X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/a0f5613a23167752c97dd17fc42cd1e4e97cf4eb..bf1f6a997ccdb115f7c77ab8ded34e80dca7abeb:/nominatim/tokenizer/legacy_tokenizer.py diff --git a/nominatim/tokenizer/legacy_tokenizer.py b/nominatim/tokenizer/legacy_tokenizer.py index d901a68d..a50dedb2 100644 --- a/nominatim/tokenizer/legacy_tokenizer.py +++ b/nominatim/tokenizer/legacy_tokenizer.py @@ -1,8 +1,17 @@ +# 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. """ Tokenizer implementing normalisation as used before Nominatim 4. """ +from typing import Optional, Sequence, List, Tuple, Mapping, Any, Callable, \ + cast, Dict, Set, Iterable from collections import OrderedDict import logging +from pathlib import Path import re import shutil from textwrap import dedent @@ -11,10 +20,12 @@ from icu import Transliterator import psycopg2 import psycopg2.extras -from nominatim.db.connection import connect +from nominatim.db.connection import connect, Connection +from nominatim.config import Configuration from nominatim.db import properties from nominatim.db import utils as db_utils from nominatim.db.sql_preprocessor import SQLPreprocessor +from nominatim.data.place_info import PlaceInfo from nominatim.errors import UsageError from nominatim.tokenizer.base import AbstractAnalyzer, AbstractTokenizer @@ -23,13 +34,13 @@ DBCFG_MAXWORDFREQ = "tokenizer_maxwordfreq" LOG = logging.getLogger() -def create(dsn, data_dir): +def create(dsn: str, data_dir: Path) -> 'LegacyTokenizer': """ Create a new instance of the tokenizer provided by this module. """ return LegacyTokenizer(dsn, data_dir) -def _install_module(config_module_path, src_dir, module_dir): +def _install_module(config_module_path: str, src_dir: Path, module_dir: Path) -> str: """ Copies the PostgreSQL normalisation module into the project directory if necessary. For historical reasons the module is saved in the '/module' subdirectory and not with the other tokenizer @@ -46,7 +57,7 @@ def _install_module(config_module_path, src_dir, module_dir): # Compatibility mode for builddir installations. if module_dir.exists() and src_dir.samefile(module_dir): LOG.info('Running from build directory. Leaving database module as is.') - return module_dir + return str(module_dir) # In any other case install the module in the project directory. if not module_dir.exists(): @@ -58,20 +69,20 @@ def _install_module(config_module_path, src_dir, module_dir): LOG.info('Database module installed at %s', str(destfile)) - return module_dir + return str(module_dir) -def _check_module(module_dir, conn): +def _check_module(module_dir: str, conn: Connection) -> None: """ Try to use the PostgreSQL module to confirm that it is correctly installed and accessible from PostgreSQL. """ with conn.cursor() as cur: try: cur.execute("""CREATE FUNCTION nominatim_test_import_func(text) - RETURNS text AS '{}/nominatim.so', 'transliteration' + RETURNS text AS %s, 'transliteration' LANGUAGE c IMMUTABLE STRICT; DROP FUNCTION nominatim_test_import_func(text) - """.format(module_dir)) + """, (f'{module_dir}/nominatim.so', )) except psycopg2.DatabaseError as err: LOG.fatal("Error accessing database module: %s", err) raise UsageError("Database module cannot be accessed.") from err @@ -83,25 +94,26 @@ class LegacyTokenizer(AbstractTokenizer): calls to the database. """ - def __init__(self, dsn, data_dir): + def __init__(self, dsn: str, data_dir: Path) -> None: self.dsn = dsn self.data_dir = data_dir - self.normalization = None + self.normalization: Optional[str] = None - def init_new_db(self, config, init_db=True): + def init_new_db(self, config: Configuration, init_db: bool = True) -> None: """ Set up a new tokenizer for the database. This copies all necessary data in the project directory to make sure the tokenizer remains stable even