Question

有没有办法可以避免这种子选择？我试图使这个查询尽可能高效，并且这样做，试图避免子选择。

我正在寻找ID_NUM表格中最新的替换行。

子选择已到位，因此我只能带回最近更换的行。这是最好的方法吗？

查询完全按照我想要的方式工作，所以请随意忽略最后一个小的子选择。

SELECT   
  A.ID_NUM,
  A.DOB,
  A.NAME,
  A.REPLACED_TMESTMP  
FROM MYTABLE A, MYTABLE B  
WHERE   
  A.ID_NUM = B.ID_NUM
  AND A.REPLACED_TMESTMP IS NOT NULL
  AND B.REPLACED_TMESTMP IS NULL
  AND A.DOB <> B.DOB
  several other conditions...

  AND A.REPLACED_TMESTMP  
    = (SELECT MAX(C.REPLACED_TMESTMP)  
      FROM MYTABLE C  
      WHERE C.ID_NUM = A.ID_NUM  
      AND A.REPLACED_TMESTMP IS NOT NULL  
      )
;

这里有一些伪代码可以帮助您理解我需要做什么

Select all from table
where a duplicate ID number exists
    that has an active row (no replaced time stamp)
    and has an inactive row(s)
    only bring back the most current replaced row for each ID num.

Answer 1

您可以将子选项重新编写为JOIN。我们在这里的一个表上有类似的设置（我们有多个＆＃34;行＆＃34;对于每个＆＃34;事物＆＃34;，但你只处理＆＃34;最新＆＃ 34;一个），我通过Visual Explain运行了一个子选择类型查询和一个JOIN类型，根据估计，JOIN将花费相当少的CPU秒来处理（当然，这只是一个估计，而你的桌子可能与我的大不相同，所以请带上一粒盐。）

我稍微更改了查询，我认为这应该包含您正在寻找的所有信息。

SELECT INACTIVE.ID_NUM
      ,INACTIVE.DOB
      ,INACTIVE.NAME,
      ,INACTIVE.REPLACED_TMESTMP  
FROM (
    SELECT *
    FROM MYTABLE
    WHERE REPLACED_TMESTMP IS NULL
) AS ACTIVE
JOIN (
    SELECT *
    FROM MYTABLE
    WHERE REPLACED_TMESTMP IS NOT NULL
) AS INACTIVE
  ON ACTIVE.ID_NUM = INACTIVE.ID_NUM
 AND ACTIVE.DOB   <> INACTIVE.DOB
  -- several other conditions...

JOIN (
    SELECT ID_NUM,
           MAX(REPLACED_TMESTMP) AS TIMESTAMP
    FROM MYTABLE
    GROUP BY ID_NUM
) MAX_REPLACE
  ON INACTIVE.ID_NUM           = MAX_REPLACE.ID_NUM
 AND INACTIVE.REPLACED_TMESTMP = MAX_REPLACE.TIMESTAMP

Answer 2

SELECT   
  A.ID_NUM,  
  MAX(A.REPLACED_TMESTMP) AS MOST_RECENT_REPLACED_ROW
FROM MYTABLE A,
     MYTABLE B
WHERE   
    A.ID_NUM = B.ID_NUM
    AND A.REPLACED_TMESTMP IS NOT NULL
    AND B.REPLACED_TMESTMP IS NULL
GROUP BY A.ID_NUM
;

Answer 3

如果您的DB2版本支持ROW_NUMBER()函数，您可以执行以下操作：

WITH ranked AS (
  SELECT
    ROW_NUMBER() OVER (
      PARTITION BY replaced.ID_NUM
      ORDER BY replaced.REPLACED_TMESTMP DESC
    ) AS rnk.
    replaced.ID_NUM,
    replaced.REPLACED_TMESTMP,
    replaced.... /* other columns as necessary */
  FROM MYTABLE replaced
    INNER JOIN MYTABLE active ON replaced.ID_NUM = active.ID_NUM
  WHERE replaced.REPLACED_TMESTMP IS NOT NULL
    AND active.REPLACED_TMESTMP IS NULL
    AND ... /* your other conditions */
)
SELECT
  ID_NUM,
  REPLACED_TMESTMP,
  ...
FROM ranked
WHERE rnk = 1
;

ROW_NUMBER()函数将数字分配给REPLACED_TMESTMP不为空的行，按该列的降序排列，按ID_NUM分区。然后，主SELECT只检索排名为1的行。

使用MAX（）时避免使用subselect

3 个答案: