如何识别在同一行的字段/列中具有重复值的行

Question

在表中，我想找到至少 2 个字段（列）具有重复的“非空”值的行。通用 SQL 解决方案将更可取，因为它可以在任何数据库中使用。如果不是这样，Oracle 和 SQL Server 就是我的目标数据库。举个例子

ID    COL1    COL2    COL3    COL4
1      11     11       11      44
2      11     22       33      44
3      11     null     33      33
4      11     null     null    44

应该返回以下行

ID    COL1    COL2    COL3    COL4
1      11     11       11      44
3      11     null     33      33

第一行有 3 个字段，重复值为 11，其他行 col3 和 col4 重复值为 33

Answer 1

蛮力方法是：

select t.*
from t
where (col1 = col2 or col1 = col3 or col1 = col4 or
       col2 = col3 or col2 = col4 or col3 = col4
      ) or
      (col1 is null and (col2 is null or col3 is null or col4 is null) or
       col2 is null and (col3 is null or col4 is null) or
       col3 is null and col4 is null
      )

这适用于任何数据库。

Answer 2

您可以在 Oracle 中使用UNPIVOT：

SQL Fiddle

Oracle 11g R2 架构设置：

CREATE TABLE table_name ( ID, COL1, COL2, COL3, COL4 ) As
SELECT 1,      11,     11,       11,      44 FROM DUAL UNION ALL
SELECT 2,      11,     22,       33,      44 FROM DUAL UNION ALL
SELECT 3,      11,     null,     33,      33 FROM DUAL UNION ALL
SELECT 4,      11,     null,     null,    44 FROM DUAL;

查询 1：

SELECT *
FROM   table_name
WHERE  id IN (
  SELECT id
  FROM   table_name
  UNPIVOT ( value FOR key IN ( COL1, COL2, COL3, COL4 ) )
  GROUP BY id, value
  HAVING COUNT( DISTINCT key ) > 1
)

Results：

| ID | COL1 |   COL2 | COL3 | COL4 |
|----|------|--------|------|------|
|  1 |   11 |     11 |   11 |   44 |
|  3 |   11 | (null) |   33 |   33 |

如果您想匹配NULL，那么只需使用UNPIVOT INCLUDE NULLS。

和 SQL Server，代码几乎相同（只需要在 UNPIVOT 上加上别名）：

SQL Fiddle

查询 1：

SELECT *
FROM   table_name
WHERE  id IN (
  SELECT id
  FROM   table_name
  UNPIVOT ( value FOR name IN ( COL1, COL2, COL3, COL4 ) ) AS u
  GROUP BY id, value
  HAVING COUNT( DISTINCT name ) > 1
)

Results：

| ID | COL1 |   COL2 | COL3 | COL4 |
|----|------|--------|------|------|
|  1 |   11 |     11 |   11 |   44 |
|  3 |   11 | (null) |   33 |   33 |

更新：

您还可以使用 Oracle 中的 *_TAB_COLUMN 字典表生成蛮力查询（SQL server 中可能有一个等效项）：

SELECT 'SELECT * FROM TABLE_NAME WHERE ('
       || LISTAGG(
            '"' || PRIOR COLUMN_NAME || '" = "' || COLUMN_NAME || '"',
            ' OR '
          ) WITHIN GROUP ( ORDER BY ROWNUM )
          || ')' AS query
FROM   USER_TAB_COLUMNS
WHERE  TABLE_NAME = 'TABLE_NAME'
AND    COLUMN_NAME LIKE 'COL%'
AND    LEVEL = 2
START WITH COLUMN_NAME LIKE 'COL%'
CONNECT BY PRIOR COLUMN_ID < COLUMN_ID;

哪些输出：

SELECT * FROM TABLE_NAME WHERE ("COL1" = "COL2" OR "COL1" = "COL3" OR "COL1" = "COL4" OR "COL2" = "COL3" OR "COL2" = "COL4" OR "COL3" = "COL4")

Answer 3

没有硬编码列名的解决方案（Sql Server）。 可以说，我们的表是 [#test]。那么我们的查询是：

;with [temp] as (
    select
         [id]           =   id
        ,[col_name1]    =   [c1].[value]('local-name(.)',   'nvarchar(256)')
        ,[col_value1]   =   [c1].[value]('.',               'nvarchar(256)')
        ,[col_name2]    =   [c2].[value]('local-name(.)',   'nvarchar(256)')
        ,[col_value2]   =   [c2].[value]('.',               'nvarchar(256)')    
    from 
        [#test] as  [t]
    cross apply
        (
            select [data] = convert(xml, (select [t].* for xml path('row') ))
        )       as  [x]
    cross apply
        [x].[data].[nodes]('row/*') as [t1]([c1])
    cross apply
        [x].[data].[nodes]('row/*') as [t2]([c2])
)
,[ids] as (
    select 
        [id]
    from 
        [temp]
    where
            ([col_name1]    <>  [col_name2] )
        and ([col_value1]   =   [col_value2])
    group by
         [id]
)
select 
    *
from
    [#test] as  [t]
inner join
    [ids]   as  [i]
on
        [t].[id] = [i].[id];

完整查询可见：https://pastebin.com/jUG5r41c

Answer 4

对于 SQL Server，我会使用 APPLY 运算符来执行此操作：

select * 
from (select *, (select COUNT(*) from (values (Col1), (Col2),.. (ColN))t(ids)) TotalCols,
                (select COUNT(distinct ids) from (values (Col1), (Col2),.. (ColN))t(ids)) DistinctCols
      from table t
    ) t
where TotalIds <> DistinctCols;