Question

我有三张桌子父母，儿童和儿童形式的孙子：

+----------------+   +----------------+   +---------------------+
| parent         |   | child          |   | grandchild          |
+----------------+   +----------------+   +---------------------+
| parent_id (PK) |   | child_id (PK)  |   | grandchild_id (PK)  |
| parent_data    |   | child_data     |   | grandchild_data     |
+----------------+   | parent_id (FK) |   | child_id (FK)       |
                     +----------------+   +---------------------+

PK =自动递增主键。
FK =外键。

我想要一个存储过程，它可以复制父表中的记录以及子表和孙表中的任何相关记录。我可以将父数据和子数据复制好，这是我正在努力的孙子表。就我而言：

CREATE FUNCTION sf_copy_parent(p_parent_id INT) RETURNS INT
BEGIN
    DECLARE new_parent_id INT;

    -- create new parent record
    INSERT INTO parent(parent_data)
        SELECT parent_data FROM parent
        WHERE parent_id=p_parent_id;
    SET new_parent_id=LAST_INSERT_ID();

    -- copy child records
    INSERT INTO child(child_data,parent_id)
        SELECT child_data,new_parent_id FROM child
        WHERE parent_id=p_parent_id;

    -- copy grandchild records ???


    -- return
    RETURN new_parent_id;
END

如果这很重要，我正在使用Mysql5.5。

Answer 1

尝试此SELECT查询（它使用'p_parent_id'和'new_parent_id'变量） -

SET @r1 = 1;
SET @child_id = NULL;
SET @r2 = 0;

SELECT c1.grandchild_data, c2.child_id FROM (
  SELECT @r1 := if(c.child_id IS NULL OR c.child_id <> @child_id, @r1 + 1, @r1) rank, @child_id := c.child_id, c.child_id, g.grandchild_data FROM child c
  JOIN grandchild g
    ON c.child_id = g.child_id
  WHERE
    c.parent_id = p_parent_id
  ORDER BY
    c.child_id, g.grandchild_id
  ) c1
JOIN (SELECT @r2 := @r2 + 1 rank, child_id FROM child WHERE parent_id = new_parent_id ORDER BY child_id) c2
  ON c1.rank = c2.rank;

如果有效，我们会将其重写为INSERT..SELECT语句，或者尝试自己动手;）

Answer 2

编辑：我已经创建了另一个我认为比这个更简单，更好的答案。

完成后的存储过程然后使用来自@Devart的答案：

CREATE FUNCTION `sp_copy`(p_parent_id INT) RETURNS int(11)
BEGIN
    DECLARE new_parent_id INT;

    -- create new parent record
    INSERT INTO parent(parent_data)
        SELECT parent_data FROM parent
        WHERE parent_id=p_parent_id;
    SET new_parent_id=LAST_INSERT_ID();

    -- copy child records
    INSERT INTO child(child_data,parent_id)
        SELECT child_data,new_parent_id FROM child
        WHERE parent_id=p_parent_id;

    -- copy grandchild records
    SET @r1 = 1;
    SET @child_id = NULL;
    SET @r2 = 0;
    INSERT INTO grandchild(grandchild_data,child_id) SELECT c1.grandchild_data, c2.child_id FROM (
    SELECT @r1 := if(c.child_id IS NULL OR c.child_id <> @child_id, @r1 + 1, @r1) rank, @child_id := c.child_id, c.child_id, g.grandchild_data FROM child c
    JOIN grandchild g
        ON c.child_id = g.child_id
    WHERE
        c.parent_id = p_parent_id
    ORDER BY
        c.child_id, g.grandchild_id
    ) c1
    JOIN (SELECT @r2 := @r2 + 1 rank, child_id FROM child WHERE parent_id = new_parent_id ORDER BY child_id) c2
        ON c1.rank = c2.rank;

    -- return new parent id
    RETURN new_parent_id;
END

Answer 3

我已经学到了更多，我相信@Devart的贡献可以简化一点。我现在使用的解决方案如下所示。

注意：如果使用的表是InnoDB而不是MyISAM，则此解决方案只能正常工作。我认为这与MyISAM的默认排序行为如何工作有关。（InnoDB使用主键，而MyISAM使用插入顺序。）特别是，对于MyISAM表，“复制孙子”部分中生成的两组排名彼此不“同步”。在相关部分中添加SORT BY子句似乎也没有什么区别。

CREATE PROCEDURE sp_copy(p_parent_id INT)
BEGIN
    DECLARE new_parent_id INT;

    -- copy parent
    INSERT INTO parent(parent_data) SELECT parent_data FROM parent WHERE parent_id=p_parent_id;
    SET new_parent_id:=LAST_INSERT_ID();

    -- copy child(s)
    INSERT INTO child(child_data, parent_id)
        SELECT child_data, new_parent_id FROM child WHERE parent_id=p_parent_id;

    -- copy grandchild(s)
    SET @rank1:=0;        
    SET @rank2:=0;            
    INSERT INTO grandchild(grandchild_data, child_id) SELECT gc.grandchild_data, c2.child_id FROM
        (SELECT child_id, @rank1:=@rank1+1 as rank FROM child WHERE parent_id=p_parent_id) c1
        INNER JOIN
        (SELECT child_id, @rank2:=@rank2+1 as rank FROM child WHERE parent_id=new_parent_id) c2 ON c1.rank=c2.rank
        INNER JOIN grandchild gc ON c1.child_id=gc.child_id;

END

此外，为了处理一个伟大的子表，可以使用复制孙子记录的相同原则。唯一额外的复杂性是在两个子查询中的每一个中添加连接。这是必需的，因为child.parent_id字段需要在WHERE子句中引用：

-- copy greatgrandchild(s)
SET @rank1:=0;        
SET @rank2:=0;            
INSERT INTO greatgrandchild(greatgrandchild_data, grandchild_id) SELECT ggc.greatgrandchild_data, gc2.grandchild_id FROM
    (SELECT grandchild_id, @rank1:=@rank1+1 as rank FROM grandchild INNER JOIN child ON child.child_id=grandchild.child_id WHERE parent_id=p_parent_id) gc1
    INNER JOIN
    (SELECT grandchild_id, @rank2:=@rank2+1 as rank FROM grandchild INNER JOIN child ON child.child_id=grandchild.child_id WHERE parent_id=new_parent_id) gc2 ON gc1.rank=gc2.rank
    INNER JOIN greatgrandchild ggc ON gc1.grandchild_id=ggc.grandchild_id;

Mysql存储过程，用于复制父/子/孙子层次结构中的记录

3 个答案: