added fix for non utf8 query event

[mastak]

I faced with this #132 problem.

Does this solution have a chance to live? (=

[julien-duponchelle]

What is the schema of the column creating the issue?

[bjoernhaeuser]

@mastak can you add a test which show cases the error (and the fix)?

[mastak]

Sorry, guys, that I'm so slowly.
The column is simple varchar, the table has latin-1 charset.
I wrote some test, but it is work only with STATEMENT based replication, and I don't know how it can be enabled. Can you help with it?

    def test_encoding_latin1_statement(self):
        db = copy.copy(self.database)
        db["charset"] = "latin1"
        self.connect_conn_control(db)

        create_query = "CREATE TABLE test (test CHAR(12)) CHARACTER SET latin1 COLLATE latin1_bin;"
        insert_query = b"INSERT INTO test VALUES('\x96')"

        self.execute(create_query)
        self.execute(insert_query)
        self.execute("COMMIT")

        self.assertIsInstance(self.stream.fetchone(), RotateEvent)
        self.assertIsInstance(self.stream.fetchone(), FormatDescriptionEvent)
        # QueryEvent for the Create Table
        self.assertIsInstance(self.stream.fetchone(), QueryEvent)

        # QueryEvent for the BEGIN
        self.assertIsInstance(self.stream.fetchone(), QueryEvent)

        event = self.stream.fetchone()
        self.assertIsInstance(event, QueryEvent)

        self.assertEqual(event.query, insert_query.decode("latin-1"))

[mastak]

Could we do something like:
at the begin of the test SET GLOBAL binlog_format = 'STATEMENT';
at the end of the test SET GLOBAL binlog_format = 'ROW';
?

[julien-duponchelle]

Clearly the fix work but will break it's non utf-8 and non latin1.

But what do you think about @baloo solution in #159 to just return a byte object so you can handle it in your code?

[mastak]

Sure, it would be good.

[mastak]

We can't return bytes (even only in case with error) because it will break existing systems.

But what if add some method like:

def _decode_query(self, query)
    return query.decode("utf-8")

and then everyone can overwrite it. What do you think?