SQLAlchemy reports "Invalid utf8mb4 character string" for BINARY column - sqlalchemy

Question

Assuming this MySQL table schema:
CREATE TABLE `user` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`uuid` binary(16) NOT NULL,
`email` varchar(255) NOT NULL,
`name` varchar(255) DEFAULT NULL,
`photo` binary(16) DEFAULT NULL,
PRIMARY KEY (`id`),
UNIQUE KEY `uuid` (`uuid`),
UNIQUE KEY `email` (`email`)
) ENGINE=InnoDB AUTO_INCREMENT=8 DEFAULT CHARSET=utf8mb4;
When I use the execute() API from SQLAlchemy connection class as such:
with self.engine.begin() as connection:
user_uuid = uuid.UUID("...")
result = connection.execute("SELECT email, name, photo FROM user WHERE uuid=%s", user_uuid.bytes)
If the UUID is F393A167-A919-4B50-BBB7-4AD356E89E6B, then SQLAlchemy prints this warning:
/site-packages/sqlalchemy/engine/default.py:450: Warning: Invalid utf8mb4 character string: 'F393A1'
The uuid column is a BINARY column, so why is SQLAlchemy considering this parameter a text one instead of a binary one and how to prevent this?

Answer 1

The explanation and solution is actually in this bug report in MySQL:
replace:
cursor.execute("""
INSERT INTO user (uuid)
VALUES (%s)
""", my_uuid)
with
cursor.execute("""
INSERT INTO user (uuid)
VALUES (_binary %s)
""", my_uuid)
Mind the underscore. It's "_binary", not "binary".
This "_binary" tells MySQL that the following string is to be interpreted as binary, not to be interpreted/validated as utf8.

Answer 2

The problem doesn't happen on Python 3, so I think that the problem is that the database driver is unable to distinguish what should be bytes given the Python 2 str type.
Regardless, it seems using SQLAlchemy core directly works correctly, presumably because it knows the column type directly.
from sqlalchemy import MetaData, Table, select
meta = MetaData()
user = Table('user', meta, autoload_with=engine)
result = select([user]).where(user.c.uuid == user_uuid.bytes)
If you wish to continue executing a string, you can cast to bytesarray like SQLAlchemy appears to be doing:
with self.engine.begin() as connection:
user_uuid = uuid.UUID("...")
result = connection.execute(
"SELECT email, name, photo FROM user WHERE uuid=%s",
bytearray(user_uuid.bytes))
Or to tell SQLAlchemy what type the bound parameter is to get this automatically:
from sqlalchemy import text, bindparam, BINARY
with self.engine.begin() as connection:
user_uuid = uuid.UUID("...")
stmt = text("SELECT email, name, photo FROM user WHERE uuid = :uuid")
stmt = stmt.bindparams(bindparam('uuid', user_uuid.bytes, type_=BINARY))
result = connection.execute(stmt)