X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/b93ec2522e25102e8a36ca78423a53b4579c68c2..6b55ab5581a3f6541d16fb7544ce8fd57dc2a673:/nominatim/tools/refresh.py diff --git a/nominatim/tools/refresh.py b/nominatim/tools/refresh.py index 33efe8f8..8c1e9d9b 100644 --- a/nominatim/tools/refresh.py +++ b/nominatim/tools/refresh.py @@ -1,31 +1,31 @@ +# 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 bringing auxiliary data in the database up-to-date. """ -import json +from typing import MutableSequence, Tuple, Any, Type, Mapping, Sequence, List, cast import logging -import re from textwrap import dedent +from pathlib import Path -from psycopg2.extras import execute_values +from psycopg2 import sql as pysql -from ..db.utils import execute_file +from nominatim.config import Configuration +from nominatim.db.connection import Connection +from nominatim.db.utils import execute_file +from nominatim.db.sql_preprocessor import SQLPreprocessor +from nominatim.version import version_str LOG = logging.getLogger() -def update_postcodes(dsn, sql_dir): - """ Recalculate postcode centroids and add, remove and update entries in the - location_postcode table. `conn` is an opne connection to the database. - """ - execute_file(dsn, sql_dir / 'update-postcodes.sql') - - -def recompute_word_counts(dsn, sql_dir): - """ Compute the frequency of full-word search terms. - """ - execute_file(dsn, sql_dir / 'words_from_search_name.sql') +OSM_TYPE = {'N': 'node', 'W': 'way', 'R': 'relation'} - -def _add_address_level_rows_from_entry(rows, entry): +def _add_address_level_rows_from_entry(rows: MutableSequence[Tuple[Any, ...]], + entry: Mapping[str, Any]) -> None: """ Converts a single entry from the JSON format for address rank descriptions into a flat format suitable for inserting into a PostgreSQL table and adds these lines to `rows`. @@ -42,133 +42,59 @@ def _add_address_level_rows_from_entry(rows, entry): for country in countries: rows.append((country, key, value, rank_search, rank_address)) -def load_address_levels(conn, table, levels): + +def load_address_levels(conn: Connection, table: str, levels: Sequence[Mapping[str, Any]]) -> None: """ Replace the `address_levels` table with the contents of `levels'. A new table is created any previously existing table is dropped. The table has the following columns: country, class, type, rank_search, rank_address """ - rows = [] + rows: List[Tuple[Any, ...]] = [] for entry in levels: _add_address_level_rows_from_entry(rows, entry) with conn.cursor() as cur: - cur.execute('DROP TABLE IF EXISTS {}'.format(table)) + cur.drop_table(table) - cur.execute("""CREATE TABLE {} (country_code varchar(2), + cur.execute(pysql.SQL("""CREATE TABLE {} ( + country_code varchar(2), class TEXT, type TEXT, rank_search SMALLINT, - rank_address SMALLINT)""".format(table)) + rank_address SMALLINT) + """).format(pysql.Identifier(table))) - execute_values(cur, "INSERT INTO {} VALUES %s".format(table), rows) + cur.execute_values(pysql.SQL("INSERT INTO {} VALUES %s") + .format(pysql.Identifier(table)), rows) - cur.execute('CREATE UNIQUE INDEX ON {} (country_code, class, type)'.format(table)) + cur.execute(pysql.SQL('CREATE UNIQUE INDEX ON {} (country_code, class, type)') + .format(pysql.Identifier(table))) conn.commit() -def load_address_levels_from_file(conn, config_file): - """ Replace the `address_levels` table with the contents of the config - file. - """ - with config_file.open('r') as fdesc: - load_address_levels(conn, 'address_levels', json.load(fdesc)) - -PLPGSQL_BASE_MODULES = ( - 'utils.sql', - 'normalization.sql', - 'ranking.sql', - 'importance.sql', - 'address_lookup.sql', - 'interpolation.sql' -) - -PLPGSQL_TABLE_MODULES = ( - ('place', 'place_triggers.sql'), - ('placex', 'placex_triggers.sql'), - ('location_postcode', 'postcode_triggers.sql') -) - -def _get_standard_function_sql(conn, config, sql_dir, enable_diff_updates, enable_debug): - """ Read all applicable SQLs containing PL/pgSQL functions, replace - placefolders and execute them. - """ - sql_func_dir = sql_dir / 'functions' - sql = '' - - # Get the basic set of functions that is always imported. - for sql_file in PLPGSQL_BASE_MODULES: - with (sql_func_dir / sql_file).open('r') as fdesc: - sql += fdesc.read() - - # Some files require the presence of a certain table - for table, fname in PLPGSQL_TABLE_MODULES: - if conn.table_exists(table): - with (sql_func_dir / fname).open('r') as fdesc: - sql += fdesc.read() - - # Replace placeholders. - sql = sql.replace('{modulepath}', - config.DATABASE_MODULE_PATH or str((config.project_dir / 'module').resolve())) - if enable_diff_updates: - sql = sql.replace('RETURN NEW; -- %DIFFUPDATES%', '--') - - if enable_debug: - sql = sql.replace('--DEBUG:', '') - - if config.get_bool('LIMIT_REINDEXING'): - sql = sql.replace('--LIMIT INDEXING:', '') - - if not config.get_bool('USE_US_TIGER_DATA'): - sql = sql.replace('-- %NOTIGERDATA% ', '') - - if not config.get_bool('USE_AUX_LOCATION_DATA'): - sql = sql.replace('-- %NOAUXDATA% ', '') - - reverse_only = 'false' if conn.table_exists('search_name') else 'true' - - return sql.replace('%REVERSE-ONLY%', reverse_only) - - -def replace_partition_string(sql, partitions): - """ Replace a partition template with the actual partition code. +def load_address_levels_from_config(conn: Connection, config: Configuration) -> None: + """ Replace the `address_levels` table with the content as + defined in the given configuration. Uses the parameter + NOMINATIM_ADDRESS_LEVEL_CONFIG to determine the location of the + configuration file. """ - for match in re.findall('^-- start(.*?)^-- end', sql, re.M | re.S): - repl = '' - for part in partitions: - repl += match.replace('-partition-', str(part)) - sql = sql.replace(match, repl) + cfg = config.load_sub_configuration('', config='ADDRESS_LEVEL_CONFIG') + load_address_levels(conn, 'address_levels', cfg) - return sql -def _get_partition_function_sql(conn, sql_dir): - """ Create functions that work on partition tables. - """ - with conn.cursor() as cur: - cur.execute('SELECT distinct partition FROM country_name') - partitions = set([0]) - for row in cur: - partitions.add(row[0]) - - with (sql_dir / 'partition-functions.src.sql').open('r') as fdesc: - sql = fdesc.read() - - return replace_partition_string(sql, sorted(partitions)) - -def create_functions(conn, config, sql_dir, - enable_diff_updates=True, enable_debug=False): +def create_functions(conn: Connection, config: Configuration, + enable_diff_updates: bool = True, + enable_debug: bool = False) -> None: """ (Re)create the PL/pgSQL functions. """ - sql = _get_standard_function_sql(conn, config, sql_dir, - enable_diff_updates, enable_debug) - sql += _get_partition_function_sql(conn, sql_dir) + sql = SQLPreprocessor(conn, config) - with conn.cursor() as cur: - cur.execute(sql) + sql.run_sql_file(conn, 'functions.sql', + disable_diff_updates=not enable_diff_updates, + debug=enable_debug) - conn.commit() WEBSITE_SCRIPTS = ( @@ -186,48 +112,138 @@ PHP_CONST_DEFS = ( ('Database_DSN', 'DATABASE_DSN', str), ('Default_Language', 'DEFAULT_LANGUAGE', str), ('Log_DB', 'LOG_DB', bool), - ('Log_File', 'LOG_FILE', str), - ('Max_Word_Frequency', 'MAX_WORD_FREQUENCY', int), + ('Log_File', 'LOG_FILE', Path), ('NoAccessControl', 'CORS_NOACCESSCONTROL', bool), ('Places_Max_ID_count', 'LOOKUP_MAX_COUNT', int), ('PolygonOutput_MaximumTypes', 'POLYGON_OUTPUT_MAX_TYPES', int), ('Search_BatchMode', 'SEARCH_BATCH_MODE', bool), ('Search_NameOnlySearchFrequencyThreshold', 'SEARCH_NAME_ONLY_THRESHOLD', str), - ('Term_Normalization_Rules', 'TERM_NORMALIZATION', str), - ('Use_Aux_Location_data', 'USE_AUX_LOCATION_DATA', bool), ('Use_US_Tiger_Data', 'USE_US_TIGER_DATA', bool), ('MapIcon_URL', 'MAPICON_URL', str), ) -def setup_website(basedir, phplib_dir, config): +def import_wikipedia_articles(dsn: str, data_path: Path, ignore_errors: bool = False) -> int: + """ Replaces the wikipedia importance tables with new data. + The import is run in a single transaction so that the new data + is replace seamlessly. + + Returns 0 if all was well and 1 if the importance file could not + be found. Throws an exception if there was an error reading the file. + """ + datafile = data_path / 'wikimedia-importance.sql.gz' + + if not datafile.exists(): + return 1 + + pre_code = """BEGIN; + DROP TABLE IF EXISTS "wikipedia_article"; + DROP TABLE IF EXISTS "wikipedia_redirect" + """ + post_code = "COMMIT" + execute_file(dsn, datafile, ignore_errors=ignore_errors, + pre_code=pre_code, post_code=post_code) + + return 0 + + +def recompute_importance(conn: Connection) -> None: + """ Recompute wikipedia links and importance for all entries in placex. + This is a long-running operations that must not be executed in + parallel with updates. + """ + with conn.cursor() as cur: + cur.execute('ALTER TABLE placex DISABLE TRIGGER ALL') + cur.execute(""" + UPDATE placex SET (wikipedia, importance) = + (SELECT wikipedia, importance + FROM compute_importance(extratags, country_code, osm_type, osm_id)) + """) + cur.execute(""" + UPDATE placex s SET wikipedia = d.wikipedia, importance = d.importance + FROM placex d + WHERE s.place_id = d.linked_place_id and d.wikipedia is not null + and (s.wikipedia is null or s.importance < d.importance); + """) + + cur.execute('ALTER TABLE placex ENABLE TRIGGER ALL') + conn.commit() + + +def _quote_php_variable(var_type: Type[Any], config: Configuration, + conf_name: str) -> str: + if var_type == bool: + return 'true' if config.get_bool(conf_name) else 'false' + + if var_type == int: + return cast(str, getattr(config, conf_name)) + + if not getattr(config, conf_name): + return 'false' + + if var_type == Path: + value = str(config.get_path(conf_name) or '') + else: + value = getattr(config, conf_name) + + quoted = value.replace("'", "\\'") + return f"'{quoted}'" + + +def setup_website(basedir: Path, config: Configuration, conn: Connection) -> None: """ Create the website script stubs. """ if not basedir.exists(): LOG.info('Creating website directory.') basedir.mkdir() - template = dedent("""\ + template = dedent(f"""\ None: + """ Mark the given OSM object for reindexing. When 'recursive' is set + to True (the default), then all dependent objects are marked for + reindexing as well. + + 'osm_type' must be on of 'N' (node), 'W' (way) or 'R' (relation). + If the given object does not exist, then nothing happens. + """ + assert osm_type in ('N', 'R', 'W') + + LOG.warning("Invalidating OSM %s %s%s.", + OSM_TYPE[osm_type], osm_id, + ' and its dependent places' if recursive else '') + + with conn.cursor() as cur: + if recursive: + sql = """SELECT place_force_update(place_id) + FROM placex WHERE osm_type = %s and osm_id = %s""" + else: + sql = """UPDATE placex SET indexed_status = 2 + WHERE osm_type = %s and osm_id = %s""" + + cur.execute(sql, (osm_type, osm_id))