Does .loc[...] behave naturally?

[kennethshsu]

I've been thinking about this for quite a while, but I also don't know if something is actually wrong. However, I do feel that when .loc[...] is called, the behavior is not natural.

Look at this example:

import chainladder as cl
abc = cl.load_sample("abc")

abc.to_frame().loc["1981",24]
abc.loc[:,:,"1981",24]

It's true that both chainladder and pandas's behaviors are identical, but why is the year quoted and the age not? 24 (when not quote), feels more like an index to me.

[jbogaardt]

I think this has to do with the dtype of the origin axis. It is a PeriodIndex and that is how it's design in pandas. If we were to use a datetime (like the valuation axis) it would be not need quotes I believe.

As an aside I find that PeriodIndex, bring a less prominent dtype leads to confusion in general. Periods also don't have a direct mapping in the arrow standard so moving to a pure date/datetime dtype might be a sensible move eventually.