Nominatim/nominatim/tools/migration.py
Sarah Hoffmann 8ed096f938 speed up reverse lookup of place nodes
Add a special index that contains the place nodes buffered by their
respective area according to their search rank. This replaces the
maximum area search for place nodes and reduces drastically the number
of place nodes that need to be retrieved.
2023-02-17 14:10:01 +01:00

388 lines
17 KiB
Python

# SPDX-License-Identifier: GPL-2.0-only
#
# This file is part of Nominatim. (https://nominatim.org)
#
# Copyright (C) 2022 by the Nominatim developer community.
# For a full list of authors see the git log.
"""
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, Connection
from nominatim.version import NominatimVersion, NOMINATIM_VERSION, parse_version
from nominatim.tools import refresh
from nominatim.tokenizer import factory as tokenizer_factory
from nominatim.errors import UsageError
LOG = logging.getLogger()
_MIGRATION_FUNCTIONS : List[Tuple[NominatimVersion, Callable[..., None]]] = []
def migrate(config: Configuration, paths: Any) -> int:
""" Check for the current database version and execute migrations,
if necesssary.
"""
with connect(config.get_libpq_dsn()) as conn:
if conn.table_exists('nominatim_properties'):
db_version_str = properties.get_property(conn, 'database_version')
else:
db_version_str = None
if db_version_str is not None:
db_version = parse_version(db_version_str)
if db_version == NOMINATIM_VERSION:
LOG.warning("Database already at latest version (%s)", db_version_str)
return 0
LOG.info("Detected database version: %s", db_version_str)
else:
db_version = _guess_version(conn)
has_run_migration = False
for version, func in _MIGRATION_FUNCTIONS:
if db_version < version or \
(db_version == (3, 5, 0, 99) and version == (3, 5, 0, 99)):
title = func.__doc__ or ''
LOG.warning("Running: %s (%s)", title.split('\n', 1)[0], version)
kwargs = dict(conn=conn, config=config, paths=paths)
func(**kwargs)
conn.commit()
has_run_migration = True
if has_run_migration:
LOG.warning('Updating SQL functions.')
refresh.create_functions(conn, config)
tokenizer = tokenizer_factory.get_tokenizer_for_db(config)
tokenizer.update_sql_functions(config)
properties.set_property(conn, 'database_version', str(NOMINATIM_VERSION))
conn.commit()
return 0
def _guess_version(conn: Connection) -> NominatimVersion:
""" 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.
"""
with conn.cursor() as cur:
# In version 3.6, the country_name table was updated. Check for that.
cnt = cur.scalar("""SELECT count(*) FROM
(SELECT svals(name) FROM country_name
WHERE country_code = 'gb')x;
""")
if cnt < 100:
LOG.fatal('It looks like your database was imported with a version '
'prior to 3.6.0. Automatic migration not possible.')
raise UsageError('Migration not possible.')
return NominatimVersion(3, 5, 0, 99)
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.
All migrations are run in the order in which they are defined in this
file. Do not run global SQL scripts for migrations as you cannot be sure
that these scripts do the same in later versions.
Functions will always be reimported in full at the end of the migration
process, so the migration functions may leave a temporary state behind
there.
"""
def decorator(func: Callable[..., None]) -> Callable[..., None]:
version = NominatimVersion(major, minor, patch, dbpatch)
_MIGRATION_FUNCTIONS.append((version, func))
return func
return decorator
@_migration(3, 5, 0, 99)
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
with the time stamp.
"""
with conn.cursor() as cur:
cur.execute("""ALTER TABLE import_status ALTER COLUMN lastimportdate
TYPE timestamp with time zone;""")
@_migration(3, 5, 0, 99)
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(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: Connection, **_: Any) -> None:
""" Transliterate housenumbers.
The database schema switched from saving raw housenumbers in
placex.housenumber to saving transliterated ones.
Note: the function create_housenumber_id() has been dropped in later
versions.
"""
with conn.cursor() as cur:
cur.execute("""CREATE OR REPLACE FUNCTION create_housenumber_id(housenumber TEXT)
RETURNS TEXT AS $$
DECLARE
normtext TEXT;
BEGIN
SELECT array_to_string(array_agg(trans), ';')
INTO normtext
FROM (SELECT lookup_word as trans,
getorcreate_housenumber_id(lookup_word)
FROM (SELECT make_standard_name(h) as lookup_word
FROM regexp_split_to_table(housenumber, '[,;]') h) x) y;
return normtext;
END;
$$ LANGUAGE plpgsql STABLE STRICT;""")
cur.execute("DELETE FROM word WHERE class = 'place' and type = 'house'")
cur.execute("""UPDATE placex
SET housenumber = create_housenumber_id(housenumber)
WHERE housenumber is not null""")
@_migration(3, 7, 0, 0)
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
when matching up boundaries and place nodes. It makes the index
idx_placex_adminname index unnecessary.
"""
with conn.cursor() as cur:
cur.execute(""" CREATE INDEX IF NOT EXISTS idx_placex_geometry_placenode ON placex
USING GIST (geometry)
WHERE osm_type = 'N' and rank_search < 26
and class = 'place' and type != 'postcode'
and linked_place_id is null""")
cur.execute(""" DROP INDEX IF EXISTS idx_placex_adminname """)
@_migration(3, 7, 0, 1)
def install_legacy_tokenizer(conn: Connection, config: Configuration, **_: Any) -> None:
""" Setup legacy tokenizer.
If no other tokenizer has been configured yet, then create the
configuration for the backwards-compatible legacy tokenizer
"""
if properties.get_property(conn, 'tokenizer') is None:
with conn.cursor() as cur:
for table in ('placex', 'location_property_osmline'):
has_column = cur.scalar("""SELECT count(*) FROM information_schema.columns
WHERE table_name = %s
and column_name = 'token_info'""",
(table, ))
if has_column == 0:
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) # type: ignore[attr-defined]
@_migration(4, 0, 99, 0)
def create_tiger_housenumber_index(conn: Connection, **_: Any) -> None:
""" Create idx_location_property_tiger_parent_place_id with included
house number.
The inclusion is needed for efficient lookup of housenumbers in
full address searches.
"""
if conn.server_version_tuple() >= (11, 0, 0):
with conn.cursor() as cur:
cur.execute(""" CREATE INDEX IF NOT EXISTS
idx_location_property_tiger_housenumber_migrated
ON location_property_tiger
USING btree(parent_place_id)
INCLUDE (startnumber, endnumber) """)
@_migration(4, 0, 99, 1)
def create_interpolation_index_on_place(conn: Connection, **_: Any) -> None:
""" Create idx_place_interpolations for lookup of interpolation lines
on updates.
"""
with conn.cursor() as cur:
cur.execute("""CREATE INDEX IF NOT EXISTS idx_place_interpolations
ON place USING gist(geometry)
WHERE osm_type = 'W' and address ? 'interpolation'""")
@_migration(4, 0, 99, 2)
def add_step_column_for_interpolation(conn: Connection, **_: Any) -> None:
""" Add a new column 'step' to the interpolations table.
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'):
return
with conn.cursor() as cur:
# Mark invalid all interpolations with no intermediate numbers.
cur.execute("""UPDATE location_property_osmline SET startnumber = null
WHERE endnumber - startnumber <= 1 """)
# Align the start numbers where odd/even does not match.
cur.execute("""UPDATE location_property_osmline
SET startnumber = startnumber + 1,
linegeo = ST_LineSubString(linegeo,
1.0 / (endnumber - startnumber)::float,
1)
WHERE (interpolationtype = 'odd' and startnumber % 2 = 0)
or (interpolationtype = 'even' and startnumber % 2 = 1)
""")
# Mark invalid odd/even interpolations with no intermediate numbers.
cur.execute("""UPDATE location_property_osmline SET startnumber = null
WHERE interpolationtype in ('odd', 'even')
and endnumber - startnumber = 2""")
# Finally add the new column and populate it.
cur.execute("ALTER TABLE location_property_osmline ADD COLUMN step SMALLINT")
cur.execute("""UPDATE location_property_osmline
SET step = CASE WHEN interpolationtype = 'all'
THEN 1 ELSE 2 END
""")
@_migration(4, 0, 99, 3)
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'):
return
with conn.cursor() as cur:
cur.execute("ALTER TABLE location_property_tiger ADD COLUMN step SMALLINT")
cur.execute("""UPDATE location_property_tiger
SET step = CASE WHEN interpolationtype = 'all'
THEN 1 ELSE 2 END
""")
@_migration(4, 0, 99, 4)
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.
"""
if not conn.table_has_column('country_name', 'derived_name'):
with conn.cursor() as cur:
cur.execute("ALTER TABLE country_name ADD COLUMN derived_name public.HSTORE")
@_migration(4, 0, 99, 5)
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.
"""
import psycopg2.extras # pylint: disable=import-outside-toplevel
tokenizer = tokenizer_factory.get_tokenizer_for_db(config)
with tokenizer.name_analyzer() as analyzer:
with conn.cursor() as cur:
psycopg2.extras.register_hstore(cur)
cur.execute("SELECT country_code, name FROM country_name")
for country_code, names in cur:
if not names:
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")
@_migration(4, 2, 99, 1)
def add_improved_geometry_reverse_placenode_index(conn: Connection, **_: Any) -> None:
""" Create improved index for reverse lookup of place nodes.
"""
with conn.cursor() as cur:
cur.execute("""CREATE INDEX IF NOT EXISTS idx_placex_geometry_reverse_lookupPlaceNode
ON placex
USING gist (ST_Buffer(geometry, reverse_place_diameter(rank_search)))
WHERE rank_address between 4 and 25 AND type != 'postcode'
AND name is not null AND linked_place_id is null AND osm_type = 'N'
""")