我有这个嵌套的 python 字典，
poke_dic = {
'Basic': {'Fire': ['Cyndaquil', 'Charmander', 'Torchic'],
'Grass': ['Chikorita', 'Bulbasaur', 'Treecko'],
'Water': ['Totodile', 'Squirtle', 'Mudkip']},
'Evo1': {'Fire': ['Quilava', 'Chameleon', 'Combusken'],
'Grass': ['Bayleef', 'Ivysaur', 'Grovyle'],
'Water': ['Croconaw', 'Wartortle', 'Marshtomp']},
'Evo2': {'Fire': ['Typhlosion', 'Charizard', 'Blaziken'],
'Grass': ['Meganium', 'Venusaur', 'Sceptile'],
'Water': ['Feraligatr', 'Blastoise', 'Swampert']}
}
当我将其转换为数据帧格式时，它会生成此表
poke_df = pandas.DataFrame(poke_dic)
poke_df
正如我们所见，列标签是一级字典的键，行标签是二级字典的键。条目的数据类型是数组。我希望分解数组并从行标签创建一个新行。
预期的输出(前几行)应该是这样的
pandas 中是否有任何命令允许我执行此操作？或者我必须首先操纵字典？提前致谢
最佳答案
选项 1
您可以从正确的数据帧开始，而不是创建数据帧和操作数据帧。这里的问题是我必须在理解中使用 enumerate 来确保索引的唯一性。如果愿意，您可以重置该级别。
pd.DataFrame({
evolution: {
(element, i): poke
for element, pokes in types.items()
for i, poke in enumerate(pokes)
}
for evolution, types in poke_dic.items()
})
Basic Evo1 Evo2
Fire 0 Cyndaquil Quilava Typhlosion
1 Charmander Chameleon Charizard
2 Torchic Combusken Blaziken
Grass 0 Chikorita Bayleef Meganium
1 Bulbasaur Ivysaur Venusaur
2 Treecko Grovyle Sceptile
Water 0 Totodile Croconaw Feraligatr
1 Squirtle Wartortle Blastoise
2 Mudkip Marshtomp Swampert
选项 2
更多的理解，但使用 pd.concat
pd.concat({
ev: pd.Series(*zip(*(
(p, e) for e, t in x.items() for p in t
))) for ev, x in poke_dic.items()
}, axis=1)
Basic Evo1 Evo2
Fire Cyndaquil Quilava Typhlosion
Fire Charmander Chameleon Charizard
Fire Torchic Combusken Blaziken
Grass Chikorita Bayleef Meganium
Grass Bulbasaur Ivysaur Venusaur
Grass Treecko Grovyle Sceptile
Water Totodile Croconaw Feraligatr
Water Squirtle Wartortle Blastoise
Water Mudkip Marshtomp Swampert
完全按照OP的要求
pd.concat({
ev: pd.Series(*zip(*(
(p, e) for e, t in x.items() for p in t
))) for ev, x in poke_dic.items()
}, axis=1).rename_axis('Type').reset_index()
Type Basic Evo1 Evo2
0 Fire Cyndaquil Quilava Typhlosion
1 Fire Charmander Chameleon Charizard
2 Fire Torchic Combusken Blaziken
3 Grass Chikorita Bayleef Meganium
4 Grass Bulbasaur Ivysaur Venusaur
5 Grass Treecko Grovyle Sceptile
6 Water Totodile Croconaw Feraligatr
7 Water Squirtle Wartortle Blastoise
8 Water Mudkip Marshtomp Swampert
选项 W/E
这些选项都不干净，所以我会坚持下去，直到感觉良好为止。
pd.concat({k: pd.DataFrame(v) for k, v in poke_dic.items()}).T.stack() \
.reset_index(1, drop=True).rename_axis('Type').reset_index()
Type Basic Evo1 Evo2
0 Fire Cyndaquil Quilava Typhlosion
1 Fire Charmander Chameleon Charizard
2 Fire Torchic Combusken Blaziken
3 Grass Chikorita Bayleef Meganium
4 Grass Bulbasaur Ivysaur Venusaur
5 Grass Treecko Grovyle Sceptile
6 Water Totodile Croconaw Feraligatr
7 Water Squirtle Wartortle Blastoise
8 Water Mudkip Marshtomp Swampert
类似@Wen的选项
pd.DataFrame(
np.column_stack([
poke_df.index.repeat(3),
np.array(poke_df.values.tolist()).transpose(0, 2, 1).reshape(-1, 3),
]),
columns=['Type'] + poke_df.columns.tolist()
)
Type Basic Evo1 Evo2
0 Fire Cyndaquil Quilava Typhlosion
1 Fire Charmander Chameleon Charizard
2 Fire Torchic Combusken Blaziken
3 Grass Chikorita Bayleef Meganium
4 Grass Bulbasaur Ivysaur Venusaur
5 Grass Treecko Grovyle Sceptile
6 Water Totodile Croconaw Feraligatr
7 Water Squirtle Wartortle Blastoise
8 Water Mudkip Marshtomp Swampert
关于python - Pandas 数据框 : convert unique row label into non-unique rows，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/49000514/