metadata pickling regression

[Mike Bayer]

from sqlalchemy import *
import pickle

metadata = MetaData()

Table('a',metadata,
     Column('id',Integer,primary_key=True),
     Column('x', Integer, Sequence("x_seq"), ),
     )

metadata2 = pickle.loads(pickle.dumps(metadata))
Table('a',metadata2,
     Column('id',Integer,primary_key=True),
     Column('x', Integer, Sequence("x_seq"), ),
     useexisting=True)

[Mike Bayer]

93da3087f1a068018a3c5a67f61dbbfa5178bdf0 , but some more thinking in de529fb81cc80cd49666a80ef60de6015ff81c1c . We change _sequences to be a dict which keys on the name/schema of the Sequence itself, and also support that the Sequence is present in _sequences from an unpickle regardless of if it was explicitly placed there. The last Sequence associated with the Table wins. The previous one is silently replaced, not unlike when Column is specified.

[Mike Bayer]

Removing milestone: 0.7.0 (automated comment)