Question

问题

我有一个递归的CTE查询，但是在创建循环时失败了。我已经修复了简单循环（例如1 - > 2 - > 1），但无法修复更复杂的循环（例如1 - > 2 - > 3 - > 2）。

查询明细

测试表有两列：Base和Parent。我想要一个包含所有祖先的清单。

如果从test2开始，我的查询将对下面的示例数据起作用，但在test1开始时则不行。

示例数据

Base    Parent
----    ------
test1   test2
test2   test3
test3   test2

SQL查询（我的尝试修复标记在评论中）

;with sample_data (Base, Parent) as (
    select 'test1', 'test2'
    union select 'test2', 'test3'
    union select 'test3', 'test2'
),
nt_list (Base, Ancestor, [level]) as (
        select  Base,
                Parent Ancestor,
                1 [level]
        from    sample_data
        where   Base = 'test1' -- START HERE
        union all
        select  ntl.Base,
                nt.Parent,
                ntl.[level] + 1 [level]
        from    nt_list ntl
        join    sample_data nt on ntl.Ancestor = nt.Base
        where   nt.Parent <> ntl.Base -- fix recursive bug (e.g. 1 -> 2 -> 1)
        -- WHAT I TRIED TO ADD BUT CANNOT: (e.g. 1 -> 2 -> 3 -> 2)
        and     nt.Parent in (select Ancestor from nt_list)
)
select  distinct
        ntl.Base,
        ntl.Ancestor
from    nt_list ntl
order by Ancestor

SQL错误：公用表表达式'nt_list'的递归成员具有多个递归引用。

Answer 1

最终版本。假设'/'永远不会成为基本名称或父名称的一部分。

;with sample_data (Base, Parent) as (
    -- TEST 1
    --        select 'test1', 'test2'
    --union   select 'test2', 'test3'
    --union   select 'test3', 'test2'
    -- TEST 2
            select 'test1', 'test2'
    union   select 'test2', 'test3'
    union   select 'test3', 'test4'
    union   select 'test3', 'test9'
    union   select 'test4', 'test5'
    union   select 'test5', 'test3'
    union   select 'test9', 'test8'
    -- TEST 3
    --        select 'test1', 'test2'
    --union   select 'test2', 'test3'
    --union   select 'test3', 'test1'
    -- TEST 4
    --        select  'test1', 'test1'
    --union   select  'test1', 'test2'
),
nt_list (Base, Ancestor, [level], [path]) as (
        select  Base,
                Parent Ancestor,
                1 [level],
                '/' + convert(varchar(max), rtrim(Base)) + '/' [path]
        from    sample_data
        where   Base = 'test1' -- START HERE
        union all
        select  ntl.Base,
                nt.Parent,
                ntl.[level] + 1 [level],
                ntl.[path] + rtrim(nt.Base) + '/'
        from    nt_list ntl
        join    sample_data nt on ntl.Ancestor = nt.Base
        where   ntl.path not like '%/' + rtrim(nt.Parent) + '/%'
)
select  distinct
        ntl.Base,
        ntl.Ancestor
from    nt_list ntl
order by Ancestor

Answer 2

您可以使用

;WITH nt_list (Base, Ancestor, [level], cycle, path)
     AS (SELECT Base,
                Parent                                                            Ancestor,
                1                                                                 [level],
                0                                                                 AS cycle,
                CAST('.' AS VARCHAR(max)) + ISNULL(Parent, '') + '.' + Base + '.' AS [path]
         FROM   NoteTest
         WHERE  Base = 'test1'
         UNION ALL
         SELECT ntl.Base,
                nt.Parent,
                ntl.[level] + 1                   [level],
                CASE
                  WHEN ntl.[path] LIKE '%.' + LTRIM(nt.Base) + '.%' THEN 1
                  ELSE 0
                END                               AS cycle,
                ntl.[path] + LTRIM(nt.Base) + '.' AS [path]
         FROM   nt_list ntl
                JOIN NoteTest nt
                  ON ntl.Ancestor = nt.Base
                     AND ntl.cycle = 0)
SELECT ntl.Base,
       ntl.Ancestor
FROM   nt_list ntl
ORDER  BY Ancestor

Answer 3

我目前的解决方法是添加级别限制（向查询添加and ntl.[level] <= 100）并让select distinct删除重复的条目。

Answer 4

这种情况很完美，但正如您对更多项目所做的那样：

    select 'test1', 'test2'
         union select 'test2', 'test3'
         union select 'test3', 'test4'
         union select 'test4', 'test5'
         union select 'test10', 'test11'
         union select 'test11', 'test30'
         ...

结果应为：

    test1   test1 <- adding this
    test1   test2
    test1   test3
    test1   test4
    test1   test5
    test10  test10 <- adding this to multiple bases
    test10  test11
    test10  test30

SQL递归CTE：通过多个递归引用来阻止递归循环

问题

查询明细

4 个答案: