合并、nvl 和 CASE 不适用于子查询中的计数 (*)

Question

我有一个 CTE 查询，我希望 NULL 行为 0。

我想知道为什么 COALESCE(tsting)、NVL(tsting2) 和 CASE(rw_cnt) 不起作用？我错过了什么吗？

with dte AS(Select store, date_id from store cross JOIN (SELECT 
(TRUNC(sysdate)-1 + (LEVEL-10)) AS DATE_ID
FROM DUAL connect by level <=( (sysdate-2)-(sysdate-10)))
store<10)

, tmp as(Select calendar_dt, str, CASE WHEN rw_cnt IS NULL THEN 0 ELSE rw_cnt 
END rw_cnt, COALESCE(rw_cnt, 0) tsting, NVL(rw_cnt, 0 ) tsting2
from (SELECT calendar_dt, str,  count(*) rw_cnt FROM table2 group by calendar_dt, str))

Select store, date_id, rw_cnt, case when rw_cnt IS NULL THEN 0 ELSE rw_cnt END testing,
tsting, tsting2 from dte LEFT OUTER JOIN tmp ON dte.date_id = tmp.calendar_dt 
AND dte.store = temp.store
order by store, date_id;

我得到如下数据：

+-------+-----------+--------+---------+--------+---------+
| STORE | DATE_ID   | RW_CNT | TESTING | TSTING | TSTING2 |
+-------+-----------+--------+---------+--------+---------+
| 3     | 18-OCT-20 | NULL   | 0       | NULL   | NULL    |
+-------+-----------+--------+---------+--------+---------+
| 3     | 19-OCT-20 | 73     | 73      | 73     | 73      |
+-------+-----------+--------+---------+--------+---------+
| 3     | 20-OCT-20 | 88     | 88      | 88     | 88      |
+-------+-----------+--------+---------+--------+---------+
| 3     | 21-OCT-20 | 63     | 63      | 63     | 63      |
+-------+-----------+--------+---------+--------+---------+
| 3     | 22-OCT-20 | 100    | 100     | 100    | 100     |
+-------+-----------+--------+---------+--------+---------+
| 3     | 23-OCT-20 | 105    | 105     | 105    | 105     |
+-------+-----------+--------+---------+--------+---------+
| 3     | 24-OCT-20 | 36     | 36      | 36     | 36      |
+-------+-----------+--------+---------+--------+---------+
| 3     | 25-OCT-20 | 3      | 3       | 3      | 3       |
+-------+-----------+--------+---------+--------+---------+
| 4     | 18-OCT-20 | NULL   | 0       | NULL   | NULL    |
+-------+-----------+--------+---------+--------+---------+
| 4     | 19-OCT-20 | 30     | 30      | 30     | 30      |
+-------+-----------+--------+---------+--------+---------+
| 4     | 20-OCT-20 | 24     | 24      | 24     | 24      |
+-------+-----------+--------+---------+--------+---------+
| 4     | 21-OCT-20 | 38     | 38      | 38     | 38      |
+-------+-----------+--------+---------+--------+---------+
| 4     | 22-OCT-20 | 21     | 21      | 21     | 21      |
+-------+-----------+--------+---------+--------+---------+
| 4     | 23-OCT-20 | 37     | 37      | 37     | 37      |
+-------+-----------+--------+---------+--------+---------+
| 4     | 24-OCT-20 | 3      | 3       | 3      | 3       |
+-------+-----------+--------+---------+--------+---------+
| 4     | 25-OCT-20 | NULL   | 0       | NULL   | NULL    |
+-------+-----------+--------+---------+--------+---------+

欢迎就查询优化提出任何建议...:)

Answer 1

您正在执行 dte 到 tmp 的 LEFT 连接，因此对于不匹配的行您会得到 nulls。

也许你认为在tmp中使用COALESCE()和NVL()最终结果将显示0而不是null，但这是不对的。
COALESCE() 和 NVL() 是 tmp 内部的，如果你从 tmp select 只有这样你就不会得到 nulls，但既然你这样做了LEFT join 并且存在不匹配的行，那么这些不匹配的行的列将由 nulls.

表示
因此，如果要删除这些 null，则必须在最后的 SELECT 语句中使用 COALESCE() and/or NVL()。

合并、nvl 和 CASE 不适用于子查询中的计数 (*)

coalesce, nvl and CASE not working for count(*) in the subquery

sql

oracle

oracle-sqldeveloper