Pandas Reverse Group来自 Cumprod

Question

I am trying to set a column to be the reverse cumprod of a series after grouping on another series in a data frame.

Attempt 1:
temp_data["factor_price"] = temp_data["factor_price"].groupby("stock_key").cumprod()[::-1]
Attempt 2:
temp_data["factor_price"] = temp_data.groupby("stock_key")["factor_price"].cumprod()[::-1]
Attempt 3: 
temp_data["factor_price"] = temp_data.groupby("stock_key")["factor_price"][::-1].cumprod()

input: temp_data = pd.DataFrame([x, x, x, y, y, y], [1, 2, 3, 0, 1, 2], columns=[
"stock_key", "factor_price"])

Output: Pd.Series(6, 6, 3, 0, 2, 2)

I have been researching a lot, and I think I could do it if I looped through the groups, but this is an exceptionally large data set and I would like to keep the code as simple as possible. Is there a chance this could be a one liner?

Thanks!

Answer 1

so I needed to apply a custom function that did the cumprod, and then convert everything to a list because of the incompatible indices with the original data frame.

temp_data["factor_price"] = temp_data.groupby("stock_key")["factor_price"].apply(lambda x: 
x[::-1].cumprod()[::-1]).to_list()

Answer 2

You can do it without a lambda function as -

temp_data[::-1].groupby('stock_key')['factor_price'].cumprod()[::-1]

The first [::-1] reverses the row order of the dataset for expected cumprod() and the second one reverses it back to get back the original order of indices.