From 2e3c5d4c5b39e29af57a9398f20fdf5cad0e9045 Mon Sep 17 00:00:00 2001 From: Sarah Hoffmann Date: Sun, 6 Jun 2021 11:00:44 +0200 Subject: [PATCH] adapt tests for ICU tokenizer --- nominatim/tokenizer/icu_name_processor.py | 4 +- nominatim/tokenizer/icu_rule_loader.py | 4 +- nominatim/tokenizer/legacy_icu_tokenizer.py | 12 +- nominatim/tokenizer/legacy_tokenizer.py | 5 +- test/bdd/steps/steps_db_ops.py | 2 +- .../test_tokenizer_icu_name_processor.py | 24 ++- test/python/test_tokenizer_icu_rule_loader.py | 9 +- test/python/test_tokenizer_legacy_icu.py | 150 ++++++++++++------ 8 files changed, 143 insertions(+), 67 deletions(-) diff --git a/nominatim/tokenizer/icu_name_processor.py b/nominatim/tokenizer/icu_name_processor.py index 0e717995..a0f22974 100644 --- a/nominatim/tokenizer/icu_name_processor.py +++ b/nominatim/tokenizer/icu_name_processor.py @@ -76,7 +76,7 @@ class ICUNameProcessor: """ Normalize the given name, i.e. remove all elements not relevant for search. """ - return self.normalizer.transliterate(name) + return self.normalizer.transliterate(name).strip() def get_variants_ascii(self, norm_name): """ Compute the spelling variants for the given normalized name @@ -108,4 +108,4 @@ class ICUNameProcessor: """ Return the normalized version of the name (including transliteration) to be applied at search time. """ - return self.search.transliterate(name) + return self.search.transliterate(' ' + name + ' ').strip() diff --git a/nominatim/tokenizer/icu_rule_loader.py b/nominatim/tokenizer/icu_rule_loader.py index 3b721169..2597656b 100644 --- a/nominatim/tokenizer/icu_rule_loader.py +++ b/nominatim/tokenizer/icu_rule_loader.py @@ -28,7 +28,7 @@ class ICURuleLoader: def get_search_rules(self): - """ Returns the ICU rules to be used during search. + """ Return the ICU rules to be used during search. The rules combine normalization, compound decomposition (including abbreviated compounds) and transliteration. """ @@ -60,7 +60,7 @@ class ICURuleLoader: return self.transliteration_rules def get_replacement_pairs(self): - """ Returns the list of possible compound decompositions with + """ Return the list of possible compound decompositions with application of abbreviations included. The result is a list of pairs: the first item is the sequence to replace, the second is a list of replacements. diff --git a/nominatim/tokenizer/legacy_icu_tokenizer.py b/nominatim/tokenizer/legacy_icu_tokenizer.py index eb850237..20932144 100644 --- a/nominatim/tokenizer/legacy_icu_tokenizer.py +++ b/nominatim/tokenizer/legacy_icu_tokenizer.py @@ -219,7 +219,7 @@ class LegacyICUNameAnalyzer: self.conn = None - def get_word_token_info(self, conn, words): + def get_word_token_info(self, words): """ 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. @@ -233,11 +233,11 @@ class LegacyICUNameAnalyzer: tokens = {} for word in words: if word.startswith('#'): - tokens[word] = ' ' + self.name_processor.get_normalized(word[1:]) + tokens[word] = ' ' + self.name_processor.get_search_normalized(word[1:]) else: - tokens[word] = self.name_processor.get_normalized(word) + tokens[word] = self.name_processor.get_search_normalized(word) - with conn.cursor() as cur: + with self.conn.cursor() as cur: cur.execute("""SELECT word_token, word_id FROM word, (SELECT unnest(%s::TEXT[]) as term) t WHERE word_token = t.term @@ -245,7 +245,7 @@ class LegacyICUNameAnalyzer: (list(tokens.values()), )) ids = {r[0]: r[1] for r in cur} - return [(k, v, ids[v]) for k, v in tokens.items()] + return [(k, v, ids.get(v, None)) for k, v in tokens.items()] @staticmethod @@ -308,7 +308,7 @@ class LegacyICUNameAnalyzer: def update_special_phrases(self, phrases, should_replace): """ Replace the search index for special phrases with the new phrases. """ - norm_phrases = set(((self.name_processor.get_search_normalized(p[0]), p[1], p[2], p[3]) + norm_phrases = set(((self.name_processor.get_normalized(p[0]), p[1], p[2], p[3]) for p in phrases)) with self.conn.cursor() as cur: diff --git a/nominatim/tokenizer/legacy_tokenizer.py b/nominatim/tokenizer/legacy_tokenizer.py index bb37115b..6040f88f 100644 --- a/nominatim/tokenizer/legacy_tokenizer.py +++ b/nominatim/tokenizer/legacy_tokenizer.py @@ -271,8 +271,7 @@ class LegacyNameAnalyzer: self.conn = None - @staticmethod - def get_word_token_info(conn, words): + def get_word_token_info(self, words): """ 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. @@ -283,7 +282,7 @@ class LegacyNameAnalyzer: The function is used for testing and debugging only and not necessarily efficient. """ - with conn.cursor() as cur: + with self.conn.cursor() as cur: cur.execute("""SELECT t.term, word_token, word_id FROM word, (SELECT unnest(%s::TEXT[]) as term) t WHERE word_token = (CASE diff --git a/test/bdd/steps/steps_db_ops.py b/test/bdd/steps/steps_db_ops.py index 8285f938..b4f0d853 100644 --- a/test/bdd/steps/steps_db_ops.py +++ b/test/bdd/steps/steps_db_ops.py @@ -214,7 +214,7 @@ def check_search_name_contents(context, exclude): for name, value in zip(row.headings, row.cells): if name in ('name_vector', 'nameaddress_vector'): items = [x.strip() for x in value.split(',')] - tokens = analyzer.get_word_token_info(context.db, items) + tokens = analyzer.get_word_token_info(items) if not exclude: assert len(tokens) >= len(items), \ diff --git a/test/python/test_tokenizer_icu_name_processor.py b/test/python/test_tokenizer_icu_name_processor.py index 9c09bccc..817f9072 100644 --- a/test/python/test_tokenizer_icu_name_processor.py +++ b/test/python/test_tokenizer_icu_name_processor.py @@ -34,6 +34,9 @@ def cfgfile(tmp_path, suffix='.yaml'): return _create_config +def get_normalized_variants(proc, name): + return proc.get_variants_ascii(proc.get_normalized(name)) + def test_simple_variants(cfgfile): fpath = cfgfile(['strasse', 'straße', 'weg'], ['strasse,straße => str', @@ -42,11 +45,11 @@ def test_simple_variants(cfgfile): rules = ICUNameProcessorRules(loader=ICURuleLoader(fpath)) proc = ICUNameProcessor(rules) - assert set(proc.get_normalized_variants("Bauwegstraße")) \ + assert set(get_normalized_variants(proc, "Bauwegstraße")) \ == {'bauweg straße', 'bauweg str'} - assert proc.get_normalized_variants("Bauwegstr") == ['bauweg str'] - assert proc.get_normalized_variants("holzweg") == ['holz weg'] - assert proc.get_normalized_variants("hallo") == ['hallo'] + assert get_normalized_variants(proc, "Bauwegstr") == ['bauweg str'] + assert get_normalized_variants(proc, "holzweg") == ['holz