and indexed_status > 0""")
+            cur.execute("DROP INDEX IF EXISTS idx_placex_pendingsector")
+
+
+@_migration(4, 2, 99, 0)
+def enable_forward_dependencies(conn: Connection, **_: Any) -> None:
+    """ Create indexes for updates with forward dependency tracking (long-running).
+    """
+    if conn.table_exists('planet_osm_ways'):
+        with conn.cursor() as cur:
+            cur.execute("""SELECT * FROM pg_indexes
+                           WHERE tablename = 'planet_osm_ways'
+                                 and indexdef LIKE '%nodes%'""")
+            if cur.rowcount == 0:
+                cur.execute("""CREATE OR REPLACE FUNCTION public.planet_osm_index_bucket(bigint[])
+                               RETURNS bigint[]
+                               LANGUAGE sql IMMUTABLE
+                                AS $function$
+                                  SELECT ARRAY(SELECT DISTINCT unnest($1) >> 5)
+                                $function$""")
+                cur.execute("""CREATE INDEX planet_osm_ways_nodes_bucket_idx
+                                 ON planet_osm_ways
+                                 USING gin (planet_osm_index_bucket(nodes))
+                                 WITH (fastupdate=off)""")
+                cur.execute("""CREATE INDEX planet_osm_rels_parts_idx
+                                 ON planet_osm_rels USING gin (parts)
+                                 WITH (fastupdate=off)""")
+                cur.execute("ANALYZE planet_osm_ways")