Is it a good idea to bann the .dt accessor from Series[Any]?

[cmp0xff]

In practice we have many usages like data["timetamp"].dt.date, or data["timetamp"] - data["timestamp_2"]).dt.floor(), where data is a DataFrame. With the most current 251201, we need to cast every and each time.

[Dr-Irv]

It's not clear to me where the cast is required in your examples.

By requiring the cast, it is forcing the user to tell us that data["timestamp"] is datetime data. I think this helps people make sure they are not doing something like data["floatdata"].dt.date which would fail at runtime (assuming floatdata is floating point data).