over updates. """ + assert config.project_dir is not None module_dir = _install_module(config.DATABASE_MODULE_PATH, config.lib_dir.module, config.project_dir / 'module') self.normalization = config.TERM_NORMALIZATION - self._install_php(config) + self._install_php(config, overwrite=True) with connect(self.dsn) as conn: _check_module(module_dir, conn) @@ -113,14 +125,22 @@ class LegacyTokenizer(AbstractTokenizer): self._init_db_tables(config) - def init_from_project(self, _): + def init_from_project(self, config: Configuration) -> None: """ Initialise the tokenizer from the project directory. """ + assert config.project_dir is not None + with connect(self.dsn) as conn: self.normalization = properties.get_property(conn, DBCFG_NORMALIZATION) + if not (config.project_dir / 'module' / 'nominatim.so').exists(): + _install_module(config.DATABASE_MODULE_PATH, + config.lib_dir.module, + config.project_dir / 'module') + + self._install_php(config, overwrite=False) - def finalize_import(self, config): + def finalize_import(self, config: Configuration) -> None: """ Do any required postprocessing to make the tokenizer data ready for use. """ @@ -129,9 +149,11 @@ class LegacyTokenizer(AbstractTokenizer): sqlp.run_sql_file(conn, 'tokenizer/legacy_tokenizer_indices.sql') - def update_sql_functions(self, config): + def update_sql_functions(self, config: Configuration) -> None: """ Reimport the SQL functions for this tokenizer. """ + assert config.project_dir is not None + with connect(self.dsn) as conn: max_word_freq = properties.get_property(conn, DBCFG_MAXWORDFREQ) modulepath = config.DATABASE_MODULE_PATH or \ @@ -142,7 +164,7 @@ class LegacyTokenizer(AbstractTokenizer): modulepath=modulepath) - def check_database(self, _): + def check_database(self, _: Configuration) -> Optional[str]: """ Check that the tokenizer is set up correctly. """ hint = """\ @@ -169,13 +191,15 @@ class LegacyTokenizer(AbstractTokenizer): return None - def migrate_database(self, config): + def migrate_database(self, config: Configuration) -> None: """ Initialise the project directory of an existing database for use with this tokenizer. This is a special migration function for updating existing databases to new software versions. """ + assert config.project_dir is not None + self.normalization = config.TERM_NORMALIZATION module_dir = _install_module(config.DATABASE_MODULE_PATH, config.lib_dir.module, @@ -186,25 +210,33 @@ class LegacyTokenizer(AbstractTokenizer): self._save_config(conn, config) - def update_statistics(self): + def update_statistics(self) -> None: """ Recompute the frequency of full words. """ with connect(self.dsn) as conn: - with conn.cursor() as cur: - cur.drop_table("word_frequencies") - LOG.info("Computing word frequencies") - cur.execute("""CREATE TEMP TABLE word_frequencies AS - SELECT unnest(name_vector) as id, count(*) - FROM search_name GROUP BY id""") - cur.execute("CREATE INDEX ON word_frequencies(id)") - LOG.info("Update word table with recomputed frequencies") - cur.execute("""UPDATE word SET search_name_count = count - FROM word_frequencies - WHERE word_token like ' %' and word_id = id""") - cur.drop_table("word_frequencies") + if conn.table_exists('search_name'): + with conn.cursor() as cur: + cur.drop_table("word_frequencies") + LOG.info("Computing word frequencies") + cur.execute("""CREATE TEMP TABLE word_frequencies AS + SELECT unnest(name_vector) as id, count(*) + FROM search_name GROUP BY id""") + cur.execute("CREATE INDEX ON word_frequencies(id)") + LOG.info("Update word table with recomputed frequencies") + cur.execute("""UPDATE word SET search_name_count = count + FROM word_frequencies + WHERE word_token like ' %' and word_id = id""") + cur.drop_table("word_frequencies") conn.commit() - def name_analyzer(self): + + def