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.
16 from nominatim.tokenizer import icu_tokenizer
17 import nominatim.tokenizer.icu_rule_loader
18 from nominatim.db import properties
19 from nominatim.db.sql_preprocessor import SQLPreprocessor
20 from nominatim.indexer.place_info import PlaceInfo
22 from mock_icu_word_table import MockIcuWordTable
25 def word_table(temp_db_conn):
26 return MockIcuWordTable(temp_db_conn)
30 def test_config(project_env, tmp_path):
31 sqldir = tmp_path / 'sql'
33 (sqldir / 'tokenizer').mkdir()
34 (sqldir / 'tokenizer' / 'icu_tokenizer.sql').write_text("SELECT 'a'")
35 shutil.copy(str(project_env.lib_dir.sql / 'tokenizer' / 'icu_tokenizer_tables.sql'),
36 str(sqldir / 'tokenizer' / 'icu_tokenizer_tables.sql'))
38 project_env.lib_dir.sql = sqldir
44 def tokenizer_factory(dsn, tmp_path, property_table,
45 sql_preprocessor, place_table, word_table):
46 (tmp_path / 'tokenizer').mkdir()
49 return icu_tokenizer.create(dsn, tmp_path / 'tokenizer')
55 def db_prop(temp_db_conn):
56 def _get_db_property(name):
57 return properties.get_property(temp_db_conn, name)
59 return _get_db_property
63 def analyzer(tokenizer_factory, test_config, monkeypatch,
64 temp_db_with_extensions, tmp_path):
65 sql = tmp_path / 'sql' / 'tokenizer' / 'icu_tokenizer.sql'
66 sql.write_text("SELECT 'a';")
68 monkeypatch.setenv('NOMINATIM_TERM_NORMALIZATION', ':: lower();')
69 tok = tokenizer_factory()
70 tok.init_new_db(test_config)
73 def _mk_analyser(norm=("[[:Punctuation:][:Space:]]+ > ' '",), trans=(':: upper()',),
74 variants=('~gasse -> gasse', 'street => st', ),
76 cfgstr = {'normalization': list(norm),
77 'sanitizers': sanitizers,
78 'transliteration': list(trans),
79 'token-analysis': [{'analyzer': 'generic',
80 'variants': [{'words': list(variants)}]}]}
81 (test_config.project_dir / 'icu_tokenizer.yaml').write_text(yaml.dump(cfgstr))
82 tok.loader = nominatim.tokenizer.icu_rule_loader.ICURuleLoader(test_config)
84 return tok.name_analyzer()
89 def sql_functions(temp_db_conn, def_config, src_dir):
90 orig_sql = def_config.lib_dir.sql
91 def_config.lib_dir.sql = src_dir / 'lib-sql'
92 sqlproc = SQLPreprocessor(temp_db_conn, def_config)
93 sqlproc.run_sql_file(temp_db_conn, 'functions/utils.sql')
94 sqlproc.run_sql_file(temp_db_conn, 'tokenizer/icu_tokenizer.sql')
95 def_config.lib_dir.sql = orig_sql
99 def getorcreate_full_word(temp_db_cursor):
100 temp_db_cursor.execute("""CREATE OR REPLACE FUNCTION getorcreate_full_word(
101 norm_term TEXT, lookup_terms TEXT[],
103 OUT partial_tokens INT[])
106 partial_terms TEXT[] = '{}'::TEXT[];
111 SELECT min(word_id) INTO full_token
112 FROM word WHERE info->>'word' = norm_term and type = 'W';
114 IF full_token IS NULL THEN
115 full_token := nextval('seq_word');
116 INSERT INTO word (word_id, word_token, type, info)
117 SELECT full_token, lookup_term, 'W',
118 json_build_object('word', norm_term, 'count', 0)
119 FROM unnest(lookup_terms) as lookup_term;
122 FOR term IN SELECT unnest(string_to_array(unnest(lookup_terms), ' ')) LOOP
124 IF NOT (ARRAY[term] <@ partial_terms) THEN
125 partial_terms := partial_terms || term;
129 partial_tokens := '{}'::INT[];
130 FOR term IN SELECT unnest(partial_terms) LOOP
131 SELECT min(word_id), max(info->>'count') INTO term_id, term_count
132 FROM word WHERE word_token = term and type = 'w';
134 IF term_id IS NULL THEN
135 term_id := nextval('seq_word');
137 INSERT INTO word (word_id, word_token, type, info)
138 VALUES (term_id, term, 'w', json_build_object('count', term_count));
141 IF NOT (ARRAY[term_id] <@ partial_tokens) THEN
