Merge pull request #2263 from AntoJvlt/special-phrases-autoupdate

Implemented auto update of special phrases while importing them

nominatim/tools/special_phrases.py

@@ -32,6 +32,17 @@ class SpecialPhrasesImporter():
         self.sanity_check_pattern = re.compile(r'^\w+$')
         self.transliterator = Transliterator.createFromRules("special-phrases normalizer",
                                                              self.config.TERM_NORMALIZATION)
+        #This set will contain all existing phrases from the word table which
+        #no longer exist on the wiki.
+        #It contain tuples with the following format: (normalized_word, class, type, operator)
+        self.words_phrases_to_delete = set()
+        #This set will contain the phrases which still exist from the wiki.
+        #It is used to prevent duplicates on the wiki by removing them from
+        #the word_phrases_to_delete only at the end.
+        self.words_phrases_still_exist = set()
+        #This set will contain all existing place_classtype tables which doesn't match any
+        #special phrases class/type on the wiki.
+        self.table_phrases_to_delete = set()
 
     def import_from_wiki(self, languages=None):
         """
@@ -41,6 +52,9 @@ class SpecialPhrasesImporter():
         if languages is not None and not isinstance(languages, list):
             raise TypeError('The \'languages\' argument should be of type list.')
 
+        self._fetch_existing_words_phrases()
+        self._fetch_existing_place_classtype_tables()
+
         #Get all languages to process.
         languages = self._load_languages() if not languages else languages
 
@@ -53,9 +67,46 @@ class SpecialPhrasesImporter():
             class_type_pairs.update(self._process_xml_content(wiki_page_xml_content, lang))
 
         self._create_place_classtype_table_and_indexes(class_type_pairs)
+        self._remove_non_existent_phrases_from_db()
         self.db_connection.commit()
         LOG.warning('Import done.')
 
+    def _fetch_existing_words_phrases(self):
+        """
+            Fetch existing special phrases from the word table.
+            Fill the word_phrases_to_delete set of the class.
+        """
+        #Only extract special phrases terms:
+        #If class=place and type=house then it is a housenumber term.
+        #If class=place and type=postcode then it is a postcode term.
+        word_query = """
+            SELECT word, class, type, operator FROM word
+            WHERE class != 'place' OR (type != 'house' AND type != 'postcode')
+        """
+        with self.db_connection.cursor() as db_cursor:
+            db_cursor.execute(SQL(word_query))
+            for row in db_cursor:
+                row[3] = '-' if row[3] is None else row[3]
+                self.words_phrases_to_delete.add(
+                    (row[0], row[1], row[2], row[3])
+                )
+
+    def _fetch_existing_place_classtype_tables(self):
+        """
+            Fetch existing place_classtype tables.
+            Fill the table_phrases_to_delete set of the class.
+        """
+        query = """
+            SELECT table_name
+            FROM information_schema.tables
+            WHERE table_schema='public'
+            AND table_name like 'place_classtype_%';
+        """
+        with self.db_connection.cursor() as db_cursor:
+            db_cursor.execute(SQL(query))
+            for row in db_cursor:
+                self.table_phrases_to_delete.add(row[0])
+
     def _load_white_and_black_lists(self):
         """
             Load white and black lists from phrases-settings.json.
@@ -122,12 +173,11 @@ class SpecialPhrasesImporter():
             phrase_class = match[1].strip()
             phrase_type = match[2].strip()
             phrase_operator = match[3].strip()
+            #Needed if some operator in the wiki are not written in english
+            phrase_operator = '-' if phrase_operator not in ('near', 'in') else phrase_operator
             #hack around a bug where building=yes was imported with quotes into the wiki
             phrase_type = re.sub(r'\"|"', '', phrase_type)
 
-            #sanity check, in case somebody added garbage in the wiki
-            self._check_sanity(lang, phrase_class, phrase_type)
-
             #blacklisting: disallow certain class/type combinations
             if (
                     phrase_class in self.black_list.keys() and
@@ -141,7 +191,22 @@ class SpecialPhrasesImporter():
             ):
                 continue
 
