Oracle SQL 按列分组，但仅当该列为空或 0

Question

简而言之，我试图按列分组，但前提是该列不是 null 或 0，在这种情况下，我仍然想获取这些行，但只是不分组。我有这张桌子some_table:

+--------+----------+---------------------+-------+
| id     | other_id | date_value          | value |
+--------+----------+---------------------+-------+
| 1      | abc      | 2011-04-20 21:03:05 | 104   |
| 2      | abc      | 2011-04-20 21:03:04 | 229   |
| 3      | xyz      | 2011-04-20 21:03:03 | 130   |
| 4      | abc      | 2011-04-20 21:02:09 | 97    |
| 5      | 0        | 2011-04-20 21:02:08 | 65    |
| 6      | xyz      | 2011-04-20 21:02:07 | 101   |
| 7      |          | 2011-04-20 21:02:07 | 200   |
| 8      | 0        | 2011-04-20 21:02:07 | 201   |
| 9      |          | 2011-04-20 21:02:07 | 202   |
+--------+----------+---------------------+-------+

我想选择行并按other_id 分组，还包括分组行的计数。此查询有效：

select id, other_id, date_value, value, cnt from
 (
   SELECT id, other_id, date_value, value, 
   ROW_NUMBER() OVER (partition by other_id order BY Date_Value desc) r,
   count(*) OVER (partition by other_id) cnt
   FROM some_table 
 )
where r = 1

结果是：

+--------+----------+---------------------+-------+-------+
| id     | other_id | date_value          | value | count |
+--------+----------+---------------------+-------+-------+
| 1      | abc      | 2011-04-20 21:03:05 | 104   | 3     |
| 3      | xyz      | 2011-04-20 21:03:03 | 130   | 2     |
| 5      | 0        | 2011-04-20 21:02:08 | 65    | 2     |
| 7      |          | 2011-04-20 21:02:07 | 200   | 2     |
+--------+----------+---------------------+-------+-------+

问题是，我不想将具有other_id 或null 或0 的行分组。所以想要的结果是这样的：

+--------+----------+---------------------+-------+-------+
| id     | other_id | date_value          | value | count |
+--------+----------+---------------------+-------+-------+
| 1      | abc      | 2011-04-20 21:03:05 | 104   | 3     |
| 3      | xyz      | 2011-04-20 21:03:03 | 130   | 2     |
| 5      | 0        | 2011-04-20 21:02:08 | 65    | 1     |
| 7      |          | 2011-04-20 21:02:07 | 200   | 1     |
| 8      | 0        | 2011-04-20 21:02:07 | 201   | 1     |
| 9      |          | 2011-04-20 21:02:07 | 202   | 1     |
+--------+----------+---------------------+-------+-------+

如您所见，除了other_id 为null 或0 时，所有内容都已分组，在这种情况下它们仍处于选中状态，但未分组。我也在尝试通过date_value 订购。

我尝试在OVER 子句中添加WHERE 子句：

OVER (partition by other_id WHERE other_id IS NOT NULL AND other_id IS NOT '0' order BY Date_Value desc)
-- this produces an error: ORA-00907: missing right parenthesis

我曾考虑进行第二个select 查询并尝试将第一个结果堆叠在第二个结果之上，但由于排序原因，我认为这不会奏效。

有没有办法像这样按条件分组？

Answer 1

这样就可以了：

SQL> select id, other_id, date_value, value, cnt from
  2   (
  3     SELECT id, other_id, date_value, value,
  4     case
  5       when other_id is null or other_id = '0' then 1
  6       else ROW_NUMBER() OVER (partition by other_id order BY Date_Value desc)
  7     end r,
  8     case
  9       when other_id is null or other_id = '0' then 1
 10       else count(*) OVER (partition by other_id)
 11     end cnt
 12     FROM some_table
 13   )
 14  where r = 1
 15  order by id;

        ID OTH DATE_VALUE               VALUE        CNT
---------- --- ------------------- ---------- ----------
         1 abc 2011-04-20 21:03:05        104          3
         3 xyz 2011-04-20 21:03:03        130          2
         5 0   2011-04-20 21:02:08         65          1
         7     2011-04-20 21:02:07        200          1
         8 0   2011-04-20 21:02:07        201          1
         9     2011-04-20 21:02:07        202          1

6 rows selected.

SQL>

Answer 2

试试，

SELECT id, other_id, date_value, value, cnt 
FROM
(
   SELECT id, other_id, date_value, value, 
          ROW_NUMBER() OVER (partition by other_id order BY Date_Value desc) r,
          count(*) OVER (partition by other_id) cnt
   FROM   some_table 
   WHERE  otherID IS NOT NULL OR
          otherID != '0'
) AS derivedTable
WHERE  r = 1