Python 按组计算日期差，第一行考虑初年的第一天。

我有一个数据框架，我需要按类型和年份分组，以计算组内的日期差异

输入

我已经尝试了下面的解决方案，它给了我DIFF的行，但是我想考虑第一组中的第一个日期在组

df1['date'] = pd.to_datetime(df1['date'])
df1['DateDiff'] = df1.groupby(['type','year']).date.diff().fillna(0)

我能够实现以下输出，这并不能解决我需要计算第一行值与一年中第一天的差值的需求

    type       date  year DateDiff
0  type1 2017-03-30  2017   0 days
1  type1 2017-05-10  2017  41 days
2  type1 2017-12-15  2017 219 days
3  type1 2018-01-15  2018   0 days
4  type1 2018-05-01  2018 106 days
5  type3 2018-01-30  2018   0 days
6  type3 2018-06-27  2018 148 days
7  type3 2019-03-20  2019   0 days
8  type3 2019-05-21  2019  62 days

期望输出为：

    type       date  year DateDiff
0  type1 2017-03-30  2017  88 days ---- (2017 - 01- 01) - (2017 - 03 - 30)
1  type1 2017-05-10  2017  41 days ---- (2017 - 03 - 30) - (2017-05-10)
2  type1 2017-12-15  2017 219 days
3  type1 2018-01-15  2018   14 days ---- (2018-01-01) - (2018-01-15)
4  type1 2018-05-01  2018 106 days
5  type3 2018-01-30  2018   29 days
6  type3 2018-06-27  2018 148 days
7  type3 2019-03-20  2019   78 days
8  type3 2019-05-21  2019  62 days

---

可能还有更优雅的方式，但请参见以下我的尝试：

import pandas as pd

df = pd.DataFrame({'type': ['type1', 'type1', 'type1','type1','type1','type3','type3','type3','type3'],
                   'date': ['2017-3-30','2017-5-10','2017-12-15','2018-01-15','2018-05-01','2018-01-30','2018-06-27','2019-03-20','2019-05-21'],
                   'year': [2017, 2017, 2017,2018,2018,2018,2018,2019,2019]})

df['date'] = pd.to_datetime(df['date'])
df = df.sort_values(by='date')
result_lst = []
for year_type, sub_df in df.groupby(['year', 'type']):
    year, type = year_type
    sub_df['shift'] = sub_df['date'].shift(1)
    sub_df.loc[sub_df.index[0], 'shift'] = pd.to_datetime(str(year), format='%Y')
    sub_df['DateDiff'] = sub_df['date'] - sub_df['shift']
    sub_df['year'] = year
    sub_df['type'] = type
    sub_df = sub_df.drop(columns=['shift'])
    result_lst.append(sub_df)
df = pd.concat(result_lst, axis=0)
print(df)

---

结果与您的目标一致。

如果您同意答案，请将其标记为正确，如果遗漏任何内容，请进行评论

import pandas as pd

df = pd.DataFrame({'type': ['type1', 'type1', 'type1','type1','type1','type3','type3','type3','type3'],
                   'date': ['2017-3-30','2017-5-10','2017-12-15','2018-01-15','2018-05-01','2018-01-30','2018-06-27','2019-03-20','2019-05-21'],
                   'year': [2017, 2017, 2017,2018,2018,2018,2018,2019,2019]})

df['date'] = pd.to_datetime(df['date'])
df = df.sort_values(by='date')
result_lst = []
for year_type, sub_df in df.groupby(['year', 'type']):
    year, type = year_type
    sub_df['shift'] = sub_df['date'].shift(1)
    sub_df.loc[sub_df.index[0], 'shift'] = pd.to_datetime(str(year), format='%Y')
    sub_df['DateDiff'] = sub_df['date'] - sub_df['shift']
    sub_df['year'] = year
    sub_df['type'] = type
    sub_df = sub_df.drop(columns=['shift'])
    result_lst.append(sub_df)
df = pd.concat(result_lst, axis=0)
print(df)