Merge remote-tracking branch 'upstream/master'

[nominatim.git] / nominatim / tools / database_import.py

diff --git a/nominatim/tools/database_import.py b/nominatim/tools/database_import.py
index 7d71386fc7921b752c10cfd57f24644aded895d2..fa60abf22a60ed7f1aad9eceebbfa2a61b1450d8 100644 (file)
--- a/nominatim/tools/database_import.py
+++ b/nominatim/tools/database_import.py
@@ -1,43 +1,50 @@
+# 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 setting up and importing a new Nominatim database.
 """
 """
 Functions for setting up and importing a new Nominatim database.
 """

+from typing import Tuple, Optional, Union, Sequence, MutableMapping, Any

 import logging
 import os
 import selectors
 import subprocess
 import logging
 import os
 import selectors
 import subprocess

-import shutil

 from pathlib import Path
 
 import psutil
 from pathlib import Path
 
 import psutil

-import psycopg2
+from psycopg2 import sql as pysql

 
 

-from ..db.connection import connect, get_pg_env
-from ..db import utils as db_utils
-from ..db.async_connection import DBConnection
-from .exec_utils import run_osm2pgsql
-from ..errors import UsageError
-from ..version import POSTGRESQL_REQUIRED_VERSION, POSTGIS_REQUIRED_VERSION
+from nominatim.config import Configuration
+from nominatim.db.connection import connect, get_pg_env, Connection
+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()
 
 
 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.
+def _require_version(module: str, actual: Tuple[int, int], expected: Tuple[int, int]) -> None:
+    """ Compares the version for the given module and raises an exception
+        if the actual version is too old.

     """
     """

-    LOG.warning('Creating database')
-    create_db(dsn, rouser)
+    if actual < expected:
+        LOG.fatal('Minimum supported version of %s is %d.%d. '
+                  'Found version %d.%d.',
+                  module, expected[0], expected[1], actual[0], actual[1])
+        raise UsageError(f'{module} is too old.')

 
 

-    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 setup_database_skeleton(dsn: str, rouser: Optional[str] = None) -> None:
+    """ Create a new database for Nominatim and populate it with the
+        essential extensions.

 
 

+        The function fails when the database already exists or Postgresql or
+        PostGIS versions are too old.

 
 

-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.

         Uses `createdb` to create the database.
 
         If 'rouser' is given, then the function also checks that the user
         Uses `createdb` to create the database.
 
         If 'rouser' is given, then the function also checks that the user


@@ -51,113 +58,50 @@ def create_db(dsn, rouser=None):
         raise UsageError('Creating new database failed.')
 
     with connect(dsn) as conn:
         raise UsageError('Creating new database failed.')
 
     with connect(dsn) as conn:

-        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.',
-                      POSTGRESQL_REQUIRED_VERSION[0], POSTGRESQL_REQUIRED_VERSION[1],
-                      postgres_version[0], postgres_version[1])
-            raise UsageError('PostgreSQL server is too old.')
+        _require_version('PostgreSQL server',
+                         conn.server_version_tuple(),
+                         POSTGRESQL_REQUIRED_VERSION)

 
         if rouser is not None:
             with conn.cursor() as cur:
                 cnt = cur.scalar('SELECT count(*) FROM pg_user where usename = %s',
                                  (rouser, ))
                 if cnt == 0:
 
         if rouser is not None:
             with conn.cursor() as cur:
                 cnt = cur.scalar('SELECT count(*) FROM pg_user where usename = %s',
                                  (rouser, ))
                 if cnt == 0:

-                    LOG.fatal("Web user '%s' does not exists. Create it with:\n"
+                    LOG.fatal("Web user '%s' does not exist. Create it with:\n"

                               "\n      createuser %s", rouser, rouser)
                     raise UsageError('Missing read-only user.')
 
                               "\n      createuser %s", rouser, rouser)
                     raise UsageError('Missing read-only user.')
 

-
-
-def setup_extensions(conn):
-    """ Set up all extensions needed for Nominatim. Also checks that the
-        versions of the extensions are sufficient.
-    """
-    with conn.cursor() as cur:
-        cur.execute('CREATE EXTENSION IF NOT EXISTS hstore')
-        cur.execute('CREATE EXTENSION IF NOT EXISTS postgis')
-    conn.commit()
-
-    postgis_version = conn.postgis_version_tuple()
-    if postgis_version < POSTGIS_REQUIRED_VERSION:
-        LOG.fatal('Minimum supported version of PostGIS is %d.%d. '
-                  'Found version %d.%d.',
-                  POSTGIS_REQUIRED_VERSION[0], POSTGIS_REQUIRED_VERSION[1],
-                  postgis_version[0], postgis_version[1])
-        raise UsageError('PostGIS version is too old.')
-
-
-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
-        for Postgresql.
-
-        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'
-
-        if not module_dir.exists() or not src_dir.samefile(module_dir):
-
-            if not module_dir.exists():
-                module_dir.mkdir()
-
-            destfile = module_dir / 'nominatim.so'
-            shutil.copy(str(src_dir / 'nominatim.so'), str(destfile))
-            destfile.chmod(0o755)
-
-            LOG.info('Database module installed at %s', str(destfile))
-        else:
-            LOG.info('Running from build directory. Leaving database module as is.')
-    else:
-        LOG.info("Using custom path for database module at '%s'", module_dir)
-
-    if conn is not None:
+        # Create extensions.

         with conn.cursor() as cur:
         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. 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')
+            cur.execute('CREATE EXTENSION IF NOT EXISTS hstore')
+            cur.execute('CREATE EXTENSION IF NOT EXISTS postgis')
+        conn.commit()

 
 

-    if ignore_partitions:
-        with connect(dsn) as conn:
-            with conn.cursor() as cur:
-                cur.execute('UPDATE country_name SET partition = 0')
-            conn.commit()
+        _require_version('PostGIS',
+                         conn.postgis_version_tuple(),
+                         POSTGIS_REQUIRED_VERSION)

 
 
 
 

-def import_osm_data(osm_file, options, drop=False, ignore_errors=False):
-    """ Import the given OSM file. 'options' contains the list of
+def import_osm_data(osm_files: Union[Path, Sequence[Path]],
+                    options: MutableMapping[str, Any],
+                    drop: bool = False, ignore_errors: bool = False) -> None:
+    """ Import the given OSM files. 'options' contains the list of

         default settings for osm2pgsql.
     """
         default settings for osm2pgsql.
     """

-    options['import_file'] = osm_file
+    options['import_file'] = osm_files

     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.
     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
+        mem = psutil.virtual_memory() # type: ignore[no-untyped-call]
+        fsize = 0
+        if isinstance(osm_files, list):
+            for fname in osm_files:
+                fsize += os.stat(str(fname)).st_size
+        else:
+            fsize = os.stat(str(osm_files)).st_size

         options['osm2pgsql_cache'] = int(min((mem.available + mem.cached) * 0.75,
                                              fsize * 2) / 1024 / 1024) + 1
 
         options['osm2pgsql_cache'] = int(min((mem.available + mem.cached) * 0.75,
                                              fsize * 2) / 1024 / 1024) + 1
 


@@ -173,25 +117,49 @@ def import_osm_data(osm_file, options, drop=False, ignore_errors=False):
         if drop:
             conn.drop_table('planet_osm_nodes')
 
         if drop:
             conn.drop_table('planet_osm_nodes')
 

-    if drop:
-        if options['flatnode_file']:
-            Path(options['flatnode_file']).unlink()
+    if drop and options['flatnode_file']:
+        Path(options['flatnode_file']).unlink()

 
 
 
 

-def truncate_data_tables(conn, max_word_frequency=None):
+def create_tables(conn: Connection, config: Configuration, reverse_only: bool = False) -> None:
+    """ 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: Connection, config: Configuration) -> None:
+    """ 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: Connection, config: Configuration) -> None:
+    """ Create tables that have explicit partitioning.
+    """
+    sql = SQLPreprocessor(conn, config)
+    sql.run_sql_file(conn, 'partition-tables.src.sql')
+
+
+def truncate_data_tables(conn: Connection) -> None:

     """ Truncate all data tables to prepare for a fresh load.
     """
     with conn.cursor() as cur:
     """ 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 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')
         cur.execute('TRUNCATE location_property_tiger')
         cur.execute('TRUNCATE location_property_osmline')
         cur.execute('TRUNCATE location_postcode')

-        cur.execute('TRUNCATE search_name')
+        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('DROP SEQUENCE IF EXISTS seq_place')
         cur.execute('CREATE SEQUENCE seq_place start 100000')
 


@@ -201,35 +169,33 @@ def truncate_data_tables(conn, max_word_frequency=None):
         for table in [r[0] for r in list(cur)]:
             cur.execute('TRUNCATE ' + table)
 
         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()
+    conn.commit()
+

 
 

-_COPY_COLUMNS = 'osm_type, osm_id, class, type, name, admin_level, address, extratags, geometry'
+_COPY_COLUMNS = pysql.SQL(',').join(map(pysql.Identifier,
+                                        ('osm_type', 'osm_id', 'class', 'type',
+                                         'name', 'admin_level', 'address',
+                                         'extratags', 'geometry')))

 
 

-def load_data(dsn, data_dir, threads):
+
+def load_data(dsn: str, threads: int) -> None:

     """ Copy data into the word and placex table.
     """
     """ 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 <threads - 1> chunks.
     place_threads = max(1, threads - 1)
     for imod in range(place_threads):
         conn = DBConnection(dsn)
         conn.connect()
     sel = selectors.DefaultSelector()
     # Then copy data from place to placex in <threads - 1> 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))
+        conn.perform(
+            pysql.SQL("""INSERT INTO placex ({columns})
+                           SELECT {columns} FROM place
+                           WHERE osm_id % {total} = {mod}
+                             AND NOT (class='place' and (type='houses' or type='postcode'))
+                             AND ST_IsValid(geometry)
+                      """).format(columns=_COPY_COLUMNS,
+                                  total=pysql.Literal(place_threads),
+                                  mod=pysql.Literal(imod)))

         sel.register(conn, selectors.EVENT_READ, conn)
 
     # Address interpolations go into another table.
         sel.register(conn, selectors.EVENT_READ, conn)
 
     # Address interpolations go into another table.


@@ -254,6 +220,27 @@ def load_data(dsn, data_dir, threads):
         print('.', end='', flush=True)
     print('\n')
 
         print('.', end='', flush=True)
     print('\n')
 

-    with connect(dsn) as conn:
-        with conn.cursor() as cur:
+    with connect(dsn) as syn_conn:
+        with syn_conn.cursor() as cur:

             cur.execute('ANALYSE')
             cur.execute('ANALYSE')

+
+
+def create_search_indices(conn: Connection, config: Configuration, drop: bool = False) -> None:
+    """ 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(pysql.SQL('DROP INDEX {}').format(pysql.Identifier(idx)))
+    conn.commit()
+
+    sql = SQLPreprocessor(conn, config)
+
+    sql.run_sql_file(conn, 'indices.sql', drop=drop)