使用“to_dict()”和“json.dump()”通过字典将数据帧转换为 JSON

Question

我正在尝试将数据帧转换为特定的 JSON 格式。我尝试分别使用 pandas 和 json 模块中的方法“to_dict()”和“json.dump()”来执行此操作，但我无法获得我想要的 JSON 格式。举例说明：

df = pd.DataFrame({
    "Location": ["1ST"] * 3 + ["2ND"] * 3,
    "Date": ["2019-01", "2019-02", "2019-03"] * 2,
    "Category": ["A", "B", "C"] * 2,
    "Number": [1, 2, 3, 4, 5, 6]
})

def dataframe_to_dictionary(df, orientation):
    dictionary = df.to_dict(orient=orientation)
    return dictionary

dict_records = dataframe_to_dictionary(df, "records")

with open("./json_records.json", "w") as json_records:
    json.dump(dict_records, json_records, indent=2)

dict_index = dataframe_to_dictionary(df, "index")

with open("./json_index.json", "w") as json_index:
    json.dump(dict_index, json_index, indent=2)

当我将“dict_records”转换为 JSON 时，我得到一个表单数组：

[
  {
    "Location": "1ST",
    "Date": "2019-01",
    "Category": "A",
    "Number": 1
  },
  {
    "Location": "1ST",
    "Date": "2019-02",
    "Category": "B",
    "Number": 2
  },
...
]

而且，当我将“dict_index”转换为 JSON 时，我得到了一个形式的对象：

{
  "0": {
    "Location": "1ST",
    "Date": "2019-01",
    "Category": "A",
    "Number": 1
  },
  "1": {
    "Location": "1ST",
    "Date": "2019-02",
    "Category": "B",
    "Number": 2
  }
...
}

但是，我正在尝试获得如下所示的格式（其中键 = 位置和值 = [{}]），如下所示。提前感谢您的帮助。

{
    1ST: [
        {
            "Date": "2019-01",
            "Category": "A",
            "Number" 1
        },
        {
            "Date": "2019-02",
            "Category": "B",
            "Number" 2
        },
        {
            "Date": "2019-03",
            "Category": "C",
            "Number" 3
        }
    ],
    2ND: [
        {},
        {},
        {}
    ]
}

Answer 1

这可以通过groupby实现：

gb = df.groupby('Location')

{k: v.drop('Location', axis=1).to_dict(orient='records') for k, v in gb}