update_word_tokens(self) -> None: + """ No house-keeping implemented for the legacy tokenizer. + """ + LOG.info("No tokenizer clean-up available.") + + + def name_analyzer(self) -> 'LegacyNameAnalyzer': """ Create a new analyzer for tokenizing names and queries using this tokinzer. Analyzers are context managers and should be used accordingly: @@ -224,19 +256,21 @@ class LegacyTokenizer(AbstractTokenizer): return LegacyNameAnalyzer(self.dsn, normalizer) - def _install_php(self, config): + def _install_php(self, config: Configuration, overwrite: bool = True) -> None: """ Install the php script for the tokenizer. """ php_file = self.data_dir / "tokenizer.php" - php_file.write_text(dedent("""\ - None: """ Set up the word table and fill it with pre-computed word frequencies. """ @@ -249,10 +283,12 @@ class LegacyTokenizer(AbstractTokenizer): db_utils.execute_file(self.dsn, config.lib_dir.data / 'words.sql') - def _save_config(self, conn, config): + def _save_config(self, conn: Connection, config: Configuration) -> None: """ Save the configuration that needs to remain stable for the given database as database properties. """ + assert self.normalization is not None + properties.set_property(conn, DBCFG_NORMALIZATION, self.normalization) properties.set_property(conn, DBCFG_MAXWORDFREQ, config.MAX_WORD_FREQUENCY) @@ -265,8 +301,8 @@ class LegacyNameAnalyzer(AbstractAnalyzer): normalization. """ - def __init__(self, dsn, normalizer): - self.conn = connect(dsn).connection + def __init__(self, dsn: str, normalizer: Any): + self.conn: Optional[Connection] = connect(dsn).connection self.conn.autocommit = True self.normalizer = normalizer psycopg2.extras.register_hstore(self.conn) @@ -274,7 +310,7 @@ class LegacyNameAnalyzer(AbstractAnalyzer): self._cache = _TokenCache(self.conn) - def close(self): + def close(self) -> None: """ Free all resources used by the analyzer. """ if self.conn: @@ -282,7 +318,7 @@ class LegacyNameAnalyzer(AbstractAnalyzer): self.conn = None - def get_word_token_info(self, words): + def get_word_token_info(self, words: Sequence[str]) -> List[Tuple[str, str, int]]: """ Return token information for the given list of words. If a word starts with # it is assumed to be a full name otherwise is a partial name. @@ -293,6 +329,7 @@ class LegacyNameAnalyzer(AbstractAnalyzer): The function is used for testing and debugging only and not necessarily efficient. """ + assert self.conn is not None with self.conn.cursor() as cur: cur.execute("""SELECT t.term, word_token, word_id FROM word, (SELECT unnest(%s::TEXT[]) as term) t @@ -308,15 +345,14 @@ class LegacyNameAnalyzer(AbstractAnalyzer): return [(r[0], r[1], r[2]) for r in cur] - def normalize(self, phrase): + def normalize(self, phrase: str) -> str: """ Normalize the given phrase, i.e. remove all properties that are irrelevant for search. """ - return self.normalizer.transliterate(phrase) + return cast(str, self.normalizer.transliterate(phrase)) - @staticmethod - def normalize_postcode(postcode): + def normalize_postcode(self, postcode: str) -> str: """ Convert the postcode to a standardized form. This function must yield exactly the same result as the SQL function @@ -325,10 +361,12 @@ class LegacyNameAnalyzer(AbstractAnalyzer): return postcode.strip().upper() - def update_postcodes_from_db(self): + def update_postcodes_from_db(self) -> None: """ Update postcode tokens in the word table from the location_postcode table. """ + assert self.conn is not None + with self.conn.cursor() as cur: # This finds us the rows in location_postcode and word that are # missing in the other table. @@ -362,9 +400,12 @@ class LegacyNameAnalyzer(AbstractAnalyzer): - def update_special_phrases(self, phrases, should_replace): + def update_special_phrases(self, phrases: Iterable[Tuple[str, str, str, str]], + should_replace: bool) -> None: """ Replace the search index for special phrases with the new phrases. """ + assert self.conn is not None + norm_phrases = set(((self.normalize(p[0]), p[1], p[2], p[3]) for p in phrases)) @@ -401,9 +442,11 @@ class LegacyNameAnalyzer(AbstractAnalyzer): len(norm_phrases), len(to_add), len(to_delete)) - def add_country_names(self, country_code, names): + def add_country_names(self, country_code: str, names: Mapping[str, str]) -> None: """ Add names for the given country to the search index. """ + assert self.conn is not None + with self.conn.cursor() as cur: cur.execute( """INSERT INTO word (word_id, word_token, country_code) @@ -415,12 +458,14 @@ class LegacyNameAnalyzer(AbstractAnalyzer): """, (country_code, list(names.values()), country_code)) - def process_place(self, place): + def process_place(self, place: PlaceInfo) -> Mapping[str, Any]: """ Determine tokenizer information about the given place. Returns a JSON-serialisable structure that will be handed into the database via the token_info field. """ + assert self.conn is not None + token_info = _TokenInfo(self._cache) names = place.name @@ -429,6 +474,7 @@ class LegacyNameAnalyzer(AbstractAnalyzer): token_info.add_names(self.conn, names) if place.is_country(): + assert place.country_code is not None self.add_country_names(place.country_code, names) address = place.address @@ -438,7 +484,8 @@ class LegacyNameAnalyzer(AbstractAnalyzer): return token_info.data - def _process_place_address(self, token_info, address): + def _process_place_address(self, token_info: '_TokenInfo', address: Mapping[str, str]) -> None: + assert self.conn is not None hnrs = [] addr_terms = [] @@ -446,15 +493,17 @@ class LegacyNameAnalyzer(AbstractAnalyzer): if key == 'postcode': # Make sure the normalized postcode is present in the word table. if re.search(r'[:,;]', value) is None: - self._cache.add_postcode(self.conn, - self.normalize_postcode(value)) + norm_pc = self.normalize_postcode(value) + token_info.set_postcode(norm_pc) + self._cache.add_postcode(self.conn, norm_pc) elif key in ('housenumber', 'streetnumber', 'conscriptionnumber'): hnrs.append(value) elif key == 'street': token_info.add_street(self.conn, value) elif key == 'place': token_info.add_place(self.conn, value) - elif not key.startswith('_') and key not in ('country', 'full'): + elif not key.startswith('_') \ + and key not in ('country', 'full', 'inclusion'): addr_terms.append((key, value)) if hnrs: @@ -468,12 +517,12 @@ class LegacyNameAnalyzer(AbstractAnalyzer): class _TokenInfo: """ Collect token information to be sent back to the database. """ - def __init__(self, cache): + def __init__(self, cache: '_TokenCache') -> None: self.cache = cache - self.data = {} + self.data: Dict[str, Any] = {} - def add_names(self, conn, names): + def add_names(self, conn: Connection, names: Mapping[str, str]) -> None: """ Add token information for the names of the place. """ with conn.cursor() as cur: @@ -482,7 +531,7 @@ class _TokenInfo: (names, )) - def add_housenumbers(self, conn, hnrs): + def add_housenumbers(self, conn: Connection, hnrs: Sequence[str]) -> None: """ Extract housenumber information from the address. """ if len(hnrs) == 1: @@ -493,7 +542,7 @@ class _TokenInfo: return # split numbers if necessary - simple_list = [] + simple_list: List[str] = [] for hnr in hnrs: simple_list.extend((x.strip() for x in re.split(r'[;,]', hnr))) @@ -501,49 +550,62 @@ class _TokenInfo: simple_list = list(set(simple_list)) with conn.cursor() as cur: - cur.execute("SELECT (create_housenumbers(%s)).* ", (simple_list, )) - self.data['hnr_tokens'], self.data['hnr'] = cur.fetchone() + cur.execute("SELECT * FROM create_housenumbers(%s)", (simple_list, )) + result = cur.fetchone() + assert result is not None + self.data['hnr_tokens'], self.data['hnr'] = result + + def set_postcode(self, postcode: str) -> None: + """ Set or replace the postcode token with the given value. + """ + self.data['postcode'] = postcode - def add_street(self, conn, street): + def add_street(self, conn: Connection, street: str) -> None: """ Add addr:street match terms. """ - def _get_street(name): + def _get_street(name: str) -> List[int]: with conn.cursor() as cur: - return cur.scalar("SELECT word_ids_from_name(%s)::text", (name, )) + return cast(List[int], + cur.scalar("SELECT word_ids_from_name(%s)::text", (name, ))) - self.data['street'] = self.cache.streets.get(street, _get_street) + tokens = self.cache.streets.get(street, _get_street) + if tokens: + self.data['street'] = tokens - def add_place(self, conn, place): + def add_place(self, conn: Connection, place: str) -> None: """ Add addr:place search and match terms. """ - def _get_place(name): + def _get_place(name: str) -> Tuple[List[int], List[int]]: with conn.cursor() as cur: cur.execute("""SELECT make_keywords(hstore('name' , %s))::text, word_ids_from_name(%s)::text""", (name, name)) - return cur.fetchone() + return cast(Tuple[List[int], List[int]], cur.fetchone()) self.data['place_search'], self.data['place_match'] = \ self.cache.places.get(place, _get_place) - def add_address_terms(self, conn, terms): + def add_address_terms(self, conn: Connection, terms: Sequence[Tuple[str, str]]) -> None: """ Add additional address terms. """ - def _get_address_term(name): + def _get_address_term(name: str) -> Tuple[List[int], List[int]]: with conn.cursor() as cur: cur.execute("""SELECT addr_ids_from_name(%s)::text, word_ids_from_name(%s)::text""", (name, name)) - return cur.fetchone() + return cast(Tuple[List[int], List[int]], cur.fetchone()) tokens = {} for key, value in terms: - tokens[key] = self.cache.address_terms.get(value, _get_address_term) + items = self.cache.address_terms.get(value, _get_address_term) + if items[0] or items[1]: + tokens[key] = items - self.data['addr'] = tokens + if tokens: + self.data['addr'] = tokens class _LRU: @@ -551,13 +613,12 @@ class _LRU: produce the item when there is a cache miss. """ - def __init__(self, maxsize=128, init_data=None): - self.data = init_data or OrderedDict() + def __init__(self, maxsize: int = 128): + self.data: 'OrderedDict[str, Any]' = OrderedDict() self.maxsize = maxsize - if init_data is not None and len(init_data) > maxsize: - self.maxsize = len(init_data) - def get(self, key, generator): + + def get(self, key: str, generator: Callable[[str], Any]) -> Any: """ Get the item with the given key from the cache. If nothing is found in the cache, generate the value through the generator function and store it in the cache. @@ -580,7 +641,7 @@ class _TokenCache: This cache is not thread-safe and needs to be instantiated per analyzer. """ - def __init__(self, conn): + def __init__(self, conn: Connection): # various LRU caches self.streets = _LRU(maxsize=256) self.places = _LRU(maxsize=128) @@ -590,18 +651,18 @@ class _TokenCache: with conn.cursor() as cur: cur.execute("""SELECT i, ARRAY[getorcreate_housenumber_id(i::text)]::text FROM generate_series(1, 100) as i""") - self._cached_housenumbers = {str(r[0]): r[1] for r in cur} + self._cached_housenumbers: Dict[str, str] = {str(r[0]): r[1] for r in cur} # For postcodes remember the ones that have already been added - self.postcodes = set() + self.postcodes: Set[str] = set() - def get_housenumber(self, number): + def get_housenumber(self, number: str) -> Optional[str]: """ Get a housenumber token from the cache. """ return self._cached_housenumbers.get(number) - def add_postcode(self, conn, postcode): + def add_postcode(self, conn: Connection, postcode: str) -> None: """ Make sure the given postcode is in the database. """ if postcode not in self.postcodes: