我有以下查询来选择LAST_UPDATE_DATE字段获取日期值大于或等于过去7天的记录的行，这很有效。
SELECT 'NEW ROW' AS 'ROW_TYPE', A.EMPLID, B.FIRST_NAME, B.LAST_NAME,
A.BANK_CD, A.ACCOUNT_NUM, ACCOUNT_TYPE, PRIORITY, A.LAST_UPDATE_DATE
FROM PS_DIRECT_DEPOSIT D
INNER JOIN PS_DIR_DEP_DISTRIB A ON A.EMPLID = D.EMPLID AND A.EFFDT = D.EFFDT
INNER JOIN PS_EMPLOYEES B ON B.EMPLID = A.EMPLID
WHERE
B.EMPL_STATUS NOT IN ('T','R','D')
AND ((A.DEPOSIT_TYPE = 'P' AND A.AMOUNT_PCT = 100)
OR A.PRIORITY = 999
OR A.DEPOSIT_TYPE = 'B')
AND A.EFFDT = (SELECT MAX(A1.EFFDT)
FROM PS_DIR_DEP_DISTRIB A1
WHERE A1.EMPLID = A.EMPLID
AND A1.EFFDT <= GETDATE())
AND D.EFF_STATUS = 'A'
AND D.EFFDT = (SELECT MAX(D1.EFFDT)
FROM PS_DIRECT_DEPOSIT D1
WHERE D1.EMPLID = D.EMPLID
AND D1.EFFDT <= GETDATE())
AND A.LAST_UPDATE_DATE >= GETDATE() - 7
我想添加到这个是为每个EMPLID添加前一个（第二个MAX）行，以便我可以输出'旧'行（在上次更新之前，最新行符合上述条件），以及我已经在查询中输出的新行。
ROW_TYPE EMPLID FIRST_NAME LAST_NAME BANK_CD ACCOUNT_NUM ACCOUNT_TYPE PRIORITY LAST_UPDATE_DATE
NEW ROW 12345 JOHN SMITH 123548999 45234879 C 999 2019-03-06 00:00:00.000
OLD ROW 12345 JOHN SMITH 214080046 92178616 C 999 2018-10-24 00:00:00.000
NEW ROW 56399 CHARLES MASTER 785816167 84314314 C 999 2019-03-07 00:00:00.000
OLD ROW 56399 CHARLES MASTER 345761227 547352 C 999 2017-05-16 00:00:00.000
所以EMPLID将由NEW ROW订购，然后是OLD ROW，如上所示。在这个例子中，'NEW ROW'获得了过去7天内的记录，如LAST_UPDATE_DATE所示。
我想得到关于如何修改查询的反馈，这样我也可以得到'旧'行（这是最小行，小于上面检索的'NEW'行）。
这是Gotham犯罪的缓慢日子，所以我给了它一个旋转。可能会工作。
但是，这不太可能开箱即用，但它应该让你开始。
你的LAST_UPDATE_DATE专栏在桌子上PS_DIR_DEP_DISTRIB，所以我们将从那里开始。首先，您想要识别过去7天内更新的所有记录，因为这些记录是您唯一感兴趣的记录。在此，我假设，我可能错了，这是自然的关键该表由EMPLID，BANK_CD和ACCOUNT_NUM组成。您将希望在几个地方的子列中包含这些列的实际自然键。也就是说，日期限制器看起来像这样：
SELECT
EMPLID
,BANK_CD
,ACCOUNT_NUM
FROM
PS_DIR_DEP_DISTRIB AS limit
WHERE
limit.LAST_UPDATE_DATE >= DATEADD(DAY, -7, CAST(GETDATE() AS DATE))
AND
limit.LAST_UPDATE_DATE <= CAST(GETDATE() AS DATE)
现在我们将它用作WHERE EXISTS子句中的相关子查询，我们将关联回基表，将自己限制为具有上周更新的自然键值的记录。我将SELECT列表改为SELECT 1，这是相关子的典型措辞，因为它在找到一个（1）时停止寻找匹配，并且实际上根本不返回任何值。
另外，由于我们正在过滤此记录集，因此我将此表的所有其他WHERE子句过滤器移动到此（即将成为）子查询中。
最后，在SELECT部分，我添加了一个DENSE_RANK强制命令记录。我们稍后使用DENSE_RANK值来仅过滤掉感兴趣的第一个（N）记录。
所以我们留下了这个：
SELECT
EMPLID
,BANK_CD
,ACCOUNT_NUM
--,ACCOUNT_TYPE --Might belong here. Can't tell without table alias in original SELECT
,PRIORITY
,EFFDT
,LAST_UPDATE_DATE
,DEPOSIT_TYPE
,AMOUNT_PCT
,DENSE_RANK() OVER (PARTITION BY --Add actual natural key columns here...
EMPLID
ORDER BY
LAST_UPDATE_DATE DESC
) AS RowNum
FROM
PS_DIR_DEP_DISTRIB AS sdist
WHERE
EXISTS
(
-- Get the set of records that were last updated in the last 7 days.
-- Correlate to the outer query so it only returns records related to this subset.
-- This uses a correlated subquery. A JOIN will work, too. Try both, pick the faster one.
-- Something like this, using the actual natural key columns in the WHERE
SELECT
1
FROM
PS_DIR_DEP_DISTRIB AS limit
WHERE
--The first two define the date range.
limit.LAST_UPDATE_DATE >= DATEADD(DAY, -7, CAST(GETDATE() AS DATE))
AND limit.LAST_UPDATE_DATE <= CAST(GETDATE() AS DATE)
AND
--And these are the correlations to the outer query.
limit.EMPLID = sdist.EMPLID
AND limit.BANK_CD = sdist.BANK_CD
AND limit.ACCOUNT_NUM = sdist.ACCOUNT_NUM
)
AND
(
dist.DEPOSIT_TYPE = 'P'
AND dist.AMOUNT_PCT = 100
)
OR dist.PRIORITY = 999
OR dist.DEPOSIT_TYPE = 'B'
用该查询将原始INNER JOIN替换为PS_DIR_DEP_DISTRIB。在SELECT列表中，第一个硬编码值现在依赖于RowNum值，因此现在是CASE表达式。在WHERE子句中，日期都是由子查询驱动的，所以它们已经消失了，有几个被折叠到子查询中，我们正在添加WHERE dist.RowNum <= 2以带回前2个记录。
（我也替换了所有表别名，所以我可以跟踪我在看什么。）
SELECT
CASE dist.RowNum
WHEN 1 THEN 'NEW ROW'
ELSE 'OLD ROW'
END AS ROW_TYPE
,dist.EMPLID
,emp.FIRST_NAME
,emp.LAST_NAME
,dist.BANK_CD
,dist.ACCOUNT_NUM
,ACCOUNT_TYPE
,dist.PRIORITY
,dist.LAST_UPDATE_DATE
FROM
PS_DIRECT_DEPOSIT AS dd
INNER JOIN
(
SELECT
EMPLID
,BANK_CD
,ACCOUNT_NUM
--,ACCOUNT_TYPE --Might belong here. Can't tell without table alias in original SELECT
,PRIORITY
,EFFDT
,LAST_UPDATE_DATE
,DEPOSIT_TYPE
,AMOUNT_PCT
,DENSE_RANK() OVER (PARTITION BY --Add actual natural key columns here...
EMPLID
ORDER BY
LAST_UPDATE_DATE DESC
) AS RowNum
FROM
PS_DIR_DEP_DISTRIB AS sdist
WHERE
EXISTS
(
-- Get the set of records that were last updated in the last 7 days.
-- Correlate to the outer query so it only returns records related to this subset.
-- This uses a correlated subquery. A JOIN will work, too. Try both, pick the faster one.
-- Something like this, using the actual natural key columns in the WHERE
SELECT
1
FROM
PS_DIR_DEP_DISTRIB AS limit
WHERE
--The first two define the date range.
limit.LAST_UPDATE_DATE >= DATEADD(DAY, -7, CAST(GETDATE() AS DATE))
AND limit.LAST_UPDATE_DATE <= CAST(GETDATE() AS DATE)
AND
--And these are the correlations to the outer query.
limit.EMPLID = sdist.EMPLID
AND limit.BANK_CD = sdist.BANK_CD
AND limit.ACCOUNT_NUM = sdist.ACCOUNT_NUM
)
AND
(
dist.DEPOSIT_TYPE = 'P'
AND dist.AMOUNT_PCT = 100
)
OR dist.PRIORITY = 999
OR dist.DEPOSIT_TYPE = 'B'
) AS dist
ON
dist.EMPLID = dd.EMPLID
AND dist.EFFDT = dd.EFFDT
INNER JOIN
PS_EMPLOYEES AS emp
ON
emp.EMPLID = dist.EMPLID
WHERE
dist.RowNum <= 2
AND
emp.EMPL_STATUS NOT IN ('T', 'R', 'D')
AND
dd.EFF_STATUS = 'A';