CODEWITHSUNDEEP
pythonpandasdatetime
Shota
Shota

Reputation: 21

Date columns and integer columns addition give error TypeError: unsupported type for timedelta days component: Series

I want to add columns startDate and x, this columns dtype is int64. Below is my code. B['startDate'] columsn dtype is datetime64[ns]

B['startDate']+datetime.timedelta(days=(30*B['x']))

But it gives me the following error:

TypeError: unsupported type for timedelta days component: Series

Upvotes: 0

Views: 177

Answers (2)

jezrael
jezrael

Reputation: 863116

Use to_timedelta here if performance is important:

df['newDate'] = df['startDate'] + pd.to_timedelta(df['x'] * 30, unit='d')
print (df)
   startDate   x    newDate
0 2020-01-01  86 2027-01-24
1 2020-01-02  58 2024-10-07
2 2020-01-03   1 2020-02-02
3 2020-01-04  95 2027-10-24
4 2020-01-05  87 2027-02-27
5 2020-01-06  45 2023-09-17
6 2020-01-07  63 2025-03-11
7 2020-01-08  92 2027-07-30
8 2020-01-09  30 2022-06-27
9 2020-01-10  22 2021-10-31

For 1k rows it is 8times faster like loop solution with map:

np.random.seed(2021)

N = 1000
df = pd.DataFrame({'startDate': pd.date_range('2020-1-1', periods=N), 
                   'x': np.random.randint(1, 100, N),})
print (df)

In [97]: %timeit df['startDate'] + pd.to_timedelta(df['x'] * 30, unit='d')
505 µs ± 18.5 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

In [98]: %timeit df['startDate'] + df['x'].map(lambda x: datetime.timedelta(days=30 * x))
4.01 ms ± 450 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

Upvotes: 1

fsl
fsl

Reputation: 3280

You have to modify your code slightly so that timedelta takes a valid days argument. Here is a toy example:

B = pd.DataFrame({'startDate': pd.date_range('2020-1-1', periods=10), 'x': np.random.randint(1, 100, 10),})                                                                                                                            
    
B['newDate'] = B['startDate'] + B['x'].map(lambda x: datetime.timedelta(days=30 * x))

Output:

   startDate   x    newDate
0 2020-01-01  89 2020-03-30
1 2020-01-02  69 2020-03-11
2 2020-01-03  84 2020-03-27
3 2020-01-04  73 2020-03-17
4 2020-01-05  34 2020-02-08
5 2020-01-06  63 2020-03-09
6 2020-01-07  17 2020-01-24
7 2020-01-08  12 2020-01-20
8 2020-01-09  15 2020-01-24
9 2020-01-10  28 2020-02-07

Upvotes: 0

Related Questions