1. We now read traditional CSV files. Since the files do not explicitly
   specify the column format, we have to guess the format of each column
   after importing.
2. For efficient serialization, table data can be pickled/unpickled
   directly.