pd.to/read_sql_table silently corrupts Categorical columns #8624
Description
In [29]: from sqlalchemy import create_engine
...: engine = create_engine('sqlite://')
...: df=pd.DataFrame([[1,'John P. Doe'],[2,'Jane Dove'],[1,'John P. Doe']],
...: columns=['person_id','person_name'])
...: df.to_sql('data1',engine)
...: df['person_name']=pd.Categorical(df.person_name)
...: df.to_sql('data2',engine)
...: print pd.read_sql_table('data1',engine)
...: print pd.read_sql_table('data2',engine)
index person_id person_name
0 0 1 John P. Doe
1 1 2 Jane Dove
2 2 1 John P. Doe
index person_id person_name
0 0 1 J
1 1 2 o
2 2 1 husing relational db to store catagorical columns in seperare tables would be very cool, and rebuilding the frame in pandas by JOIN from the multiple tables would save time on the wire. also memory if categorical was build directly.