Sql server 2005 TSQL递归-表中具有双重值的最大递归错误

相关问题：

我要问第二个关于递归的问题，因为这一次它是不同的（在我看来），所以请不要删除它

我有一个包含值的表：

ORDER_ID   Previous_STATUS_ID   Next_STATUS_ID  Create_Date
2          null                 1               '2012-01-02'
2          1                    2               '2012-01-03'
2          2                    3               '2012-01-04'
2          3                    1               '2012-01-05'
3          1                    2               '2012-01-06'
2          2                    3               '2012-01-10'
2          3                    5               '2012-01-13'
2          5                    1               '2012-01-22'
2          1                    2               '2012-01-22'

这很好：

with change_tree as 
(
  SELECT order_id,
         previous_status_id, 
         next_status_id,
         cast(next_status_id as varchar(max)) as status_path
  FROM status_change
  WHERE previous_status_id = 5
    AND order_id = 2

  UNION ALL 

  SELECT sc.order_id,
         sc.previous_status_id,
         sc.next_status_id,
         ct.status_path + ',' + cast(sc.next_status_id as varchar(max))
  FROM status_change sc 
    JOIN change_tree ct ON ct.next_status_id = sc.previous_status_id AND ct.order_id = sc.order_id
)
SELECT *
FROM change_tree
WHERE status_path = '5,1';

但如果我这样修改它：

with change_tree as 
(
  SELECT sc.order_id,
         sc.previous_status_id, 
         sc.next_status_id,
         cast(sc.next_status_id as varchar(max)) as status_path,
         sc.Create_Date as StartDate,
         sc.Create_Date as EndDate
  FROM status_change sc
  WHERE previous_status_id = 1
    AND order_id = 2

  UNION ALL 

  SELECT sc.order_id,
         sc.previous_status_id,
         sc.next_status_id,
         ct.status_path + ',' + cast(sc.next_status_id as varchar(max))
         sc.Create_Date as StartDate,
         st.Create_Date as EndDate
  FROM status_change sc 
    JOIN change_tree ct ON ct.next_status_id = sc.previous_status_id AND ct.order_id = sc.order_id
)
SELECT *
FROM change_tree
WHERE status_path = '1,2,3';

我得到最大递归错误

这是我希望得到的数据：

ORDER_ID   StartDate        EndDate
2          '2012-01-02'     '2012-01-04' 
2          '2012-01-05'     '2012-01-10' 

这将作为报告进行。所以我想把结果存储在另一个表中，晚上只做那些新的或状态序列已经结束的订单

因此，在我的报告表中，我必须有这两条记录，在将下一条记录添加到status_change表之后，如下所示：

ORDER_ID   Previous_STATUS_ID   Next_STATUS_ID  Create_Date
2          2                    3               '2012-02-25'

我的过程（函数应仅添加到报告表的最后一个序列）。

---

希望您能理解我的观点：）

您在循环中被捕获，因为首先您选择了此行：

2          1                    2               '2012-01-22'

然后

然后

然后一次又一次地排第一行。您需要检查您的数据是否与您的报告足够一致

或者，您可能需要在CTE中添加以下条件：

ct.previous_status_id > sc.previous_status_id 

或者，您可以通过控制递归级别的addind参数更改脚本，如下所示：

with change_tree as  
( 
  SELECT sc.order_id, 
         sc.previous_status_id,  
         sc.next_status_id, 
         cast(sc.next_status_id as varchar(max)) as status_path, 
         sc.Create_Date as StartDate, 
         sc.Create_Date as EndDate,
         1 AS deep 
  FROM status_change sc 
  WHERE previous_status_id = 1 
    AND order_id = 2 

  UNION ALL  

  SELECT sc.order_id, 
         sc.previous_status_id, 
         sc.next_status_id, 
         ct.status_path + ',' + cast(sc.next_status_id as varchar(max)) 
         sc.Create_Date as StartDate, 
         st.Create_Date as EndDate,
         ct.deep + 1 AS deep
  FROM status_change sc  
    JOIN change_tree ct ON ct.next_status_id = sc.previous_status_id AND ct.order_id = sc.order_id 
  WHERE deep < 3
) 
SELECT * 
FROM change_tree 
WHERE status_path = '1,2,3'; 

---

这正是我现在的处境。我来核对你的答案。还有一件事-如何以增量方式完成该报告？这样我的程序就不必每天晚上检查整个状态变更表了@安德烈·古里诺维奇认为，这可能是另一个问题的主题。但你为什么担心你的桌子会被完全重新加工呢？有这么大吗？处理需要很多时间吗？我有大约800k+行和大约30个模式要搜索（不同的状态和路径组合）。所以我认为，对每一行进行每一次组合都会给我带来大约2400万次操作。但订单仍然被添加到表中，我的模式数量也在增加。报告类型为“按需”，所以用户点击按钮，片刻后他应该会得到该报告。这就是为什么我认为聚合表是必要的。如果我错了，请纠正我：）@Andrey GurinovCan您的旧数据已更改？我的意思是，增量过程应该只向聚合表添加新记录，还是应该更新/删除旧记录？对于30种模式，我认为我建议的最后一个选项将是最有用的。您可以在CTE和外部选择中使用OR运算符组合不同的状态路径组合。

ct.previous_status_id > sc.previous_status_id 

with change_tree as  
( 
  SELECT sc.order_id, 
         sc.previous_status_id,  
         sc.next_status_id, 
         cast(sc.next_status_id as varchar(max)) as status_path, 
         sc.Create_Date as StartDate, 
         sc.Create_Date as EndDate,
         1 AS deep 
  FROM status_change sc 
  WHERE previous_status_id = 1 
    AND order_id = 2 

  UNION ALL  

  SELECT sc.order_id, 
         sc.previous_status_id, 
         sc.next_status_id, 
         ct.status_path + ',' + cast(sc.next_status_id as varchar(max)) 
         sc.Create_Date as StartDate, 
         st.Create_Date as EndDate,
         ct.deep + 1 AS deep
  FROM status_change sc  
    JOIN change_tree ct ON ct.next_status_id = sc.previous_status_id AND ct.order_id = sc.order_id 
  WHERE deep < 3
) 
SELECT * 
FROM change_tree 
WHERE status_path = '1,2,3'; 

with change_tree as  
( 
  SELECT sc.order_id, 
         sc.previous_status_id,  
         sc.next_status_id, 
         cast(sc.next_status_id as varchar(max)) as status_path, 
         sc.Create_Date as StartDate, 
         sc.Create_Date as EndDate
  FROM status_change sc 
  WHERE previous_status_id = 1 
    AND order_id = 2 

  UNION ALL  

  SELECT sc.order_id, 
         sc.previous_status_id, 
         sc.next_status_id, 
         ct.status_path + ',' + cast(sc.next_status_id as varchar(max)) 
         sc.Create_Date as StartDate, 
         st.Create_Date as EndDate
  FROM status_change sc  
    JOIN change_tree ct ON ct.next_status_id = sc.previous_status_id AND ct.order_id = sc.order_id 
  WHERE 
     '1,2,3' LIKE ct.status_path + ',' + cast(sc.next_status_id as varchar(max)) + '%'
) 
SELECT * 
FROM change_tree 
WHERE status_path = '1,2,3';