根据聚合数据的结果对数据帧进行排序（即，根据较短列表的index（））

Question

我有一个数据表，我想在显示前排序。排序基于聚合。在下面的例子中，我已经计算了一些汽车，我想按每个制造商的汽车总数来订购数据框，按升序排列，例如“奔驰”，然后是“宝马”，然后是“奥迪”。

我想保留所有原始列-聚合仅用于确定排序顺序。

import pandas as pd

df = pd.DataFrame({
    "manufacturer": ["bmw", "bmw", "bmw", "audi", "audi", "audi", "audi", "mercedes", "mercedes"],
    "color": ["red", "blue", "silver", "red", "blue", "black", "blue", "green", "white"],
    "count": [5, 4, 7, 1, 1, 9, 7, 5, 3],
    "misc": ["a", "bc", "def", "gh", "ijk", "l", "mnop", "qrst", "uvwxyz"],
})

# order manufacturers by increasing total car counts
order = list(df.groupby("manufacturer").agg({"count": sum}).sort_values(by="count", ascending=True).index)

# apply ordering to original dataframe based on manufacturer's position in 'order' list
df.sort_values(by="manufacturer", key=lambda x: x.apply(order.index))

有没有一种更像蟒蛇/熊猫的方法来做到这一点，因为上面的感觉是我为一些应该简单的事情付出了太多的努力。

Answer 1

让我们尝试使用argsort

out = df.iloc[df.groupby("manufacturer")['count'].transform('sum').argsort()]
Out[207]: 
  manufacturer   color  count    misc
7     mercedes   green      5    qrst
8     mercedes   white      3  uvwxyz
0          bmw     red      5       a
1          bmw    blue      4      bc
2          bmw  silver      7     def
3         audi     red      1      gh
4         audi    blue      1     ijk
5         audi   black      9       l
6         audi    blue      7    mnop