X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/2f6e4edcdb0f919f6a0481dd02664e4b2e507aaa..f8b5a63de39691ea6e7fcb9180a856f0d2650999:/nominatim/tokenizer/legacy_icu_tokenizer.py diff --git a/nominatim/tokenizer/legacy_icu_tokenizer.py b/nominatim/tokenizer/legacy_icu_tokenizer.py index f3eb7b4e..12ee0404 100644 --- a/nominatim/tokenizer/legacy_icu_tokenizer.py +++ b/nominatim/tokenizer/legacy_icu_tokenizer.py @@ -3,7 +3,6 @@ Tokenizer implementing normalisation as used before Nominatim 4 but using libICU instead of the PostgreSQL module. """ from collections import Counter -import io import itertools import logging import re @@ -14,6 +13,7 @@ import psycopg2.extras 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.tokenizer.icu_rule_loader import ICURuleLoader from nominatim.tokenizer.icu_name_processor import ICUNameProcessor, ICUNameProcessorRules @@ -123,7 +123,7 @@ class LegacyICUTokenizer: """ return LegacyICUNameAnalyzer(self.dsn, ICUNameProcessor(self.naming_rules)) - + # pylint: disable=missing-format-attribute def _install_php(self, phpdir): """ Install the php script for the tokenizer. """ @@ -163,22 +163,28 @@ class LegacyICUTokenizer: words = Counter() name_proc = ICUNameProcessor(self.naming_rules) with conn.cursor(name="words") as cur: - cur.execute("SELECT svals(name) as v, count(*) FROM place GROUP BY v") + 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)): - for term in word.split(): - words[term] += cnt + if ' ' in word: + terms.update(word.split()) + for term in terms: + words[term] += cnt # copy them back into the word table - copystr = io.StringIO(''.join(('{}\t{}\n'.format(*args) for args in words.items()))) - + with CopyBuffer() as copystr: + for args in words.items(): + copystr.add(*args) - with conn.cursor() as cur: - copystr.seek(0) - cur.copy_from(copystr, 'word', columns=['word_token', 'search_name_count']) - cur.execute("""UPDATE word SET word_id = nextval('seq_word') - WHERE word_id is null""") + with conn.cursor() as cur: + copystr.copy_out(cur, 'word', + columns=['word_token', 'search_name_count']) + cur.execute("""UPDATE word SET word_id = nextval('seq_word') + WHERE word_id is null""") conn.commit() @@ -265,7 +271,6 @@ class LegacyICUNameAnalyzer: table. """ to_delete = [] - copystr = io.StringIO() with self.conn.cursor() as cur: # This finds us the rows in location_postcode and word that are # missing in the other table. @@ -278,26 +283,25 @@ class LegacyICUNameAnalyzer: ON pc = word) x WHERE pc is null or word is null""") - for postcode, word in cur: - if postcode is None: - to_delete.append(word) - else: - copystr.write(postcode) - copystr.write('\t ') - copystr.write(self.name_processor.get_search_normalized(postcode)) - copystr.write('\tplace\tpostcode\t0\n') + with CopyBuffer() as copystr: + for postcode, word in cur: + if postcode is None: + to_delete.append(word) + else: + copystr.add( + postcode, + ' ' + self.name_processor.get_search_normalized(postcode), + 'place', 'postcode', 0) - if to_delete: - cur.execute("""DELETE FROM WORD - WHERE class ='place' and type = 'postcode' - and word = any(%s) - """, (to_delete, )) + if to_delete: + cur.execute("""DELETE FROM WORD + WHERE class ='place' and type = 'postcode' + and word = any(%s) + """, (to_delete, )) - if copystr.getvalue(): - copystr.seek(0) - cur.copy_from(copystr, 'word', - columns=['word', 'word_token', 'class', 'type', - 'search_name_count']) + copystr.copy_out(cur, 'word', + columns=['word', 'word_token', 'class', 'type', + 'search_name_count']) def update_special_phrases(self, phrases, should_replace): @@ -331,34 +335,24 @@ class LegacyICUNameAnalyzer: """ to_add = new_phrases - existing_phrases - copystr = io.StringIO() added = 0 - for word, cls, typ, oper in to_add: - term = self.name_processor.get_search_normalized(word) - if term: - copystr.write(word) - copystr.write('\t ') - copystr.write(term) - copystr.write('\t') - copystr.write(cls) - copystr.write('\t') - copystr.write(typ) - copystr.write('\t') - copystr.write(oper if oper in ('in', 'near') else '\\N') - copystr.write('\t0\n') - added += 1 - - - if copystr.tell() > 0: - copystr.seek(0) - cursor.copy_from(copystr, 'word', + with CopyBuffer() as copystr: + for word, cls, typ, oper in to_add: + term = self.name_processor.get_search_normalized(word) + if term: + copystr.add(word, ' ' + term, cls, typ, + oper if oper in ('in', 'near') else None, 0) + added += 1 + + copystr.copy_out(cursor, 'word', columns=['word', 'word_token', 'class', 'type', 'operator', 'search_name_count']) return added - def _remove_special_phrases(self, cursor, new_phrases, existing_phrases): + @staticmethod + def _remove_special_phrases(cursor, new_phrases, existing_phrases): """ Remove all phrases from the databse that are no longer in the new phrase list. """ @@ -417,33 +411,36 @@ class LegacyICUNameAnalyzer: self.add_country_names(country_feature.lower(), names) address = place.get('address') - if address: - hnrs = [] - addr_terms = [] - for key, value in address.items(): - if key == 'postcode': - self._add_postcode(value) - elif key in ('housenumber', 'streetnumber', 'conscriptionnumber'): - hnrs.append(value) - elif key == 'street': - token_info.add_street(*self._compute_name_tokens({'name': value})) - elif key == 'place': - token_info.add_place(*self._compute_name_tokens({'name': value})) - elif not key.startswith('_') and \ - key not in ('country', 'full'): - addr_terms.append((key, *self._compute_name_tokens({'name': value}))) - - if hnrs: - hnrs = self._split_housenumbers(hnrs) - token_info.add_housenumbers(self.conn, [self._make_standard_hnr(n) for n in hnrs]) - - if addr_terms: - token_info.add_address_terms(addr_terms) + self._process_place_address(token_info, address) return token_info.data + def _process_place_address(self, token_info, address): + hnrs = [] + addr_terms = [] + for key, value in address.items(): + if key == 'postcode': + self._add_postcode(value) + elif key in ('housenumber', 'streetnumber', 'conscriptionnumber'): + hnrs.append(value) + elif key == 'street': + token_info.add_street(*self._compute_name_tokens({'name': value})) + elif key == 'place': + token_info.add_place(*self._compute_name_tokens({'name': value})) + elif not key.startswith('_') and \ + key not in ('country', 'full'): + addr_terms.append((key, *self._compute_name_tokens({'name': value}))) + + if hnrs: + hnrs = self._split_housenumbers(hnrs) + token_info.add_housenumbers(self.conn, [self._make_standard_hnr(n) for n in hnrs]) + + if addr_terms: + token_info.add_address_terms(addr_terms) + + def _compute_name_tokens(self, names): """ Computes the full name and partial name tokens for the given dictionary of names. @@ -457,6 +454,9 @@ class LegacyICUNameAnalyzer: full, part = self._cache.names.get(norm_name, (None, None)) if full is None: variants = self.name_processor.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)) @@ -476,12 +476,13 @@ class LegacyICUNameAnalyzer: given dictionary of names. """ full_names = set() - for name in (n for ns in names.values() for n in re.split('[;,]', ns)): - full_names.add(name.strip()) + for name in (n.strip() for ns in names.values() for n in re.split('[;,]', ns)): + if name: + full_names.add(name) - brace_idx = name.find('(') - if brace_idx >= 0: - full_names.add(name[:brace_idx].strip()) + brace_idx = name.find('(') + if brace_idx >= 0: + full_names.add(name[:brace_idx].strip()) return full_names