Sql 选择不同的

我有这个表

列\u选项卡

：

INDEX_OWNER     INDEX_NAME           TEXT
------------------------------------------------------------------------
DUMMY           PK_FUL_REQUEST       CUSTOMERID, ID, INSTITUTIONID
NIR             PK_FUL_REQUEST       CUSTOMERID, ID, INSTITUTIONID
NIR             IX_REQUEST_HASH      CUSTOMERID, HASH_KEY, INSTITUTIONID
DUMMY           UQ_REQUEST_HASH      CUSTOMERID, HASH_KEY, INSTITUTIONID

我想得到文本列相等但索引名称不相等的行

我可以用两张相同的桌子，但我可以用一张桌子吗

预期结果是：

    INDEX_OWNER     INDEX_NAME           TEXT
------------------------------------------------------------------------
NIR             IX_REQUEST_HASH      CUSTOMERID, HASH_KEY, INSTITUTIONID
DUMMY           UQ_REQUEST_HASH      CUSTOMERID, HASH_KEY, INSTITUTIONID

对它的查询可以通过以下方式实现：

select index_owner, index_name, text
  from columns_tab a, columns_tab b
 where a.text = b.text
   and a.index_name != b.index_name

我不想从同一张表中选择两次就可以实现它。

给你：

select owner, index_name, text
  from (select t.*,
               count(*) over(partition by text, index_name) both_match,
               count(*) over(partition by text) text_match
          from COLUMNS_TAB t)
 where text_match > 1
   and both_match = 1;

但必须指出，如果您在性能方面两次避免查询表，那么我的方法不一定会表现得更好。您应该比较两个查询在实际数据上的执行计划。

另一个答案：

Oracle 11g R2架构设置：

CREATE TABLE columns_tab ( INDEX_OWNER,     INDEX_NAME,           TEXT ) AS
          SELECT 'DUMMY', 'PK_FUL_REQUEST',  'CUSTOMERID, ID, INSTITUTIONID' FROM DUAL
UNION ALL SELECT 'NIR',   'PK_FUL_REQUEST',  'CUSTOMERID, ID, INSTITUTIONID' FROM DUAL
UNION ALL SELECT 'NIR',   'IX_REQUEST_HASH', 'CUSTOMERID, HASH_KEY, INSTITUTIONID' FROM DUAL
UNION ALL SELECT 'DUMMY', 'UQ_REQUEST_HASH', 'CUSTOMERID, HASH_KEY, INSTITUTIONID' FROM DUAL;

WITH prev_next AS (
  SELECT t.*,
         LAG( index_name  ) OVER ( PARTITION BY text ORDER BY index_name ) AS p_name,
         LEAD( index_name ) OVER ( PARTITION BY text ORDER BY index_name ) AS n_name
  FROM   columns_tab t
)
SELECT index_owner,
       index_name,
       text
FROM   prev_next
WHERE  index_name <> p_name
OR     index_name <> n_name

| INDEX_OWNER |      INDEX_NAME |                                TEXT |
|-------------|-----------------|-------------------------------------|
|         NIR | IX_REQUEST_HASH | CUSTOMERID, HASH_KEY, INSTITUTIONID |
|       DUMMY | UQ_REQUEST_HASH | CUSTOMERID, HASH_KEY, INSTITUTIONID |

查询1：

CREATE TABLE columns_tab ( INDEX_OWNER,     INDEX_NAME,           TEXT ) AS
          SELECT 'DUMMY', 'PK_FUL_REQUEST',  'CUSTOMERID, ID, INSTITUTIONID' FROM DUAL
UNION ALL SELECT 'NIR',   'PK_FUL_REQUEST',  'CUSTOMERID, ID, INSTITUTIONID' FROM DUAL
UNION ALL SELECT 'NIR',   'IX_REQUEST_HASH', 'CUSTOMERID, HASH_KEY, INSTITUTIONID' FROM DUAL
UNION ALL SELECT 'DUMMY', 'UQ_REQUEST_HASH', 'CUSTOMERID, HASH_KEY, INSTITUTIONID' FROM DUAL;

WITH prev_next AS (
  SELECT t.*,
         LAG( index_name  ) OVER ( PARTITION BY text ORDER BY index_name ) AS p_name,
         LEAD( index_name ) OVER ( PARTITION BY text ORDER BY index_name ) AS n_name
  FROM   columns_tab t
)
SELECT index_owner,
       index_name,
       text
FROM   prev_next
WHERE  index_name <> p_name
OR     index_name <> n_name

| INDEX_OWNER |      INDEX_NAME |                                TEXT |
|-------------|-----------------|-------------------------------------|
|         NIR | IX_REQUEST_HASH | CUSTOMERID, HASH_KEY, INSTITUTIONID |
|       DUMMY | UQ_REQUEST_HASH | CUSTOMERID, HASH_KEY, INSTITUTIONID |

---

那张表上有主键吗？你期望的结果是什么？你想得到所有可能的组合吗，比如1不等于2，1不等于3，2不等于3，等等？或者，您希望只获取按文本排序的结果，然后比较每个邻居索引名称吗？在预期结果示例中，您如何知道选择哪一行？你的标准是什么？我的意思是，每一对中都有两个可能的候选人。我想要他们两个。看看我写的查询。我想要同样的东西，但是只查询一次表（如果可能的话）。