使用pandas normalize将json嵌套到csv

Question

使用给定的脚本，我能够获得输出，如我在截图中所示， 但是有一个名为 cve.description.description_data 的列，它再次采用json格式。我也想提取这些数据。

import json 
import pandas as pd 
from pandas.io.json import json_normalize 
#load json object
with open('nvdcve-1.0-modified.json') as f:
   d = json.load(f)
#tells us parent node is 'programs'
nycphil = json_normalize(d['CVE_Items'])
nycphil.head(3)
works_data = json_normalize(data=d['CVE_Items'], record_path='cve')
works_data.head(3)
nycphil.to_csv("test4.csv")

如果我更改works_data = json_normalize(data=d['CVE_Items'], record_path='cve.descr')，则会出现此错误：

  

“result = result [spec] KeyError：'cve.description'”

JSON格式如下：

{
   "CVE_data_type":"CVE",
   "CVE_data_format":"MITRE",
   "CVE_data_version":"4.0",
   "CVE_data_numberOfCVEs":"1000",
   "CVE_data_timestamp":"2018-04-04T00:00Z",
   "CVE_Items":[
      {
         "cve":{
            "data_type":"CVE",
            "data_format":"MITRE",
            "data_version":"4.0",
            "CVE_data_meta":{
               "ID":"CVE-2001-1594",
               "ASSIGNER":"cve@mitre.org"
            },
            "affects":{
               "vendor":{
                  "vendor_data":[
                     {
                        "vendor_name":"gehealthcare",
                        "product":{
                           "product_data":[
                              {
                                 "product_name":"entegra_p&r",
                                 "version":{
                                    "version_data":[
                                       {
                                          "version_value":"*"
                                       }
                                    ]
                                 }
                              }
                           ]
                        }
                     }
                  ]
               }
            },
            "problemtype":{
               "problemtype_data":[
                  {
                     "description":[
                        {
                           "lang":"en",
                           "value":"CWE-255"
                        }
                     ]
                  }
               ]
            },
            "references":{
               "reference_data":[
                  {
                     "url":"http://apps.gehealthcare.com/servlet/ClientServlet/2263784.pdf?DOCCLASS=A&REQ=RAC&DIRECTION=2263784-100&FILENAME=2263784.pdf&FILEREV=5&DOCREV_ORG=5&SUBMIT=+  ACCEPT+"
                  },
                  {
                     "url":"http://www.forbes.com/sites/thomasbrewster/2015/07/10/vulnerable- "
                  },
                  {
                     "url":"https://ics-cert.us-cert.gov/advisories/ICSMA-18-037-02"
                  },
                  {
                     "url":"https://twitter.com/digitalbond/status/619250429751222277"
                  }
               ]
            },
            "description":{
               "description_data":[
                  {
                     "lang":"en",
                     "value":"GE Healthcare eNTEGRA P&R has a password of (1) value."
                  }
               ]
            }
         },
         "configurations":{
            "CVE_data_version":"4.0",
            "nodes":[
               {
                  "operator":"OR",
                  "cpe":[
                     {
                        "vulnerable":true,
                        "cpe22Uri":"cpe:/a:gehealthcare:entegra_p%26r",
                        "cpe23Uri":"cpe:2.3:a:gehealthcare:entegra_p\\&r:*:*:*:*:*:*:*:*"
                     }
                  ]
               }
            ]
         },
         "impact":{
            "baseMetricV2":{
               "cvssV2":{
                  "version":"2.0",
                  "vectorString":"(AV:N/AC:L/Au:N/C:C/I:C/A:C)",
                  "accessVector":"NETWORK",
                  "accessComplexity":"LOW",
                  "authentication":"NONE",
                  "confidentialityImpact":"COMPLETE",
                  "integrityImpact":"COMPLETE",
                  "availabilityImpact":"COMPLETE",
                  "baseScore":10.0
               },
               "severity":"HIGH",
               "exploitabilityScore":10.0,
               "impactScore":10.0,
               "obtainAllPrivilege":false,
               "obtainUserPrivilege":false,
               "obtainOtherPrivilege":false,
               "userInteractionRequired":false
            }
         },
         "publishedDate":"2015-08-04T14:59Z",
         "lastModifiedDate":"2018-03-28T01:29Z"
      }
   ]
}

我想要压扁所有数据。

Answer 1

假设在行和所有其他元数据重复之间描绘了多个URL，请考虑使用递归函数调用来提取嵌套json对象 d 中的每个键值对。

递归函数将调用global来更新所需的全局对象，以便绑定到pd.DataFrame()调用的字典列表中。最后一个循环结束更新递归函数的字典 inner ，以集成不同的URL（存储在 multi 中）

import json 
import pandas as pd 

# load json object
with open('nvdcve-1.0-modified.json') as f:
   d = json.load(f)

multi = []; inner = {}

def recursive_extract(i):
    global multi, inner

    if type(i) is list:
        if len(i) == 1:
            for k,v in i[0].items():
                if type(v) in [list, dict]:
                    recursive_extract(v)
                else:                
                    inner[k] = v
        else:
            multi = i

    if type(i) is dict:
        for k,v in i.items():
            if type(v) in [list, dict]:
                recursive_extract(v)
            else:                
                inner[k] = v

recursive_extract(d['CVE_Items'])

data_dict = []
for i in multi:    
    tmp = inner.copy()
    tmp.update(i)
    data_dict.append(tmp)

df = pd.DataFrame(data_dict)
df.to_csv('Output.csv')

输出 （除了网址，所有列都相同，为了强调而加宽）