142 partial_tokens := partial_tokens || term_id;
152 def test_init_new(tokenizer_factory, test_config, db_prop):
153 tok = tokenizer_factory()
154 tok.init_new_db(test_config)
156 assert db_prop(nominatim.tokenizer.icu_rule_loader.DBCFG_IMPORT_NORM_RULES) \
157 .startswith(':: lower ();')
160 def test_init_word_table(tokenizer_factory, test_config, place_row, temp_db_cursor):
161 place_row(names={'name' : 'Test Area', 'ref' : '52'})
162 place_row(names={'name' : 'No Area'})
163 place_row(names={'name' : 'Holzstrasse'})
165 tok = tokenizer_factory()
166 tok.init_new_db(test_config)
168 assert temp_db_cursor.table_exists('word')
171 def test_init_from_project(test_config, tokenizer_factory):
172 tok = tokenizer_factory()
173 tok.init_new_db(test_config)
175 tok = tokenizer_factory()
176 tok.init_from_project(test_config)
178 assert tok.loader is not None
181 def test_update_sql_functions(db_prop, temp_db_cursor,
182 tokenizer_factory, test_config, table_factory,
184 tok = tokenizer_factory()
185 tok.init_new_db(test_config)
187 table_factory('test', 'txt TEXT')
189 func_file = test_config.lib_dir.sql / 'tokenizer' / 'icu_tokenizer.sql'
190 func_file.write_text("""INSERT INTO test VALUES (1133)""")
192 tok.update_sql_functions(test_config)
194 test_content = temp_db_cursor.row_set('SELECT * FROM test')
195 assert test_content == set((('1133', ), ))
198 def test_finalize_import(tokenizer_factory, temp_db_conn,
199 temp_db_cursor, test_config, sql_preprocessor_cfg):
200 func_file = test_config.lib_dir.sql / 'tokenizer' / 'legacy_tokenizer_indices.sql'
201 func_file.write_text("""CREATE FUNCTION test() RETURNS TEXT
202 AS $$ SELECT 'b'::text $$ LANGUAGE SQL""")
204 tok = tokenizer_factory()
205 tok.init_new_db(test_config)
207 tok.finalize_import(test_config)
209 temp_db_cursor.scalar('SELECT test()') == 'b'
212 def test_check_database(test_config, tokenizer_factory,
213 temp_db_cursor, sql_preprocessor_cfg):
214 tok = tokenizer_factory()
215 tok.init_new_db(test_config)
217 assert tok.check_database(test_config) is None
220 def test_update_statistics_reverse_only(word_table, tokenizer_factory):
221 tok = tokenizer_factory()
222 tok.update_statistics()
225 def test_update_statistics(word_table, table_factory, temp_db_cursor, tokenizer_factory):
226 word_table.add_full_word(1000, 'hello')
227 table_factory('search_name',
228 'place_id BIGINT, name_vector INT[]',
230 tok = tokenizer_factory()
232 tok.update_statistics()
234 assert temp_db_cursor.scalar("""SELECT count(*) FROM word
236 (info->>'count')::int > 0""") > 0
239 def test_normalize_postcode(analyzer):
240 with analyzer() as anl:
241 anl.normalize_postcode('123') == '123'
242 anl.normalize_postcode('ab-34 ') == 'AB-34'
243 anl.normalize_postcode('38 Б') == '38 Б'
246 def test_update_postcodes_from_db_empty(analyzer, table_factory, word_table):
247 table_factory('location_postcode', 'postcode TEXT',
248 content=(('1234',), ('12 34',), ('AB23',), ('1234',)))
250 with analyzer() as anl:
251 anl.update_postcodes_from_db()
253 assert word_table.count() == 3
254 assert word_table.get_postcodes() == {'1234', '12 34', 'AB23'}
257 def test_update_postcodes_from_db_add_and_remove(analyzer, table_factory, word_table):
258 table_factory('location_postcode', 'postcode TEXT',
259 content=(('1234',), ('45BC', ), ('XX45', )))
260 word_table.add_postcode(' 1234', '1234')
261 word_table.add_postcode(' 5678', '5678')
263 with analyzer() as anl:
264 anl.update_postcodes_from_db()
266 assert word_table.count() == 3
267 assert word_table.get_postcodes() == {'1234', '45BC', 'XX45'}
270 def test_update_special_phrase_empty_table(analyzer, word_table):
271 with analyzer() as anl:
272 anl.update_special_phrases([
273 ("König bei", "amenity", "royal", "near"),
274 ("Könige ", "amenity", "royal", "-"),
275 ("street", "highway", "primary", "in")
278 assert word_table.get_special() \
279 == {('KÖNIG BEI', 'König bei', 'amenity', 'royal', 'near'),
280 ('KÖNIGE', 'Könige', 'amenity', 'royal', None),
281 ('STREET', 'street', 'highway', 'primary', 'in')}
284 def test_update_special_phrase_delete_all(analyzer, word_table):
285 word_table.add_special('FOO', 'foo', 'amenity', 'prison', 'in')
286 word_table.add_special('BAR', 'bar', 'highway', 'road', None)
288 assert word_table.count_special() == 2
290 with analyzer() as anl:
291 anl.update_special_phrases([], True)
293 assert word_table.count_special() == 0
296 def test_update_special_phrases_no_replace(analyzer, word_table):
297 word_table.add_special('FOO', 'foo', 'amenity', 'prison', 'in')
298 word_table.add_special('BAR', 'bar', 'highway', 'road', None)
300 assert word_table.count_special() == 2
302 with analyzer() as anl:
303 anl.update_special_phrases([], False)
305 assert word_table.count_special() == 2
308 def test_update_special_phrase_modify(analyzer, word_table):
309 word_table.add_special('FOO', 'foo', 'amenity', 'prison', 'in')
310 word_table.add_special('BAR', 'bar', 'highway', 'road', None)
312 assert word_table.count_special() == 2
314 with analyzer() as anl:
315 anl.update_special_phrases([
316 ('prison', 'amenity', 'prison', 'in'),
317 ('bar', 'highway', 'road', '-'),
318 ('garden', 'leisure', 'garden', 'near')
321 assert word_table.get_special() \
322 == {('PRISON', 'prison', 'amenity', 'prison', 'in'),
323 ('BAR', 'bar', 'highway', 'road', None),
324 ('GARDEN', 'garden', 'leisure', 'garden', 'near')}
327 def test_add_country_names_new(analyzer, word_table):
328 with analyzer() as anl:
329 anl.add_country_names('es', {'name': 'Espagña', 'name:en': 'Spain'})
331 assert word_table.get_country() == {('es', 'ESPAGÑA'), ('es', 'SPAIN')}
334 def test_add_country_names_extend(analyzer, word_table):
335 word_table.add_country('ch', 'SCHWEIZ')
337 with analyzer() as anl:
338 anl.add_country_names('ch', {'name': 'Schweiz', 'name:fr': 'Suisse'})
340 assert word_table.get_country() == {('ch', 'SCHWEIZ'), ('ch', 'SUISSE')}
343 class TestPlaceNames:
345 @pytest.fixture(autouse=True)
346 def setup(self, analyzer, sql_functions):
347 sanitizers = [{'step': 'split-name-list'},
348 {'step': 'strip-brace-terms'}]
349 with analyzer(sanitizers=sanitizers) as anl:
354 def expect_name_terms(self, info, *expected_terms):
355 tokens = self.analyzer.get_word_token_info(expected_terms)
357 assert token[2] is not None, "No token for {0}".format(token)
359 assert eval(info['names']) == set((t[2] for t in tokens))
362 def process_named_place(self, names):
363 return self.analyzer.process_place(PlaceInfo({'name': names}))
366 def test_simple_names(self):
367 info = self.process_named_place({'name': 'Soft bAr', 'ref': '34'})
369 self.expect_name_terms(info, '#Soft bAr', '#34', 'Soft', 'bAr', '34')
372 @pytest.mark.parametrize('sep', [',' , ';'])
373 def test_names_with_separator(self, sep):
374 info = self.process_named_place({'name': sep.join(('New York', 'Big Apple'))})
376 self.expect_name_terms(info, '#New York', '#Big Apple',
377 'new', 'york', 'big', 'apple')
380 def test_full_names_with_bracket(self):
381 info = self.process_named_place({'name': 'Houseboat (left)'})
383 self.expect_name_terms(info, '#Houseboat (left)', '#Houseboat',
387 def test_country_name(self, word_table):
388 place = PlaceInfo({'name' : {'name': 'Norge'},
389 'country_code': 'no',
392 'type': 'administrative'})
394 info = self.analyzer.process_place(place)
396 self.expect_name_terms(info, '#norge', 'norge')
397 assert word_table.get_country() == {('no', 'NORGE')}
400 class TestPlaceAddress:
402 @pytest.fixture(autouse=True)
403 def setup(self, analyzer, sql_functions):
404 hnr = {'step': 'clean-housenumbers',
405 'filter-kind': ['housenumber', 'conscriptionnumber', 'streetnumber']}
406 with analyzer(trans=(":: upper()", "'🜵' > ' '"), sanitizers=[hnr]) as anl:
412 def getorcreate_hnr_id(self, temp_db_cursor):
413 temp_db_cursor.execute("""CREATE OR REPLACE FUNCTION getorcreate_hnr_id(lookup_term TEXT)
414 RETURNS INTEGER AS $$
415 SELECT -nextval('seq_word')::INTEGER; $$ LANGUAGE SQL""")
418 def process_address(self, **kwargs):
419 return self.analyzer.process_place(PlaceInfo({'address': kwargs}))
422 def name_token_set(self, *expected_terms):
423 tokens = self.analyzer.get_word_token_info(expected_terms)
425 assert token[2] is not None, "No token for {0}".format(token)
427 return set((t[2] for t in tokens))
430 @pytest.mark.parametrize('pcode', ['12345', 'AB 123', '34-345'])
431 def test_process_place_postcode(self, word_table, pcode):
432 self.process_address(postcode=pcode)
434 assert word_table.get_postcodes() == {pcode, }
437 @pytest.mark.parametrize('pcode', ['12:23', 'ab;cd;f', '123;836'])
438 def test_process_place_bad_postcode(self, word_table, pcode):
439 self.process_address(postcode=pcode)
441 assert not word_table.get_postcodes()
444 @pytest.mark.parametrize('hnr', ['123a', '1', '101'])
445 def test_process_place_housenumbers_simple(self, hnr, getorcreate_hnr_id):
446 info = self.process_address(housenumber=hnr)
448 assert info['hnr'] == hnr.upper()
449 assert info['hnr_tokens'] == "{-1}"
452 def test_process_place_housenumbers_duplicates(self, getorcreate_hnr_id):
453 info = self.process_address(housenumber='134',
454 conscriptionnumber='134',
457 assert set(info['hnr'].split(';')) == set(('134', '99A'))
458 assert info['hnr_tokens'] == "{-1,-2}"
461 def test_process_place_housenumbers_cached(self, getorcreate_hnr_id):
462 info = self.process_address(housenumber="45")
463 assert info['hnr_tokens'] == "{-1}"
465 info = self.process_address(housenumber="46")
466 assert info['hnr_tokens'] == "{-2}"
468 info = self.process_address(housenumber="41;45")
469 assert eval(info['hnr_tokens']) == {-1, -3}
471 info = self.process_address(housenumber="41")
472 assert eval(info['hnr_tokens']) == {-3}
475 def test_process_place_street(self):
476 self.analyzer.process_place(PlaceInfo({'name': {'name' : 'Grand Road'}}))
477 info = self.process_address(street='Grand Road')
479 assert eval(info['street']) == self.name_token_set('#Grand Road')
482 def test_process_place_nonexisting_street(self):
483 info = self.process_address(street='Grand Road')
485 assert 'street' not in info
488 def test_process_place_multiple_street_tags(self):
489 self.analyzer.process_place(PlaceInfo({'name': {'name' : 'Grand Road',
491 info = self.process_address(**{'street': 'Grand Road',
492 'street:sym_ul': '05989'})
494 assert eval(info['street']) == self.name_token_set('#Grand Road', '#05989')
497 def test_process_place_street_empty(self):
498 info = self.process_address(street='🜵')
500 assert 'street' not in info
503 def test_process_place_street_from_cache(self):
504 self.analyzer.process_place(PlaceInfo({'name': {'name' : 'Grand Road'}}))
505 self.process_address(street='Grand Road')
507 # request address again
508 info = self.process_address(street='Grand Road')
510 assert eval(info['street']) == self.name_token_set('#Grand Road')
513 def test_process_place_place(self):
514 info = self.process_address(place='Honu Lulu')
516 assert eval(info['place']) == self.name_token_set('HONU', 'LULU')
519 def test_process_place_place_extra(self):
520 info = self.process_address(**{'place:en': 'Honu Lulu'})
522 assert 'place' not in info
525 def test_process_place_place_empty(self):
526 info = self.process_address(place='🜵')
528 assert 'place' not in info
531 def test_process_place_address_terms(self):
532 info = self.process_address(country='de', city='Zwickau', state='Sachsen',
533 suburb='Zwickau', street='Hauptstr',
534 full='right behind the church')
536 city = self.name_token_set('ZWICKAU')
537 state = self.name_token_set('SACHSEN')
539 result = {k: eval(v) for k,v in info['addr'].items()}
541 assert result == {'city': city, 'suburb': city, 'state': state}
544 def test_process_place_multiple_address_terms(self):
545 info = self.process_address(**{'city': 'Bruxelles', 'city:de': 'Brüssel'})
547 result = {k: eval(v) for k,v in info['addr'].items()}
549 assert result == {'city': self.name_token_set('Bruxelles')}
552 def test_process_place_address_terms_empty(self):
553 info = self.process_address(country='de', city=' ', street='Hauptstr',
554 full='right behind the church')
556 assert 'addr' not in info
559 class TestUpdateWordTokens:
561 @pytest.fixture(autouse=True)
562 def setup(self, tokenizer_factory, table_factory, placex_table, word_table):
563 table_factory('search_name', 'place_id BIGINT, name_vector INT[]')
564 self.tok = tokenizer_factory()
568 def search_entry(self, temp_db_cursor):
569 place_id = itertools.count(1000)
572 temp_db_cursor.execute("INSERT INTO search_name VALUES (%s, %s)",
573 (next(place_id), list(args)))
578 @pytest.mark.parametrize('hnr', ('1a', '1234567', '34 5'))
579 def test_remove_unused_housenumbers(self, word_table, hnr):
580 word_table.add_housenumber(1000, hnr)
582 assert word_table.count_housenumbers() == 1
583 self.tok.update_word_tokens()
584 assert word_table.count_housenumbers() == 0
587 def test_keep_unused_numeral_housenumbers(self, word_table):
588 word_table.add_housenumber(1000, '5432')
590 assert word_table.count_housenumbers() == 1
591 self.tok.update_word_tokens()
592 assert word_table.count_housenumbers() == 1
595 def test_keep_housenumbers_from_search_name_table(self, word_table, search_entry):
596 word_table.add_housenumber(9999, '5432a')
597 word_table.add_housenumber(9991, '9 a')
598 search_entry(123, 9999, 34)
600 assert word_table.count_housenumbers() == 2
601 self.tok.update_word_tokens()
602 assert word_table.count_housenumbers() == 1
605 def test_keep_housenumbers_from_placex_table(self, word_table, placex_table):
606 word_table.add_housenumber(9999, '5432a')
607 word_table.add_housenumber(9990, '34z')
608 placex_table.add(housenumber='34z')
609 placex_table.add(housenumber='25432a')
611 assert word_table.count_housenumbers() == 2
612 self.tok.update_word_tokens()
613 assert word_table.count_housenumbers() == 1
616 def test_keep_housenumbers_from_placex_table_hnr_list(self, word_table, placex_table):
617 word_table.add_housenumber(9991, '9 b')
618 word_table.add_housenumber(9990, '34z')
619 placex_table.add(housenumber='9 a;9 b;9 c')
621 assert word_table.count_housenumbers() == 2
622 self.tok.update_word_tokens()
623 assert word_table.count_housenumbers() == 1