Как подсчитать количество дней отсутствия на складе?

Вы можете использовать это:

df["Current_Stock"] = (
    df.assign(Current_Stock=df["Current_Stock"].mask(df["Current_Stock"].le(0)))
    .groupby("SKU")["Current_Stock"]
    .transform(lambda x: x.bfill().isna())
)
df = df.groupby("SKU")["Current_Stock"].sum().reset_index(name = "Days_Out_of_Stock")

  SKU  Days_Out_of_Stock
0   A                  3
1   B                  4
2   C                  0
3   D                  0
4   E                  0
5   F                  1

Определите дни без запаса (le ), затем выполните специальный groupby.agg с обратным cummin , чтобы сохранить только самый последний, и сумму:

out = (df.loc[::-1, 'Current_Stock'].le(0)
         .groupby(df['SKU'])
         .agg(lambda x: x.cummin().sum())
         .reset_index(name='Days_Out_of_Stock')
      )

Вариант с именованной агрегацией:

out = (df.groupby('SKU', as_index=False)
         .agg(**{'Days_Out_of_Stock': ('Current_Stock',
                                       lambda x: x.le(0)[::-1].cummin().sum())})
      )

Выход:

  SKU  Days_Out_of_Stock
0   A                  3
1   B                  4
2   C                  0
3   D                  0
4   E                  0
5   F                  1

Промежуточные, чтобы продемонстрировать логику:

   SKU  Current_Stock Date_Updated  le(0)  rev_cummin
0    A              2    2024/7/26  False       False
1    A              0    2024/7/27   True        True
2    A              0    2024/7/28   True        True
3    A              0    2024/7/29   True        True
4    B              0    2024/7/26   True        True
5    B              0    2024/7/27   True        True
6    B             -1    2024/7/28   True        True
7    B             -1    2024/7/29   True        True
8    C              2    2024/7/26  False       False
9    C             -1    2024/7/27   True       False
10   C              0    2024/7/28   True       False
11   C              1    2024/7/29  False       False
12   D              2    2024/7/26  False       False
13   D              2    2024/7/27  False       False
14   D              3    2024/7/28  False       False
15   D              1    2024/7/29  False       False
16   E              2    2024/7/26  False       False
17   E              0    2024/7/27   True       False
18   E              0    2024/7/28   True       False
19   E              1    2024/7/29  False       False
20   F              2    2024/7/26  False       False
21   F             -1    2024/7/27   True       False
22   F              1    2024/7/28  False       False
23   F              0    2024/7/29   True        True

import pandas as pd
import numpy as np

# Sample data
data = {'id': ['A', 'A', 'A', 'A',
                'B', 'B', 'B', 'B',
                'C', 'C', 'C', 'C',
                'D', 'D', 'D', 'D',
                'E', 'E', 'E', 'E',
                'F', 'F', 'F', 'F'],
        'current_stock': [2, 0, 0, 0,
                          0, 0, -1, -1,
                          2, -1, 0, 1,
                          2, 2, 3, 1,
                          2, 0, 0, 1,
                          2, -1, 1, 0],
        'date_updated': ['1990/7/26', '1990/7/27', '1990/7/28', '1990/7/29',
                         '1990/7/26', '1990/7/27', '1990/7/28', '1990/7/29',
                         '1990/7/26', '1990/7/27', '1990/7/28', '1990/7/29',
                         '1990/7/26', '1990/7/27', '1990/7/28', '1990/7/29',
                         '1990/7/26', '1990/7/27', '1990/7/28', '1990/7/29',
                         '1990/7/26', '1990/7/27', '1990/7/28', '1990/7/29']}
df = pd.DataFrame(data)
df['date_updated'] = pd.to_datetime(df['date_updated'])

def f(df):
   
    
    # Create a boolean mask for rows where current stock is less than 1 (out of stock)
    out_of_stock_mask = df['current_stock'].values < 1
    
    # Create a boolean mask for rows where current stock is greater than or equal to 1 (in stock)
    in_stock_mask = df['current_stock'].values >= 1
    
    # If all rows are out of stock, return the total number of rows
    if out_of_stock_mask.all():
        return out_of_stock_mask.sum()
    
    # Find the index of the last row where stock was in stock
    last_restock_index = np.where(in_stock_mask)[0][-1]
    
    # Calculate the number of days the stock was out of stock after the last restock
    out_of_stock_total_days = out_of_stock_mask[last_restock_index + 1:].sum()  
    
    return out_of_stock_total_days

# Group the DataFrame by 'id' and apply the function 'f' to each group
res = df.groupby('id').apply(f,include_groups =False).reset_index(name='Days_Out_of_Stock')
print(res)
'''
  id  Days_Out_of_Stock
0  A                  3
1  B                  4
2  C                  0
3  D                  0
4  E                  0
5  F                  1

'''