针对结构不良 API 响应的 OPENJSON() 优化

OPENJSON() optimization for poorly structured API response

我正在尝试使用 TSheets API 将数据提取到我的内部数据库中,但是返回的响应方式让我无法弄清楚如何有效地将其放入 table结构。

出于演示目的,我已经简化了响应,但它基本上看起来像这样:

{
    "results": {
        "users": {
            "12345": {
                "id": 12345,
                "first_name": "Demo",
                "last_name": "User",
                "username": "demo@gmail.com",
                "email": "demo@gmail.com"
            },
            "321123": {
                "id": 321123,
                "first_name": "John",
                "last_name": "Wayne",
                "username": "notreal@email.com",
                "email": "notreal@email.com"
            },
            "98765": {
                "id": 98765,
                "first_name": "Suzie",
                "last_name": "Q",
                "username": "email@company.com",
                "email": "email@company.com"
            }
        }
    },
    "more": false
}

每个用户都列为单独的 属性,而不是一组用户,其 ID 作为 属性 的名称。他们在所有端点上都使用这种模式,所以我需要一种方法来了解响应的结构,以便像我习惯的那样查询它。

我写了一个语句,使用动态 sql 将其放入 table 结构中,但我想知道是否有人更擅长 JSON 函数可以提出一个更好的解决方案。

这是我的 SQL 代码...

GO

--// simplifed version of the actual json response for security and demo purposes
DECLARE @user_response NVARCHAR(MAX) = N'

{
    "results": {
        "users": {
            "12345": {
                "id": 12345,
                "first_name": "Demo",
                "last_name": "User",
                "username": "demo@gmail.com",
                "email": "demo@gmail.com"
            },
            "321123": {
                "id": 321123,
                "first_name": "John",
                "last_name": "Wayne",
                "username": "notreal@email.com",
                "email": "notreal@email.com"
            },
            "98765": {
                "id": 98765,
                "first_name": "Suzie",
                "last_name": "Q",
                "username": "email@company.com",
                "email": "email@company.com"
            }
        }
    },
    "more": false
}
'

--// put users object into variable
DECLARE @users NVARCHAR(MAX) = (
    SELECT users.users
    FROM OPENJSON(@user_response)
        WITH (results NVARCHAR(MAX) AS JSON
                , more VARCHAR(20)) as body

            CROSS APPLY OPENJSON(results)
                WITH (users NVARCHAR(MAX) AS JSON) as users
)

--// extract the keys from the users object
DECLARE @keys TABLE ([key] VARCHAR(100), [index] INT)

INSERT INTO @keys
SELECT [key], ROW_NUMBER() OVER (ORDER BY [key]) 'index'
FROM OPENJSON(@users)

--// initialize looping variables
DECLARE @i INT = 1
    , @key VARCHAR(100)
    , @sql NVARCHAR(MAX)

SELECT @sql = 'DECLARE @user_response NVARCHAR(MAX) = N''' + @user_response + ''''

--// loop through keys and UNION individual queries on the original json response
WHILE (@i <= (SELECT MAX([index]) FROM @keys))
BEGIN

    SELECT @key = (SELECT [key] FROM @keys WHERE [index] = @i)

    SELECT @sql = @sql + CASE WHEN @i = 1 THEN '' ELSE 'UNION' END + '
        SELECT user_data.*
        FROM OPENJSON(@user_response)
            WITH (results NVARCHAR(MAX) AS JSON)

                CROSS APPLY OPENJSON(results)
                    WITH (users NVARCHAR(MAX) AS JSON)

                    CROSS APPLY OPENJSON(users)
                        WITH ([' + @key + '] NVARCHAR(MAX) AS JSON)

                    CROSS APPLY OPENJSON([' + @key + '])
                        WITH (id INT
                                , first_name VARCHAR(100)
                                , last_name VARCHAR(100)
                                , username VARCHAR(200)
                                , email VARCHAR(200)) as [user_data]'


    SELECT @i = @i + 1

END

--// execute final dynamic query
EXEC sp_executesql @sql

这条语句的结果集如下所示:

|id   |first_name|last_name|username         |email            |
|-----|----------|---------|-----------------|-----------------|
|98765|Suzie     |Q        |email@company.com|email@company.com|
|321123|John     |Wayne    |notreal@email.com|notreal@email.com|
|12345|Demo      |User     |demo@gmail.com   |demo@gmail.com   |

提前感谢您的想法和反馈。

像这样:

select u.*
from openjson(@user_response, '$.results.users') d
cross apply 
openjson(d.value)
with
(
   id         int          '$.id',
   first_name varchar(200) '$.first_name',
   last_name  varchar(200) '$.last_name',
   username   varchar(200) '$.username',
   email      varchar(200) '$.email'
) u
 

如果你需要它,你可以通过 d.[key]

获取对象名称