Question

我有一个如下所示的多索引数据框：

             2019-01-08 2019-01-15  2019-01-22  2019-01-29  2019-02-05  
6392    height        3         6           5            3          3
        length        3         3           5            9          3
6393 
        height        1         6           1            4          3
        length        5         3           2            3          3

我想将其转换为类似于以下内容的JSON。

{
    "6392": {
        "2019-01-08": [{
            "height": 3
            "length": 3
        }],
        "2019-01-15": [{
            "height": 
            "length": 3
        }],
        "2019-012-22": [{
            "height": 5
            "length": 5
        }],
            ...
    }, 
    "6393": {
        "2019-01-08": [{
            "height": 1
            "length": 5
        }],
        "2019-01-15": [{
            "height": 6
            "length": 3
        }],
        "2019-012-22": [{
            "height": 1
            "length": 2
        }],
            ...
}

我尝试了类似df.to_json(orient='index')的操作，该操作返回错误。而且使用reset_index()不会返回我想要的层次！

感谢您的帮助。

Answer 1

根据Quang的建议，我将按照这种方式处理您的实际数据集：

import numpy as np
import pandas as pd
arrays = [np.array(['bar', 'bar', 'baz', 'baz', 'foo', 'foo', 'qux', 'qux']),
          np.array(['one', 'two', 'one', 'two', 'one', 'two', 'one', 'two'])]
df = pd.DataFrame(np.random.randn(8,4), index=arrays, columns=['col1','col2','col3','col4'])
D = df.groupby(level=0).apply(lambda df: df.xs(df.name).to_dict()).to_dict()

输出此字典：


{'bar': {'col1': {'one': -0.9687674292695906, 'two': -0.7892120308117504},
  'col2': {'one': -0.08468610899521901, 'two': -0.8123345931126713},
  'col3': {'one': 0.8136040202024982, 'two': 1.4254756109087028},
  'col4': {'one': -0.5631944934736082, 'two': -1.0686604230467418}},
 'baz': {'col1': {'one': -0.8329277599190955, 'two': -0.797572943803082},
  'col2': {'one': -1.18912237734452, 'two': -0.6222985373781997},
  'col3': {'one': -0.6307550007277682, 'two': -0.43423342334272047},
  'col4': {'one': -0.8090341502048565, 'two': 1.7846384031629874}},
 'foo': {'col1': {'one': 0.17441065807207026, 'two': -0.142104023898428},
  'col2': {'one': 0.4865273350791687, 'two': 1.4119728392158484},
  'col3': {'one': -1.7834681421564647, 'two': 0.9228194356473829},
  'col4': {'one': -0.7426715146036388, 'two': 0.32663534732439187}},
 'qux': {'col1': {'one': -0.32243916994536376, 'two': -0.4490530023512784},
  'col2': {'one': 0.31957291028411916, 'two': -1.6707253441375334},
  'col3': {'one': 0.2794431740425791, 'two': 1.0928413422340624},
  'col4': {'one': -0.818204166504019, 'two': -1.2567773847741046}}}

可以使用以下命令将其转换为json文件：

import json
with open('/path/to/file.json', 'w') as json_file:
    json.dump(D, json_file)

即：

{
   "bar":{
      "col1":{
         "one":-0.9687674292695906,
         "two":-0.7892120308117504
      },
      "col2":{
         "one":-0.08468610899521901,
         "two":-0.8123345931126713
      },
      "col3":{
         "one":0.8136040202024982,
         "two":1.4254756109087028
      },
      "col4":{
         "one":-0.5631944934736082,
         "two":-1.0686604230467418
      }
   },
   "baz":{
      "col1":{
         "one":-0.8329277599190955,
         "two":-0.797572943803082
      },
      "col2":{
         "one":-1.18912237734452,
         "two":-0.6222985373781997
      },
      "col3":{
         "one":-0.6307550007277682,
         "two":-0.43423342334272047
      },
      "col4":{
         "one":-0.8090341502048565,
         "two":1.7846384031629874
      }
   },
   ...

距离您的需求足够近吗？

Answer 2

这是您的数据集：

df = pd.DataFrame({'2019-01-08': [3, 3, 1, 5], '2019-01-15': [6,3,6,3]},
                  index=[[6392, 6392, 6393, 6393], ['height', 'length', 'height', 'length']])
df
#               2019-01-08  2019-01-15
# 6392  height  3           6
#       length  3           3
# 6393  height  1           6
#       length  5           3

，这将在Quang的建议下完成所需的JSON转换：

D = (df
  .groupby(level=0)
  .apply(lambda df: df.xs(df.name).to_dict())
  .to_dict()
)

D
# {6392: {'2019-01-08': {'height': 3, 'length': 3},
#  '2019-01-15': {'height': 6, 'length': 3}},
# 6393: {'2019-01-08': {'height': 1, 'length': 5},
#  '2019-01-15': {'height': 6, 'length': 3}}}

如果您坚持将内在字典包裹在列表中，那就这样做

for k in D:
    for m in D[k]:
        D[k][m] = [D[k][m]]
D
# {6392: {'2019-01-08': [{'height': 3, 'length': 3}],
#  '2019-01-15': [{'height': 6, 'length': 3}]},
# 6393: {'2019-01-08': [{'height': 1, 'length': 5}],
#  '2019-01-15': [{'height': 6, 'length': 3}]}}

熊猫多索引DataFrame到JSON

2 个答案: