ENH: new orient setting for read_json to support common API format

[ryancasburn-KAI]

Is your feature request related to a problem?

I see many APIs return results in the form:

{
"results": [
    {"a" :1,
     "b":2
    },
    {
     "a" :3,
     "b":4
    }
  ]
}

This format isn't directly supported by pandas. The data is in the "records" orient, but there is an extra layer. Currently to load this file I am using the requests module to load from https, then using the json module to strip out the outer layer, then feeding this data to pd.read_json as text. This feels like overkill, since pandas can read from https, but for this format (which is common for APIs) I need multiple other packages and lines of code.

This will change a three import, multi-line issue into a single import, single line solution.

Describe the solution you'd like

While I initially describe this as a new orient, I don't think that is the best way to implement this. I believe the read_json function should have a new parameter (such as "strip_layer") which will be the value of that outer layer. In the example above that would be "results". I make this suggestion as what is inside the outer layer could be several different orients, so we need to leave that as a possibility. This is something that happens first, then the data is processed.

API breaking implications

Need to consider what this means for chunking.

Additional context

My current code:

import pandas as pd
import requests
import json

data = requests.get(url).json()
data = data["results"]
data = json.dumps(data)
data = pd.read_json(data)

versus my desired code with this improvement:

import pandas as pd

data = pd.read_json(url, strip_layer="results")

Might I suggest this gets added to the IO Method Robustness/Input Types Project?

[attack68]

you don't have to dump and re-read, what about:

data = requests.get(url).json()
data = pd.DataFrame(data["results"])

But I see your point..

[ryancasburn-KAI]

True, I wasn't the most efficient in my example. Still need another package either way.