-            #add class/type to the pairs dict
+            #Check if the phrase already exists in the database.
+            if (
+                    (normalized_label, phrase_class, phrase_type, phrase_operator)
+                    in self.words_phrases_to_delete
+            ):
+                #Remove this phrase from the ones to delete as it still exist on the wiki.
+                self.words_phrases_still_exist.add(
+                    (normalized_label, phrase_class, phrase_type, phrase_operator)
+                )
+                class_type_pairs.add((phrase_class, phrase_type))
+                #Dont need to add this phrase as it already exists in the word table.
+                continue
+
+            #sanity check, in case somebody added garbage in the wiki
+            self._check_sanity(lang, phrase_class, phrase_type)
+
             class_type_pairs.add((phrase_class, phrase_type))
 
             self._process_amenity(
@@ -191,6 +256,15 @@ class SpecialPhrasesImporter():
             phrase_class = pair[0]
             phrase_type = pair[1]
 
+            table_name = 'place_classtype_{}_{}'.format(phrase_class, phrase_type)
+
+            if table_name in self.table_phrases_to_delete:
+                #Remove this table from the ones to delete as it match a class/type
+                #still existing on the special phrases of the wiki.
+                self.table_phrases_to_delete.remove(table_name)
+                #So dont need to create the table and indexes.
+                continue
+
             #Table creation
             self._create_place_classtype_table(sql_tablespace, phrase_class, phrase_type)
 
@@ -251,6 +325,41 @@ class SpecialPhrasesImporter():
                               .format(Identifier(table_name),
                                       Identifier(self.config.DATABASE_WEBUSER)))
 
+    def _remove_non_existent_phrases_from_db(self):
+        """
+            Remove special phrases which doesn't exist on the wiki anymore.
+            Delete from the word table and delete the place_classtype tables.
+        """
+        LOG.warning('Cleaning database...')
+        self.words_phrases_to_delete = self.words_phrases_to_delete - self.words_phrases_still_exist
+        #Array containing all queries to execute. Contain tuples of format (query, parameters)
+        queries_parameters = []
+
+        #Delete phrases from the word table which are not on the wiki anymore.
+        for phrase_to_delete in self.words_phrases_to_delete:
+            if phrase_to_delete[3] == '-':
+                query = """
+                    DELETE FROM word WHERE word = %s AND class = %s AND type = %s AND operator IS null
+                """
+                parameters = (phrase_to_delete[0], phrase_to_delete[1], phrase_to_delete[2], )
+                queries_parameters.append((query, parameters))
+            else:
+                query = """
+                    DELETE FROM word WHERE word = %s AND class = %s AND type = %s AND operator = %s
+                """
+                parameters = (phrase_to_delete[0], phrase_to_delete[1],
+                              phrase_to_delete[2], phrase_to_delete[3], )
+                queries_parameters.append((query, parameters))
+
+        #Delete place_classtype tables corresponding to class/type which are not on the wiki anymore
+        for table in self.table_phrases_to_delete:
+            query = SQL('DROP TABLE IF EXISTS {}').format(Identifier(table))
+            queries_parameters.append((query, ()))
+
+        with self.db_connection.cursor() as db_cursor:
+            for query, parameters in queries_parameters:
+                db_cursor.execute(query, parameters)
+
     def _convert_php_settings_if_needed(self, file_path):
         """
             Convert php settings file of special phrases to json file if it is still in php format.

test/python/test_tools_import_special_phrases.py

@@ -2,6 +2,7 @@
     Tests for import special phrases methods
     of the class SpecialPhrasesImporter.
 """
+from mocks import MockParamCapture
 from nominatim.errors import UsageError
 from pathlib import Path
 import tempfile
@@ -11,6 +12,72 @@ from nominatim.tools.special_phrases import SpecialPhrasesImporter
 
 TEST_BASE_DIR = Path(__file__) / '..' / '..'
 
