sql-server-2005 - 连接最近的日期(无子查询)

Question

我有这个不太标准化的表格:

ItemName    Type    DateTransferred

Hand Drill  IN  2012-01-16 11:06:10.077
Hand Drill  OUT 2012-01-16 11:06:16.563
Hand Drill  IN  2012-01-16 11:06:26.780
Grinder     IN  2012-01-16 11:06:33.917
Hand Drill  OUT 2012-01-16 11:06:45.443

创建查询:

CREATE TABLE [dbo].[TransferLog](
    [ItemName] [nvarchar](50) NOT NULL,
    [Type] [nvarchar](3) NOT NULL,
    [DateTransferred] [datetime] NOT NULL
) ON [PRIMARY]


ALTER TABLE [dbo].[TransferLog] 
ADD  CONSTRAINT [DF_TransferLog_DateTransferred]  
    DEFAULT (getdate()) FOR [DateTransferred]

基本上，上表记录了项目团队从仓库借出的元素(类型:IN)和归还的元素(类型:OUT)。

我想要实现的是获取所有借用的设备，何时借用(IN)以及何时归还(OUT)。当尝试将“借用交易”与其相应的“返回交易”匹配时会出现问题，因为它们唯一的关系是 ItemName:

选择所有“借用交易”:

select tIn.ItemName, tIn.DateTransferred as DateBorrowed
from transferLog as tIn
where tIn.[type] = 'IN'

结果:

ItemName    DateBorrowed

Hand Drill  2012-01-16 11:06:10.077
Hand Drill  2012-01-16 11:06:26.780
Grinder     2012-01-16 11:06:33.917

尝试选择所有“借用交易”及其对应的“归还交易”:

select tIn.ItemName, tIn.DateTransferred as DateBorrowed, 
    tOut.DateTransferred as   DateReturned
from transferLog as tIn
left join transferLog as tOut
on tIn.ItemName = tOut.ItemName
    and tOut.[type] = 'OUT'
where tIn.[type] = 'IN'

结果:

ItemName    DateBorrowed                    DateReturned

Hand Drill  2012-01-16 11:06:10.077     2012-01-16 11:06:16.563
Hand Drill  2012-01-16 11:06:10.077     2012-01-16 11:06:45.443
Hand Drill  2012-01-16 11:06:26.780     2012-01-16 11:06:16.563
Hand Drill  2012-01-16 11:06:26.780     2012-01-16 11:06:45.443
Grinder     2012-01-16 11:06:33.917     NULL

注意，每个“借入交易”只能有一个或没有对应的“归还交易”，以上结果将每个“借入交易”与每个“归还交易”匹配，只要它们具有相同的 ItemName。结果应该是:

ItemName    DateBorrowed                    DateReturned

Hand Drill  2012-01-16 11:06:10.077     2012-01-16 11:06:16.563
Hand Drill  2012-01-16 11:06:26.780     2012-01-16 11:06:45.443
Grinder     2012-01-16 11:06:33.917     NULL

现在，我正在考虑如何将“归还交易”与大于且最接近“借用交易”的 DateTransferred 的 DateTransferred 进行匹配。像这样的东西:

select tIn.ItemName, tIn.DateTransferred as DateBorrowed, 
    tOut.DateTransferred as DateReturned
from transferLog as tIn
left join transferLog as tOut
on tIn.ItemName = tOut.ItemName
and tOut.[type] = 'OUT'
and 
        tOut.DateTransferred > tIn.DateTransferred 
        -- AND NEAREST tOut.DateTransferred TO tIn.DateTransferred
where tIn.[type] = 'IN'

我读了这个( SQL Join on Nearest less than date )和这个( Join tables on nearest date in the past, in MySQL )，但是子查询对我来说是一个困难的选择，因为我需要的查询结果只是另一个查询的一部分，我担心它会影响性能。

Answer 1

下面确实使用了子查询(更准确地说，是一个公共(public)表表达式)，但它应该足够有效:

;
WITH ranked AS (
  SELECT
    *,
    rnk = ROW_NUMBER() OVER (PARTITION BY ItemName, Type ORDER BY DateTransferred)
  FROM TransferLog
)
SELECT
  r_in.ItemName,
  r_in.DateTransferred AS DateBorrowed,
  r_out.DateTransferred AS DateReturned
FROM ranked r_in
  LEFT JOIN ranked r_out ON r_out.Type = 'OUT'
    AND r_in.ItemName = r_out.ItemName
    AND r_in.rnk = r_out.rnk
WHERE r_in.Type = 'IN'

如您所见，其想法是分别对 IN 和 OUT 行进行排名，然后将前者与后者进行匹配(使用外连接，因为最后一个 IN 项可以没有匹配项)。

引用文献:

• Ranking Functions (Transact-SQL)

• ROW_NUMBER (Transact-SQL)

• WITH common_table_expression (Transact-SQL)