Question

我有一张名为timedetail的表，其中输入了工作时间（doecode w）或休假时间（doecode vac）。这是我到目前为止所做的，但它也正在恢复工作时间。

TimeDetail

empnum  tsdate     doecode
123     1/1/2013   VAC
123     1/2/2013   VAC
123     1/3/2013   VAC
123     1/4/2013   VAC
123     1/5/2013   VAC
123     1/6/2013   W
123     1/7/2013   W
123     1/8/2013   W
123     1/9/2013   VAC
111     1/1/2013   W
111     1/2/2013   VAC
111     1/3/2013   W
111     1/4/2013   VAC
111     1/5/2013   VAC

empnum 123应该返回但不是111

WITH R AS (
SELECT
   empnum, tsdate,
   ROW_NUMBER() OVER (PARTITION BY empnum ORDER BY tsdate)
   - ROW_NUMBER() OVER (PARTITION BY empnum ORDER BY tsdate, doecode) as grp
FROM
   timedetail t
WHERE
   doecode in ('VAC') 
--   and not exists (
--      select * from timedetail b where doecode < ' '
--      and b.tsdate = t.tsdate and b.empnum = t.empnum
--   )
)
SELECT empnum, MIN(tsdate) AS fdt, MAX(tsdate) AS tdt, COUNT(*) AS no_of_days
FROM R  
GROUP BY empnum, grp
HAVING COUNT(*) > 4
ORDER BY empnum, MIN(tsdate), MAX(tsdate)

Answer 1

鉴于忽略组（VAC或W）中的差距并容忍重复值的新要求，您提供的查询有两个问题。

仅按empnum分区两次。这不会返回有用的结果。相反，您需要第二个按empnum, doecode进行分区。
由于需要按doecode进行分区，因此您无法从派生表中排除任何doecode值，并且必须将WHERE doecode = 'VAC'移出到主查询中。

我已在以下查询中解决了这些问题：

WITH R AS (
   SELECT
      T.*,
      Grp = 
         DENSE_RANK() OVER (PARTITION BY T.empnum ORDER BY T.tsdate)
         - DENSE_RANK() OVER (PARTITION BY T.empnum, T.doecode ORDER BY T.tsdate)
         -- the subtracted value has to partition by doecode
   FROM
      dbo.TimeDetail T
)
SELECT
   empnum,
   FromDate = MIN(tsdate),
   ToDate = MAX(tsdate),
   DayCount = COUNT(DISTINCT tsdate)
FROM R
WHERE doecode = 'VAC' -- must be out here
GROUP BY empnum, grp
HAVING COUNT(DISTINCT tsdate) >= 5 -- distinct dates
ORDER BY
   empnum,
   FromDate
;

See this working in a SQL Fiddle

我使用>= 5而不是> 4，因为我认为它更清楚地表达了意图。

使用以下“棘手”的测试数据：

empnum tsdate     doecode
------ ---------- -------
123    2013-01-01 VAC
123    2013-01-02 VAC
123    2013-01-03 VAC
123    2013-01-03 VAC -- duplicate row
123    2013-01-04 VAC
123    2013-01-04 VAC -- duplicate row
123    2013-01-07 VAC -- skipped a weekend
111    2013-01-01 W
111    2013-01-02 VAC
111    2013-01-03 W
111    2013-01-04 VAC
111    2013-01-06 W
111    2013-01-07 W
111    2013-01-08 W
111    2013-01-09 W
111    2013-01-10 W

这将返回：

empnum FromDate   ToDate     DayCount
------ ---------- ---------- --------
123    2013-01-01 2013-01-07 5

对于记录，如果您的数据是连续的且无法复制，则只需要一个Row_number功能。对顺序项进行分组就像Itzik Ben-Gan的分组岛解决方案一样简单：

WITH R AS (
   SELECT
      T.*,
      Grp = DateAdd(day, 
         -ROW_NUMBER() OVER (PARTITION BY T.empnum, T.doecode ORDER BY T.tsdate),
         T.tsdate
      )
   FROM
      dbo.TimeDetail T
   WHERE
      doecode in ('VAC') 
)
SELECT
   empnum,
   FromDate = MIN(tsdate),
   ToDate = MAX(tsdate),
   DayCount = COUNT(*)
FROM R  
GROUP BY empnum, grp
HAVING COUNT(*) >= 5
ORDER BY
   empnum,
   FromDate
;

Sql server查询以查明员工何时缺席5天或更长时间

1 个答案: