Came up on discourse. Someone has a DataFrame,

df = DataFrame(y = [1, 2, 3, 4, 5, 6])

and they want to add a new column z where z = rolling(sum, df.y, 2).

Unfortunately, this gives a 5 element vector, so you can't use transform from DataFrames.

Can we add a keyword argument to add padding? Could hurt performance because it would change the type of output... but could be a good option to have.