1 # SPDX-License-Identifier: GPL-2.0-only
3 # This file is part of Nominatim. (https://nominatim.org)
5 # Copyright (C) 2022 by the Nominatim developer community.
6 # For a full list of authors see the git log.
8 Tests for ICU tokenizer.
15 from nominatim.tokenizer import icu_tokenizer
16 import nominatim.tokenizer.icu_rule_loader
17 from nominatim.db import properties
18 from nominatim.db.sql_preprocessor import SQLPreprocessor
19 from nominatim.data.place_info import PlaceInfo
21 from mock_icu_word_table import MockIcuWordTable
24 def word_table(temp_db_conn):
25 return MockIcuWordTable(temp_db_conn)
29 def test_config(project_env, tmp_path):
30 sqldir = tmp_path / 'sql'
32 (sqldir / 'tokenizer').mkdir()
33 (sqldir / 'tokenizer' / 'icu_tokenizer.sql').write_text("SELECT 'a'")
35 project_env.lib_dir.sql = sqldir
41 def tokenizer_factory(dsn, tmp_path, property_table,
42 sql_preprocessor, place_table, word_table):
43 (tmp_path / 'tokenizer').mkdir()
46 return icu_tokenizer.create(dsn, tmp_path / 'tokenizer')
52 def db_prop(temp_db_conn):
53 def _get_db_property(name):
54 return properties.get_property(temp_db_conn, name)
56 return _get_db_property
60 def analyzer(tokenizer_factory, test_config, monkeypatch,
61 temp_db_with_extensions, tmp_path):
62 sql = tmp_path / 'sql' / 'tokenizer' / 'icu_tokenizer.sql'
63 sql.write_text("SELECT 'a';")
65 monkeypatch.setenv('NOMINATIM_TERM_NORMALIZATION', ':: lower();')
66 tok = tokenizer_factory()
67 tok.init_new_db(test_config)
70 def _mk_analyser(norm=("[[:Punctuation:][:Space:]]+ > ' '",), trans=(':: upper()',),
71 variants=('~gasse -> gasse', 'street => st', ),
72 sanitizers=[], with_housenumber=False,
74 cfgstr = {'normalization': list(norm),
75 'sanitizers': sanitizers,
76 'transliteration': list(trans),
77 'token-analysis': [{'analyzer': 'generic',
78 'variants': [{'words': list(variants)}]}]}
80 cfgstr['token-analysis'].append({'id': '@housenumber',
81 'analyzer': 'housenumbers'})
83 cfgstr['token-analysis'].append({'id': '@postcode',
84 'analyzer': 'postcodes'})
85 (test_config.project_dir / 'icu_tokenizer.yaml').write_text(yaml.dump(cfgstr))
86 tok.loader = nominatim.tokenizer.icu_rule_loader.ICURuleLoader(test_config)
88 return tok.name_analyzer()
93 def sql_functions(temp_db_conn, def_config, src_dir):
94 orig_sql = def_config.lib_dir.sql
95 def_config.lib_dir.sql = src_dir / 'lib-sql'
96 sqlproc = SQLPreprocessor(temp_db_conn, def_config)
97 sqlproc.run_sql_file(temp_db_conn, 'functions/utils.sql')
98 sqlproc.run_sql_file(temp_db_conn, 'tokenizer/icu_tokenizer.sql')
99 def_config.lib_dir.sql = orig_sql
103 def getorcreate_full_word(temp_db_cursor):
104 temp_db_cursor.execute("""CREATE OR REPLACE FUNCTION getorcreate_full_word(
105 norm_term TEXT, lookup_terms TEXT[],
107 OUT partial_tokens INT[])
110 partial_terms TEXT[] = '{}'::TEXT[];
115 SELECT min(word_id) INTO full_token
116 FROM word WHERE info->>'word' = norm_term and type = 'W';
118 IF full_token IS NULL THEN
119 full_token := nextval('seq_word');
120 INSERT INTO word (word_id, word_token, type, info)
121 SELECT full_token, lookup_term, 'W',
122 json_build_object('word', norm_term, 'count', 0)
123 FROM unnest(lookup_terms) as lookup_term;
126 FOR term IN SELECT unnest(string_to_array(unnest(lookup_terms), ' ')) LOOP
128 IF NOT (ARRAY[term] <@ partial_terms) THEN
129 partial_terms := partial_terms || term;
133 partial_tokens := '{}'::INT[];
134 FOR term IN SELECT unnest(partial_terms) LOOP
135 SELECT min(word_id), max(info->>'count') INTO term_id, term_count
136 FROM word WHERE word_token = term and type = 'w';
138 IF term_id IS NULL THEN
139 term_id := nextval('seq_word');
141 INSERT INTO word (word_id, word_token, type, info)
142 VALUES (term_id, term, 'w', json_build_object('count', term_count));
145 IF NOT (ARRAY[term_id] <@ partial_tokens) THEN
146 partial_tokens := partial_tokens || term_id;
156 def test_init_new(tokenizer_factory, test_config, db_prop):
157 tok = tokenizer_factory()
158 tok.init_new_db(test_config)
160 assert db_prop(nominatim.tokenizer.icu_rule_loader.DBCFG_IMPORT_NORM_RULES) \
161 .startswith(':: lower ();')
164 def test_init_word_table(tokenizer_factory, test_config, place_row, temp_db_cursor):
165 place_row(names={'name' : 'Test Area', 'ref' : '52'})
166 place_row(names={'name' : 'No Area'})
167 place_row(names={'name' : 'Holzstrasse'})
169 tok = tokenizer_factory()
170 tok.init_new_db(test_config)
172 assert temp_db_cursor.table_exists('word')
175 def test_init_from_project(test_config, tokenizer_factory):
176 tok = tokenizer_factory()
177 tok.init_new_db(test_config)
179 tok = tokenizer_factory()
180 tok.init_from_project(test_config)
182 assert tok.loader is not None
185 def test_update_sql_functions(db_prop, temp_db_cursor,
186 tokenizer_factory, test_config, table_factory,
188 tok = tokenizer_factory()
189 tok.init_new_db(test_config)
191 table_factory('test', 'txt TEXT')
193 func_file = test_config.lib_dir.sql / 'tokenizer' / 'icu_tokenizer.sql'
194 func_file.write_text("""INSERT INTO test VALUES (1133)""")
196 tok.update_sql_functions(test_config)
198 test_content = temp_db_cursor.row_set('SELECT * FROM test')
199 assert test_content == set((('1133', ), ))
202 def test_finalize_import(tokenizer_factory, temp_db_conn,
203 temp_db_cursor, test_config, sql_preprocessor_cfg):
204 tok = tokenizer_factory()
205 tok.init_new_db(test_config)
207 assert not temp_db_conn.index_exists('idx_word_word_id')
209 tok.finalize_import(test_config)
211 assert temp_db_conn.index_exists('idx_word_word_id')
214 def test_check_database(test_config, tokenizer_factory,
215 temp_db_cursor, sql_preprocessor_cfg):
216 tok = tokenizer_factory()
217 tok.init_new_db(test_config)
219 assert tok.check_database(test_config) is None
222 def test_update_statistics_reverse_only(word_table, tokenizer_factory, test_config):
223 tok = tokenizer_factory()
224 tok.update_statistics(test_config)
227 def test_update_statistics(word_table, table_factory, temp_db_cursor,
228 tokenizer_factory, test_config):
229 word_table.add_full_word(1000, 'hello')
230 word_table.add_full_word(1001, 'bye')
231 table_factory('search_name',
232 'place_id BIGINT, name_vector INT[], nameaddress_vector INT[]',
233 [(12, [1000], [1001])])
234 tok = tokenizer_factory()
236 tok.update_statistics(test_config)
238 assert temp_db_cursor.scalar("""SELECT count(*) FROM word
239 WHERE type = 'W' and word_id = 1000 and
240 (info->>'count')::int > 0""") == 1
241 assert temp_db_cursor.scalar("""SELECT count(*) FROM word
242 WHERE type = 'W' and word_id = 1001 and
243 (info->>'addr_count')::int > 0""") == 1
246 def test_normalize_postcode(analyzer):
247 with analyzer() as anl:
248 anl.normalize_postcode('123') == '123'
249 anl.normalize_postcode('ab-34 ') == 'AB-34'
250 anl.normalize_postcode('38 Б') == '38 Б'
255 @pytest.fixture(autouse=True)
256 def setup(self, analyzer, sql_functions):
257 sanitizers = [{'step': 'clean-postcodes'}]
258 with analyzer(sanitizers=sanitizers, with_postcode=True) as anl:
263 def process_postcode(self, cc, postcode):
264 return self.analyzer.process_place(PlaceInfo({'country_code': cc,
265 'address': {'postcode': postcode}}))
268 def test_update_postcodes_from_db_empty(self, table_factory, word_table):
269 table_factory('location_postcode', 'country_code TEXT, postcode TEXT',
270 content=(('de', '12345'), ('se', '132 34'),
271 ('bm', 'AB23'), ('fr', '12345')))
273 self.analyzer.update_postcodes_from_db()
275 assert word_table.count() == 5
276 assert word_table.get_postcodes() == {'12345', '132 34@132 34', 'AB 23@AB 23'}
279 def test_update_postcodes_from_db_ambigious(self, table_factory, word_table):
280 table_factory('location_postcode', 'country_code TEXT, postcode TEXT',
281 content=(('in', '123456'), ('sg', '123456')))
283 self.analyzer.update_postcodes_from_db()
285 assert word_table.count() == 3
286 assert word_table.get_postcodes() == {'123456', '123456@123 456'}
289 def test_update_postcodes_from_db_add_and_remove(self, table_factory, word_table):
290 table_factory('location_postcode', 'country_code TEXT, postcode TEXT',
291 content=(('ch', '1234'), ('bm', 'BC 45'), ('bm', 'XX45')))
292 word_table.add_postcode(' 1234', '1234')
293 word_table.add_postcode(' 5678', '5678')
295 self.analyzer.update_postcodes_from_db()
297 assert word_table.count() == 5
298 assert word_table.get_postcodes() == {'1234', 'BC 45@BC 45', 'XX 45@XX 45'}
301 def test_process_place_postcode_simple(self, word_table):
302 info = self.process_postcode('de', '12345')
304 assert info['postcode'] == '12345'
306 assert word_table.get_postcodes() == {'12345', }
309 def test_process_place_postcode_with_space(self, word_table):
310 info = self.process_postcode('in', '123 567')
312 assert info['postcode'] == '123567'
314 assert word_table.get_postcodes() == {'123567@123 567', }
318 def test_update_special_phrase_empty_table(analyzer, word_table):
319 with analyzer() as anl:
320 anl.update_special_phrases([
321 ("König bei", "amenity", "royal", "near"),
322 ("Könige ", "amenity", "royal", "-"),
323 ("street", "highway", "primary", "in")
326 assert word_table.get_special() \
327 == {('KÖNIG BEI', 'König bei', 'amenity', 'royal', 'near'),
328 ('KÖNIGE', 'Könige', 'amenity', 'royal', None),
329 ('STREET', 'street', 'highway', 'primary', 'in')}
332 def test_update_special_phrase_delete_all(analyzer, word_table):
333 word_table.add_special('FOO', 'foo', 'amenity', 'prison', 'in')
334 word_table.add_special('BAR', 'bar', 'highway', 'road', None)
336 assert word_table.count_special() == 2
338 with analyzer() as anl:
339 anl.update_special_phrases([], True)
341 assert word_table.count_special() == 0
344 def test_update_special_phrases_no_replace(analyzer, word_table):
345 word_table.add_special('FOO', 'foo', 'amenity', 'prison', 'in')
346 word_table.add_special('BAR', 'bar', 'highway', 'road', None)
348 assert word_table.count_special() == 2
350 with analyzer() as anl:
351 anl.update_special_phrases([], False)
353 assert word_table.count_special() == 2
356 def test_update_special_phrase_modify(analyzer, word_table):
357 word_table.add_special('FOO', 'foo', 'amenity', 'prison', 'in')
358 word_table.add_special('BAR', 'bar', 'highway', 'road', None)
360 assert word_table.count_special() == 2
362 with analyzer() as anl:
363 anl.update_special_phrases([
364 ('prison', 'amenity', 'prison', 'in'),
365 ('bar', 'highway', 'road', '-'),
366 ('garden', 'leisure', 'garden', 'near')
369 assert word_table.get_special() \
370 == {('PRISON', 'prison', 'amenity', 'prison', 'in'),
371 ('BAR', 'bar', 'highway', 'road', None),
372 ('GARDEN', 'garden', 'leisure', 'garden', 'near')}
375 def test_add_country_names_new(analyzer, word_table):
376 with analyzer() as anl:
377 anl.add_country_names('es', {'name': 'Espagña', 'name:en': 'Spain'})
379 assert word_table.get_country() == {('es', 'ESPAGÑA'), ('es', 'SPAIN')}
382 def test_add_country_names_extend(analyzer, word_table):
383 word_table.add_country('ch', 'SCHWEIZ')
385 with analyzer() as anl:
386 anl.add_country_names('ch', {'name': 'Schweiz', 'name:fr': 'Suisse'})
388 assert word_table.get_country() == {('ch', 'SCHWEIZ'), ('ch', 'SUISSE')}
391 class TestPlaceNames:
393 @pytest.fixture(autouse=True)
394 def setup(self, analyzer, sql_functions):
395 sanitizers = [{'step': 'split-name-list'},
396 {'step': 'strip-brace-terms'}]
397 with analyzer(sanitizers=sanitizers) as anl:
402 def expect_name_terms(self, info, *expected_terms):
403 tokens = self.analyzer.get_word_token_info(expected_terms)
405 assert token[2] is not None, "No token for {0}".format(token)
407 assert eval(info['names']) == set((t[2] for t in tokens))
410 def process_named_place(self, names):
411 return self.analyzer.process_place(PlaceInfo({'name': names}))
414 def test_simple_names(self):
415 info = self.process_named_place({'name': 'Soft bAr', 'ref': '34'})
417 self.expect_name_terms(info, '#Soft bAr', '#34', 'Soft', 'bAr', '34')
420 @pytest.mark.parametrize('sep', [',' , ';'])
421 def test_names_with_separator(self, sep):
422 info = self.process_named_place({'name': sep.join(('New York', 'Big Apple'))})
424 self.expect_name_terms(info, '#New York', '#Big Apple',
425 'new', 'york', 'big', 'apple')
428 def test_full_names_with_bracket(self):
429 info = self.process_named_place({'name': 'Houseboat (left)'})
431 self.expect_name_terms(info, '#Houseboat (left)', '#Houseboat',
435 def test_country_name(self, word_table):
436 place = PlaceInfo({'name' : {'name': 'Norge'},
437 'country_code': 'no',
440 'type': 'administrative'})
442 info = self.analyzer.process_place(place)
444 self.expect_name_terms(info, '#norge', 'norge')
445 assert word_table.get_country() == {('no', 'NORGE')}
448 class TestPlaceAddress:
450 @pytest.fixture(autouse=True)
451 def setup(self, analyzer, sql_functions):
452 hnr = {'step': 'clean-housenumbers',
453 'filter-kind': ['housenumber', 'conscriptionnumber', 'streetnumber']}
454 with analyzer(trans=(":: upper()", "'🜵' > ' '"), sanitizers=[hnr]) as anl:
460 def getorcreate_hnr_id(self, temp_db_cursor):
461 temp_db_cursor.execute("""CREATE OR REPLACE FUNCTION getorcreate_hnr_id(lookup_term TEXT)
462 RETURNS INTEGER AS $$
463 SELECT -nextval('seq_word')::INTEGER; $$ LANGUAGE SQL""")
466 def process_address(self, **kwargs):
467 return self.analyzer.process_place(PlaceInfo({'address': kwargs}))
470 def name_token_set(self, *expected_terms):
471 tokens = self.analyzer.get_word_token_info(expected_terms)
473 assert token[2] is not None, "No token for {0}".format(token)
475 return set((t[2] for t in tokens))
478 @pytest.mark.parametrize('pcode', ['12345', 'AB 123', '34-345'])
479 def test_process_place_postcode(self, word_table, pcode):
480 self.process_address(postcode=pcode)
482 assert word_table.get_postcodes() == {pcode, }
485 @pytest.mark.parametrize('hnr', ['123a', '1', '101'])
486 def test_process_place_housenumbers_simple(self, hnr, getorcreate_hnr_id):
487 info = self.process_address(housenumber=hnr)
489 assert info['hnr'] == hnr.upper()
490 assert info['hnr_tokens'] == "{-1}"
493 def test_process_place_housenumbers_duplicates(self, getorcreate_hnr_id):
494 info = self.process_address(housenumber='134',
495 conscriptionnumber='134',
498 assert set(info['hnr'].split(';')) == set(('134', '99A'))
499 assert info['hnr_tokens'] == "{-1,-2}"
502 def test_process_place_housenumbers_cached(self, getorcreate_hnr_id):
503 info = self.process_address(housenumber="45")
504 assert info['hnr_tokens'] == "{-1}"
506 info = self.process_address(housenumber="46")
507 assert info['hnr_tokens'] == "{-2}"
509 info = self.process_address(housenumber="41;45")
510 assert eval(info['hnr_tokens']) == {-1, -3}
512 info = self.process_address(housenumber="41")
513 assert eval(info['hnr_tokens']) == {-3}
516 def test_process_place_street(self):
517 self.analyzer.process_place(PlaceInfo({'name': {'name' : 'Grand Road'}}))
518 info = self.process_address(street='Grand Road')
520 assert eval(info['street']) == self.name_token_set('#Grand Road')
523 def test_process_place_nonexisting_street(self):
524 info = self.process_address(street='Grand Road')
526 assert info['street'] == '{}'
529 def test_process_place_multiple_street_tags(self):
530 self.analyzer.process_place(PlaceInfo({'name': {'name' : 'Grand Road',
532 info = self.process_address(**{'street': 'Grand Road',
533 'street:sym_ul': '05989'})
535 assert eval(info['street']) == self.name_token_set('#Grand Road', '#05989')
538 def test_process_place_street_empty(self):
539 info = self.process_address(street='🜵')
541 assert info['street'] == '{}'
544 def test_process_place_street_from_cache(self):
545 self.analyzer.process_place(PlaceInfo({'name': {'name' : 'Grand Road'}}))
546 self.process_address(street='Grand Road')
548 # request address again
549 info = self.process_address(street='Grand Road')
551 assert eval(info['street']) == self.name_token_set('#Grand Road')
554 def test_process_place_place(self):
555 info = self.process_address(place='Honu Lulu')
557 assert eval(info['place']) == self.name_token_set('HONU', 'LULU')
560 def test_process_place_place_extra(self):
561 info = self.process_address(**{'place:en': 'Honu Lulu'})
563 assert 'place' not in info
566 def test_process_place_place_empty(self):
567 info = self.process_address(place='🜵')
569 assert 'place' not in info
572 def test_process_place_address_terms(self):
573 info = self.process_address(country='de', city='Zwickau', state='Sachsen',
574 suburb='Zwickau', street='Hauptstr',
575 full='right behind the church')
577 city = self.name_token_set('ZWICKAU')
578 state = self.name_token_set('SACHSEN')
580 result = {k: eval(v) for k,v in info['addr'].items()}
582 assert result == {'city': city, 'suburb': city, 'state': state}
585 def test_process_place_multiple_address_terms(self):
586 info = self.process_address(**{'city': 'Bruxelles', 'city:de': 'Brüssel'})
588 result = {k: eval(v) for k,v in info['addr'].items()}
590 assert result == {'city': self.name_token_set('Bruxelles')}
593 def test_process_place_address_terms_empty(self):
594 info = self.process_address(country='de', city=' ', street='Hauptstr',
595 full='right behind the church')
597 assert 'addr' not in info
600 class TestPlaceHousenumberWithAnalyser:
602 @pytest.fixture(autouse=True)
603 def setup(self, analyzer, sql_functions):
604 hnr = {'step': 'clean-housenumbers',
605 'filter-kind': ['housenumber', 'conscriptionnumber', 'streetnumber']}
606 with analyzer(trans=(":: upper()", "'🜵' > ' '"), sanitizers=[hnr], with_housenumber=True) as anl:
612 def getorcreate_hnr_id(self, temp_db_cursor):
613 temp_db_cursor.execute("""CREATE OR REPLACE FUNCTION create_analyzed_hnr_id(norm_term TEXT, lookup_terms TEXT[])
614 RETURNS INTEGER AS $$
615 SELECT -nextval('seq_word')::INTEGER; $$ LANGUAGE SQL""")
618 def process_address(self, **kwargs):
619 return self.analyzer.process_place(PlaceInfo({'address': kwargs}))
622 def name_token_set(self, *expected_terms):
623 tokens = self.analyzer.get_word_token_info(expected_terms)
625 assert token[2] is not None, "No token for {0}".format(token)
627 return set((t[2] for t in tokens))
630 @pytest.mark.parametrize('hnr', ['123 a', '1', '101'])
631 def test_process_place_housenumbers_simple(self, hnr, getorcreate_hnr_id):
632 info = self.process_address(housenumber=hnr)
634 assert info['hnr'] == hnr.upper()
635 assert info['hnr_tokens'] == "{-1}"
638 def test_process_place_housenumbers_duplicates(self, getorcreate_hnr_id):
639 info = self.process_address(housenumber='134',
640 conscriptionnumber='134',
643 assert set(info['hnr'].split(';')) == set(('134', '99 A'))
644 assert info['hnr_tokens'] == "{-1,-2}"
647 def test_process_place_housenumbers_cached(self, getorcreate_hnr_id):
648 info = self.process_address(housenumber="45")
649 assert info['hnr_tokens'] == "{-1}"
651 info = self.process_address(housenumber="46")
652 assert info['hnr_tokens'] == "{-2}"
654 info = self.process_address(housenumber="41;45")
655 assert eval(info['hnr_tokens']) == {-1, -3}
657 info = self.process_address(housenumber="41")
658 assert eval(info['hnr_tokens']) == {-3}
661 class TestUpdateWordTokens:
663 @pytest.fixture(autouse=True)
664 def setup(self, tokenizer_factory, table_factory, placex_table, word_table):
665 table_factory('search_name', 'place_id BIGINT, name_vector INT[]')
666 self.tok = tokenizer_factory()
670 def search_entry(self, temp_db_cursor):
671 place_id = itertools.count(1000)
674 temp_db_cursor.execute("INSERT INTO search_name VALUES (%s, %s)",
675 (next(place_id), list(args)))
680 @pytest.fixture(params=['simple', 'analyzed'])
681 def add_housenumber(self, request, word_table):
682 if request.param == 'simple':
684 word_table.add_housenumber(hid, hnr)
685 elif request.param == 'analyzed':
687 word_table.add_housenumber(hid, [hnr])
692 @pytest.mark.parametrize('hnr', ('1a', '1234567', '34 5'))
693 def test_remove_unused_housenumbers(self, add_housenumber, word_table, hnr):
694 word_table.add_housenumber(1000, hnr)
696 assert word_table.count_housenumbers() == 1
697 self.tok.update_word_tokens()
698 assert word_table.count_housenumbers() == 0
701 def test_keep_unused_numeral_housenumbers(self, add_housenumber, word_table):
702 add_housenumber(1000, '5432')
704 assert word_table.count_housenumbers() == 1
705 self.tok.update_word_tokens()
706 assert word_table.count_housenumbers() == 1
709 def test_keep_housenumbers_from_search_name_table(self, add_housenumber, word_table, search_entry):
710 add_housenumber(9999, '5432a')
711 add_housenumber(9991, '9 a')
712 search_entry(123, 9999, 34)
714 assert word_table.count_housenumbers() == 2
715 self.tok.update_word_tokens()
716 assert word_table.count_housenumbers() == 1
719 def test_keep_housenumbers_from_placex_table(self, add_housenumber, word_table, placex_table):
720 add_housenumber(9999, '5432a')
721 add_housenumber(9990, '34z')
722 placex_table.add(housenumber='34z')
723 placex_table.add(housenumber='25432a')
725 assert word_table.count_housenumbers() == 2
726 self.tok.update_word_tokens()
727 assert word_table.count_housenumbers() == 1
730 def test_keep_housenumbers_from_placex_table_hnr_list(self, add_housenumber, word_table, placex_table):
731 add_housenumber(9991, '9 b')
732 add_housenumber(9990, '34z')
733 placex_table.add(housenumber='9 a;9 b;9 c')
735 assert word_table.count_housenumbers() == 2
736 self.tok.update_word_tokens()
737 assert word_table.count_housenumbers() == 1