I decided to create a brand new project to migrate schema and data from mysql (for now) to PostgreSQL.
I've named pg_chamelion as the database from can be anything. As I said I'm starting with mysql but when I'll get an usable product I will write libraries for sqllite and other dbms.
I'm using sqlalchemy, quick and dirty.
I'm not a great fan of those database abstraction layers; they are making developers ignorant about a wonderful language called SQL and generating monsters IMHO.
Anyway, the metadata package is idiots proof, I've written a conect-read-transform library in barely 120 rows and now I'm able to dump the ddl on file or write directly over a postgresql connection.
The next step is to move data across the databases; I'll give freedom of choice, statements or dump reload using copy.
My old script used statements and incredibly worked fine even for medium size databases, but if I want my new library to be usable for the new buzzword, bigdata I need an high performance strategy or this will remain a nice toy.
The project can be pulled here, GPL V3 licence.