+# 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 but using
libICU instead of the PostgreSQL module.
"""
-from collections import Counter
import itertools
import json
import logging
from textwrap import dedent
from nominatim.db.connection import connect
-from nominatim.db.properties import set_property, get_property
from nominatim.db.utils import CopyBuffer
from nominatim.db.sql_preprocessor import SQLPreprocessor
from nominatim.indexer.place_info import PlaceInfo
self.dsn = dsn
self.data_dir = data_dir
self.loader = None
- self.term_normalization = None
def init_new_db(self, config, init_db=True):
"""
self.loader = ICURuleLoader(config)
- self.term_normalization = config.TERM_NORMALIZATION
-
self._install_php(config.lib_dir.php)
self._save_config()
with connect(self.dsn) as conn:
self.loader.load_config_from_db(conn)
- self.term_normalization = get_property(conn, DBCFG_TERM_NORMALIZATION)
- def finalize_import(self, _):
+ def finalize_import(self, config):
""" Do any required postprocessing to make the tokenizer data ready
for use.
"""
+ with connect(self.dsn) as conn:
+ sqlp = SQLPreprocessor(conn, config)
+ sqlp.run_sql_file(conn, 'tokenizer/legacy_tokenizer_indices.sql')
def update_sql_functions(self, config):
def check_database(self, config):
""" Check that the tokenizer is set up correctly.
"""
+ # Will throw an error if there is an issue.
self.init_from_project(config)
- if self.term_normalization is None:
- return "Configuration for tokenizer 'icu' are missing."
- return None
+ def update_statistics(self):
+ """ Recompute frequencies for all name words.
+ """
+ with connect(self.dsn) as conn:
+ 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 info = info || jsonb_build_object('count', count)
+ FROM word_frequencies WHERE word_id = id""")
+ cur.drop_table("word_frequencies")
+ conn.commit()
+
+
+ def _cleanup_housenumbers(self):
+ """ Remove unused house numbers.
+ """
+ with connect(self.dsn) as conn:
+ if not conn.table_exists('search_name'):
+ return
+ with conn.cursor(name="hnr_counter") as cur:
+ cur.execute("""SELECT word_id, word_token FROM word
+ WHERE type = 'H'
+ AND NOT EXISTS(SELECT * FROM search_name
+ WHERE ARRAY[word.word_id] && name_vector)
+ AND (char_length(word_token) > 6
+ OR word_token not similar to '\\d+')
+ """)
+ candidates = {token: wid for wid, token in cur}
+ with conn.cursor(name="hnr_counter") as cur:
+ cur.execute("""SELECT housenumber FROM placex
+ WHERE housenumber is not null
+ AND (char_length(housenumber) > 6
+ OR housenumber not similar to '\\d+')
+ """)
+ for row in cur:
+ for hnr in row[0].split(';'):
+ candidates.pop(hnr, None)
+ LOG.info("There are %s outdated housenumbers.", len(candidates))
+ if candidates:
+ with conn.cursor() as cur:
+ cur.execute("""DELETE FROM word WHERE word_id = any(%s)""",
+ (list(candidates.values()), ))
+ conn.commit()
+
+
+
+ def update_word_tokens(self):
+ """ Remove unused tokens.
+ """
+ LOG.warning("Cleaning up housenumber tokens.")
+ self._cleanup_housenumbers()
+ LOG.warning("Tokenizer house-keeping done.")
def name_analyzer(self):
php_file.write_text(dedent(f"""\
<?php
@define('CONST_Max_Word_Frequency', 10000000);
- @define('CONST_Term_Normalization_Rules', "{self.term_normalization}");
+ @define('CONST_Term_Normalization_Rules', "{self.loader.normalization_rules}");
@define('CONST_Transliteration', "{self.loader.get_search_rules()}");
require_once('{phpdir}/tokenizer/icu_tokenizer.php');"""))
"""
with connect(self.dsn) as conn:
self.loader.save_config_to_db(conn)
- set_property(conn, DBCFG_TERM_NORMALIZATION, self.term_normalization)
def _init_db_tables(self, config):
sqlp.run_sql_file(conn, 'tokenizer/icu_tokenizer_tables.sql')
conn.commit()
- LOG.warning("Precomputing word tokens")
-
- # get partial words and their frequencies
- words = self._count_partial_terms(conn)
-
- # copy them back into the word table
- with CopyBuffer() as copystr:
- for term, cnt in words.items():
- copystr.add('w', term, json.dumps({'count': cnt}))
-
- with conn.cursor() as cur:
- copystr.copy_out(cur, 'word',
- columns=['type', 'word_token', 'info'])
- cur.execute("""UPDATE word SET word_id = nextval('seq_word')
- WHERE word_id is null and type = 'w'""")
-
- conn.commit()
-
- def _count_partial_terms(self, conn):
- """ Count the partial terms from the names in the place table.
