MySQL - 按时间获取最后一个条目(或者如果为空则首先获取)

时间:2011-12-16 02:12:48

标签: mysql sql-server cross-apply

我有以下SQL Server 2008查询:

SELECT T.*,Data.Value FROM [Table] T OUTER APPLY      

(SELECT TOP 1 E.Value FROM [Table2] E     
ORDER BY CASE WHEN T.TDateTime >= E.EDateTime then 1 else 2 end,
ABS(DateDiff(ss,T.TDateTime,E.EDatetime))) AS Data

这基本上会获得E中每条记录的最后T值,但如果T中的记录位于E中的第一条记录之前,则会获得E中的第一条记录。

MySQL中的等价物是什么?

修改


这是我的架构和数据:

DROP TABLE IF EXISTS `data`;
CREATE TABLE `data` (
  `ID` int(11) NOT NULL AUTO_INCREMENT,
  `DataDateTime` datetime DEFAULT NULL,
  `Value` int(11) DEFAULT NULL,
  PRIMARY KEY (`ID`)
) ENGINE=InnoDB AUTO_INCREMENT=9 DEFAULT CHARSET=utf8;


LOCK TABLES `data` WRITE;
INSERT INTO `data` VALUES (1,'2012-02-01 00:00:00',1),(2,'2012-03-01 01:00:00',2),(3,'2012-04-01 02:00:00',3),(4,'2012-05-01 03:00:00',4),(5,'2012-06-01 04:00:00',5),(6,'2012-07-01 05:00:00',6),(7,'2012-08-01 06:00:00',7),(8,'2012-09-01 07:00:00',8);
UNLOCK TABLES;


DROP TABLE IF EXISTS `t`;
CREATE TABLE `t` (
  `ID` int(11) NOT NULL AUTO_INCREMENT,
  `TDateTime` datetime DEFAULT NULL,
  PRIMARY KEY (`ID`)
) ENGINE=InnoDB AUTO_INCREMENT=9 DEFAULT CHARSET=utf8;

LOCK TABLES `t` WRITE;
INSERT INTO `t` VALUES (1,'2012-01-01 00:00:00'),(2,'2012-02-01 00:00:00'),(3,'2012-02-01 12:00:00'),(4,'2012-03-01 00:00:00'),(5,'2012-04-01 00:00:00'),(6,'2012-05-01 12:00:00'),(7,'2012-06-01 00:00:00'), (8,'2012-07-01 00:00:00');
UNLOCK TABLES;

SQLFiddle:

CREATE TABLE `data` (
  `ID` int(11) NOT NULL AUTO_INCREMENT,
  `DataDateTime` datetime DEFAULT NULL,
  `Value` int(11) DEFAULT NULL,
  PRIMARY KEY (`ID`)
);

INSERT INTO `data` (`DataDateTime`, `Value`) VALUES 
('2012-02-01 00:00:00',1),
('2012-03-01 01:00:00',2),
('2012-04-01 02:00:00',3),
('2012-05-01 03:00:00',4),
('2012-06-01 04:00:00',5),
('2012-07-01 05:00:00',6),
('2012-08-01 06:00:00',7),
('2012-09-01 07:00:00',8);


CREATE TABLE `t` (
  `ID` int(11) NOT NULL AUTO_INCREMENT,
  `TDateTime` datetime DEFAULT NULL,
  PRIMARY KEY (`ID`)
);

INSERT INTO `t` (`TDateTime`) VALUES 
('2012-01-01 00:00:00'),
('2012-02-01 00:00:00'),
('2012-02-01 12:00:00'),
('2012-03-01 00:00:00'),
('2012-04-01 00:00:00'),
('2012-05-01 12:00:00'),
('2012-06-01 00:00:00'),
('2012-07-01 00:00:00');

我想要的输出:

T.ID, T.TDateTime, Data.DataDateTime, Data.Value
1, 2012-01-01 00:00:00, 2012-02-01 00:00:00, 1
2, 2012-02-01 00:00:00, 2012-02-01 00:00:00, 1
3, 2012-02-01 12:00:00, 2012-02-01 00:00:00, 1
4, 2012-03-01 00:00:00, 2012-02-01 00:00:00, 1
5, 2012-04-01 00:00:00, 2012-03-01 01:00:00, 2
6, 2012-05-01 12:00:00, 2012-05-01 03:00:00, 4
7, 2012-06-01 00:00:00, 2012-05-01 03:00:00, 4
8, 2012-07-01 00:00:00, 2012-06-01 04:00:00, 5

4 个答案:

答案 0 :(得分:2)

这是我的提交:)

select *, if(Segment1Time<=ifnull(Segment2Time,Segment1Time),
             Segment1Value,
             Segment2Value) Value
from
(
  select *,
    (select DataDateTime from `data` where DataDateTime<=t.TDateTime order by DataDateTime desc limit 1) Segment1Time,
    (select Value from `data` where DataDateTime<=t.TDateTime order by DataDateTime desc limit 1) Segment1Value,
    (select DataDateTime from `data` where DataDateTime> t.TDateTime order by DataDateTime limit 1) Segment2Time,
    (select Value from `data` where DataDateTime> t.TDateTime order by DataDateTime limit 1) Segment2Value
  from `t` t
) X
order by tdatetime;

这是查询的EXPLAIN。这个查询的好处在于,data.datadatetime上的索引与4个子查询呈线性关系,所有子查询都生成单个SEEK,而不必通过所有记录对它们进行排名。从理论上讲,data表越大,它应该越好。

ID  SELECT_TYPE TABLE   TYPE    POSSIBLE_KEYS   KEY KEY_LEN REF ROWS    EXTRA
1   PRIMARY <derived2>  ALL (null)  (null)  (null)  (null)  8   Using filesort
2   DERIVED t   ALL (null)  (null)  (null)  (null)  8   
6   DEPENDENT SUBQUERY  data    ALL DataDateTime    (null)  (null)  (null)  8   Using where; Using filesort
5   DEPENDENT SUBQUERY  data    index   DataDateTime    DataDateTime    9   (null)  1   Using where; Using index
4   DEPENDENT SUBQUERY  data    ALL DataDateTime    (null)  (null)  (null)  8   Using where; Using filesort
3   DEPENDENT SUBQUERY  data    index   DataDateTime    DataDateTime    9   (null)  1   Using where; Using index

答案 1 :(得分:2)

我花了一些时间来了解这些要求。最后开始使用OUTER APPLY分析您的基本查询并将其更改为:

SELECT t.*, data.*
  FROM t
  JOIN (SELECT t_ID = t.ID, 
        data_ID = ISNULL((SELECT TOP 1 ID FROM data WHERE data.DataDateTime <= t.TDateTime ORDER BY DataDateTime DESC),
                         (SELECT TOP 1 ID FROM data WHERE data.DataDateTime > t.TDateTime ORDER BY DataDateTime ASC))
   FROM t) lnk
    ON lnk.t_ID = t.ID
  JOIN data
    ON data.ID = lnk.data_ID
  ORDER BY t.ID

执行计划显示它效率较低,但这让我感到惊讶。但是,在DataDateTime上添加一个索引会大大改变你的MSSQL版本可能派上用场的东西吗?!

无论如何,从这里开始我在MySQL中创建了这个:

SELECT t.*, data.*
  FROM t
  JOIN (SELECT t.ID t_ID, 
           COALESCE((SELECT ID FROM data WHERE data.DataDateTime <= t.TDateTime ORDER BY DataDateTime DESC LIMIT 1),
                           (SELECT ID FROM data WHERE data.DataDateTime > t.TDateTime ORDER BY DataDateTime ASC LIMIT 1)) data_ID
   FROM t) lnk
    ON lnk.t_ID = t.ID
  JOIN data
    ON data.ID = lnk.data_ID
  ORDER BY t.ID

似乎做了预期的事情......

答案 2 :(得分:1)

答案 3 :(得分:1)

您可以交叉加入tdata,然后根据(相当于)data中的t排列每ORDER BY行的DATEDIFF(ss, dt1, dt2)行SQL Server查询。

UNIX_TIMESTAMP(dt2) - UNIX_TIMESTAMP(dt1)部分可以替换为SELECT ID, TDateTime, DataDateTime, Value FROM ( SELECT ID, TDateTime, DataDateTime, Value, @rnk := @rnk * (@lastid = ID) + 1 AS rnk, @lastid := ID FROM ( SELECT t.ID, t.TDateTime, data.DataDateTime, data.Value FROM t CROSS JOIN data, (SELECT @lastid := 0, @rnk := 0) s ORDER BY t.ID, (t.TDateTime >= data.DataDateTime) DESC, ABS(UNIX_TIMESTAMP(t.TDateTime) - UNIX_TIMESTAMP(data.DataDateTime)) ) s ) s WHERE rnk = 1 ; 。排名可以使用变量来实现。这是我尝试解决方案:

{{1}}

您可以找到有效的演示at SQL Fiddle