2

解决了

万岁 StackOverlow！

当我离开时，人们留下了 2 个解决方案（谢谢大家——为两个工作解决方案分配业力的协议是什么？）

这是我回来发布的解决方案。它源自另一个 StackOver 解决方案：

如何使用聚合函数在 MySQL 查询中获取分组记录的第一条和最后一条记录？

...我的适应是：

SELECT
DATE_FORMAT(`DTE`, "%m/%d/%Y") AS trading_day,
MIN(`PRICE`) AS min_price,
MAX(`PRICE`) AS max_price,
SUBSTRING_INDEX( 
    GROUP_CONCAT(
        PRICE
        ORDER BY DTE ASC
        )
, ',', 1 ) AS opn_price,
SUBSTRING_INDEX( 
    GROUP_CONCAT(
        PRICE
        ORDER BY DTE DESC
        )
, ',', 1 ) AS cls_price 
FROM `CHART_DATA` 
GROUP BY trading_day
;

上面“Q”开头的数据是我试图结束的数据。希望这对其他人有所帮助，因为我怀疑我的日志相当普遍。

我敢打赌，这三种解决方案中的一种具有性能优势。如果有人碰巧知道 MySQL 的内部工作原理和查询优化并愿意推荐“首选”解决方案，那么将来了解这将很有用。

最终解决

更新#2

尝试使用这个从另一个方向来实现它：

http://forums.mysql.com/read.php?65,363723,363723

我得到：

SELECT
DATE_FORMAT(`DTE`, "%m/%d/%Y") AS trading_day,
MIN(`PRICE`) AS min_price,
MAX(`PRICE`) AS max_price,
(SELECT opn_price FROM 
    (SELECT 
    DATE_FORMAT(`DTE`, "%m/%d/%Y") AS a_day,    
    PRICE AS opn_price,
    MIN(DTE) AS opn
    FROM `CHART_DATA` 
    GROUP BY a_day
    ORDER BY opn ASC LIMIT 1) AS tblO) AS opnqt,
(SELECT cls_price FROM 
    (SELECT 
    DATE_FORMAT(`DTE`, "%m/%d/%Y") AS a_day,    
    PRICE AS cls_price,
    MIN(DTE) AS cls
    FROM `CHART_DATA` 
    GROUP BY a_day
    ORDER BY cls DESC LIMIT 1) AS tblC) AS clsqt    
FROM `CHART_DATA` cht
GROUP BY trading_day;

这与下面第一个 UPDATE 中的查询有类似的功能障碍；返回的 'clsqt' (cls_price) 值是在数据中找到的最后收盘价。布莱赫。

另外，我们开始再次进入“极其复杂”的查询空间，这对性能没有好处。

但是，如果有人看到 'clsqt' 值的修复，我会很高兴地接受它并在以后解决性能问题。:)

结束更新 #2

更新

如此接近......这就是我今天的位置：

SELECT 
DATE_FORMAT(`cht1`.`DTE`, "%m/%d/%Y") AS trading_day,
MIN(`cht1`.`PRICE`) AS min_price,
MAX(`cht1`.`PRICE`) AS max_price,
    MIN(cht1.DTE) AS opn_date1,
    MIN(cht1.DTE) AS opn_date2,
    `cht2`.`PRICE` AS opn_price,
        MAX(cht1.DTE) AS cls_date1,
        MAX(cht3.DTE) AS cls_date3,
        `cht3`.`PRICE` AS cls_price
FROM `CHART_DATA` cht1
     LEFT JOIN `CHART_DATA` cht2 
     ON cht2.DTE = cht1.DTE
     LEFT JOIN `CHART_DATA` cht3 
     ON cht3.DTE = cht1.DTE
GROUP BY trading_day
HAVING opn_date1 = opn_date2
AND cls_date1 = cls_date3
;

这将正确地重新调整所有内容，但正确的“cls_price”（“cls_price”的返回值与“opn_price”的值相同）。

但是，“cls_date1”和“cls_date3”是正确的值，所以我必须接近。

有人看到我不是什么吗？

结束更新

关于左连接和自连接，我一直在研究 SO……而且我必须承认我并没有在摸索。

我发现这个“Q”似乎非常接近我想要的： left join with condition for right table in mysql

我要做的是从单个表中检索开盘价、收盘价、最低价和最高价天数（下面的示例数据）。

最小值和最大值很简单：

SELECT 
DATE_FORMAT(`DTE`, "%m/%d/%Y") AS trading_day,
MIN(`PRICE`) AS min_price,
MAX(`PRICE`) AS max_price
FROM `CHART_DATA` 
GROUP BY trading_day;

我希望按日期分组返回结果，例如：

'trading_day' 'opn_price' 'min_price' 'max_price' 'cls_price'

好的，所以我只用一个连接来尝试“婴儿步骤”......

SELECT 
DATE_FORMAT(`cht1`.`DTE`, "%m/%d/%Y") AS trading_day,
MIN(`cht1`.`PRICE`) AS min_price,
MAX(`cht1`.`PRICE`) AS max_price,
`cht2`.`PRICE` AS opn_price
FROM `CHART_DATA` cht1
 LEFT JOIN `CHART_DATA` cht2 
 ON cht2.DTE = MIN(cht1.DTE)
GROUP BY trading_day;

...我收到消息“无效使用组功能”

当然，删除“GROUP BY”没有帮助，因为我需要返回聚合列。

我有一个非常复杂的解决方案，可以获得打开和关闭结果，但不是最小值和最大值——它们位于不同的结果集中。我觉得我已经使这比必要的更复杂，并且我可以理解上面提到的“Q”中引用的自连接发生了什么，我的整体编码将无法估量地即兴发挥。但在过去的周末，我花了大约 12 个小时在这上面，而且比以往任何时候都更加困惑。

在这一点上，欢迎所有的洞察力、解释和观察......

/* SAMPLE TABLE AND DATA */

CREATE TABLE `CHART_DATA` (
  `ID` varchar(10) DEFAULT NULL,
  `DTE` datetime DEFAULT NULL,
  `PRICE` double DEFAULT NULL
) ENGINE=MyISAM DEFAULT CHARSET=latin1;

/*Data for the table `CHART_DATA` */

INSERT INTO `chart_data`
            (`id`,`dte`,`price`)
VALUES      ('1','2011-01-01 00:10:00',0.65),
            ('2','2011-01-01 06:10:00',0.92),
            ('3','2011-01-01 12:10:00',0.59),
            ('4','2011-01-01 18:10:00',0.16),
            ('5','2011-01-02 00:10:00',0.28),
            ('6','2011-01-02 06:10:00',0.12),
            ('7','2011-01-02 12:10:00',0.92),
            ('8','2011-01-02 18:10:00',0.1),
            ('9','2011-01-03 00:10:00',0.34),
            ('10','2011-01-03 06:10:00',0.79),
            ('11','2011-01-03 12:10:00',1.23),
            ('12','2011-01-03 18:10:00',1.24),
            ('13','2011-01-04 00:10:00',1.12),
            ('14','2011-01-04 06:10:00',0.8),
            ('15','2011-01-04 12:10:00',0.65),
            ('16','2011-01-04 18:10:00',0.78),
            ('17','2011-01-05 00:10:00',0.65),
            ('18','2011-01-05 06:10:00',1.19),
            ('19','2011-01-05 12:10:00',0.89),
            ('20','2011-01-05 18:10:00',1.05),
            ('21','2011-01-06 00:10:00',0.29),
            ('22','2011-01-06 06:10:00',0.43),
            ('23','2011-01-06 12:10:00',0.26),
            ('24','2011-01-06 18:10:00',0.34),
            ('25','2011-01-07 00:10:00',0.22),
            ('26','2011-01-07 06:10:00',0.37),
            ('27','2011-01-07 12:10:00',1.22),
            ('28','2011-01-07 18:10:00',1.16),
            ('29','2011-01-08 00:10:00',0.3),
            ('30','2011-01-08 06:10:00',1.17),
            ('31','2011-01-08 12:10:00',0.62),
            ('32','2011-01-08 18:10:00',0.86),
            ('33','2011-01-09 00:10:00',0.84),
            ('34','2011-01-09 06:10:00',1.11),
            ('35','2011-01-09 12:10:00',0.92),
            ('36','2011-01-09 18:10:00',1.03),
            ('37','2011-01-10 00:10:00',1.13),
            ('38','2011-01-10 06:10:00',0.58),
            ('39','2011-01-10 12:10:00',1.03),
            ('40','2011-01-10 18:10:00',0.21),
            ('41','2011-01-11 00:10:00',0.12),
            ('42','2011-01-11 06:10:00',1.01),
            ('43','2011-01-11 12:10:00',0.19),
            ('44','2011-01-11 18:10:00',1.14),
            ('45','2011-01-12 00:10:00',0.55),
            ('46','2011-01-12 06:10:00',0.75),
            ('47','2011-01-12 12:10:00',0.66),
            ('48','2011-01-12 18:10:00',1.1),
            ('49','2011-01-13 00:10:00',0.68),
            ('50','2011-01-13 06:10:00',0.3),
            ('51','2011-01-13 12:10:00',0.9),
            ('52','2011-01-13 18:10:00',0.88),
            ('53','2011-01-14 00:10:00',0.64),
            ('54','2011-01-14 06:10:00',1.06),
            ('55','2011-01-14 12:10:00',1.12),
            ('56','2011-01-14 18:10:00',0.76),
            ('57','2011-01-15 00:10:00',0.18),
            ('58','2011-01-15 06:10:00',1.08),
            ('59','2011-01-15 12:10:00',0.66),
            ('60','2011-01-15 18:10:00',0.38),
            ('61','2011-01-16 00:10:00',1),
            ('62','2011-01-16 06:10:00',1.18),
            ('63','2011-01-16 12:10:00',1.15),
            ('64','2011-01-16 18:10:00',0.58),
            ('65','2011-01-17 00:10:00',1.04),
            ('66','2011-01-17 06:10:00',0.81),
            ('67','2011-01-17 12:10:00',0.35),
            ('68','2011-01-17 18:10:00',0.91),
            ('69','2011-01-18 00:10:00',0.14),
            ('70','2011-01-18 06:10:00',0.13),
            ('71','2011-01-18 12:10:00',1.03),
            ('72','2011-01-18 18:10:00',0.16),
            ('73','2011-01-19 00:10:00',1.05),
            ('74','2011-01-19 06:10:00',1.13),
            ('75','2011-01-19 12:10:00',1.21),
            ('76','2011-01-19 18:10:00',0.34),
            ('77','2011-01-20 00:10:00',0.63),
            ('78','2011-01-20 06:10:00',0.62),
            ('79','2011-01-20 12:10:00',0.19),
            ('80','2011-01-20 18:10:00',1.21),
            ('81','2011-01-21 00:10:00',0.83),
            ('82','2011-01-21 06:10:00',0.99),
            ('83','2011-01-21 12:10:00',0.83),
            ('84','2011-01-21 18:10:00',0.21),
            ('85','2011-01-22 00:10:00',0.8),
            ('86','2011-01-22 06:10:00',0.69),
            ('87','2011-01-22 12:10:00',0.87);
4

