python - 如何通过 DataFrame 压扁 Pandas group？

Question

我有一个按日期和“结果”分组的 Pandas DataFrame:

api_logs.groupby([api_logs.index.date, 'Outcome']).size()

            Outcome2017-04-22  Success      72017-04-24  Failure     32            Success     592017-04-25  Failure     23            Success     912017-04-26  Failure      1            Success     592017-04-27  Failure      3            Success      12017-04-28  Failure      1            Success      22017-04-29  Success      32017-05-03  Failure     382017-05-04  Failure      6            Success    727

如何将嵌套数据展平，使其结构如下？

            Failure    Success2017-04-22                   72017-04-24       32         592017-04-25       23         912017-04-26        1         592017-04-27        3          12017-04-28        1          22017-04-29        32017-05-03       382017-05-04        6        727

我的最终目标是将失败和成功一起绘制在图表中，所以也许有完全不同的方法？

Answer 1

使用unstack reshape :

df = api_logs.groupby([api_logs.index.date, 'Outcome']).size().unstack()
print (df)
Outcome     Failure  Success
2017-04-22      NaN      7.0
2017-04-24     32.0     59.0
2017-04-25     23.0     91.0
2017-04-26      1.0     59.0
2017-04-27      3.0      1.0
2017-04-28      1.0      2.0
2017-04-29      NaN      3.0
2017-05-03     38.0      NaN
2017-05-04      6.0    727.0

也可以通过参数 fill_value 将 NaN 替换为 0:

df = api_logs.groupby([api_logs.index.date, 'Outcome']).size().unstack(fill_value=0)

print (df)
Outcome     Failure  Success
2017-04-22        0        7
2017-04-24       32       59
2017-04-25       23       91
2017-04-26        1       59
2017-04-27        3        1
2017-04-28        1        2
2017-04-29        0        3
2017-05-03       38        0
2017-05-04        6      727