I having a nested json object, I am able to parse and flatten it to a single level dataframe by preserving hierarchy. Now I need to generate hierarchical data frame need some help on that.

我有一个嵌套的json对象，我能够通过保留层次结构将其解析和扁平化为单层数据帧。现在我需要生成分层数据帧，需要一些帮助。

Sample Object:

示例对象：

{"rr":{"bp": {
"0": "0 - 10",
"1": "10 - 20",
"2": "20 - 30"
},
"al": {
"0": 11.8,
"1": 77.2,
"2": 98.4
}}
}

{“rr”：{“BP”：{“0”：“0-10”，“1”：“10-20”，“2”：“20-30”}，“al”：{“0”：11.8，“1”：77.2，“2”：98.4}}

flattened Dataframe:
  rr.bp.0  rr.bp.1  rr.bp.2  rr.al.0  rr.al.1  rr.al.2
0  0 - 10  10 - 20  20 - 30     11.8     77.2     98.4

expected hierarchy dataframe 
Header1     Header2 0   1   2   3   4   5   6   7   8
rr      bp  0 - 10  10 - 20 20 - 30                      
rr      al  11.8    77.2    98.4

expected hierarchy dataframe

预期的层次结构数据帧

trying something like this

尝试像这样的事情

``
for key, value in flattened_data.items():
    keys = key.split(".")
    column_headers = keys[:-1]
    index_label = keys[-1]
    columns = pd.MultiIndex.from_tuples([tuple(column_headers)], names=column_headers)
    temp_df = pd.DataFrame([value],columns=columns)
    temp_df.index = [index_label]
df = pd.concat([df,temp_df])
``

but everything is coming as NaN values

但一切都是以NaN价值观为基础的

Update
I want the output to be dynamic . You can assume that my key names are in order and depth is preserved in flattened dataframe.

更新我希望输出是动态的。您可以假定我的密钥名称是有序的，并且深度被保留在扁平的数据帧中。

Thanks @Timeless. Based on your answer I tried this, it is working
df will be a single level flattened df
example: For this json {"rr":{"bp": { "0": "0 - 10", "1": "10 - 20", "2": "20 - 30" }, "al": { "0": 11.8, "1": 77.2, "2": 98.4 }} }
df will be flattened Dataframe:
rr.bp.0 rr.bp.1 rr.bp.2 rr.al.0 rr.al.1 rr.al.2
0 0 - 10 10 - 20 20 - 30 11.8 77.2 98.4

谢谢@Timeless。根据你的回答我试了这个，它工作的df将是一个单层扁平化的df例子：对于这个json {“rr”：{“bp”：{“0”：“0 - 10”，“1”：“10 - 20”，“2”：“20 - 30”}，“al”：{“0”：11.8，“1”：77.2，“2”：98.4 } df将变平数据帧：rr.bp.0 rr.bp.1 rr.bp.2 rr.al.0 rr.al.1 rr.al.2 0 0 - 10 10 - 20 20 - 30 11.8 77.2 98.4

If more nested keys are there then more dots and keys are added in the same order

如果存在更多嵌套键，则会按相同顺序添加更多点和键

max_depth = 0
max_depth_list = []
for col in flatten_copy.columns:
    max_depth_list.append(len(re.findall('\.', col)))
max_depth = max(max_depth_list)
df = (
    pd.DataFrame(df).pipe(lambda x: x.set_axis(x.columns.str.split(".", expand=True), axis=1))
    .stack(list(range(max_depth))).droplevel(0).rename_axis([f"Header{i}" for i in range(max_depth)]).reset_index()
)

Your expected output is ambiguous or at least doesn't match with the title of your question.

您的预期输出不明确，或者至少与您问题的标题不匹配。

I suppose that you're expecting a DataFrame like this one :

我想您正在期待这样的DataFrame：

sample_obj = {
    "rr": {
        "bp": {"0": "0 - 10", "1": "10 - 20", "2": "20 - 30"},
        "al": { "0": 11.8, "1": 77.2, "2": 98.4}
          }
}

df = (
    pd.DataFrame(sample_obj).stack()
        .apply(pd.Series) # with a FutureWarning in 2.1.0
        .swaplevel().rename_axis(["Header1", "Header2"]).reset_index()
)

Output :

输出：

print(df)

  Header1 Header2       0        1        2
0      rr      al    11.8     77.2     98.4
1      rr      bp  0 - 10  10 - 20  20 - 30

UPDATE :

更新：

If you start from flattened_data, you can use this :

如果从FLATEED_DATA开始，则可以使用以下命令：

flattened_data = {
    'rr.bp.0': {0: '0 - 10'},
    'rr.bp.1': {0: '10 - 20'},
    'rr.bp.2': {0: '20 - 30'},
    'rr.al.0': {0: 11.8},
    'rr.al.1': {0: 77.2},
    'rr.al.2': {0: 98.4}
} # which is the result of pd.json_normalize(sample_obj).to_dict()

df = (
    pd.DataFrame(flattened_data)
        .pipe(lambda x: x.set_axis(
            x.columns.str.split(".", expand=True), axis=1))
        .stack([0, 1]).droplevel(0)
        .rename_axis(["Header1", "Header2"])
        .reset_index()
)