python - 根据另一列向日期添加一天

Question

我有一个问题，我正在组合两个数据集，它们对午夜有不同的定义。因此，我想为每次出现的午夜向其中一个数据集添加一天，以便它们都遵循相同的日期设置。

我按以下方式安排了我的日期和时间:

df['Date'] = pd.to_datetime(df['Date']).dt.strftime('%d/%m/%Y')
df['Hour'] = pd.to_datetime(df['Hour']).dt.strftime('%H:%M')

然后我试图修改任何出现的 00:00:00 以使其晚一天:

df.loc['Hour' == '00:00:00', 'Date'] = pd.DatetimeIndex(df.Date) + timedelta(days=1)

但是我不断提出一个 KeyError:

raise KeyError("cannot use a single bool to index into setitem")
KeyError: 'cannot use a single bool to index into setitem

如有任何帮助，我们将不胜感激。

目标:输入:

Date        |  Hour
---------------------
19/06/2016  |  23:30
19/06/2016  |  23:45
19/06/2016  |  00:00
20/06/2016  |  00:15
20/06/2016  |  00:30

输出:

    Date        |  Hour
    ---------------------
    19/06/2016  |  23:30
    19/06/2016  |  23:45
    20/06/2016  |  00:00
    20/06/2016  |  00:15
    20/06/2016  |  00:30

Answer 1

您可以使用 mask检查午夜并添加一天:

f['Date'] = pd.to_datetime(f['Date'])
m = f['Hour'] == '00:00'
f['Date'] = f['Date'].mask(m, f['Date'] + pd.Timedelta(1, unit='d')).dt.strftime('%d/%m/%Y')

loc 的解决方案:

m = f['Hour'] == '00:00'
dates = pd.to_datetime(f['Date'])
f.loc[m, 'Date'] = (dates + pd.Timedelta(1, unit='d')).dt.strftime('%d/%m/%Y')
#alternative
#f.loc[m, 'Date'] = (dates[m] + pd.Timedelta(1, unit='d')).dt.strftime('%d/%m/%Y')

解决方案 numpy.where :

m = f['Hour'] == '00:00'
dates = (pd.to_datetime(f['Date']) + pd.Timedelta(1, unit='d')).dt.strftime('%d/%m/%Y')
f['Date'] = np.where(m, dates, f['Date']) 

---

print (f)
         Date   Hour
0  19/06/2016  23:30
1  19/06/2016  23:45
2  20/06/2016  00:00
3  20/06/2016  00:15
4  20/06/2016  00:30

代码内:

for fname in glob.glob(path):
    fname = fname.replace(r'\2016', '/2016')
    f = pd.DataFrame(pd.read_csv(fname))
    f = f.replace({'Hour': {'24:00:00': '00:00'}})
    f['Date'] = pd.to_datetime(f['Date']).dt.strftime('%d/%m/%Y')
    f['Hour'] = pd.to_datetime(f['Hour']).dt.strftime('%H:%M')

    m = f['Hour'] == '00:00'
    dates = (pd.to_datetime(f['Date']) + pd.Timedelta(1, unit='d')).dt.strftime('%d/%m/%Y')
    f['Date'] = np.where(m, dates, f['Date'])

    print(fname)
    if a == 0:
        f_2016['Date'] = f['Date']
        f_2016['Hour'] = f['Hour']
        a = 1
    f_2016 = pd.merge(f_2016, f, on=['Date', 'Hour'])
    print(pd.DataFrame.head(f_2016, n=100))