• 数据

SQL> SELECT * FROM demo_a;CUSTOMER        TOTAL
---------- ----------
A                 100200SQL> SELECT * FROM demo_b;CUSTOMER           RN        QTY
---------- ---------- ----------
A                   1         30
A                   2         60
A                   3         101        1902         10

• 目标

以CUST字段内容关联A、B两表：相等或同时为空，并按B表RN字段的顺序累计QTY数值，截止到不超过A表对应TOTAL值的行

• 问题

先将两表关联起来

SQL> SELECT *2    FROM demo_a a,3         (SELECT t.*, SUM(qty) over(PARTITION BY customer ORDER BY rn) sum_qty4            FROM demo_b t) b5   WHERE (a.customer IS NULL AND b.customer IS NULL OR6         a.customer = b.customer)7  ;CUSTOMER        TOTAL CUSTOMER           RN        QTY    SUM_QTY
---------- ---------- ---------- ---------- ---------- ----------
A                 100 A                   1         30         30
A                 100 A                   2         60         90
A                 100 A                   3         10        100200                     1        190        190200                     2         10        200

整理得到递归条件为：当SUM_QTY小于TOTAL时，则RN应当加1

SQL> SELECT *2    FROM demo_a a,3         (SELECT t.*, SUM(qty) over(PARTITION BY customer ORDER BY rn) sum_qty4            FROM demo_b t) b5   WHERE (a.customer IS NULL AND b.customer IS NULL OR6         a.customer = b.customer)7   START WITH b.rn = 18  CONNECT BY (PRIOR a.customer IS NULL AND a.customer IS NULL OR9             PRIOR a.customer = a.customer)10         AND PRIOR b.sum_qty < PRIOR a.total11         AND b.rn = PRIOR b.rn + 112  ;CUSTOMER        TOTAL CUSTOMER           RN        QTY    SUM_QTY
---------- ---------- ---------- ---------- ---------- ----------

没有结果，若是将Where条件改为a.customer = b.customer则可正常得出结果，但这会漏掉CUST为空的数据。难道递归查询中不可以使用OR来关联表吗

• 原因

观察执行计划

---------------------------------------------------------------------------------------------------
| Id  | Operation                                | Name   | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                         |        |       |       |     6 (100)|          |
|*  1 |  FILTER                                  |        |       |       |            |          |
|*  2 |   CONNECT BY NO FILTERING WITH START-WITH|        |       |       |            |          |
|*  3 |    HASH JOIN                             |        |     5 |   430 |     5  (20)| 00:00:01 |
|   4 |     TABLE ACCESS FULL                    | DEMO_A |     2 |    60 |     2   (0)| 00:00:01 |
|   5 |     VIEW                                 |        |     5 |   280 |     3  (34)| 00:00:01 |
|   6 |      WINDOW SORT                         |        |     5 |   215 |     3  (34)| 00:00:01 |
|   7 |       TABLE ACCESS FULL                  | DEMO_B |     5 |   215 |     2   (0)| 00:00:01 |
---------------------------------------------------------------------------------------------------Predicate Information (identified by operation id):
---------------------------------------------------1 - filter(("A"."CUSTOMER" IS NULL AND "B"."CUSTOMER" IS NULL))2 - filter("B"."RN"=1)3 - access("A"."CUSTOMER"="B"."CUSTOMER")

可见两表关联处(id=3)仅使用了谓词：a.customer=b.customer，而a.customer is null and b.customer is null则成为了最终结果集的一个筛选谓词(id=1)

究其原因，在递归查询的Where条件中，最先执行的是表关联条件，其它筛选条件则是在递归后执行。此处所指的表关联条件必须是同时出现两个表的表达式，a.customer is null and b.customer is null由于仍可分为两个表达式，而每个表达式中只出现了一个表，故作为了最后执行的筛选条件

此例程中表关联条件可以使用NVL处理以使两表同时出现

SQL> SELECT *2    FROM demo_a a,3         (SELECT t.*, SUM(qty) over(PARTITION BY customer ORDER BY rn) sum_qty4            FROM demo_b t) b5   WHERE nvl(a.customer, 'NULL') = nvl(b.customer, 'NULL')6   START WITH b.rn = 17  CONNECT BY (PRIOR a.customer IS NULL AND a.customer IS NULL OR8             PRIOR a.customer = a.customer)9         AND PRIOR b.sum_qty < PRIOR a.total10         AND b.rn = PRIOR b.rn + 111  ;CUSTOMER        TOTAL CUSTOMER           RN        QTY    SUM_QTY
---------- ---------- ---------- ---------- ---------- ----------
A                 100 A                   1         30         30
A                 100 A                   2         60         90
A                 100 A                   3         10        100200                     1        190        190200                     2         10        200