Динамический ROOT с SQL Server для JSON PATH

Question

У меня есть таблица ключевых слов, которую я хочу отобразить в объект Json:

ID  Keyword     Text    Value
4   Category    A       10
5   Category    B       20
1   Season      Winter  1
2   Season      Spring  2
3   Season      Summer  3
9   Season      Fall    4
6   UnitType    Ft      Feet
7   UnitType    Set     Set
8   UnitType    $   Dollar

, которая группирует каждую категорию в свой собственный объект.

Мне удалось создать это:

{"Keywords":[
{"Keyword":"Category","Values":[{"Value":"10","Text":"A"},{"Value":"20","Text":"B"}]},
{"Keyword":"Season","Values":[{"Value":"1","Text":"Winter"},{"Value":"2","Text":"Spring"},{"Value":"3","Text":"Summer"},{"Value":"4","Text":"Fall"}]},
{"Keyword":"UnitType","Values":[{"Value":"Dollar","Text":"$"},{"Value":"Feet","Text":"Ft"},{"Value":"Set","Text":"Set"}]}
]}

используя:

SELECT T.Keyword AS 'Keyword', 
(SELECT [subT].[Value] AS 'Value', [subT].[Text] AS 'Text' 
FROM tblKeywords subT WHERE subT.Keyword=T.Keyword 
ORDER BY [subT].[Value]
FOR JSON PATH) AS 'Values'
FROM tblKeywords T
GROUP BY T.Keyword
FOR JSON PATH, ROOT('Keywords')

Но я хотел бы иметь возможность установить ROOT для подзапроса в качестве самого ключевого слова.Это вообще возможно?

Желаемый результат:

{"Keywords":[
{"Category":[{"Value":"10","Text":"A"},{"Value":"20","Text":"B"}]},
{"Season":[{"Value":"1","Text":"Winter"},{"Value":"2","Text":"Spring"},{"Value":"3","Text":"Summer"},{"Value":"4","Text":"Fall"}]},
{"UnitType":[{"Value":"Dollar","Text":"$"},{"Value":"Feet","Text":"Ft"},{"Value":"Set","Text":"Set"}]}
]}

Answer 1

Другой возможный подход с использованием комбинации FOR JSON и манипуляции со строками, но без динамического оператора:

Введите:

CREATE TABLE #tblKeywords (
   Id int, 
   [Keyword] nvarchar(50), 
   [Text] nvarchar(50), 
   [Value] nvarchar(50)
);
INSERT INTO #tblKeywords
   (ID, [Keyword], [Text], [Value])
VALUES
   (4, 'Category', 'A',      '10'),
   (5, 'Category', 'B',      '20'),
   (1, 'Season',   'Winter', '1'),
   (2, 'Season',   'Spring', '2'),
   (3, 'Season',   'Summer', '3'),
   (9, 'Season',   'Fall',   '4'),
   (6, 'UnitType', 'Ft',     'Feet'),
   (7, 'UnitType', 'Set',    'Set'),
   (8, 'UnitType', '$',      'Dollar')

T-SQL:

SELECT CONCAT(
   N'{"Keywords":[',
   STUFF(
      (
      SELECT DISTINCT CONCAT(N',{"', k.[Keyword], '":', c.[Json], N'}')
      FROM #tblKeywords k
      CROSS APPLY (
         SELECT [Value], [Text]
         FROM #tblKeywords
         WHERE [Keyword] = k.[Keyword]
         FOR JSON PATH
      ) c([Json])
      FOR XML PATH('')
      ), 1, 1, N''
   ),  
   N']}'
) AS JsonOutput

Выход:

JsonOutput

{"Keywords":[{"Category":[{"Value":"10","Text":"A"},{"Value":"20","Text":"B"}]},{"Season":[{"Value":"1","Text":"Winter"},{"Value":"2","Text":"Spring"},{"Value":"3","Text":"Summer"},{"Value":"4","Text":"Fall"}]},{"UnitType":[{"Value":"Feet","Text":"Ft"},{"Value":"Set","Text":"Set"},{"Value":"Dollar","Text":"$"}]}]}

Answer 2

Исходные данные:

DROP TABLE IF EXISTS #tblKeywords;
CREATE TABLE #tblKeywords (ID INT, Keyword NVARCHAR(255), [Text] NVARCHAR(255), Value NVARCHAR(255));
INSERT INTO #tblKeywords(ID,Keyword,Text,[Value])VALUES
     (4,'Category','A','10')
    ,(5,'Category','B','20')
    ,(1,'Season','Winter','1')
    ,(2,'Season','Spring','2')
    ,(3,'Season','Summer','3')
    ,(9,'Season','Fall','4')
    ,(6,'UnitType','Ft','Feet')
    ,(7,'UnitType','Set','Set')
    ,(8,'UnitType','$','Dollar')
;

Я знаю, динамический SQL, однако:

DECLARE @DynSql NVARCHAR(MAX) = (
    SELECT 'SELECT ' + STUFF((
            SELECT DISTINCT N',(SELECT [subT].[Value] AS [Value], [subT].[Text] AS [Text] 
            FROM #tblKeywords subT 
            WHERE subT.Keyword = ''' + t.Keyword + N''' 
            ORDER BY [subT].[Value]
            FOR JSON PATH
            ) AS [' + t.Keyword + N']' + CHAR(13) + '   '
            FROM #tblKeywords t
            FOR XML PATH(''),TYPE).value('(./text())[1]','NVARCHAR(MAX)'),1,1,'')
        + 'FOR JSON PATH, ROOT(''Keywords'')'
)
;
--PRINT @DynSql;
EXEC (@DynSql);