ENH: Better read_json error when handling bad keys

[jtratner]

(in json data with orient=split)

Fixes #4730.

[jtratner]

@Komnomnomnom does this look okay to you? Just changes the exception to be slightly clearer...

[jtratner]

E.g.

import pandas as pd
from pandas.compat import StringIO
import json
d = {u'columns': [0],
 u'data': [[0], [1], [2], [3], [4], [5], [6], [7], [8], [9]],
 u'index': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
 u'badkey': [5]}
s = StringIO(json.dumps(d))
pd.read_json(s, orient='split')

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "pandas/io/json.py", line 182, in read_json
    date_unit).parse()
  File "pandas/io/json.py", line 242, in parse
    self._parse_no_numpy()
  File "pandas/io/json.py", line 478, in _parse_no_numpy
    self.obj = _run_with_kwarg_exception(DataFrame, decoded, dtype=None)
  File "pandas/io/json.py", line 391, in _run_with_kwarg_exception
    raise ValueError(msg)
ValueError: JSON data had unexpected key: 'badkey'

[jtratner]

and I think it's clear enough in Python 3 too:

Traceback (most recent call last):
  File "./pandas/io/json.py", line 382, in _run_with_kwarg_exception
    return func(**kwargs)
TypeError: __init__() got an unexpected keyword argument 'badkey'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "./pandas/io/json.py", line 182, in read_json
    date_unit).parse()
  File "./pandas/io/json.py", line 242, in parse
    self._parse_no_numpy()
  File "./pandas/io/json.py", line 478, in _parse_no_numpy
    self.obj = _run_with_kwarg_exception(DataFrame, decoded, dtype=None)
  File "./pandas/io/json.py", line 391, in _run_with_kwarg_exception
    raise ValueError(msg)
ValueError: JSON data had unexpected key: 'badkey'

[Komnomnomnom]

I agree that the exception is much clearer. Rather than relying on the text of the message in the raised TypeError, which seems kind of fragile, would it be better to instead check that the decoded keys are only those encoded by split and raise a ValueError directly if not?

columns, index, data for DataFrame
name, index, data for Series

[jtratner]

@Komnomnomnom thanks for the cleaner matchgroup usage. I did it this way only to make it easier to maintain...but I guess we can just define what the keys are in json once, so no big problem.

[jtratner]

@Komnomnomnom any other elements need guards for bad keys? Otherwise, this is probably good.

[Komnomnomnom]

Nope split should be the only orient using named keys. Thanks @jtratner!