SQLAlchemy says "The table already exists" even though it is not

I would like to use SQLAlchemy ORM not only to query the database but also to create tables. So I am connecting to an empty schema in my database:

engine = create_engine(connection_string)
Base = declarative_base()

      

        
        
        
      

    

I have several classes that are similar to the following example:

class SomeClass(Base):
    __tablename__= 'SomeClass'
    id = Column(Integer, primary_key=True, auto_increment=True)
    name = Column(String(50))

    def __init__(self, name):
        self.name = name

      

        
        
        
      

    

To create the database, I call:

Base.metadata.create_all(engine)

      

        
        
        
      

    

This works great - FIRST time. If I now delete the tables manually from the database OR by calling Base.metadata.drop_all()

and restarting the script (hence calling create_all

again), I get an error that my tables already exist:

sqlalchemy.exc.OperationalError: (OperationalError) (1050, "Table 'someclass' already exists") 

      

        
        
        
      

    

I can clearly say that it is not. I can create a table with the same name using CREATE TABLE from my CLI tool and also, if I now stop the database daemon and start it again, Iwill be able to create tables again using create_all

.

So, I think it has something to do with the connection between SQLAlchemy and the database. Obviosly SQLAlchemy thinks the table still exists after I dumped it, which is not correct. But if I try to query this none existing table, it tells me that the table does not exist.

I googled around but couldn't find an explanation for this behavior. I use SQLAlchemy too often, so maybe I'm just not deep enough in this to understand the problem.

Anyway, I'd appreciate a hint.

I am using Python 2.7.3 and MySQL 5.1.47 and SQLAlchemy 0.8.0b2

PS: Using Postgres I have never suffered from such a problem.