+def test_fetch_existing_words_phrases_basic(special_phrases_importer, word_table,
+                                            temp_db_conn):
+    """
+        Check for the fetch_existing_words_phrases() method.
+        It should return special phrase term added to the word
+        table.
+    """
+    with temp_db_conn.cursor() as temp_db_cursor:
+        query ="""
+            INSERT INTO word VALUES(99999, 'lookup_token', 'normalized_word',
+            'class', 'type', null, 0, 'near');
+        """
+        temp_db_cursor.execute(query)
+
+    assert not special_phrases_importer.words_phrases_to_delete
+    special_phrases_importer._fetch_existing_words_phrases()
+    contained_phrase = special_phrases_importer.words_phrases_to_delete.pop()
+    assert contained_phrase == ('normalized_word', 'class', 'type', 'near')
+
+def test_fetch_existing_words_phrases_housenumber(special_phrases_importer, word_table,
+                                                  temp_db_conn):
+    """
+        Check for the fetch_existing_words_phrases() method.
+        It should return nothing as the term added correspond
+        to a housenumber term.
+    """
+    with temp_db_conn.cursor() as temp_db_cursor:
+        query ="""
+            INSERT INTO word VALUES(99999, 'lookup_token', 'normalized_word',
+            'place', 'house', null, 0, 'near');
+        """
+        temp_db_cursor.execute(query)
+
+    special_phrases_importer._fetch_existing_words_phrases()
+    assert not special_phrases_importer.words_phrases_to_delete
+
+def test_fetch_existing_words_phrases_postcode(special_phrases_importer, word_table,
+                                               temp_db_conn):
+    """
+        Check for the fetch_existing_words_phrases() method.
+        It should return nothing as the term added correspond
+        to a postcode term.
+    """
+    with temp_db_conn.cursor() as temp_db_cursor:
+        query ="""
+            INSERT INTO word VALUES(99999, 'lookup_token', 'normalized_word',
+            'place', 'postcode', null, 0, 'near');
+        """
+        temp_db_cursor.execute(query)
+
+    special_phrases_importer._fetch_existing_words_phrases()
+    assert not special_phrases_importer.words_phrases_to_delete
+
+def test_fetch_existing_place_classtype_tables(special_phrases_importer, temp_db_conn):
+    """
+        Check for the fetch_existing_place_classtype_tables() method.
+        It should return the table just created.
+    """
+    with temp_db_conn.cursor() as temp_db_cursor:
+        query = 'CREATE TABLE place_classtype_testclasstypetable()'
+        temp_db_cursor.execute(query)
+
+    special_phrases_importer._fetch_existing_place_classtype_tables()
+    contained_table = special_phrases_importer.table_phrases_to_delete.pop()
+    assert contained_table == 'place_classtype_testclasstypetable'
+
 def test_check_sanity_class(special_phrases_importer):
     """
         Check for _check_sanity() method.
@@ -80,7 +147,7 @@ def test_convert_settings_giving_json(special_phrases_importer):
     assert returned == json_file
 
 def test_process_amenity_with_operator(special_phrases_importer, getorcreate_amenityoperator_funcs,
-                                       word_table, temp_db_conn):
+                                       temp_db_conn):
     """
         Test that _process_amenity() execute well the 
         getorcreate_amenityoperator() SQL function and that
@@ -188,13 +255,72 @@ def test_process_xml_content(temp_db_conn, def_config, special_phrases_importer,
     assert check_amenities_without_op(temp_db_conn)
     assert results[class_test] and type_test in results.values()
 
