如何使用 pandas 从嵌套的 JSON 数组中提取值

Question

我有一个大的 JSON 文件（40 万行）。我正在尝试隔离以下内容：

政策-“描述”

策略项 - “用户”和“数据库值”

JSON 文件 - https://pastebin.com/hv8mLfgx

Pandas 的预期输出：https://imgur.com/a/FVcNGsZ

“Policy Items”之后的所有内容在整个文件中都以完全相同的方式重新迭代。我试过下面的代码来隔离“用户”。它似乎不起作用，我正在尝试将所有这些转储到 CSV 中。

Edit* 这是我尝试尝试的解决方案，但无法使其正常工作 - Deeply nested JSON response to pandas dataframe

from pandas.io.json import json_normalize as Jnormal
import json
import pprint, csv
import re

with open("Ranger_Policies_20190204_195010.json") as file:
    jsonDF = json.load(file)
    for item in jsonDF['policies'][0]['policyItems'][0]:
        print ('{} - {} - {}'.format(jsonDF['users']))

编辑 2：我有一些工作代码能够获取一些用户，但它不能获取所有用户。 25 个中只有 11 个。

from pandas.io.json import json_normalize as Jnormal
import json
import pprint, csv
import re

with open("Ranger_Policies_20190204_195010.json") as file:
    jsonDF = json.load(file)
    pNode = Jnormal(jsonDF['policies'][0]['policyItems'], record_path='users')
    print(pNode.head(500))

编辑 3：这是最终的工作副本，但是我仍然没有复制所有 TABLE 数据。我设置了一个循环来简单地忽略所有内容。捕获所有内容，我会在 Excel 中对其进行排序，有人知道为什么我无法捕获所有 TABLE 值吗？

    json_data = json.load(file)
    with open("test.csv", 'w', newline='') as fd:
        wr = csv.writer(fd)
        wr.writerow(('Database name', 'Users', 'Description', 'Table'))
        for policy in json_data['policies']:
            desc = policy['description']
            db_values = policy['resources']['database']['values']
            db_tables = policy['resources']['table']['values']
            for item in policy['policyItems']:
                users = item['users']
                for dbT in db_tables:
                    for user in users:
                        for db in db_values:
                            _ = wr.writerow((db, user, desc, dbT))```

Answer 1

Pandas 在这里太过分了：csv 标准模块就足够了。您只需迭代策略以提取数据库值的描述，然后在 policyItems 上提取用户：

with open("Ranger_Policies_20190204_195010.json") as file:
    jsonDF = json.load(file)
with open("outputfile.csv", newline='') as fd:
    wr = csv.writer(fd)
    _ = wr.writerow(('Database name', 'Users', 'Description'))
    for policy in js['policies']:
        desc = policy['description']
        db_values = policy['resources']['database']['values']
        for item in policy['policyItems']:
            users = item['users']
            for user in users:
                for db in db_values:
                    if db != '*':
                        _ = wr.writerow((db, user, desc))

Answer 2

这是一种方法，假设您的 json 数据位于名为 json_data 的变量中

from itertools import product

def make_dfs(data):
    cols = ['db_name', 'user', 'description']

    for item in data.get('policies'):
        description = item.get('description')
        users = item.get('policyItems', [{}])[0].get('users', [None])
        db_name = item.get('resources', {}).get('database', {}).get('values', [None])
        db_name = [name for name in db_name if name != '*']
        prods = product(db_name, users, [description])
        yield pd.DataFrame.from_records(prods, columns=cols)

df = pd.concat(make_dfs(json_data), ignore_index=True)

print(df)

   db_name          user                               description
0    m2_db          hive  Policy for all - database, table, column
1    m2_db  rangerlookup  Policy for all - database, table, column
2    m2_db     ambari-qa  Policy for all - database, table, column
3    m2_db          af34  Policy for all - database, table, column
4    m2_db          g748  Policy for all - database, table, column
5    m2_db          hdfs  Policy for all - database, table, column
6    m2_db          dh10  Policy for all - database, table, column
7    m2_db          gs22  Policy for all - database, table, column
8    m2_db          dh27  Policy for all - database, table, column
9    m2_db          ct52  Policy for all - database, table, column
10   m2_db  livy_pyspark  Policy for all - database, table, column

在Python 3.5.1 和pandas==0.23.4 上测试