Comment échanger les indices d'une dataframe à plusieurs indices avec pandas ?

Exemples de comment échanger les indices d'une dataframe à plusieurs indices avec pandas :

Créer une datafrane avec plusieurs indices

Pour créer une datafrane avec plusieurs indices avec pandas, une solution est d'utiliser MultiIndex / advanced indexing, voir aussi how to create a dataframe with multiple indexes with pandas

import pandas as pd

iterables = [["Land", "Ocean", "Snow/Ice"], ["Liquid", "Ice"]]

index = pd.MultiIndex.from_product(iterables, names=["Id1", "Id2"])

df = pd.DataFrame({'Count A': [12., 70., 30., 20., 17.0, 45.0], 'Count B': [12., 70., 30., 20., 17.0, 45.0]}, index=index)

donne

                                 Count A  Count B
Id1      Id2                     
Land     Liquid     12.0     12.0
                 Ice        70.0     70.0
Ocean    Liquid     30.0     30.0
                 Ice        20.0     20.0
Snow/Ice Liquid     17.0     17.0
                 Ice        45.0     45.0

Échanger les indices d'une dataframe

Pour échanger id1 et id2, une solution est d'utiliser pandas.DataFrame.swaplevel

df = df.swaplevel()

donne alors

                                 Count A  Count B
Id2    Id1                       
Liquid Land         12.0     12.0
Ice    Land         70.0     70.0
Liquid Ocean        30.0     30.0
Ice    Ocean        20.0     20.0
Liquid Snow/Ice     17.0     17.0
Ice    Snow/Ice     45.0     45.0

Note:  pour rendre la dataframe plus facile à lire, une solution consiste à utiliser [pandas.DataFrame.sort_index](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.sort_index.html)

    df = df.sort_index()

donne alors

                                 Count A  Count B
Id2    Id1                       
Ice    Land         70.0     70.0
             Ocean        20.0     20.0
             Snow/Ice     45.0     45.0
Liquid Land         12.0     12.0
             Ocean        30.0     30.0
             Snow/Ice     17.0     17.0

Un autre exemple avec 3 indices

Autre exemple avec un dataframe à 3 indices

import pandas as pd

iterables = [["Land", "Ocean"], ["Liquid", "Ice"], ["Successful", "Unsuccessful"]]

index = pd.MultiIndex.from_product(iterables, names=["Id1", "Id2", "Id3"])

df = pd.DataFrame({'Count A': [12., 70., 30., 20., 17.0, 45.0, 42., 7.], 'Count B': [12., 70., 30., 20., 17.0, 45.0, 42., 7.]}, index=index)

donne

                           Count A  Count B
Id1   Id2    Id3                           
Land  Liquid Successful       12.0     12.0
             Unsuccessful     70.0     70.0
      Ice    Successful       30.0     30.0
             Unsuccessful     20.0     20.0
Ocean Liquid Successful       17.0     17.0
             Unsuccessful     45.0     45.0
      Ice    Successful       42.0     42.0
             Unsuccessful      7.0      7.0

Note: avec plus de 2 niveaux d'indices, par défaut swaplevel() permutera les deux derniers niveaux (par exemple id2 et id3 ici):

df = df.swaplevel()

donne

                           Count A  Count B
Id1   Id3          Id2                     
Land  Successful   Liquid     12.0     12.0
      Unsuccessful Liquid     70.0     70.0
      Successful   Ice        30.0     30.0
      Unsuccessful Ice        20.0     20.0
Ocean Successful   Liquid     17.0     17.0
      Unsuccessful Liquid     45.0     45.0
      Successful   Ice        42.0     42.0
      Unsuccessful Ice         7.0      7.0

    df = df.sort_index()

donne

                           Count A  Count B
Id1   Id3          Id2                     
Land  Successful   Ice        30.0     30.0
                   Liquid     12.0     12.0
      Unsuccessful Ice        20.0     20.0
                   Liquid     70.0     70.0
Ocean Successful   Ice        42.0     42.0
                   Liquid     17.0     17.0
      Unsuccessful Ice         7.0      7.0
                   Liquid     45.0     45.0

Choisissez les indices à échanger

Recréons la dataframe :

import pandas as pd

iterables = [["Land", "Ocean"], ["Liquid", "Ice"], ["Successful", "Unsuccessful"]]

index = pd.MultiIndex.from_product(iterables, names=["Id1", "Id2", "Id3"])

df = pd.DataFrame({'Count A': [12., 70., 30., 20., 17.0, 45.0, 42., 7.], 'Count B': [12., 70., 30., 20., 17.0, 45.0, 42., 7.]}, index=index)

donne

                                                     Count A  Count B
Id1   Id2    Id3                           
Land  Liquid Successful       12.0     12.0
                         Unsuccessful     70.0     70.0
            Ice    Successful       30.0     30.0
                         Unsuccessful     20.0     20.0
Ocean Liquid Successful       17.0     17.0
                         Unsuccessful     45.0     45.0
            Ice    Successful       42.0     42.0
                         Unsuccessful      7.0      7.0

Pour échanger id1 et id2, une solution est de faire

df = df.swaplevel(0,1)

donne alors

                                                     Count A  Count B
Id2    Id1   Id3                           
Liquid Land  Successful       12.0     12.0
                         Unsuccessful     70.0     70.0
Ice    Land  Successful       30.0     30.0
                         Unsuccessful     20.0     20.0
Liquid Ocean Successful       17.0     17.0
                         Unsuccessful     45.0     45.0
Ice    Ocean Successful       42.0     42.0
                         Unsuccessful      7.0      7.0

et

df = df.sort_index()

donnera

                                                     Count A  Count B
Id2    Id1   Id3                           
Ice    Land  Successful       30.0     30.0
                         Unsuccessful     20.0     20.0
             Ocean Successful       42.0     42.0
                         Unsuccessful      7.0      7.0
Liquid Land  Successful       12.0     12.0
                         Unsuccessful     70.0     70.0
             Ocean Successful       17.0     17.0
                         Unsuccessful     45.0     45.0

Références

• pandas.DataFrame.swaplevel
• pandas.DataFrame.sort_index