Why Create ?

Firstly, Json-like Object here, simply speaking, are dicts, or list of dicts. For example:

TEST_CASE1 = [{'A1': 1, 'A2': 2, 'A3': 3},
              {'A1': [4, 5, 6], 'A2': 7, 'A3': 'x'}]

TEST_CASE2 = {'A1': [{'B1': 4, 'B2': 5},
                     {'B1': 6, 'B3': 7}],
              'A2': {'C1': [8, 9],
                     'C2': [10, 11]},
              'A3': {'A1': {'B4': 12},
                    {'A4': 10}}

They are collections.MutableMapping or collections.MutableSequence objects. It is hard to retrieve data from those sufficiently complex one(e.g., json response from maps web service Api).

{
    "status": 0,
    "type": 2,
    "result": {
        "origin": {
            "city_id": 131,
            "city_name": "北京市",
            "location": {
                "lng": 116.3081418168,
                "lat": 40.056871561538
            }
        },
        "destination": {
            "city_id": 289,
            "city_name": "上海市",
            "location": {
                "lng": 121.50581834502,
                "lat": 31.222960154925
            }
        },
        "total": 49,
        "routes": [
            {
                "distance": 51837,
                "duration": 22544,
                "arrive_time": "2018-08-14 23:18:00",
                "price": 553,
                "price_detail": [],
                "steps": [
                    [
                        {
                            "distance": 708,
                            "duration": 603,
                            "instructions": "步行708米",
...

We usually meet these tasks or more,

• Task1: simplify these nested of nested (loop N times) object.

# flatten TEST_CASE1
{'0_A1': 1,
 '0_A2': 2,
 '0_A3': 3,
 '1_A1_0': 4,
 '1_A1_1': 5,
 '1_A1_2': 6,
 '1_A2': 7,
 '1_A3': 'x'}

• Task2: filter by keys thoroughly with other conditions.

# drop "A1" from TEST_CASE1
[{'A2': 2, 'A3': 3}, {'A2': 7, 'A3': 'x'}]

• Taks3: extract every nodes which has subkey requirement.

# select "A1" from TEST_CASE2
[[{'B1': 4, 'B2': 5}, {'B1': 6, 'B3': 7}], {'B4': 12}]
# select "A1" which has subkey "B1" from TEST_CASE2
[{'B1': 4, 'B2': 5}, {'B1': 6, 'B3': 7}]

• Task4: replace some value without replacement.

# update "A1" with 10086 from TEST_CASE1
[{'A1': 10086, 'A2': 2, 'A3': 3},
 {'A1': 10086, 'A2': 7, 'A3': 'x'}]
# or update each sub-element
[{'A1': 10086, 'A2': 2, 'A3': 3},
 {'A1': [10086, 10086, 10086], 'A2': 7, 'A3': 'x'}]

• Task5: sort whole object to make look pretty.

# sort by value thoroughly from TEST_CASE1 descending
[{'A3': 3, 'A2': 2, 'A1': 1},
 {'A3': 'x', 'A2': 7, 'A1': [4, 5, 6]}]

• Task6: convert to pandas as it is so popular for data scientist.

# convert TEST_CASE1 to DataFrame
          A1  A2 A3
0          1   2  3
1  [4, 5, 6]   7  x