1 # SPDX-License-Identifier: GPL-2.0-only
3 # This file is part of Nominatim. (https://nominatim.org)
5 # Copyright (C) 2022 by the Nominatim developer community.
6 # For a full list of authors see the git log.
8 Functions for setting up and importing a new Nominatim database.
10 from typing import Tuple, Optional, Union, Sequence, MutableMapping, Any
15 from pathlib import Path
18 from psycopg2 import sql as pysql
20 from nominatim.config import Configuration
21 from nominatim.db.connection import connect, get_pg_env, Connection
22 from nominatim.db.async_connection import DBConnection
23 from nominatim.db.sql_preprocessor import SQLPreprocessor
24 from nominatim.tools.exec_utils import run_osm2pgsql
25 from nominatim.errors import UsageError
26 from nominatim.version import POSTGRESQL_REQUIRED_VERSION, POSTGIS_REQUIRED_VERSION
28 LOG = logging.getLogger()
30 def _require_version(module: str, actual: Tuple[int, int], expected: Tuple[int, int]) -> None:
31 """ Compares the version for the given module and raises an exception
32 if the actual version is too old.
35 LOG.fatal('Minimum supported version of %s is %d.%d. '
36 'Found version %d.%d.',
37 module, expected[0], expected[1], actual[0], actual[1])
38 raise UsageError(f'{module} is too old.')
41 def setup_database_skeleton(dsn: str, rouser: Optional[str] = None) -> None:
42 """ Create a new database for Nominatim and populate it with the
45 The function fails when the database already exists or Postgresql or
46 PostGIS versions are too old.
48 Uses `createdb` to create the database.
50 If 'rouser' is given, then the function also checks that the user
51 with that given name exists.
53 Requires superuser rights by the caller.
55 proc = subprocess.run(['createdb'], env=get_pg_env(dsn), check=False)
57 if proc.returncode != 0:
58 raise UsageError('Creating new database failed.')
60 with connect(dsn) as conn:
61 _require_version('PostgreSQL server',
62 conn.server_version_tuple(),
63 POSTGRESQL_REQUIRED_VERSION)
65 if rouser is not None:
66 with conn.cursor() as cur:
67 cnt = cur.scalar('SELECT count(*) FROM pg_user where usename = %s',
70 LOG.fatal("Web user '%s' does not exist. Create it with:\n"
71 "\n createuser %s", rouser, rouser)
72 raise UsageError('Missing read-only user.')
75 with conn.cursor() as cur:
76 cur.execute('CREATE EXTENSION IF NOT EXISTS hstore')
77 cur.execute('CREATE EXTENSION IF NOT EXISTS postgis')
79 postgis_version = conn.postgis_version_tuple()
80 if postgis_version[0] >= 3:
81 cur.execute('CREATE EXTENSION IF NOT EXISTS postgis_raster')
85 _require_version('PostGIS',
86 conn.postgis_version_tuple(),
87 POSTGIS_REQUIRED_VERSION)
90 def import_osm_data(osm_files: Union[Path, Sequence[Path]],
91 options: MutableMapping[str, Any],
92 drop: bool = False, ignore_errors: bool = False) -> None:
93 """ Import the given OSM files. 'options' contains the list of
94 default settings for osm2pgsql.
96 options['import_file'] = osm_files
97 options['append'] = False
98 options['threads'] = 1
100 if not options['flatnode_file'] and options['osm2pgsql_cache'] == 0:
101 # Make some educated guesses about cache size based on the size
102 # of the import file and the available memory.
103 mem = psutil.virtual_memory()
105 if isinstance(osm_files, list):
106 for fname in osm_files:
107 fsize += os.stat(str(fname)).st_size
109 fsize = os.stat(str(osm_files)).st_size
110 options['osm2pgsql_cache'] = int(min((mem.available + mem.cached) * 0.75,
111 fsize * 2) / 1024 / 1024) + 1
113 run_osm2pgsql(options)
115 with connect(options['dsn']) as conn:
116 if not ignore_errors:
117 with conn.cursor() as cur:
118 cur.execute('SELECT * FROM place LIMIT 1')
119 if cur.rowcount == 0:
120 raise UsageError('No data imported by osm2pgsql.')
123 conn.drop_table('planet_osm_nodes')
125 if drop and options['flatnode_file']:
126 Path(options['flatnode_file']).unlink()
129 def create_tables(conn: Connection, config: Configuration, reverse_only: bool = False) -> None:
130 """ Create the set of basic tables.
131 When `reverse_only` is True, then the main table for searching will
132 be skipped and only reverse search is possible.
134 sql = SQLPreprocessor(conn, config)
135 sql.env.globals['db']['reverse_only'] = reverse_only
137 sql.run_sql_file(conn, 'tables.sql')
140 def create_table_triggers(conn: Connection, config: Configuration) -> None:
141 """ Create the triggers for the tables. The trigger functions must already
142 have been imported with refresh.create_functions().
144 sql = SQLPreprocessor(conn, config)
145 sql.run_sql_file(conn, 'table-triggers.sql')
148 def create_partition_tables(conn: Connection, config: Configuration) -> None:
149 """ Create tables that have explicit partitioning.
151 sql = SQLPreprocessor(conn, config)
152 sql.run_sql_file(conn, 'partition-tables.src.sql')
155 def truncate_data_tables(conn: Connection) -> None:
156 """ Truncate all data tables to prepare for a fresh load.
158 with conn.cursor() as cur:
159 cur.execute('TRUNCATE placex')
160 cur.execute('TRUNCATE place_addressline')
161 cur.execute('TRUNCATE location_area')
162 cur.execute('TRUNCATE location_area_country')
163 cur.execute('TRUNCATE location_property_tiger')
164 cur.execute('TRUNCATE location_property_osmline')
165 cur.execute('TRUNCATE location_postcode')
166 if conn.table_exists('search_name'):
167 cur.execute('TRUNCATE search_name')
168 cur.execute('DROP SEQUENCE IF EXISTS seq_place')
169 cur.execute('CREATE SEQUENCE seq_place start 100000')
171 cur.execute("""SELECT tablename FROM pg_tables
172 WHERE tablename LIKE 'location_road_%'""")
174 for table in [r[0] for r in list(cur)]:
175 cur.execute('TRUNCATE ' + table)
180 _COPY_COLUMNS = pysql.SQL(',').join(map(pysql.Identifier,
181 ('osm_type', 'osm_id', 'class', 'type',
182 'name', 'admin_level', 'address',
183 'extratags', 'geometry')))
186 def load_data(dsn: str, threads: int) -> None:
187 """ Copy data into the word and placex table.
189 sel = selectors.DefaultSelector()
190 # Then copy data from place to placex in <threads - 1> chunks.
191 place_threads = max(1, threads - 1)
192 for imod in range(place_threads):
193 conn = DBConnection(dsn)
196 pysql.SQL("""INSERT INTO placex ({columns})
197 SELECT {columns} FROM place
198 WHERE osm_id % {total} = {mod}
199 AND NOT (class='place' and (type='houses' or type='postcode'))
200 AND ST_IsValid(geometry)
201 """).format(columns=_COPY_COLUMNS,
202 total=pysql.Literal(place_threads),
203 mod=pysql.Literal(imod)))
204 sel.register(conn, selectors.EVENT_READ, conn)
206 # Address interpolations go into another table.
207 conn = DBConnection(dsn)
209 conn.perform("""INSERT INTO location_property_osmline (osm_id, address, linegeo)
210 SELECT osm_id, address, geometry FROM place
211 WHERE class='place' and type='houses' and osm_type='W'
212 and ST_GeometryType(geometry) = 'ST_LineString'
214 sel.register(conn, selectors.EVENT_READ, conn)
216 # Now wait for all of them to finish.
217 todo = place_threads + 1
219 for key, _ in sel.select(1):
225 print('.', end='', flush=True)
228 with connect(dsn) as syn_conn:
229 with syn_conn.cursor() as cur:
230 cur.execute('ANALYSE')
233 def create_search_indices(conn: Connection, config: Configuration,
234 drop: bool = False, threads: int = 1) -> None:
235 """ Create tables that have explicit partitioning.
238 # If index creation failed and left an index invalid, they need to be
239 # cleaned out first, so that the script recreates them.
240 with conn.cursor() as cur:
241 cur.execute("""SELECT relname FROM pg_class, pg_index
242 WHERE pg_index.indisvalid = false
243 AND pg_index.indexrelid = pg_class.oid""")
244 bad_indices = [row[0] for row in list(cur)]
245 for idx in bad_indices:
246 LOG.info("Drop invalid index %s.", idx)
247 cur.execute(pysql.SQL('DROP INDEX {}').format(pysql.Identifier(idx)))
250 sql = SQLPreprocessor(conn, config)
252 sql.run_parallel_sql_file(config.get_libpq_dsn(),
253 'indices.sql', min(8, threads), drop=drop)