Fixed comments handling

mysql_ch_replicator/converter.py

@@ -1075,8 +1075,51 @@ def convert_drop_table_query(self, mysql_query):
         raise Exception('not implement')
 
     def _strip_comments(self, create_statement):
-        pattern = r'\bCOMMENT(?:\s*=\s*|\s+)([\'"])(?:\\.|[^\\])*?\1'
-        return re.sub(pattern, '', create_statement, flags=re.IGNORECASE)
+        """
+        Strip COMMENT clauses from CREATE TABLE statements.
+        Handles MySQL-style quote escaping where quotes are doubled ('' or "").
+        """
+        result = []
+        i = 0
+        while i < len(create_statement):
+            # Look for COMMENT keyword (case insensitive)
+            if (i + 7 < len(create_statement) and 
+                create_statement[i:i+7].upper() == 'COMMENT' and
+                (i == 0 or not create_statement[i-1].isalnum()) and
+                (i + 7 >= len(create_statement) or not create_statement[i+7].isalnum())):
+
+                # Skip COMMENT keyword
+                i += 7
+
+                # Skip whitespace and optional '='
+                while i < len(create_statement) and create_statement[i].isspace():
+                    i += 1
+                if i < len(create_statement) and create_statement[i] == '=':
+                    i += 1
+                    while i < len(create_statement) and create_statement[i].isspace():
+                        i += 1
+
+                # Find the quoted string
+                if i < len(create_statement) and create_statement[i] in ('"', "'"):
+                    quote_char = create_statement[i]
+                    i += 1  # Skip opening quote
+
+                    # Find the closing quote, handling escaped quotes
+                    while i < len(create_statement):
+                        if create_statement[i] == quote_char:
+                            # Check if this is an escaped quote (doubled)
+                            if i + 1 < len(create_statement) and create_statement[i + 1] == quote_char:
+                                i += 2  # Skip both quotes
+                            else:
+                                i += 1  # Skip closing quote
+                                break
+                        else:
+                            i += 1
+            else:
+                result.append(create_statement[i])
+                i += 1
+
+        return ''.join(result)
 
     def parse_mysql_table_structure(self, create_statement, required_table_name=None):
         create_statement = self._strip_comments(create_statement)

test_mysql_ch_replicator.py

@@ -492,13 +492,13 @@ def test_multi_column_erase():
 
     prepare_env(cfg, mysql, ch)
 
-    mysql.execute(f'''
+    mysql.execute(f"""
 CREATE TABLE `{TEST_TABLE_NAME}` (
-    departments int(11) NOT NULL,
-    termine int(11) NOT NULL,
+    departments int(11) NOT NULL COMMENT '事件类型，可选值: ''SYSTEM'', ''BUSINESS''',
+    termine int(11) NOT NULL COMMENT '事件类型，可选值: ''SYSTEM'', ''BUSINESS''',
     PRIMARY KEY (departments,termine)
 )
-''')
+""")
 
