OPENJSON 与动态 JSON 键值对

Question

我们有一个 JSON 文档，其中周数是动态键。我们想将它们加载到关系 table.

如果我们硬编码周数，我们就能够实现关系结果集，如下所示。但是，它看起来像是一种带有硬编码值的迂回方法。我们想让它动态化。

TSQL 中有没有办法将键值对动态映射为关系 table？

DECLARE @json NVARCHAR(MAX) = N'[
  {
    "ID": "1",
    "Measure": "Current Sales",
    "2019Week12": "33",
    "2019Week13": "33",
    "2019Week14": "34"
  },
  {
    "ID": "2",
    "Measure": "Current Sales",
    "2019Week12": "",
    "2019Week13": "10",
    "2019Week14": "60"
  }]';

SELECT ID,Measure, WeekNumber, Sales
FROM
(   SELECT * FROM OPENJSON(@json)
    with 
    ( ID int '$.ID',
    Measure VARCHAR(30) '$.Measure',
    [2019Week12] INT '$."2019Week12"',
    [2019Week13] INT '$."2019Week13"',
    [2019Week14] INT '$."2019Week14"'
    )
) as p
UNPIVOT
(
Sales FOR WeekNumber IN ([2019Week12],[2019Week13],[2019Week14]) 
) as unpvt

我们得到的结果集是：

+----+---------------+------------+-------+
| ID |    Measure    | WeekNumber | Sales |
+----+---------------+------------+-------+
|  1 | Current Sales | 2019Week12 |    33 |
|  1 | Current Sales | 2019Week13 |    33 |
|  1 | Current Sales | 2019Week14 |    34 |
|  2 | Current Sales | 2019Week12 |     0 |
|  2 | Current Sales | 2019Week13 |    10 |
|  2 | Current Sales | 2019Week14 |    60 |
+----+---------------+------------+-------+

Answer 1

您没有说明预期的输出。我得到的是：您想获得与上面相同的结果，而无需按字面意思指定名称。我希望我没听错：

SELECT   JSON_VALUE(A.[value],'$.ID') AS ID
        ,JSON_VALUE(A.[value],'$.Measure') AS Measure
        ,B.[key] AS [varName]
        ,B.[value] AS [varValue]  
        ,ROW_NUMBER() OVER(PARTITION BY JSON_VALUE(A.[value],'$.ID') ORDER BY B.[key]) RowIndex
FROM OPENJSON(@json) A
CROSS APPLY OPENJSON(A.[value]) B
WHERE b.[key] NOT IN('ID','Measure');

结果

+----+---------------+------------+----------+----------+
| ID | Measure       | varName    | varValue | RowIndex |
+----+---------------+------------+----------+----------+
| 1  | Current Sales | 2019Week12 | 33       | 1        |
+----+---------------+------------+----------+----------+
| 1  | Current Sales | 2019Week13 | 33       | 2        |
+----+---------------+------------+----------+----------+
| 1  | Current Sales | 2019Week14 | 34       | 3        |
+----+---------------+------------+----------+----------+
| 2  | Current Sales | 2019Week12 |          | 1        |
+----+---------------+------------+----------+----------+
| 2  | Current Sales | 2019Week13 | 10       | 2        |
+----+---------------+------------+----------+----------+
| 2  | Current Sales | 2019Week14 | 60       | 3        |
+----+---------------+------------+----------+----------+

简而言之：

• 我们使用 OPENJSON() 深入研究您的 json 字符串。这将 return 派生集 A.
中包含的两个对象
• 现在我们再次使用OPENJSON()传入A.[value]，也就是json对象本身。
• 这将 return 所有包含的项目，但我们在 WHERE.
中抑制 ID 和 Measurement
• 我们使用 JSON_VALUE().
直接从 A.[value] 获取 ID 和 Measurement 这两个特殊列

更新

一个增强可能是这样的：

SELECT   C.ID
        ,C.varName AS [varName]
        ,TRY_CAST(LEFT(C.varName,4) AS INT) AS MeasureYear
        ,TRY_CAST(RIGHT(C.varName,2) AS INT) AS MeasureWeek
        ,C.varContent AS [varValue]  
        ,ROW_NUMBER() OVER(PARTITION BY C.ID ORDER BY C.varName) RowIndex
FROM OPENJSON(@json) A
CROSS APPLY OPENJSON(A.[value]) B
CROSS APPLY (SELECT JSON_VALUE(A.[value],'$.ID') AS ID
                   ,JSON_VALUE(A.[value],'$.Measure') AS Measure
                   ,B.[key] AS varName
                   ,B.[value] AS varContent) C
WHERE C.varName NOT IN('ID','Measure');

想法：再添加一个 APPLY 允许 return 值作为 常规列 。这使得继续处理这些值变得更加容易，并使这些内容更具可读性。