用相同的值填充列，但组中的第一行除外

Question

I have a df:

import pandas as pd
df = pd.DataFrame({'user_id': [1,1,2,1,2,1,2,3], 'movie_id': ['35','120','898','546','989','42','546','35'], 
'time':['1.7','2.1','1.3','2.4','1.4','7.0','2.1','1.1']})

that looks like this:

user_id  movie_id  time
1          35      1,7
1         120      2.1
2         898      1.3
1         546      2.4
2         989      1.4
1         42       7.0
2         546      2.1
3         35.      1.1

my goal is to group by user_id, sort by time and fill new column with '1' except 1st row inside each group - column 'time' displays number of seconds that has been elapsed from the last click. eventually I should get such output with indicators for the last movie the user has rated before the active one:

user_id  movie_id  time  last_rated
1          35      1.7      0
1         120      2.1      1
2         898      1.3      0
1         546      2.4      1
2         989      1.4      1
1         42       7.0      1
2         546      2.1      1
3         35       1.1      0

I've experimented with group_by, shift, cumsum but still can't get the desired output.. any help would be very appreciated!

Answer 1

Can use cumcount() and np.where()

df['last_rated'] = np.where(df.groupby('user_id').cumcount() == 0, 0, 1)

or (as per @coldspeed below)

df.groupby('user_id').cumcount().astype(bool).astype(int)

Outputs

    user_id   movie_id  time    last_rated
0   1         35          1.7   0
1   1         120         2.1   1
2   2         898         1.3   0
3   1         546         2.4   1
4   2         989         1.4   1
5   1         42          7.0   1
6   2         546         2.1   1
7   3         35          1.1   0

---

You can use sort_values upfront to assure you have your sorted condition right. But if you want to keep your df as is, you can sort inside the groups:

g = df.groupby('user_id', as_index=False).apply(lambda x: x.sort_values(by='time')).groupby('user_id').cumcount().reset_index(level=0,drop=True)

df['l'] = (g/g).fillna(0)

Answer 2

You can use GroupBy + transform with min to calculate a series of minimum values by user_id . Then check for equality against df['time'] and convert from bool to int .

g = df.groupby('user_id')['time'].transform('min')
df['last_rated'] = (df['time'] != g).astype(int)

Assuming your dataframe is already sorted by time for each user_id , you can more efficiently use GroupBy with 'first' :

g = df.groupby('user_id')['time'].transform('first')

Result:

print(df)

   user_id movie_id time  last_rated
0        1       35  1.7           0
1        1      120  2.1           1
2        2      898  1.3           0
3        1      546  2.4           1
4        2      989  1.4           1
5        1       42  7.0           1
6        2      546  2.1           1
7        3       35  1.1           0