weg'] + assert get_normalized_variants(proc, "hallo") == ['hallo'] def test_multiple_replacements(cfgfile): @@ -55,6 +58,17 @@ def test_multiple_replacements(cfgfile): rules = ICUNameProcessorRules(loader=ICURuleLoader(fpath)) proc = ICUNameProcessor(rules) - assert set(proc.get_normalized_variants("Saint Johns Street")) == \ + assert set(get_normalized_variants(proc, "Saint Johns Street")) == \ {'saint johns street', 's johns street', 'st johns street', 'saint johns st', 's johns st', 'st johns st'} + + +def test_search_normalized(cfgfile): + fpath = cfgfile(['street'], ['street => s,st', 'master => mstr']) + + rules = ICUNameProcessorRules(loader=ICURuleLoader(fpath)) + proc = ICUNameProcessor(rules) + + assert proc.get_search_normalized('Master Street') == 'master street' + assert proc.get_search_normalized('Earnes St') == 'earne s st' + assert proc.get_search_normalized('Nostreet') == 'no street' diff --git a/test/python/test_tokenizer_icu_rule_loader.py b/test/python/test_tokenizer_icu_rule_loader.py index d89e13b5..20b127f3 100644 --- a/test/python/test_tokenizer_icu_rule_loader.py +++ b/test/python/test_tokenizer_icu_rule_loader.py @@ -69,7 +69,10 @@ def test_get_synonym_pairs(cfgfile): repl = loader.get_replacement_pairs() - assert repl == [(' strasse ', {' strasse ', ' str ', ' st '}), - ('strasse ', {' strasse ', ' str ', ' st '}), - ('weg ', {' weg '})] + assert sorted(((a, sorted(b)) for a, b in repl)) == \ + sorted([(' strasse ', [' st ', ' str ', ' strasse ']), + ('strasse ', [' st ', ' str ', ' strasse ']), + ('st ' , [' st ']), + ('str ' , [' str ']), + ('weg ', [' weg '])]) diff --git a/test/python/test_tokenizer_legacy_icu.py b/test/python/test_tokenizer_legacy_icu.py index d8ca2f22..f7558dac 100644 --- a/test/python/test_tokenizer_legacy_icu.py +++ b/test/python/test_tokenizer_legacy_icu.py @@ -2,10 +2,13 @@ Tests for Legacy ICU tokenizer. """ import shutil +import yaml import pytest from nominatim.tokenizer import legacy_icu_tokenizer +from nominatim.tokenizer.icu_name_processor import ICUNameProcessorRules +from nominatim.tokenizer.icu_rule_loader import ICURuleLoader from nominatim.db import properties @@ -40,16 +43,10 @@ def tokenizer_factory(dsn, tmp_path, property_table, @pytest.fixture def db_prop(temp_db_conn): def _get_db_property(name): - return properties.get_property(temp_db_conn, - getattr(legacy_icu_tokenizer, name)) + return properties.get_property(temp_db_conn, name) return _get_db_property -@pytest.fixture -def tokenizer_setup(tokenizer_factory, test_config): - tok = tokenizer_factory() - tok.init_new_db(test_config) - @pytest.fixture def analyzer(tokenizer_factory, test_config, monkeypatch, @@ -62,9 +59,16 @@ def analyzer(tokenizer_factory, test_config, monkeypatch, tok.init_new_db(test_config) monkeypatch.undo() - def _mk_analyser(trans=':: upper();', abbr=(('STREET', 'ST'), )): - tok.transliteration = trans - tok.abbreviations = abbr + def _mk_analyser(norm=("[[:Punctuation:][:Space:]]+ > ' '",), trans=(':: upper()',), + suffixes=('gasse', ), abbr=('street => st', )): + cfgfile = tmp_path / 'analyser_test_config.yaml' + with cfgfile.open('w') as stream: + cfgstr = {'normalization' : list(norm), + 'transliteration' : list(trans), + 'compound_suffixes' : list(suffixes), + 'abbreviations' : list(abbr)} + yaml.dump(cfgstr, stream) + tok.naming_rules = ICUNameProcessorRules(loader=ICURuleLoader(cfgfile)) return tok.name_analyzer() @@ -72,10 +76,54 @@ def analyzer(tokenizer_factory, test_config, monkeypatch, @pytest.fixture -def getorcreate_term_id(temp_db_cursor): - temp_db_cursor.execute("""CREATE OR REPLACE FUNCTION getorcreate_term_id(lookup_term TEXT) - RETURNS INTEGER AS $$ - SELECT nextval('seq_word')::INTEGER; $$ LANGUAGE SQL""") +def getorcreate_full_word(temp_db_cursor): + temp_db_cursor.execute("""CREATE OR REPLACE FUNCTION getorcreate_full_word( + norm_term TEXT, lookup_terms TEXT[], + OUT full_token INT, + OUT partial_tokens INT[]) + AS $$ +DECLARE + partial_terms TEXT[] = '{}'::TEXT[]; + term TEXT; + term_id INTEGER; + term_count INTEGER; +BEGIN + SELECT min(word_id) INTO full_token + FROM word WHERE word = norm_term and class is null and country_code is null; + + IF full_token IS NULL THEN + full_token := nextval('seq_word'); + INSERT INTO word (word_id, word_token, word, search_name_count) + SELECT full_token, ' ' || lookup_term, norm_term, 0 FROM unnest(lookup_terms) as lookup_term; + END IF; + + FOR term IN SELECT unnest(string_to_array(unnest(lookup_terms), ' ')) LOOP + term := trim(term); + IF NOT (ARRAY[term] <@ partial_terms) THEN + partial_terms := partial_terms || term; + END IF; + END LOOP; + + partial_tokens := '{}'::INT[]; + FOR term IN SELECT unnest(partial_terms) LOOP + SELECT min(word_id), max(search_name_count) INTO term_id, term_count + FROM word WHERE word_token = term and class is null and country_code is null; + + IF term_id IS NULL THEN + term_id := nextval('seq_word'); + term_count := 0; + INSERT INTO word (word_id, word_token, search_name_count) + VALUES (term_id, term, 0); + END IF; + + IF NOT (ARRAY[term_id] <@ partial_tokens) THEN + partial_tokens := partial_tokens || term_id; + END IF; + END LOOP; +END; +$$ +LANGUAGE plpgsql; + """) @pytest.fixture @@ -91,19 +139,23 @@ def test_init_new(tokenizer_factory, test_config, monkeypatch, db_prop): tok = tokenizer_factory() tok.init_new_db(test_config) - assert db_prop('DBCFG_NORMALIZATION') == ':: lower();' - assert db_prop('DBCFG_TRANSLITERATION') is not None - assert db_prop('DBCFG_ABBREVIATIONS') is not None + assert db_prop(legacy_icu_tokenizer.DBCFG_TERM_NORMALIZATION) == ':: lower();' + assert db_prop(legacy_icu_tokenizer.DBCFG_MAXWORDFREQ) is not None -def test_init_from_project(tokenizer_setup, tokenizer_factory): +def test_init_from_project(monkeypatch, test_config, tokenizer_factory): + monkeypatch.setenv('NOMINATIM_TERM_NORMALIZATION', ':: lower();') + monkeypatch.setenv('NOMINATIM_MAX_WORD_FREQUENCY', '90300') tok = tokenizer_factory() + tok.init_new_db(test_config) + monkeypatch.undo() + tok = tokenizer_factory() tok.init_from_project() - assert tok.normalization is not None - assert tok.transliteration is not None - assert tok.abbreviations is not None + assert tok.naming_rules is not None + assert tok.term_normalization == ':: lower();' + assert tok.max_word_frequency == '90300' def test_update_sql_functions(db_prop, temp_db_cursor, @@ -114,7 +166,7 @@ def test_update_sql_functions(db_prop, temp_db_cursor, tok.init_new_db(test_config) monkeypatch.undo() - assert db_prop('DBCFG_MAXWORDFREQ') == '1133' + assert db_prop(legacy_icu_tokenizer.DBCFG_MAXWORDFREQ) == '1133' table_factory('test', 'txt TEXT') @@ -127,16 +179,8 @@ def test_update_sql_functions(db_prop, temp_db_cursor, assert test_content == set((('1133', ), )) -def test_make_standard_word(analyzer): - with analyzer(abbr=(('STREET', 'ST'), ('tiny', 't'))) as anl: - assert anl.make_standard_word('tiny street') == 'TINY ST' - - with analyzer(abbr=(('STRASSE', 'STR'), ('STR', 'ST'))) as anl: - assert anl.make_standard_word('Hauptstrasse') == 'HAUPTST' - - def test_make_standard_hnr(analyzer): - with analyzer(abbr=(('IV', '4'),)) as anl: + with analyzer(abbr=('IV => 4',)) as anl: assert anl._make_standard_hnr('345') == '345' assert anl._make_standard_hnr('iv') == 'IV' @@ -176,7 +220,7 @@ def test_update_special_phrase_empty_table(analyzer, word_table): assert word_table.get_special() \ == {(' KÖNIG BEI', 'könig bei', 'amenity', 'royal', 'near'), (' KÖNIGE', 'könige', 'amenity', 'royal', None), - (' ST', 'street', 'highway', 'primary', 'in')} + (' STREET', 'street', 'highway', 'primary', 'in')} def test_update_special_phrase_delete_all(analyzer, word_table): @@ -222,26 +266,42 @@ def test_update_special_phrase_modify(analyzer, word_table): (' GARDEN', 'garden', 'leisure', 'garden', 'near')} -def test_process_place_names(analyzer, getorcreate_term_id): - with analyzer() as anl: - info = anl.process_place({'name' : {'name' : 'Soft bAr', 'ref': '34'}}) +class TestPlaceNames: - assert info['names'] == '{1,2,3,4,5}' + @pytest.fixture(autouse=True) + def setup(self, analyzer, getorcreate_full_word): + with analyzer() as anl: + self.analyzer = anl + yield anl -@pytest.mark.parametrize('sep', [',' , ';']) -def test_full_names_with_separator(analyzer, getorcreate_term_id, sep): - with analyzer() as anl: - names = anl._compute_full_names({'name' : sep.join(('New York', 'Big Apple'))}) + def expect_name_terms(self, info, *expected_terms): + tokens = self.analyzer.get_word_token_info(expected_terms) + for token in tokens: + assert token[2] is not None, "No token for {0}".format(token) - assert names == set(('NEW YORK', 'BIG APPLE')) + assert eval(info['names']) == set((t[2] for t in tokens)) -def test_full_names_with_bracket(analyzer, getorcreate_term_id): - with analyzer() as anl: - names = anl._compute_full_names({'name' : 'Houseboat (left)'}) + def test_simple_names(self): + info = self.analyzer.process_place({'name' : {'name' : 'Soft bAr', 'ref': '34'}}) + + self.expect_name_terms(info, '#Soft bAr', '#34','Soft', 'bAr', '34') + + + @pytest.mark.parametrize('sep', [',' , ';']) + def test_names_with_separator(self, sep): + info = self.analyzer.process_place({'name' : {'name' : sep.join(('New York', 'Big Apple'))}}) + + self.expect_name_terms(info, '#New York', '#Big Apple', + 'new', 'york', 'big', 'apple') + + + def test_full_names_with_bracket(self): + info = self.analyzer.process_place({'name' : {'name' : 'Houseboat (left)'}}) - assert names == set(('HOUSEBOAT (LEFT)', 'HOUSEBOAT')) + self.expect_name_terms(info, '#Houseboat (left)', '#Houseboat', + 'houseboat', 'left') @pytest.mark.parametrize('pcode', ['12345', 'AB 123', '34-345']) -- 2.43.2