Question

我试图通过跟随这个pandas tutorial来读取嵌套字典的json，问题是我的一些嵌套列表/字典是NaN所以如果我尝试调用normalize函数我得到一个can't find Key Error，因为它仅存在于字典较高级别的某些元素中。

这是我的数据：

q
Out[235]: 
[{u'Code': u'GE',
  u'datetime': u'2011-11-14T19:30:03-05:00[US/Eastern]'},
 {u'Code': u'PP',
  u'datetime': u'2012-21-14T18:50-05:00[US/Eastern]'},
 {u'Code': u'IO',
  u'Summary': [{u'prod': u'book',
    u'num': 81.04,
    u'devil': 17},
   {u'prod': u'game',
    u'num': 191.5,
    u'devil': 10},
   {u'prod': u'desk',
    u'num': 55.5,
    u'devil': -6},
   {u'angel': u'ipo',
    u'num': 503.0,
    u'devil': 1}],
  u'datetime': u'2013-10-14T16:30-05:00[US/Eastern]'},
 {u'Code': u'BI',
  u'datetime': u'2014-11-14T12:30-05:00[US/Eastern]'},
 {u'Code': u'EZ',
  u'datetime': u'2015-12-14T10:00-05:00[US/Eastern]'},
 {u'Code': u'JC',
  u'datetime': u'2016-10-14T08:30:01-05:00[US/Eastern]'},
 {u'Code': u'WX',
  u'Summary': [{u'angel': u'yut',
    u'num': 0,
    u'prod': u'read',
    u'devil': 0.0},
   {u'angel': u'fgf',
    u'prod': u'fart',
    u'devil': 0.0},
   {u'prod': u'red',
    u'num': 673,
    u'angel': u'deft',
    u'devil': 0},
   { u'devil': 0,
    u'prod': u'dog'},
   {u'angel': u'hut',
    u'devil': 99}],
  u'datetime': u'2017-10-13T05:00:02-05:00[US/Eastern]'}]

我可以在像这样的数据框中半观看它：

    pd.DataFrame(q)
    Out[229]: 
             Code                                            Summary                        datetime
    0          GE                                                NaN  2011-11-11T19:30:03-05:00[US/Eastern]
    1          PP                                                NaN  2012-12-25T18:50-05:00[US/Eastern]
    2          IO  [{u'prod': u'book', u'angel': u'I...               2013-11-04T16:30-05:00[US/Eastern]
    3          BI                                                NaN  2014-12-14T08:30:01-05:00[US/Eastern]
    4          JC                                                NaN  2016-11-14T04:30-05:00[US/Eastern]
    5          WX  [{u'prod': u'orange', u'devil': -2, u's...         2017-10-13T03:30:08-05:00[US/Eastern]

如上所述，运行pd.io.json.json_normalize(q, 'Summary',['Code', 'datetime'])会产生KeyError: 'Summary'

我该如何解决这个问题？理想情况下，我希望NaN单元格值不存在。

Answer 1

IIUC：

In [94]: (json_normalize([x for x in q if x.get('Summary')],
                         'Summary',
                         ['Code', 'datetime'])
    ...:                .append(pd.DataFrame([x for x in q if not x.get('Summary')])))
    ...:
Out[94]:
  Code angel                               datetime  devil     num  prod
0   IO   NaN     2013-10-14T16:30-05:00[US/Eastern]   17.0   81.04  book
1   IO   NaN     2013-10-14T16:30-05:00[US/Eastern]   10.0  191.50  game
2   IO   NaN     2013-10-14T16:30-05:00[US/Eastern]   -6.0   55.50  desk
3   IO   ipo     2013-10-14T16:30-05:00[US/Eastern]    1.0  503.00   NaN
4   WX   yut  2017-10-13T05:00:02-05:00[US/Eastern]    0.0    0.00  read
5   WX   fgf  2017-10-13T05:00:02-05:00[US/Eastern]    0.0     NaN  fart
6   WX  deft  2017-10-13T05:00:02-05:00[US/Eastern]    0.0  673.00   red
7   WX   NaN  2017-10-13T05:00:02-05:00[US/Eastern]    0.0     NaN   dog
8   WX   hut  2017-10-13T05:00:02-05:00[US/Eastern]   99.0     NaN   NaN
0   GE   NaN  2011-11-14T19:30:03-05:00[US/Eastern]    NaN     NaN   NaN
1   PP   NaN     2012-21-14T18:50-05:00[US/Eastern]    NaN     NaN   NaN
2   BI   NaN     2014-11-14T12:30-05:00[US/Eastern]    NaN     NaN   NaN
3   EZ   NaN     2015-12-14T10:00-05:00[US/Eastern]    NaN     NaN   NaN
4   JC   NaN  2016-10-14T08:30:01-05:00[US/Eastern]    NaN     NaN   NaN

或使用pd.concat()：

In [95]: pd.concat([json_normalize([x for x in q if x.get('Summary')],
    ...:                           'Summary',
    ...:                           ['Code', 'datetime']),
    ...:            pd.DataFrame([x for x in q if not x.get('Summary')])],
    ...:           ignore_index=True)
    ...:
Out[95]:
   Code angel                               datetime  devil     num  prod
0    IO   NaN     2013-10-14T16:30-05:00[US/Eastern]   17.0   81.04  book
1    IO   NaN     2013-10-14T16:30-05:00[US/Eastern]   10.0  191.50  game
2    IO   NaN     2013-10-14T16:30-05:00[US/Eastern]   -6.0   55.50  desk
3    IO   ipo     2013-10-14T16:30-05:00[US/Eastern]    1.0  503.00   NaN
4    WX   yut  2017-10-13T05:00:02-05:00[US/Eastern]    0.0    0.00  read
5    WX   fgf  2017-10-13T05:00:02-05:00[US/Eastern]    0.0     NaN  fart
6    WX  deft  2017-10-13T05:00:02-05:00[US/Eastern]    0.0  673.00   red
7    WX   NaN  2017-10-13T05:00:02-05:00[US/Eastern]    0.0     NaN   dog
8    WX   hut  2017-10-13T05:00:02-05:00[US/Eastern]   99.0     NaN   NaN
9    GE   NaN  2011-11-14T19:30:03-05:00[US/Eastern]    NaN     NaN   NaN
10   PP   NaN     2012-21-14T18:50-05:00[US/Eastern]    NaN     NaN   NaN
11   BI   NaN     2014-11-14T12:30-05:00[US/Eastern]    NaN     NaN   NaN
12   EZ   NaN     2015-12-14T10:00-05:00[US/Eastern]    NaN     NaN   NaN
13   JC   NaN  2016-10-14T08:30:01-05:00[US/Eastern]    NaN     NaN   NaN

Pandas用NaN条目读取嵌套的json

1 个答案: