在 oracle 的单行中选择不同的列值

Question

我有以下要求。

Column1 Column2 Column3 Column4 Column5 Column6
Abc     02      03      02      05      07

当我选择这一行时，我应该得到如下值

Abc 02 03 05 07 <null> 

应消除重复项，但列应保持原样。

感谢您的帮助！！

谢谢， 赛

Answer 1

您将数据存储在应该在行中的列中。数据结构应该是这样的：

column1 number value
  ABC     1      02
  ABC     2      03
  ABC     3      02
. . .

那么您的查询将很容易。您可以通过 unpivoting 获得所需的值：

select distinct column1, value
from (select column1, column2 as valuefrom t union all
      select column1, column3 from t union all
      select column1, column4 . . .
     ) x;

您可以使用以下方法将它们作为列表获取：

select column1, listagg(value, ', ') within group (order by value)
from (select column1, column2 as valuefrom t union
      select column1, column3 from t union
      select column1, column4 . . .
     ) x
group by column1;

Answer 2

好的。我将所有内容（包括基表）放入一个查询中，以便其他人可以使用它 - 测试它，找到更好的方法等。在现实生活中，我称之为“t”（第二个 cte 或公用表表达式）是起始表和我所说的“c”（第一个 cte）应该作为表或视图存在于某处，或者它可以像我一样动态构建。要求：必须事先知道表 t 中所有列的数量、名称和顺序，它们不能是动态的（除非想使用动态 SQL 重写）。

还要注意，如果我可以假设所有列名都是大写的，那会容易得多。我不想假设它，所以使用大写（...）的代码看起来更复杂 - 因为 Oracle 会在取消透视时将列名更改为大写（否则我将不得不做更多的工作...... )

我以新的“表格”格式返回结果（如果 t 有多于一行，则可以使用该格式 - 输出是相同数量的行）。我没有写成“update t...”格式；如果需要，这应该相对容易做到。我不知道“我必须改变值”是否意味着到位（即更新），或者它是否意味着输出与输入相比。

输入表（仅一行）如下所示：

COLUMN1    COLUMN2    COLUMN3    COLUMN4    COLUMN5    COLUMN6
---------- ---------- ---------- ---------- ---------- ----------
Abc        02         03         02         05         07

输出：

COLUMN1    COLUMN2    COLUMN3    COLUMN4    COLUMN5    COLUMN6
---------- ---------- ---------- ---------- ---------- ----------
Abc        02         03         05         07

还有代码：

with c (col_rank, col_name) as (select 1, upper('Column1') from dual union all
                                select 2, upper('Column2') from dual union all
                                select 3, upper('Column3') from dual union all
                                select 4, upper('Column4') from dual union all
                                select 5, upper('Column5') from dual union all
                                select 6, upper('Column6') from dual          ),
     t (Column1, Column2, Column3, Column4, Column5, Column6) as 
                         (select 'Abc', '02', '03', '02', '05', '07' from dual),
     x as (select * from t unpivot (col_value for col_name in (Column1, 
                                 Column2, Column3, Column4, Column5, Column6))),
     y as (select col_rank, col_name, col_value from x natural join c),
     z as (select col_rank, col_value, 
                  case when col_value in (select y1.col_value from y y1 
                                          where y1.col_rank < y2.col_rank) then 0 
                                                          else 1 end flag from y y2),
     u as (select col_value, 
                  count(flag) over (order by col_rank rows between unbounded preceding 
                                       and current row) val_rank from z where flag = 1)
select * from (select col_name, col_value from c left outer join u 
                                                  on c.col_rank = u.val_rank )
pivot (min(col_value) for col_name in (upper('Column1') Column1, upper('Column2') Column2,
                                       upper('Column3') Column3, upper('Column4') Column4, 
                                       upper('Column5') Column5, upper('Column6') Column6))

Answer 3

Oracle 设置：

CREATE TABLE table_name ( col1, col2, col3, col4, col5, col6 ) AS
SELECT 'abc', '02', '03', '02', '05', '07' FROM DUAL UNION ALL
SELECT 'abc', '05', '04', '02', '03', '01' FROM DUAL UNION ALL
SELECT 'abc', '01', '01', '01', '01', '01' FROM DUAL;

CREATE TYPE stringlist AS TABLE OF VARCHAR2(100);
/

查询 - 不使用用户定义的函数：

SELECT MAX( col1 ) AS col1,
       MAX( CASE rn WHEN 1 THEN value END ) AS col2,
       MAX( CASE rn WHEN 2 THEN value END ) AS col3,
       MAX( CASE rn WHEN 3 THEN value END ) AS col4,
       MAX( CASE rn WHEN 4 THEN value END ) AS col5,
       MAX( CASE rn WHEN 5 THEN value END ) AS col6
FROM   (
  SELECT col1,
         rid,
         ROW_NUMBER() OVER (PARTITION BY t.rid ORDER BY v.column_value) AS rn,
         v.COLUMN_VALUE AS value
  FROM   ( SELECT col1,
                  ROWID AS rid, 
                  SET( stringlist( col2, col3, col4, col5, col6 ) ) AS cols
           FROM   table_name ) t,
           TABLE( t.cols ) v
)
GROUP BY rid;

查询 - 使用 UDF：

CREATE FUNCTION nth_item(
  collection STRINGLIST,
  n          INT
) RETURN VARCHAR2 DETERMINISTIC
AS
BEGIN
  IF collection IS NULL OR n < 1 OR n > collection.COUNT THEN
    RETURN NULL;
  END IF;
  RETURN collection(n);
END;
/

然后：

SELECT col1,
       nth_item( cols, 1 ) AS col2,
       nth_item( cols, 2 ) AS col3,
       nth_item( cols, 4 ) AS col4,
       nth_item( cols, 4 ) AS col5,
       nth_item( cols, 5 ) AS col6
FROM   (
  SELECT  col1,
          SET( stringlist( col2, col3, col4, col5, col6 ) ) AS cols
  FROM    table_name
);

输出（对于两个查询）：

COL1 COL2 COL3 COL4 COL5 COL6
---- ---- ---- ---- ---- ----
abc  02   03   05   07
abc  01   02   03   04   05
abc  01

Answer 4

with
  tbl as (
    select  'Abc' Column1, '02' Column2, '03' Column3,  '02' Column4, '03' Column5, '07' Column6, 1 rrowid from dual union all 
    select  'Dfg' Column1, '10' Column2, '10' Column3,  '10' Column4, '10' Column5, '10' Column6, 2 rrowid from dual
  ),
  unpv as (
    select *
    from tbl
    unpivot (val FOR col IN (column1 AS 1, column2 AS 2, column3 AS 3, column4 AS 4, column5 as 5, column6 as 6))
  ),
  clc as (
    select rrowid, val, row_number() over (partition by rrowid order by col) col
    from (
      select rrowid, col,
        decode(row_number() over (partition by rrowid, val order by col), 1, val, null) val
      from unpv
    )
    where not val is null)
select * from clc
pivot  (max(val) AS val FOR (col) IN (1 AS a, 2 AS b, 3 AS c, 4 as d, 5 as e, 6 as f));

结果是：

    RROWID A_VAL B_VAL C_VAL D_VAL E_VAL F_VAL
---------- ----- ----- ----- ----- ----- -----
         1 Abc   02    03    07          
         2 Dfg   10