+def test_remove_non_existent_phrases_from_db(special_phrases_importer, default_phrases,
+                                             temp_db_conn):
+    """
+        Check for the remove_non_existent_phrases_from_db() method.
+
+        It should removed entries from the word table which are contained
+        in the words_phrases_to_delete set and not those also contained
+        in the words_phrases_still_exist set.
+
+        place_classtype tables contained in table_phrases_to_delete should
+        be deleted.
+    """
+    with temp_db_conn.cursor() as temp_db_cursor:
+        to_delete_phrase_tuple = ('normalized_word', 'class', 'type', 'near')
+        to_keep_phrase_tuple = (
+            'normalized_word_exists', 'class_exists', 'type_exists', 'near'
+        )
+        special_phrases_importer.words_phrases_to_delete = {
+            to_delete_phrase_tuple,
+            to_keep_phrase_tuple
+        }
+        special_phrases_importer.words_phrases_still_exist = {
+            to_keep_phrase_tuple
+        }
+        special_phrases_importer.table_phrases_to_delete = {
+            'place_classtype_testclasstypetable_to_delete'
+        }
+
+        query_words = 'SELECT word, class, type, operator FROM word;'
+        query_tables = """
+            SELECT table_name
+            FROM information_schema.tables
+            WHERE table_schema='public'
+            AND table_name like 'place_classtype_%';
+        """
+
+        special_phrases_importer._remove_non_existent_phrases_from_db()
+
+        temp_db_cursor.execute(query_words)
+        words_result = temp_db_cursor.fetchall()
+        temp_db_cursor.execute(query_tables)
+        tables_result = temp_db_cursor.fetchall()
+        assert len(words_result) == 1 and words_result[0] == [
+            'normalized_word_exists', 'class_exists', 'type_exists', 'near'
+        ]
+        assert (len(tables_result) == 1 and
+            tables_result[0][0] == 'place_classtype_testclasstypetable_to_keep'
+        )
+
 def test_import_from_wiki(monkeypatch, temp_db_conn, def_config, special_phrases_importer, placex_table, 
-                          getorcreate_amenity_funcs, getorcreate_amenityoperator_funcs):
+                          getorcreate_amenity_funcs, getorcreate_amenityoperator_funcs, word_table):
     """
         Check that the main import_from_wiki() method is well executed.
         It should create the place_classtype table, the place_id and centroid indexes,
         grand access to the web user and executing the SQL functions for amenities.
     """
+    mock_fetch_existing_words_phrases = MockParamCapture()
+    mock_fetch_existing_place_classtype_tables = MockParamCapture()
+    mock_remove_non_existent_phrases_from_db = MockParamCapture()
+
+    monkeypatch.setattr('nominatim.tools.special_phrases.SpecialPhrasesImporter._fetch_existing_words_phrases',
+                        mock_fetch_existing_words_phrases)
+    monkeypatch.setattr('nominatim.tools.special_phrases.SpecialPhrasesImporter._fetch_existing_place_classtype_tables',
+                        mock_fetch_existing_place_classtype_tables)
+    monkeypatch.setattr('nominatim.tools.special_phrases.SpecialPhrasesImporter._remove_non_existent_phrases_from_db',
+                        mock_remove_non_existent_phrases_from_db)
     monkeypatch.setattr('nominatim.tools.special_phrases.SpecialPhrasesImporter._get_wiki_content', mock_get_wiki_content)
     special_phrases_importer.import_from_wiki(['en'])
 
@@ -206,6 +332,9 @@ def test_import_from_wiki(monkeypatch, temp_db_conn, def_config, special_phrases
     assert check_grant_access(temp_db_conn, def_config.DATABASE_WEBUSER, class_test, type_test)
     assert check_amenities_with_op(temp_db_conn)
     assert check_amenities_without_op(temp_db_conn)
+    assert mock_fetch_existing_words_phrases.called == 1
+    assert mock_fetch_existing_place_classtype_tables.called == 1
+    assert mock_remove_non_existent_phrases_from_db.called == 1
 
 def mock_get_wiki_content(lang):
     """
@@ -305,6 +434,18 @@ def temp_phplib_dir_with_migration():
 
         yield Path(phpdir)
 
+@pytest.fixture
+def default_phrases(word_table, temp_db_cursor):
+    temp_db_cursor.execute("""
+        INSERT INTO word VALUES(99999, 'lookup_token', 'normalized_word',
+        'class', 'type', null, 0, 'near');
+
+        INSERT INTO word VALUES(99999, 'lookup_token', 'normalized_word_exists',
+        'class_exists', 'type_exists', null, 0, 'near');
+
+        CREATE TABLE place_classtype_testclasstypetable_to_delete();
+        CREATE TABLE place_classtype_testclasstypetable_to_keep();""")
+
 @pytest.fixture
 def make_strandard_name_func(temp_db_cursor):
     temp_db_cursor.execute("""