按 parents 分页时，显示他们的 children

Question

我有一个包含 parent 和 child 行的结果集。（Child 行从来没有 childs）。我需要对其进行分页（考虑排序）以便：

首先需要select只有分页页上的parent（例如，当page size=10时，它必须包含<=10parent行），并且然后 "stick" 给他们 child 那些在分页页面中的 parents。

源结果集如下所示：

+----+-----------+-------------+
| ID | PARENT_ID | SORT_COLUMN |
+----+-----------+-------------+
|  1 |           | y           |
|  2 |         1 | z           |
|  3 |           | u           |
|  4 |           | q           |
|  5 |         4 | o           |
|  6 |         4 | p           |
|  7 |           | c           |
+----+-----------+-------------+

~想要的结果：

+----+-----------+-------------+----+----------+
| ID | PARENT_ID | SORT_COLUMN | RN | RN_CHILD |
+----+-----------+-------------+----+----------+
|  7 |           | c           |  1 |          |
|  4 |           | q           |  2 |          |
|  5 |         4 | o           |  2 |        1 |
|  6 |         4 | p           |  2 |        2 |
|  3 |           | u           |  3 |          |
|  1 |           | y           |  4 |          |
|  2 |         1 | z           |  4 |        1 |
+----+-----------+-------------+----+----------+

现在我是这样做的：

with
  cte as
    (select 1 as id, null as parent_id, 'y' as sort_column from dual
     union all
     select 2 as id, 1 as parent_id, 'z' as sort_column from dual
     union all
     select 3 as id, null as parent_id, 'u' as sort_column from dual
     union all
     select 4 as id, null as parent_id, 'q' as sort_column from dual
     union all
     select 5 as id, 4 as parent_id, 'o' as sort_column from dual
     union all
     select 6 as id, 4 as parent_id, 'p' as sort_column from dual
     union all
     select 7 as id, null as parent_id, 'c' as sort_column from dual)
select 
  *
from
  (select
    t.*, 
    dense_rank() over (order by 
                         case when t.parent_id is null 
                           then 
                             t.sort_column 
                           else 
                             (select t2.sort_column from cte t2 where t2.id = t.parent_id) 
                         end) as RN,
    case 
      when parent_id is null 
        then
          null 
        else 
          row_number() over (partition by t.parent_id order by t.sort_column) 
    end as RN_CHILD
from cte t)
--where RN between :x and :y
order by RN, RN_CHILD nulls first

但我认为无需对结果集进行不必要的额外访问即可完成此操作。 (select t2.sort_column from cte t2 where t2.id = t.parent_id).

怎么做？

UPD： parents 必须按 sort_column 排序，并且 childs 在 parents 中也必须是按 sort_column.

排序

Answer 1

首先，值语句有一个技巧，其次，您可以使用左连接和合并来获得您想要的内容，如下所示：

with
  cte(id,parent_id,sort_column) as
    (
      VALUES 
        (1, null, 'y' ),
        (2, 1   , 'z' ),
        (3, null, 'u' ),
        (4, null, 'q' ),
        (5, 4   , 'o' ),
        (6, 4   , 'p' ),
        (7, null, 'c' )
    ),
  cte_branch as
  (
     SELECT coalesce(parent.id, cte.id) as branch_id, cte.id, cte.parent_id, cte.sort_column, 
            row_number over (partition by coalesce(parent.id, cte.id) order by cte.sort_column) as rn
     FROM cte
     left join cte parent on cte.parent_id = parent.id
  )
select * from cte_branch
order by rn, id nulls first

Answer 2

您需要对 RN 使用 DENSE_RANK，对 CHILD_RN 使用 ROW_NUMBER，如下所示：

with
  cte(id,parent_id,sort_column) as
    ( 
        SELECT 1, null, 'y' FROM DUAL UNION ALL
        SELECT 2, 1   , 'z' FROM DUAL UNION ALL
        SELECT 3, null, 'u' FROM DUAL UNION ALL
        SELECT 4, null, 'q' FROM DUAL UNION ALL
        SELECT 5, 4   , 'o' FROM DUAL UNION ALL
        SELECT 6, 4   , 'p' FROM DUAL UNION ALL
        SELECT 7, null, 'c' FROM DUAL
    )
SELECT
    C.*,
    DENSE_RANK() OVER(
        ORDER BY
            (CASE
                WHEN C.PARENT_ID IS NOT NULL THEN C.PARENT_ID
                ELSE C.ID
            END) DESC NULLS FIRST
    ) AS RN,
    CASE
        WHEN C.PARENT_ID IS NOT NULL THEN ROW_NUMBER() OVER(
            PARTITION BY C.PARENT_ID
            ORDER BY
                C.ID
        )
    END AS CHILD_RN
FROM
    CTE C
ORDER BY
    RN;

输出：

        ID  PARENT_ID S         RN   CHILD_RN
---------- ---------- - ---------- ----------
         7            c          1           
         4            q          2           
         5          4 o          2          1
         6          4 p          2          2
         3            u          3           
         2          1 z          4          1
         1            y          4           

7 rows selected.

干杯！！

Answer 3

在我的例子中，可以使用 connect_by_root 子句代替对结果集的额外访问：

（另外，我注意到我原来的 SQL 包含一个错误 - 具有相同 sort_column 值的 parents 具有相同的 RN 值，已修复这里）

with
  cte(id, parent_id, sort_column) as
    (select 1, null, 'y' from dual union all
     select 2, 1,    'z' from dual union all
     select 3, null, 'u' from dual union all
     select 4, null, 'q' from dual union all
     select 5, 4,    'o' from dual union all
     select 6, 4,    'p' from dual union all
     select 7, null, 'c' from dual)
select 
  *
from
  (select
    t.*, 
    case when t.parent_id is null
      then
        row_number() over (partition by parent_id order by t.sort_column)
      else
        dense_rank() over (order by connect_by_root sort_column)
    end as RN as RN,
    case 
      when parent_id is null 
        then
          null 
        else 
          row_number() over (partition by t.parent_id order by t.sort_column) 
    end as RN_CHILD
from cte t
connect by prior id = parent_id
start with parent_id is null)
--where RN between :x and :y
order by RN, RN_CHILD nulls first

按 parents 分页时，显示他们的 children

When paginating by parents, display their children

sql

oracle

pagination

hierarchical-data

window-functions