使用Python和Pandas从嵌套JSON反序列化数据

我有嵌套Json中的时间序列数据，我正努力将其放入扁平的数据帧中

输入数据 数据如下：

预期产量 扁平的熊猫数据框：国家|日期|病例|死亡|恢复

我试过的 现在，我可以使用df=pd.json\u normalizejson\u数据，max\u level=1，但这就留下了嵌入列表。 我也可以使用df=pd.json_normalizejson_数据，但这只会为每个日期创建一个新的列，随着时间的推移，这是不可持续的

---

一定有一种优雅的方式来做这件事。最后一种方法是编写Python循环。

这是数据的一个子集，针对阿富汗国家json数据中的第一个条目：

content = [{"country":"Afghanistan","province":None,"timeline":{"cases":{"3/13/20":7,"3/14/20":11,"3/15/20":16,"3/16/20":21,"3/17/20":22,"3/18/20":22,"3/19/20":22,"3/20/20":24,"3/21/20":24,"3/22/20":40,"3/23/20":40,"3/24/20":74,"3/25/20":84,"3/26/20":94,"3/27/20":110,"3/28/20":110,"3/29/20":120,"3/30/20":170,"3/31/20":174,"4/1/20":237,"4/2/20":273,"4/3/20":281,"4/4/20":299,"4/5/20":349,"4/6/20":367,"4/7/20":423,"4/8/20":444,"4/9/20":484,"4/10/20":521,"4/11/20":555},"deaths":{"3/13/20":0,"3/14/20":0,"3/15/20":0,"3/16/20":0,"3/17/20":0,"3/18/20":0,"3/19/20":0,"3/20/20":0,"3/21/20":0,"3/22/20":1,"3/23/20":1,"3/24/20":1,"3/25/20":2,"3/26/20":4,"3/27/20":4,"3/28/20":4,"3/29/20":4,"3/30/20":4,"3/31/20":4,"4/1/20":4,"4/2/20":6,"4/3/20":6,"4/4/20":7,"4/5/20":7,"4/6/20":11,"4/7/20":14,"4/8/20":14,"4/9/20":15,"4/10/20":15,"4/11/20":18},"recovered":{"3/13/20":0,"3/14/20":0,"3/15/20":0,"3/16/20":1,"3/17/20":1,"3/18/20":1,"3/19/20":1,"3/20/20":1,"3/21/20":1,"3/22/20":1,"3/23/20":1,"3/24/20":1,"3/25/20":2,"3/26/20":2,"3/27/20":2,"3/28/20":2,"3/29/20":2,"3/30/20":2,"3/31/20":5,"4/1/20":5,"4/2/20":10,"4/3/20":10,"4/4/20":10,"4/5/20":15,"4/6/20":18,"4/7/20":18,"4/8/20":29,"4/9/20":32,"4/10/20":32,"4/11/20":32}}}]

一种方法是读入时间线数据，然后将国家和省份数据读入数据框：

res = pd.DataFrame(content[0]['timeline']).assign(country = content[0]['country'],
                                                  province = content[0]['province']
                                                  )

res.head()


         cases    deaths    recovered   country    province
3/13/20   7          0        0        Afghanistan  None
3/14/20   11         0        0        Afghanistan  None
3/15/20   16         0        0        Afghanistan  None
3/16/20   21         0        1        Afghanistan  None
3/17/20   22         0        1        Afghanistan  None

---

请注意，整个数据都包装在一个列表中，因此索引为0。

共享您的问题中的json数据示例，以便有人可以使用并提供可能的解决方案。必须有一种优雅的方法来实现这一点。最后的办法是编写Python循环。我认为最优雅的方式可能是循环。@sammywemmy原始json URL是个问题！我不仅提供了原始json，还提供了下载它的代码。你应该做的是共享一个样本，而不是整个数据。你已经下载了数据，剪下一小部分代表你想要达到的目标，并在你的问题中分享。不要误解我。很酷你分享了你所做的数据。无论谁想回答你的问题，处理其中的一个子集都会更容易，并允许你对整个数据集推断出任何答案。这是有用的+1向上投票。但我仍然没有解决我的问题的办法。如何应用这个中间步骤来交付最终的数据帧？我们的整个json被包装在一个列表中，因此如果你将这个特定的代码应用到它，其中内容代表整个json，那么你应该拥有我们的数据帧。这将是相当多的，并希望，应该适合记忆。划掉它，它应该放入内存就是这样。results=pd.DataFrame for i in rangelenjson_data:df=pd.DataFramejson_data[i]['timeline'].assigncountry=json_data[i]['country']，province=json_data[i]['province']results=results.appenddf，ignore_index=False

res = pd.DataFrame(content[0]['timeline']).assign(country = content[0]['country'],
                                                  province = content[0]['province']
                                                  )

res.head()


         cases    deaths    recovered   country    province
3/13/20   7          0        0        Afghanistan  None
3/14/20   11         0        0        Afghanistan  None
3/15/20   16         0        0        Afghanistan  None
3/16/20   21         0        1        Afghanistan  None
3/17/20   22         0        1        Afghanistan  None