如何在 Oracle SQL 中尽可能高效地构建层次结构

Question

这是在 Oracle SQL 中使用他们的 BI Publisher 数据模型工具

假设你有两张桌子：

hierarchy_table acct_description_table

在这两个表中你有这些列：

hierarchy_table
+------------+-----------+-------+
| primarykey | parentkey | depth |
+------------+-----------+-------+

acct_description_table
+------------+-------------+
| acct_value | description |
+------------+-------------+

你想用这些数据建立一个像这样的层次结构（省略描述）：

+----------------+----------------+----------------+
|  acct_depth_0  |  acct_depth_1  |  acct_depth_2  | . . .
+----------------+----------------+----------------+
|     450000     |     450040     |                | . . .
+----------------+----------------+----------------+
|     450000     |     450050     |     450051     | . . .
+----------------+----------------+----------------+

实现这一目标的最佳方法是什么？到目前为止，我有以下 SQL：

select distinct
  t1.acct,
  t1.desc,
  t2.acct,
  t2.desc,
  t3.acct,
  t3.desc,
  t4.acct,
  t4.desc,
  t5.acct,
  t5.desc,
  t6.acct,
  t6.desc
from (select tree.primarykey acct, act.description desc 
      from hierarchy_table tree, acct_description_table act 
      where 1=1
        and tree.depth = 0
        and act.acct_value = tree.primarykey
     ) t1
left join (select tree.primarykey acct, tree.parentkey parent, act.description desc 
           from hierarchy_table tree, acct_description_table act 
           where 1=1
             and tree.depth = 1
             and act.acct_value = tree.primarykey
     ) t2
  on 1=1 and t1.acct = t2.parent
...
...
...
left join (select tree.primarykey acct, tree.parentkey parent, act.description desc 
           from hierarchy_table tree, acct_description_table act 
           where 1=1
             and tree.depth = 5
             and act.acct_value = tree.primarykey
     ) t6
  on 1=1 and t5.acct = t6.parent

正如您在这样的查询中看到的，我们将执行 5 次不同的 left join 操作来完成表格。我研究了递归以帮助加快此查询。但是，我不确定如何对其进行编码以满足我们对每个深度对应的列的需求。

有没有人做过类似的事情？或者有谁知道我们可以比由 5 个不同的左连接组成的当前查询更快地做到这一点？谢谢！

Answer 1

Oracle 的CONNECT BY 功能适用于层次结构并且非常高效。以下是您可以将其用于数据的一种方式：

with hierarchy_table ( primarykey, parentkey, depth ) as 
-- This is test data, you would not have this WITH clause...
( SELECT '450000', null, 0 FROM DUAL UNION ALL
  SELECT '450040', '450000', 1 FROM DUAL UNION ALL
  SELECT '450050', '450000', 1 FROM DUAL UNION ALL  
  SELECT '450051', '450050', 2 FROM DUAL ),
    acct_description_table ( acct_value, description ) AS
-- This is test data, you would not have this WITH clause...
( SELECT '450000', 'Acct #450000' FROM DUAL UNION ALL
  SELECT '450040', 'Acct #450040' FROM DUAL UNION ALL
  SELECT '450050', 'Acct #450050' FROM DUAL UNION ALL  
  SELECT '450051', 'Acct #450051' FROM DUAL )
-- Real query starts here
SELECT REGEXP_SUBSTR(SYS_CONNECT_BY_PATH(ht.primarykey,':'),'[^:]+',1,1) acct1,
       REGEXP_SUBSTR(SYS_CONNECT_BY_PATH(ht.primarykey,':'),'[^:]+',1,2) acct2,
       REGEXP_SUBSTR(SYS_CONNECT_BY_PATH(ht.primarykey,':'),'[^:]+',1,3) acct3,
       REGEXP_SUBSTR(SYS_CONNECT_BY_PATH(ht.primarykey,':'),'[^:]+',1,4) acct4,
       REGEXP_SUBSTR(SYS_CONNECT_BY_PATH(ht.primarykey,':'),'[^:]+',1,5) acct5,
       REGEXP_SUBSTR(SYS_CONNECT_BY_PATH(ht.primarykey,':'),'[^:]+',1,6) acct6,
       adt.description
FROM    hierarchy_table ht
INNER JOIN acct_description_table adt ON adt.acct_value = ht.primarykey
WHERE CONNECT_BY_ISLEAF = 1
START WITH ht.parentkey IS NULL
CONNECT BY ht.parentkey = PRIOR ht.primarykey
AND ht.depth = PRIOR ht.depth + 1  -- There is actually no need for the "DEPTH" column in your table.
AND LEVEL <= 6;

+--------+--------+--------+-------+-------+-------+--------------+
| ACCT1  | ACCT2  | ACCT3  | ACCT4 | ACCT5 | ACCT6 | DESCRIPTION  |
+--------+--------+--------+-------+-------+-------+--------------+
| 450000 | 450040 |        |       |       |       | Acct #450040 |
| 450000 | 450050 | 450051 |       |       |       | Acct #450051 |
+--------+--------+--------+-------+-------+-------+--------------+