如何使用Oracle SQL查找連續第二高的值？

Question

我有一排這樣的數據庫：

1|abc|10|30|12

最大值是30，第二高是12。如何獲取表中每一行的第二個值？

Answer 1

這將適用於任意數量的列，只需確保將它們添加到以下查詢中標記為merge_col的串聯列表中：

select col1, col2, col3, col4, col5, second_highest
  from (select x.*,
               regexp_substr(merge_col, '[^|]+', 1, levels.column_value) as second_highest,
               row_number() over(partition by x.col1 order by to_number(regexp_substr(merge_col, '[^|]+', 1, levels.column_value)) desc) as rn
          from (select t.*, col3 || '|' || col4 || '|' || col5 as merge_col
                  from tbl t) x,
               table(cast(multiset
                          (select level
                             from dual
                           connect by level <=
                                      length(regexp_replace(merge_col,
                                                            '[^|]+')) + 1) as
                          sys.OdciNumberList)) levels)
 where rn = 2

小提琴測試： http ://sqlfiddle.com/#!4/b446f/2/0

換句話說，對於其他列，請更改：

col3 || '|' || col4 || '|' || col5 as merge_col

至：

col3 || '|' || col4 || '|' || col5 || '|' || col6 .........  as merge_col

不管有多少列，都可以代替......

Answer 2

有多種方法可以實現這一目標。 嘗試這個例子：

SELECT MAX('column_name') FROM 'table_name'
WHERE 'column_name' NOT IN (SELECT MAX('column_name') FROM 'table_name' )

您基本上從查詢中排除了最高編號，然后從其余查詢中選擇最高編號。

Answer 3

假設值都不同：

select t.*,
       (case when col1 <> greatest(col1, col2, col3) and
                  col1 <> least(col1, col2, col3)
             then col1
             when col2 <> greatest(col1, col2, col3) and
                  col2 <> least(col1, col2, col3)
             then col2
             else col3
        end) as secondgreatest
from table t;

Answer 4

如果可以將所有列連接為以逗號分隔的列表，則可以執行以下操作（假設您正在尋找第二高的數值 ）：

WITH d1 AS (
    SELECT keycolumn1, keycolumn2, col1 || ',' || col2 || ',' || ... || ',' || colx AS other_columns
      FROM mytable
), d2 AS (
    SELECT keycolumn1, keycolumn2, LEVEL AS pos
         , TO_NUMBER(REGEXP_SUBSTR(other_columns, '[^,]+', 1, LEVEL)) AS cvalue
      FROM d1
   CONNECT BY REGEXP_SUBSTR(other_columns, '[^,]+', 1, LEVEL) IS NOT NULL
       AND PRIOR keycolumn1 = keycolumn1
       AND PRIOR keycolumn2 = keycolumn2
       AND PRIOR DBMS_RANDOM.VALUE IS NOT NULL
)
SELECT keycolumn1, keycolumn2, pos, cvalue FROM (
    SELECT keycolumn1, keycolumn2, pos, cvalue
         , ROW_NUMBER() OVER ( PARTITION BY keycolumn1, keycolumn2 ORDER BY cvalue DESC ) AS rn
      FROM d2
) WHERE rn = 2;

上面的代碼將返回給定的一組“鍵”列值的第二高值，以及在其中找到該值的列的位置。 我使用串聯和CONNECT BY取消透視表。