2 Functions for setting up and importing a new Nominatim database.
8 from pathlib import Path
11 import psycopg2.extras
12 from psycopg2 import sql as pysql
14 from nominatim.db.connection import connect, get_pg_env
15 from nominatim.db import utils as db_utils
16 from nominatim.db.async_connection import DBConnection
17 from nominatim.db.sql_preprocessor import SQLPreprocessor
18 from nominatim.tools.exec_utils import run_osm2pgsql
19 from nominatim.errors import UsageError
20 from nominatim.version import POSTGRESQL_REQUIRED_VERSION, POSTGIS_REQUIRED_VERSION
22 LOG = logging.getLogger()
24 def setup_database_skeleton(dsn, data_dir, no_partitions, rouser=None):
25 """ Create a new database for Nominatim and populate it with the
26 essential extensions and data.
28 LOG.warning('Creating database')
29 create_db(dsn, rouser)
31 LOG.warning('Setting up database')
32 with connect(dsn) as conn:
33 setup_extensions(conn)
35 LOG.warning('Loading basic data')
36 import_base_data(dsn, data_dir, no_partitions)
39 def create_db(dsn, rouser=None):
40 """ Create a new database for the given DSN. Fails when the database
41 already exists or the PostgreSQL version is too old.
42 Uses `createdb` to create the database.
44 If 'rouser' is given, then the function also checks that the user
45 with that given name exists.
47 Requires superuser rights by the caller.
49 proc = subprocess.run(['createdb'], env=get_pg_env(dsn), check=False)
51 if proc.returncode != 0:
52 raise UsageError('Creating new database failed.')
54 with connect(dsn) as conn:
55 postgres_version = conn.server_version_tuple()
56 if postgres_version < POSTGRESQL_REQUIRED_VERSION:
57 LOG.fatal('Minimum supported version of Postgresql is %d.%d. '
58 'Found version %d.%d.',
59 POSTGRESQL_REQUIRED_VERSION[0], POSTGRESQL_REQUIRED_VERSION[1],
60 postgres_version[0], postgres_version[1])
61 raise UsageError('PostgreSQL server is too old.')
63 if rouser is not None:
64 with conn.cursor() as cur:
65 cnt = cur.scalar('SELECT count(*) FROM pg_user where usename = %s',
68 LOG.fatal("Web user '%s' does not exists. Create it with:\n"
69 "\n createuser %s", rouser, rouser)
70 raise UsageError('Missing read-only user.')
74 def setup_extensions(conn):
75 """ Set up all extensions needed for Nominatim. Also checks that the
76 versions of the extensions are sufficient.
78 with conn.cursor() as cur:
79 cur.execute('CREATE EXTENSION IF NOT EXISTS hstore')
80 cur.execute('CREATE EXTENSION IF NOT EXISTS postgis')
83 postgis_version = conn.postgis_version_tuple()
84 if postgis_version < POSTGIS_REQUIRED_VERSION:
85 LOG.fatal('Minimum supported version of PostGIS is %d.%d. '
86 'Found version %d.%d.',
87 POSTGIS_REQUIRED_VERSION[0], POSTGIS_REQUIRED_VERSION[1],
88 postgis_version[0], postgis_version[1])
89 raise UsageError('PostGIS version is too old.')
92 def import_base_data(dsn, sql_dir, ignore_partitions=False):
93 """ Create and populate the tables with basic static data that provides
94 the background for geocoding. Data is assumed to not yet exist.
96 db_utils.execute_file(dsn, sql_dir / 'country_name.sql')
97 db_utils.execute_file(dsn, sql_dir / 'country_osm_grid.sql.gz')
100 with connect(dsn) as conn:
101 with conn.cursor() as cur:
102 cur.execute('UPDATE country_name SET partition = 0')
106 def import_osm_data(osm_file, options, drop=False, ignore_errors=False):
107 """ Import the given OSM file. 'options' contains the list of
108 default settings for osm2pgsql.
110 options['import_file'] = osm_file
111 options['append'] = False
112 options['threads'] = 1
114 if not options['flatnode_file'] and options['osm2pgsql_cache'] == 0:
115 # Make some educated guesses about cache size based on the size
116 # of the import file and the available memory.
117 mem = psutil.virtual_memory()
118 fsize = os.stat(str(osm_file)).st_size
119 options['osm2pgsql_cache'] = int(min((mem.available + mem.cached) * 0.75,
120 fsize * 2) / 1024 / 1024) + 1
122 run_osm2pgsql(options)
124 with connect(options['dsn']) as conn:
125 if not ignore_errors:
126 with conn.cursor() as cur:
127 cur.execute('SELECT * FROM place LIMIT 1')
128 if cur.rowcount == 0:
129 raise UsageError('No data imported by osm2pgsql.')
132 conn.drop_table('planet_osm_nodes')
134 if drop and options['flatnode_file']:
135 Path(options['flatnode_file']).unlink()
138 def create_tables(conn, config, reverse_only=False):
139 """ Create the set of basic tables.
140 When `reverse_only` is True, then the main table for searching will
141 be skipped and only reverse search is possible.
143 sql = SQLPreprocessor(conn, config)
144 sql.env.globals['db']['reverse_only'] = reverse_only
146 sql.run_sql_file(conn, 'tables.sql')
149 def create_table_triggers(conn, config):
150 """ Create the triggers for the tables. The trigger functions must already
151 have been imported with refresh.create_functions().
153 sql = SQLPreprocessor(conn, config)
154 sql.run_sql_file(conn, 'table-triggers.sql')
157 def create_partition_tables(conn, config):
158 """ Create tables that have explicit partitioning.
160 sql = SQLPreprocessor(conn, config)
161 sql.run_sql_file(conn, 'partition-tables.src.sql')
164 def truncate_data_tables(conn):
165 """ Truncate all data tables to prepare for a fresh load.
167 with conn.cursor() as cur:
168 cur.execute('TRUNCATE placex')
169 cur.execute('TRUNCATE place_addressline')
170 cur.execute('TRUNCATE location_area')
171 cur.execute('TRUNCATE location_area_country')
172 cur.execute('TRUNCATE location_property_tiger')
173 cur.execute('TRUNCATE location_property_osmline')
174 cur.execute('TRUNCATE location_postcode')
175 if conn.table_exists('search_name'):
176 cur.execute('TRUNCATE search_name')
177 cur.execute('DROP SEQUENCE IF EXISTS seq_place')
178 cur.execute('CREATE SEQUENCE seq_place start 100000')
180 cur.execute("""SELECT tablename FROM pg_tables
181 WHERE tablename LIKE 'location_road_%'""")
183 for table in [r[0] for r in list(cur)]:
184 cur.execute('TRUNCATE ' + table)
189 _COPY_COLUMNS = pysql.SQL(',').join(map(pysql.Identifier,
190 ('osm_type', 'osm_id', 'class', 'type',
191 'name', 'admin_level', 'address',
192 'extratags', 'geometry')))
195 def load_data(dsn, threads):
196 """ Copy data into the word and placex table.
198 sel = selectors.DefaultSelector()
199 # Then copy data from place to placex in <threads - 1> chunks.
200 place_threads = max(1, threads - 1)
201 for imod in range(place_threads):
202 conn = DBConnection(dsn)
205 pysql.SQL("""INSERT INTO placex ({columns})
206 SELECT {columns} FROM place
207 WHERE osm_id % {total} = {mod}
208 AND NOT (class='place' and (type='houses' or type='postcode'))
209 AND ST_IsValid(geometry)
210 """).format(columns=_COPY_COLUMNS,
211 total=pysql.Literal(place_threads),
212 mod=pysql.Literal(imod)))
213 sel.register(conn, selectors.EVENT_READ, conn)
215 # Address interpolations go into another table.
216 conn = DBConnection(dsn)
218 conn.perform("""INSERT INTO location_property_osmline (osm_id, address, linegeo)
219 SELECT osm_id, address, geometry FROM place
220 WHERE class='place' and type='houses' and osm_type='W'
221 and ST_GeometryType(geometry) = 'ST_LineString'
223 sel.register(conn, selectors.EVENT_READ, conn)
225 # Now wait for all of them to finish.
226 todo = place_threads + 1
228 for key, _ in sel.select(1):
234 print('.', end='', flush=True)
237 with connect(dsn) as conn:
238 with conn.cursor() as cur:
239 cur.execute('ANALYSE')
242 def create_search_indices(conn, config, drop=False):
243 """ Create tables that have explicit partitioning.
246 # If index creation failed and left an index invalid, they need to be
247 # cleaned out first, so that the script recreates them.
248 with conn.cursor() as cur:
249 cur.execute("""SELECT relname FROM pg_class, pg_index
250 WHERE pg_index.indisvalid = false
251 AND pg_index.indexrelid = pg_class.oid""")
252 bad_indices = [row[0] for row in list(cur)]
253 for idx in bad_indices:
254 LOG.info("Drop invalid index %s.", idx)
255 cur.execute('DROP INDEX "{}"'.format(idx))
258 sql = SQLPreprocessor(conn, config)
260 sql.run_sql_file(conn, 'indices.sql', drop=drop)
263 def create_country_names(conn, tokenizer, languages=None):
264 """ Add default country names to search index. `languages` is a comma-
265 separated list of language codes as used in OSM. If `languages` is not
266 empty then only name translations for the given languages are added
270 languages = languages.split(',')
272 def _include_key(key):
273 return key == 'name' or \
274 (key.startswith('name:') and (not languages or key[5:] in languages))
276 with conn.cursor() as cur:
277 psycopg2.extras.register_hstore(cur)
278 cur.execute("""SELECT country_code, name FROM country_name
279 WHERE country_code is not null""")
281 with tokenizer.name_analyzer() as analyzer:
282 for code, name in cur:
283 names = {'countrycode': code}
285 names['short_name'] = 'UK'
287 names['short_name'] = 'United States'
289 # country names (only in languages as provided)
291 names.update(((k, v) for k, v in name.items() if _include_key(k)))
293 analyzer.add_country_names(code, names)