pandas 相当于 fct_reorder

Question

有没有一种方法可以根据 pandas 数据框列与同一数据框中另一个分类列的关系对其进行重新排序，类似于 R 中 forcats 包中的 fct_reorder？我的一个朋友想要运行一个 python 脚本来绘制 plotnine 中的情节。

可以在下面找到 reprex 数据框：

Group   Name    Height
0   3   Abigail 151.09962170955896
1   2   Amelia  144.53368144215813
2   1   Ava 150.84441176683055
3   2   Charlotte   144.2526003986535
4   3   Emily   150.01613555140298
5   1   Emma    127.9293425061458
6   3   Evelyn  154.35548000906718
7   3   Harper  155.22807300246453
8   1   Isabella    116.54302297370651
9   2   Mia 155.0605589215757
10  1   Olivia  142.7742924211066
11  2   Sophia  154.2912468881105

我也做了一个csv文件供下载： https://github.com/Biomiha/factors/blob/master/Fct_reorder_reprex.csv

将其作为小标题读入 R session：

df <- structure(list(Group = c(3, 2, 1, 2, 3, 1, 3, 3, 1, 2, 1, 2), 
Name = c("Abigail", "Amelia", "Ava", "Charlotte", "Emily", 
"Emma", "Evelyn", "Harper", "Isabella", "Mia", "Olivia", 
"Sophia"), Height = c(151.099621709559, 144.533681442158, 
150.844411766831, 144.252600398653, 150.016135551403, 127.929342506146, 
154.355480009067, 155.228073002465, 116.543022973707, 155.060558921576, 
142.774292421107, 154.29124688811)), class = c("spec_tbl_df",  "tbl_df", "tbl", "data.frame"), row.names = c(NA, -12L), spec = structure(list(
cols = list(Group = structure(list(), class = c("collector_double", 
"collector")), Name = structure(list(), class = c("collector_character", 
"collector")), Height = structure(list(), class = c("collector_double", 
"collector"))), default = structure(list(), class = c("collector_guess", 
"collector")), skip = 1), class = "col_spec"))

并将其作为 Pandas DataFrame 读入 python session，复制上面的 table 并粘贴使用：

df = pd.read_clipboard()

我的R代码是：

library(tidyverse)

# The unordered plot that is the default looks like:
plot_without <- df %>%
  dplyr::mutate(Group = as.factor(Group)) %>% 
  ggplot(aes(x = Name, y = Height, fill = Group)) +
  geom_bar(stat = "identity") +
  labs(title = "Plot without ordering")
plot_without

# To order the 'Name' variable, using fct_reorder (this is what I want but from python):
plot_with <- df %>%
  dplyr::mutate(Group = as.factor(Group),
                Name = fct_reorder(Name, Group, identity)) %>% 
  ggplot(aes(x = Name, y = Height, fill = Group)) +
  geom_bar(stat = "identity") +
  labs(title = "Ordered plot")
plot_with

到目前为止等效的 python 代码是：

import sys
import pandas as pd
from plotnine import *
df=pd.read_csv('Fct_reorder_reprex.csv')
df['Group'] = df['Group'].astype('category')
ggplot(df) + geom_bar(aes(x = 'Name', y = 'Height', fill = 'Group', col = 'Group'), stat = 'identity') + labs(title='Python unordered plot')

plotnine 输出如下所示：

问题是，我如何告诉 pandas 根据 Group 列对 Name 列重新排序（即将颜色组合在一起）？

Answer 1

两年过去了，现在我们在python中找到了完美的解决方案：

import pandas as pd
from datar.all import f, mutate, fct_reorder, as_factor, identity
from plotnine import ggplot, geom_bar, labs, aes

df = pd.read_csv("https://github.com/Biomiha/factors/raw/master/Fct_reorder_reprex.csv")
df

     Group       Name      Height
   <int64>   <object>   <float64>
0        3    Abigail  151.099622
1        2     Amelia  144.533681
2        1        Ava  150.844412
3        2  Charlotte  144.252600
4        3      Emily  150.016136
5        1       Emma  127.929343
6        3     Evelyn  154.355480
7        3     Harper  155.228073
8        1   Isabella  116.543023
9        2        Mia  155.060559
10       1     Olivia  142.774292
11       2     Sophia  154.291247

plot_without = (
    df
    >> mutate(Group=as_factor(f.Group))
    >> ggplot(aes(x="Name", y="Height", fill="Group"))
    + geom_bar(stat="identity")
    + labs(title="Plot without ordering")
)
plot_without

plot_with = (
    df
    >> mutate(Group=as_factor(f.Group), Name=fct_reorder(f.Name, f.Group, _fun=identity))
    >> ggplot(aes(x="Name", y="Height", fill="Group"))
    + geom_bar(stat="identity")
    + labs(title="Plot without ordering")
)

plot_with

pandas 相当于 fct_reorder

pandas equivalent of fct_reorder

python

r

pandas

categorical-data