SQL Server 2016 - 从单个表将逻辑层次结构构建到 JSON 结构中

Question

我有一个表格，其中包含与一个非常大的文档有关的信息。该表如下所示：

ID      | Title         | Parent_ID     | <other columns>
--------+---------------+---------------+-------------------
0       | Root          | null          | ...
1       | Introduction  | 0             | ...
2       | Glossary      | 1             | ...
3       | Audience      | 1             | ...
4       | "A"           | 2             | ...
5       | "B"           | 2             | ...
6       | "C"           | 2             | ...

结果 JSON 应如下所示（为清楚起见，省略了 <other columns> 部分）：

{"ID"        : 0        ,
 "Title"     : "Root"   ,
 "Contents"  : [{"ID"        : 1             ,
                 "Title"     : "Introduction",
                 "Contents"  : [{"ID"          : 2           ,
                                 "Title"       : "Glossary"  ,
                                 "Contents"    : [{"ID"       : 4       ,
                                                   "Title"    : "A"     ,
                                                   "Contents" : []       },
                                                  {"ID"       : 5       ,
                                                   "Title"    : "B"     ,
                                                   "Contents" : []       },
                                                  {"ID"       : 6       ,
                                                   "Title"    : "C"     ,
                                                   "Contents" : []       }]
                                },
                                {"ID"       : 3          ,
                                 "Title"    : "Audience" ,
                                 "Contents" : []
                                }
                               ]
                },
                ....
               ]
}

我确实有一个简单的（递归）过程可以处理这个问题，但希望有一种更简单的方法使用 DBMS 的 JSON 功能（也许使用 CTE？）。

Answer 1

如果已知父/子关系的最大深度？
然后你可以像这个例子一样把它拉下来：

dbfiddle here

的测试

测试数据：

CREATE TABLE documentdetails 
(
   ID INT PRIMARY KEY NOT NULL,
   Title VARCHAR(30) NOT NULL,
   Parent_ID INT,
   FOREIGN KEY (Parent_ID) REFERENCES documentdetails (ID) 
);

INSERT INTO documentdetails (ID, Title, Parent_ID) VALUES 
(1, 'Root', null), (2, 'Introduction', 1), (3, 'Glossary', 1),
(4, 'Audience', 1), (5, 'A', 2), (6,'B', 2), (7, 'C', 2), 
(8, 'Foo', null), (9, 'Bar Intro', 8), (10, 'Glossy stuff', 8), (11, 'What The Fook', 8),
(12, 'Yo', 9), (13, 'Ai', 10), (14, 'Potato', 11);

查询：

SELECT 
root.ID, 
root.Title,
(  
   SELECT lvl0.ID, lvl0.Title, 0 as Depth,
   (  
      SELECT lvl1.ID, lvl1.Title, 1 as Depth,
      ( 
         SELECT lvl2.ID, lvl2.Title, 2 as Depth,
         ( 
            SELECT lvl3.ID, lvl3.Title, 3 as Depth
            FROM documentdetails lvl3
            WHERE lvl3.Parent_ID = lvl2.ID
            FOR JSON PATH
         ) AS Contents
         FROM documentdetails lvl2
         WHERE lvl2.Parent_ID = lvl1.ID
         FOR JSON PATH
      ) AS Contents
      FROM documentdetails lvl1
      WHERE lvl1.Parent_ID = lvl0.ID
      FOR JSON PATH
   ) AS Contents
   FROM documentdetails lvl0
   WHERE lvl0.ID = root.ID
   FOR JSON PATH
) AS Contents
FROM documentdetails root
WHERE root.Parent_ID IS NULL;

结果：

ID  Title   Contents
--  -----   --------
1   Root    [{"ID":1,"Title":"Root","Depth":0,"Contents":[{"ID":2,"Title":"Introduction","Depth":1,"Contents":[{"ID":5,"Title":"A","Depth":2},{"ID":6,"Title":"B","Depth":2},{"ID":7,"Title":"C","Depth":2}]},{"ID":3,"Title":"Glossary","Depth":1},{"ID":4,"Title":"Audience","Depth":1}]}]
8   Foo     [{"ID":8,"Title":"Foo","Depth":0,"Contents":[{"ID":9,"Title":"Bar Intro","Depth":1,"Contents":[{"ID":12,"Title":"Yo","Depth":2}]},{"ID":10,"Title":"Glossy stuff","Depth":1,"Contents":[{"ID":13,"Title":"Ai","Depth":2}]},{"ID":11,"Title":"What The Fook","Depth":1,"Contents":[{"ID":14,"Title":"Potato","Depth":2}]}]}]

如果您不知道表格中的最大深度？
这是使用递归 CTE 来了解这一点的 SQL。

WITH RCTE AS
(
   SELECT ID as rootID, 0 as lvl, ID, Parent_ID
   FROM documentdetails
   WHERE Parent_ID IS NULL

   UNION ALL

   SELECT r.rootID, lvl + 1, t.ID, t.Parent_ID
   FROM RCTE r
   JOIN documentdetails t ON t.Parent_ID = r.ID
)
SELECT rootID, MAX(lvl) as Depth, COUNT(*) as Nodes
FROM RCTE
GROUP BY rootID
ORDER BY MAX(lvl) DESC, COUNT(*) DESC;

或者相反，从孩子那里播种。
（如果 ID 是 PRIMARY KEY，那么这可能会更快，因为 ID 上的 JOIN）

WITH RCTE AS
(
   SELECT ID as baseID, 0 as lvl, ID, Parent_ID
   FROM documentdetails
   WHERE Parent_ID IS NOT NULL

   UNION ALL

   SELECT r.baseID, lvl + 1, t.ID, t.Parent_ID
   FROM RCTE r
   JOIN documentdetails t ON t.ID = r.Parent_ID
)
SELECT ID as rootID, MAX(lvl) as Depth
FROM RCTE 
WHERE Parent_ID IS NULL
GROUP BY ID
ORDER BY MAX(lvl) DESC, COUNT(*) DESC;