X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/f6e894a53af83a69f553555cb4a6248d57a58391..616789bbccd4b35cc643b20c5af3c9835eb22643:/nominatim/tools/database_import.py diff --git a/nominatim/tools/database_import.py b/nominatim/tools/database_import.py index 89e01f66..25efedb9 100644 --- a/nominatim/tools/database_import.py +++ b/nominatim/tools/database_import.py @@ -2,16 +2,40 @@ Functions for setting up and importing a new Nominatim database. """ import logging +import os +import selectors import subprocess import shutil +from pathlib import Path -from ..db.connection import connect, get_pg_env -from ..db import utils as db_utils -from ..errors import UsageError -from ..version import POSTGRESQL_REQUIRED_VERSION, POSTGIS_REQUIRED_VERSION +import psutil +import psycopg2 + +from nominatim.db.connection import connect, get_pg_env +from nominatim.db import utils as db_utils +from nominatim.db.async_connection import DBConnection +from nominatim.db.sql_preprocessor import SQLPreprocessor +from nominatim.tools.exec_utils import run_osm2pgsql +from nominatim.errors import UsageError +from nominatim.version import POSTGRESQL_REQUIRED_VERSION, POSTGIS_REQUIRED_VERSION LOG = logging.getLogger() +def setup_database_skeleton(dsn, data_dir, no_partitions, rouser=None): + """ Create a new database for Nominatim and populate it with the + essential extensions and data. + """ + LOG.warning('Creating database') + create_db(dsn, rouser) + + LOG.warning('Setting up database') + with connect(dsn) as conn: + setup_extensions(conn) + + LOG.warning('Loading basic data') + import_base_data(dsn, data_dir, no_partitions) + + def create_db(dsn, rouser=None): """ Create a new database for the given DSN. Fails when the database already exists or the PostgreSQL version is too old. @@ -28,7 +52,7 @@ def create_db(dsn, rouser=None): raise UsageError('Creating new database failed.') with connect(dsn) as conn: - postgres_version = conn.server_version_tuple() # pylint: disable=E1101 + postgres_version = conn.server_version_tuple() if postgres_version < POSTGRESQL_REQUIRED_VERSION: LOG.fatal('Minimum supported version of Postgresql is %d.%d. ' 'Found version %d.%d.', @@ -37,7 +61,7 @@ def create_db(dsn, rouser=None): raise UsageError('PostgreSQL server is too old.') if rouser is not None: - with conn.cursor() as cur: # pylint: disable=E1101 + with conn.cursor() as cur: cnt = cur.scalar('SELECT count(*) FROM pg_user where usename = %s', (rouser, )) if cnt == 0: @@ -65,7 +89,7 @@ def setup_extensions(conn): raise UsageError('PostGIS version is too old.') -def install_module(src_dir, project_dir, module_dir): +def install_module(src_dir, project_dir, module_dir, conn=None): """ Copy the normalization module from src_dir into the project directory under the '/module' directory. If 'module_dir' is set, then use the module from there instead and check that it is accessible @@ -73,6 +97,9 @@ def install_module(src_dir, project_dir, module_dir): The function detects when the installation is run from the build directory. It doesn't touch the module in that case. + + If 'conn' is given, then the function also tests if the module + can be access via the given database. """ if not module_dir: module_dir = project_dir / 'module' @@ -92,30 +119,222 @@ def install_module(src_dir, project_dir, module_dir): else: LOG.info("Using custom path for database module at '%s'", module_dir) - return module_dir - - -def check_module_dir_path(conn, path): - """ Check that the normalisation module can be found and executed - from the given path. - """ - with conn.cursor() as cur: - cur.execute("""CREATE FUNCTION nominatim_test_import_func(text) - RETURNS text AS '{}/nominatim.so', 'transliteration' - LANGUAGE c IMMUTABLE STRICT; - DROP FUNCTION nominatim_test_import_func(text) - """.format(path)) + if conn is not None: + with conn.cursor() as cur: + try: + cur.execute("""CREATE FUNCTION nominatim_test_import_func(text) + RETURNS text AS '{}/nominatim.so', 'transliteration' + LANGUAGE c IMMUTABLE STRICT; + DROP FUNCTION nominatim_test_import_func(text) + """.format(module_dir)) + except psycopg2.DatabaseError as err: + LOG.fatal("Error accessing database module: %s", err) + raise UsageError("Database module cannot be accessed.") from err def import_base_data(dsn, sql_dir, ignore_partitions=False): """ Create and populate the tables with basic static data that provides - the background for geocoding. + the background for geocoding. Data is assumed to not yet exist. """ db_utils.execute_file(dsn, sql_dir / 'country_name.sql') db_utils.execute_file(dsn, sql_dir / 'country_osm_grid.sql.gz') if ignore_partitions: with connect(dsn) as conn: - with conn.cursor() as cur: # pylint: disable=E1101 + with conn.cursor() as cur: cur.execute('UPDATE country_name SET partition = 0') - conn.commit() # pylint: disable=E1101 + conn.commit() + + +def import_osm_data(osm_file, options, drop=False, ignore_errors=False): + """ Import the given OSM file. 'options' contains the list of + default settings for osm2pgsql. + """ + options['import_file'] = osm_file + options['append'] = False + options['threads'] = 1 + + if not options['flatnode_file'] and options['osm2pgsql_cache'] == 0: + # Make some educated guesses about cache size based on the size + # of the import file and the available memory. + mem = psutil.virtual_memory() + fsize = os.stat(str(osm_file)).st_size + options['osm2pgsql_cache'] = int(min((mem.available + mem.cached) * 0.75, + fsize * 2) / 1024 / 1024) + 1 + + run_osm2pgsql(options) + + with connect(options['dsn']) as conn: + if not ignore_errors: + with conn.cursor() as cur: + cur.execute('SELECT * FROM place LIMIT 1') + if cur.rowcount == 0: + raise UsageError('No data imported by osm2pgsql.') + + if drop: + conn.drop_table('planet_osm_nodes') + + if drop: + if options['flatnode_file']: + Path(options['flatnode_file']).unlink() + + +def create_tables(conn, config, reverse_only=False): + """ Create the set of basic tables. + When `reverse_only` is True, then the main table for searching will + be skipped and only reverse search is possible. + """ + sql = SQLPreprocessor(conn, config) + sql.env.globals['db']['reverse_only'] = reverse_only + + sql.run_sql_file(conn, 'tables.sql') + + +def create_table_triggers(conn, config): + """ Create the triggers for the tables. The trigger functions must already + have been imported with refresh.create_functions(). + """ + sql = SQLPreprocessor(conn, config) + sql.run_sql_file(conn, 'table-triggers.sql') + + +def create_partition_tables(conn, config): + """ Create tables that have explicit partitioning. + """ + sql = SQLPreprocessor(conn, config) + sql.run_sql_file(conn, 'partition-tables.src.sql') + + +def truncate_data_tables(conn, max_word_frequency=None): + """ Truncate all data tables to prepare for a fresh load. + """ + with conn.cursor() as cur: + cur.execute('TRUNCATE word') + cur.execute('TRUNCATE placex') + cur.execute('TRUNCATE place_addressline') + cur.execute('TRUNCATE location_area') + cur.execute('TRUNCATE location_area_country') + cur.execute('TRUNCATE location_property') + cur.execute('TRUNCATE location_property_tiger') + cur.execute('TRUNCATE location_property_osmline') + cur.execute('TRUNCATE location_postcode') + if conn.table_exists('search_name'): + cur.execute('TRUNCATE search_name') + cur.execute('DROP SEQUENCE IF EXISTS seq_place') + cur.execute('CREATE SEQUENCE seq_place start 100000') + + cur.execute("""SELECT tablename FROM pg_tables + WHERE tablename LIKE 'location_road_%'""") + + for table in [r[0] for r in list(cur)]: + cur.execute('TRUNCATE ' + table) + + if max_word_frequency is not None: + # Used by getorcreate_word_id to ignore frequent partial words. + cur.execute("""CREATE OR REPLACE FUNCTION get_maxwordfreq() + RETURNS integer AS $$ + SELECT {} as maxwordfreq; + $$ LANGUAGE SQL IMMUTABLE + """.format(max_word_frequency)) + conn.commit() + +_COPY_COLUMNS = 'osm_type, osm_id, class, type, name, admin_level, address, extratags, geometry' + +def load_data(dsn, data_dir, threads): + """ Copy data into the word and placex table. + """ + # Pre-calculate the most important terms in the word list. + db_utils.execute_file(dsn, data_dir / 'words.sql') + + sel = selectors.DefaultSelector() + # Then copy data from place to placex in chunks. + place_threads = max(1, threads - 1) + for imod in range(place_threads): + conn = DBConnection(dsn) + conn.connect() + conn.perform("""INSERT INTO placex ({0}) + SELECT {0} FROM place + WHERE osm_id % {1} = {2} + AND NOT (class='place' and type='houses') + AND ST_IsValid(geometry) + """.format(_COPY_COLUMNS, place_threads, imod)) + sel.register(conn, selectors.EVENT_READ, conn) + + # Address interpolations go into another table. + conn = DBConnection(dsn) + conn.connect() + conn.perform("""INSERT INTO location_property_osmline (osm_id, address, linegeo) + SELECT osm_id, address, geometry FROM place + WHERE class='place' and type='houses' and osm_type='W' + and ST_GeometryType(geometry) = 'ST_LineString' + """) + sel.register(conn, selectors.EVENT_READ, conn) + + # Now wait for all of them to finish. + todo = place_threads + 1 + while todo > 0: + for key, _ in sel.select(1): + conn = key.data + sel.unregister(conn) + conn.wait() + conn.close() + todo -= 1 + print('.', end='', flush=True) + print('\n') + + with connect(dsn) as conn: + with conn.cursor() as cur: + cur.execute('ANALYSE') + + +def create_search_indices(conn, config, drop=False): + """ Create tables that have explicit partitioning. + """ + + # If index creation failed and left an index invalid, they need to be + # cleaned out first, so that the script recreates them. + with conn.cursor() as cur: + cur.execute("""SELECT relname FROM pg_class, pg_index + WHERE pg_index.indisvalid = false + AND pg_index.indexrelid = pg_class.oid""") + bad_indices = [row[0] for row in list(cur)] + for idx in bad_indices: + LOG.info("Drop invalid index %s.", idx) + cur.execute('DROP INDEX "{}"'.format(idx)) + conn.commit() + + sql = SQLPreprocessor(conn, config) + + sql.run_sql_file(conn, 'indices.sql', drop=drop) + +def create_country_names(conn, config): + """ Create search index for default country names. + """ + + with conn.cursor() as cur: + cur.execute("""SELECT getorcreate_country(make_standard_name('uk'), 'gb')""") + cur.execute("""SELECT getorcreate_country(make_standard_name('united states'), 'us')""") + cur.execute("""SELECT COUNT(*) FROM + (SELECT getorcreate_country(make_standard_name(country_code), + country_code) FROM country_name WHERE country_code is not null) AS x""") + cur.execute("""SELECT COUNT(*) FROM + (SELECT getorcreate_country(make_standard_name(name->'name'), country_code) + FROM country_name WHERE name ? 'name') AS x""") + sql_statement = """SELECT COUNT(*) FROM (SELECT getorcreate_country(make_standard_name(v), + country_code) FROM (SELECT country_code, skeys(name) + AS k, svals(name) AS v FROM country_name) x WHERE k""" + + languages = config.LANGUAGES + + if languages: + sql_statement = "{} IN (".format(sql_statement) + delim = '' + for language in languages.split(','): + sql_statement = "{}{}'name:{}'".format(sql_statement, delim, language) + delim = ', ' + sql_statement = '{})'.format(sql_statement) + else: + sql_statement = "{} LIKE 'name:%'".format(sql_statement) + sql_statement = "{}) v".format(sql_statement) + cur.execute(sql_statement) + conn.commit()