 
     mysql.execute(f"INSERT INTO `{TEST_TABLE_NAME}` (departments, termine) VALUES (10, 20);", commit=True)
@@ -3082,3 +3082,143 @@ def test_resume_initial_replication_with_ignore_deletes():
     finally:
         # Clean up temp config file
         os.unlink(config_file)
+
+
+@pytest.mark.parametrize("input_sql,expected_output", [
+    # Basic single quote comment
+    (
+        "CREATE TABLE test (id int NOT NULL COMMENT 'Simple comment', name varchar(255))",
+        "CREATE TABLE test (id int NOT NULL , name varchar(255))"
+    ),
+    # Basic double quote comment
+    (
+        "CREATE TABLE test (id int NOT NULL COMMENT \"Simple comment\", name varchar(255))",
+        "CREATE TABLE test (id int NOT NULL , name varchar(255))"
+    ),
+    # Comment with escaped single quotes (the original bug case)
+    (
+        "CREATE TABLE test (id int NOT NULL COMMENT '事件类型，可选值: ''SYSTEM'', ''BUSINESS''', name varchar(255))",
+        "CREATE TABLE test (id int NOT NULL , name varchar(255))"
+    ),
+    # Comment with escaped double quotes
+    (
+        "CREATE TABLE test (id int NOT NULL COMMENT \"Value can be: \"\"ACTIVE\"\" or \"\"INACTIVE\"\"\", name varchar(255))",
+        "CREATE TABLE test (id int NOT NULL , name varchar(255))"
+    ),
+    # Multiple comments in same table
+    (
+        """CREATE TABLE test (
+            id int NOT NULL COMMENT 'Primary key',
+            name varchar(255) COMMENT 'User name',
+            status enum('active','inactive') COMMENT 'Status with ''quotes'''
+        )""",
+        """CREATE TABLE test (
+            id int NOT NULL ,
+            name varchar(255) ,
+            status enum('active','inactive') 
+        )"""
+    ),
+    # Comment with COMMENT = syntax
+    (
+        "CREATE TABLE test (id int NOT NULL COMMENT = 'Primary key', name varchar(255))",
+        "CREATE TABLE test (id int NOT NULL , name varchar(255))"
+    ),
+    # Comment with mixed quotes and special characters
+    (
+        "CREATE TABLE test (id int COMMENT 'Mixed: ''single'', \"double\", and `backtick`', name text)",
+        "CREATE TABLE test (id int , name text)"
+    ),
+    # Multiline comment
+    (
+        """CREATE TABLE test (
+            id int NOT NULL COMMENT 'This is a
+            multiline comment
+            with newlines',
+            name varchar(255)
+        )""",
+        """CREATE TABLE test (
+            id int NOT NULL ,
+            name varchar(255)
+        )"""
+    ),
+    # Comment with Unicode characters
+    (
+        "CREATE TABLE test (id int COMMENT '用户ID - 主键', name varchar(255) COMMENT 'Имя пользователя')",
+        "CREATE TABLE test (id int , name varchar(255) )"
+    ),
+    # No comments (should remain unchanged)
+    (
+        "CREATE TABLE test (id int NOT NULL, name varchar(255))",
+        "CREATE TABLE test (id int NOT NULL, name varchar(255))"
+    ),
+    # Comment at table level
+    (
+        "CREATE TABLE test (id int NOT NULL, name varchar(255)) COMMENT 'Table comment'",
+        "CREATE TABLE test (id int NOT NULL, name varchar(255)) "
+    ),
+    # Complex case with multiple escaped quotes and special characters
+    (
+        """CREATE TABLE test (
+            departments int(11) NOT NULL COMMENT '事件类型，可选值: ''SYSTEM'', ''BUSINESS''',
+            termine int(11) NOT NULL COMMENT '事件类型，可选值: ''SYSTEM'', ''BUSINESS''',
+            PRIMARY KEY (departments,termine)
+        )""",
+        """CREATE TABLE test (
+            departments int(11) NOT NULL ,
+            termine int(11) NOT NULL ,
+            PRIMARY KEY (departments,termine)
+        )"""
+    ),
+    # Comment with JSON-like content
+    (
+        "CREATE TABLE test (config json COMMENT '{\"type\": \"config\", \"values\": [\"a\", \"b\"]}', id int)",
+        "CREATE TABLE test (config json , id int)"
+    ),
+    # Comment with SQL injection-like content (should be safely handled)
+    (
+        "CREATE TABLE test (id int COMMENT 'DROP TABLE users; --', name varchar(255))",
+        "CREATE TABLE test (id int , name varchar(255))"
+    ),
+    # Empty comment
+    (
+        "CREATE TABLE test (id int COMMENT '', name varchar(255))",
+        "CREATE TABLE test (id int , name varchar(255))"
+    ),
+    # Comment with only spaces
+    (
+        "CREATE TABLE test (id int COMMENT '   ', name varchar(255))",
+        "CREATE TABLE test (id int , name varchar(255))"
+    ),
+    # Case insensitive COMMENT keyword
+    (
+        "CREATE TABLE test (id int comment 'lowercase', name varchar(255) Comment 'Mixed case')",
+        "CREATE TABLE test (id int , name varchar(255) )"
+    ),
+])
+def test_strip_comments_function(input_sql, expected_output):
+    """
+    Test the _strip_comments function with various realistic scenarios.
+
+    This test covers:
+    - Basic single and double quoted comments
+    - Escaped quotes within comments (MySQL style with doubled quotes)
+    - Multiple comments in the same table
+    - COMMENT = syntax
+    - Multiline comments with newlines
+    - Unicode characters in comments
+    - Table-level comments
+    - Complex real-world scenarios
+    - Edge cases like empty comments and case variations
+    """
+    from mysql_ch_replicator.converter import MysqlToClickhouseConverter
+
+    converter = MysqlToClickhouseConverter()
+    result = converter._strip_comments(input_sql)
+
+    # Normalize whitespace for comparison (remove extra spaces that might be left behind)
+    def normalize_whitespace(text):
+        import re
+        # Replace multiple spaces with single space, but preserve newlines
+        return re.sub(r'[ \t]+', ' ', text).strip()
+
+    assert normalize_whitespace(result) == normalize_whitespace(expected_output), f"Failed for input: {input_sql}"