如何将嵌套字典转换为数据框

Posted 2023-03-12

【中文标题】如何将嵌套字典转换为数据框

【英文标题】：How to convert a nested dict into dataframe

【发布时间】：2021-08-30 20:52:37

【问题描述】：

假设我有一个 API 响应：

    "fact": 
        "UP": [
            "SCODE": "CNB",
            "SNAME": "Kanpur Central"
        , 
            "SCODE": "JHS",
            "SNAME": "Jhansi Junction"
        ],
        "MP": [
            "SCODE": "BPL",
            "SNAME": "Bhopal Junction"
        , 
            "SCODE": "JBP",
            "SNAME": "Jabalpur Junction"
        ]
    

我必须将其转换为如下所示的数据框（预期输出）：

fact    SCODE   SNAME
UP      CNB     Kanpur Central
UP      JHS     Jhansi Junction
MP      BPL     Bhopal Junction
MP      JBP     Jabalpur Junction

我的努力：我尝试使用 json_normalize() 但没有达到预期的输出：

pd.json_normalize(response).apply(pd.Series.explode)

【参考方案1】：

一种选择是用 python 重塑：

df = pd.DataFrame(['fact': k, **item
                   for k, lst in response['fact'].items()
                   for item in lst])

  fact SCODE              SNAME
0   UP   CNB     Kanpur Central
1   UP   JHS    Jhansi Junction
2   MP   BPL    Bhopal Junction
3   MP   JBP  Jabalpur Junction

---

pandas 选项通过explode + apply pd.Series：

df = (
    pd.DataFrame(response)['fact']
        .explode()
        .apply(pd.Series)
        .rename_axis('fact')
        .reset_index()
)

  fact SCODE              SNAME
0   MP   BPL    Bhopal Junction
1   MP   JBP  Jabalpur Junction
2   UP   CNB     Kanpur Central
3   UP   JHS    Jhansi Junction

【参考方案2】：

使用来自 OP 的 response。 您必须创建另一个结构，因为json_normalize 与字典列表一起使用，并且其中必须包含fact：

new_response = ["fact": rfact, **r for rfact in response["fact"] for r in response["fact"][rfact]]

最后，你只需要应用函数：

final_result = pd.json_normalize(new_response)

  fact SCODE              SNAME
0   UP   CNB     Kanpur Central
1   UP   JHS    Jhansi Junction
2   MP   BPL    Bhopal Junction
3   MP   JBP  Jabalpur Junction

【参考方案3】：

不如直接在字典中工作效率高（所选答案做得很好）：

    data = 
    "fact": 
        "UP": [
            "SCODE": "CNB",
            "SNAME": "Kanpur Central"
        , 
            "SCODE": "JHS",
            "SNAME": "Jhansi Junction"
        ],
        "MP": [
            "SCODE": "BPL",
            "SNAME": "Bhopal Junction"
        , 
            "SCODE": "JBP",
            "SNAME": "Jabalpur Junction"
        ]
    


keys = data['fact']

   (pd.concat([jn(data['fact'][key]) for key in keys], 
              keys = keys)
      .droplevel(-1)
      .rename_axis(index='fact')
      .reset_index()
     )



fact SCODE              SNAME
0   UP   CNB     Kanpur Central
1   UP   JHS    Jhansi Junction
2   MP   BPL    Bhopal Junction
3   MP   JBP  Jabalpur Junction