Python CategoricalDtype自定义排序实现原理解析

#%%

import pandas as pd
from datetime import datetime

city =["上海","北京","深圳","杭州","苏州","青岛","大连","齐齐哈尔","大理","丽江",
     "天津","济南","南京","广州","无锡","连云港","张家界"]

#创建自动从list中选取valuse值的get_list函数
#replace=True代表允许选出的元素重复
def get_list(items,size=20):
  return pd.Series(items).sample(n=size,replace=True).to_list()

#通过get_list自动生成数据，最终生成一个DataFrame
df = pd.DataFrame({
   "城市":get_list(city),
   "仓位":get_list(["经济舱","商务舱","头等舱"]),
   "航线":get_list(["单程","往返"]),
   "日期": get_list([datetime(2020,8,1),datetime(2020,8,2),
           datetime(2020,8,3),datetime(2020,8,4)]),
   "时间": get_list(["09:00 - 12:00",
            "13:00 - 15:30",
            "06:30 - 15:00",
            "18:00 - 21:00",
            "20:00 - 23:20",
            "10:00 - 15:00"]),
   "航空公司": get_list(["东方航空","南方航空","深圳航空","山东航空","中国航空"]),
   "出票数量":get_list([10,15,20,25,30,35,40,45,50,55,60]),
  })
#%%
df

#%%
#自定义type，以及自定义排序的顺序
my_type = pd.CategoricalDtype(
  categories=["头等舱","商务舱","经济舱"],
  ordered=True
)
df["仓位"] = df["仓位"].astype(my_type) #将指定列转成自定义的type
df.dtypes
#%%
#通过透视表统计数据
tb = pd.pivot_table(
  df,
  index=["城市","仓位","航线","日期","时间"],
  values="出票数量",
  aggfunc=sum
)
tb