- """
- words = Counter()
- name_proc = self.loader.make_token_analysis()
-
- with conn.cursor(name="words") as cur:
- cur.execute(""" SELECT v, count(*) FROM
- (SELECT svals(name) as v FROM place)x
- WHERE length(v) < 75 GROUP BY v""")
-
- for name, cnt in cur:
- terms = set()
- for word in name_proc.get_variants_ascii(name_proc.get_normalized(name)):
- if ' ' in word:
- terms.update(word.split())
- for term in terms:
- words[term] += cnt
-
- return words
-
class LegacyICUNameAnalyzer(AbstractAnalyzer):
""" The legacy analyzer uses the ICU library for splitting names.
def _search_normalized(self, name):
""" Return the search token transliteration of the given name.
"""
- return self.token_analysis.get_search_normalized(name)
+ return self.token_analysis.search.transliterate(name).strip()
def _normalized(self, name):
""" Return the normalized version of the given name with all
non-relevant information removed.
"""
- return self.token_analysis.get_normalized(name)
+ return self.token_analysis.normalizer.transliterate(name).strip()
def get_word_token_info(self, words):
def _process_place_address(self, token_info, address):
- hnrs = []
+ hnrs = set()
addr_terms = []
+ streets = []
for item in address:
if item.kind == 'postcode':
self._add_postcode(item.name)
- elif item.kind in ('housenumber', 'streetnumber', 'conscriptionnumber'):
- hnrs.append(item.name)
+ elif item.kind == 'housenumber':
+ norm_name = self._make_standard_hnr(item.name)
+ if norm_name:
+ hnrs.add(norm_name)
elif item.kind == 'street':
- token_info.add_street(self._compute_partial_tokens(item.name))
+ streets.extend(self._retrieve_full_tokens(item.name))
elif item.kind == 'place':
- token_info.add_place(self._compute_partial_tokens(item.name))
- elif not item.kind.startswith('_') and \
+ if not item.suffix:
+ token_info.add_place(self._compute_partial_tokens(item.name))
+ elif not item.kind.startswith('_') and not item.suffix and \
item.kind not in ('country', 'full'):
addr_terms.append((item.kind, self._compute_partial_tokens(item.name)))
if hnrs:
- hnrs = self._split_housenumbers(hnrs)
- token_info.add_housenumbers(self.conn, [self._make_standard_hnr(n) for n in hnrs])
+ token_info.add_housenumbers(self.conn, hnrs)
if addr_terms:
token_info.add_address_terms(addr_terms)
+ if streets:
+ token_info.add_street(streets)
+
+
def _compute_partial_tokens(self, name):
""" Normalize the given term, split it into partial words and return
then token list for them.
return tokens
+ def _retrieve_full_tokens(self, name):
+ """ Get the full name token for the given name, if it exists.
+ The name is only retrived for the standard analyser.
+ """
+ norm_name = self._search_normalized(name)
+
+ # return cached if possible
+ if norm_name in self._cache.fulls:
+ return self._cache.fulls[norm_name]
+
+ with self.conn.cursor() as cur:
+ cur.execute("SELECT word_id FROM word WHERE word_token = %s and type = 'W'",
+ (norm_name, ))
+ full = [row[0] for row in cur]
+
+ self._cache.fulls[norm_name] = full
+
+ return full
+
+
def _compute_name_tokens(self, names):
""" Computes the full name and partial name tokens for the given
dictionary of names.
partial_tokens = set()
for name in names:
+ analyzer_id = name.get_attr('analyzer')
norm_name = self._normalized(name.name)
- full, part = self._cache.names.get(norm_name, (None, None))
+ if analyzer_id is None:
+ token_id = norm_name
+ else:
+ token_id = f'{norm_name}@{analyzer_id}'
+
+ full, part = self._cache.names.get(token_id, (None, None))
if full is None:
- variants = self.token_analysis.get_variants_ascii(norm_name)
+ variants = self.token_analysis.analysis[analyzer_id].get_variants_ascii(norm_name)
if not variants:
continue
with self.conn.cursor() as cur:
cur.execute("SELECT (getorcreate_full_word(%s, %s)).*",
- (norm_name, variants))
+ (token_id, variants))
full, part = cur.fetchone()
- self._cache.names[norm_name] = (full, part)
+ self._cache.names[token_id] = (full, part)
full_tokens.add(full)
partial_tokens.update(part)
self._cache.postcodes.add(postcode)
- @staticmethod
- def _split_housenumbers(hnrs):
- if len(hnrs) > 1 or ',' in hnrs[0] or ';' in hnrs[0]:
- # split numbers if necessary
- simple_list = []
- for hnr in hnrs:
- simple_list.extend((x.strip() for x in re.split(r'[;,]', hnr)))
-
- if len(simple_list) > 1:
- hnrs = list(set(simple_list))
- else:
- hnrs = simple_list
-
- return hnrs
-
-
-
-
class _TokenInfo:
""" Collect token information to be sent back to the database.
"""
def add_street(self, tokens):
""" Add addr:street match terms.
"""
- if tokens:
- self.data['street'] = self._mk_array(tokens)
+ self.data['street'] = self._mk_array(tokens)
def add_place(self, tokens):
def __init__(self):
self.names = {}
self.partials = {}
+ self.fulls = {}
self.postcodes = set()
self.housenumbers = {}
projects / nominatim.git / blobdiff