Pandas replace column value with Last available value

user_10083382 0
Viewed 83

For sample data frame that can be derived using code below, I want to update the column Offset_Date such that for any date in column Offset_Date that is not within column Date I want to replace that date in Offset_Date with last available value in column Date.

data = {"date": ['2021-01-01', '2021-01-03', '2021-01-04', '2021-01-05',
                 '2021-01-07', '2021-01-09', '2021-01-10', '2021-01-11'],

        "offset_date": ['2021-01-02', '2021-01-04', '2021-01-05',
                        '2021-01-06', '2021-01-08', '2021-01-10',
                        '2021-01-11', '2021-01-12']}

test_df = pd.DataFrame(data)
test_df['date'] = pd.to_datetime(test_df['date'])
test_df['offset_date'] = pd.to_datetime(test_df['offset_date'])

To explain further in 1st row of above data frame date 2021-01-02 is not within column date so I want to replace that value with last available value in column date i.e. 2021-01-01.

I want to perform a vectorized approach so I tried the following, which lead to incorrect results.

test_df['offset_date_upd'] = np.where(test_df['offset_date'] in test_df['date'].values,
                                      test_df['offset_date'], 
                                      test_df[test_df['date'] <= test_df['offset_date']].values.max())

How can I get the below desired output using a vectorized approach?

Desired Output

+------------+-------------+
|    Date    | Offset_Date |
+------------+-------------+
| 2021-01-01 | 2021-01-01  |
| 2021-03-01 | 2021-04-01  |
| 2021-04-01 | 2021-05-01  |
| 2021-05-01 | 2021-05-01  |
| 2021-07-01 | 2021-07-01  |
| 2021-09-01 | 2021-10-01  |
| 2021-10-01 | 2021-11-01  |
| 2021-11-01 | 2021-11-01  |
+------------+-------------+
pandas vectorization conditional-statements date python
3 Answers

IIUC, you can use isin, mask, and fillna:

test_df['offset_date'] = (test_df['offset_date']
                          .where(test_df['offset_date'].isin(test_df['date']),
                                 test_df['date'])
                         )

output:

        date offset_date
0 2021-01-01  2021-01-01
1 2021-01-03  2021-01-04
2 2021-01-04  2021-01-05
3 2021-01-05  2021-01-05
4 2021-01-07  2021-01-07
5 2021-01-09  2021-01-10
6 2021-01-10  2021-01-11
7 2021-01-11  2021-01-11
user_163434640

The bellow approach should works for your case

test_df["offset_date"] = np.where(
    test_df.offset_date.isin(test_df.date),
    test_df.offset_date,
    test_df.date
)
user_98922080

This is the purpose of Pandas' merge_asof function.
We have to be specific about which columns are going where. This will work as a left join and in this case, we want 'offset_date' to represent the "left". Then, for each value in 'offset_date', we look for the greatest value in 'date' that does not exceed that value in 'offset_date'.

The one gotcha in this approach is that both columns need to be sorted. If this is an issue with the real data, let me know and I'll add a section at the bottom that deals with this.

new_offset_date = pd.merge_asof(
    test_df[['offset_date']],  # limit `test_df` to just the column I need for the merge
    test_df[['date']],         # limit `test_df` to just the other column I need
    left_on=['offset_date'],   # name the columns since they have different names
    right_on=['date']          # name the other column as well
)['date']

Now the new dates are in a pandas.Series, we can see it if we use the assign method which DOES NOT overwrite your dataframe and you'll need to use test_df = test_df.assign(offset_date=new_offset_date) to actually persist the new dates in the dataframe.

test_df.assign(offset_date=new_offset_date)

        date offset_date
0 2021-01-01  2021-01-01
1 2021-01-03  2021-01-04
2 2021-01-04  2021-01-05
3 2021-01-05  2021-01-05
4 2021-01-07  2021-01-07
5 2021-01-09  2021-01-10
6 2021-01-10  2021-01-11
7 2021-01-11  2021-01-11
user_23366540
Related