2 回答 2

1 
SELECT
  a.trading_day, a.min_price, a.max_price, 
  b.price as opn_price, 
  c.price as cls_price
FROM
  (SELECT 
     DATE_FORMAT(`DTE`, "%m/%d/%Y") AS trading_day,
     MIN(`PRICE`) AS min_price,
     MAX(`PRICE`) AS max_price,
     MIN(`dte`) AS open_date,
     MAX(`dte`) AS close_date
   FROM `CHART_DATA`
   GROUP BY trading_day) a
LEFT JOIN
  `CHART_DATA` b ON b.dte = a.open_date
LEFT JOIN
  `CHART_DATA` c ON c.dte = a.close_date

注意：如果您的打开或关闭条目与另一行具有完全相同的日期/时间值（即在打开之后立即发生的交易，或在关闭之前立即发生的交易），此解决方案可能会出现一些问题。为了解决这个问题，我建议您添加一个序列号，该序列号以保证唯一性和随时间递增的方式存储。如果你这样做，那么你将使用序列值来dte替换我在我的示例中用作连接字段的位置。open_dateclose_date

于 2011-09-20T20:56:27.307 回答
1

所以你要：

  • C对应于给定日期组中所有行的别名，因此您可以使用MAX()和MIN()覆盖该组中的行。
  • C2对应于给定日期的最后一行的别名。
  • C3对应于比同一天晚的一行的别名C2。如果没有找到，即为C3.*NULL，C2则为当天最新的。

这通常被标记为greatest-n-per-group查询，并且在 Stack Overflow 上经常出现。这是我为您的测试数据测试的解决方案，但您可以按照我添加到您的问题的标签进行其他解决方案和讨论。

编辑：我错过了开盘价和收盘价的要求。以下为编辑。

SELECT DATE_FORMAT(C.`DTE`, '%m/%d/%Y') AS trading_day, 
  MIN(C.`PRICE`) AS min_price, 
  MAX(C.`PRICE`) AS max_price, 
  Copen.`PRICE` AS opening_price,
  Cclose.`PRICE` AS closing_price 
FROM `CHART_DATA` AS C 
INNER JOIN `CHART_DATA` AS Cclose 
  ON DAY(C.`DTE`) = DAY(Cclose.`DTE`) 
LEFT OUTER JOIN `CHART_DATA` AS Cclose_later 
  ON DAY(C.`DTE`) = DAY(Cclose_later.`DTE`) AND Cclose.`DTE` < Cclose_later.`DTE`
INNER JOIN `CHART_DATA` AS Copen 
  ON DAY(C.`DTE`) = DAY(Copen.`DTE`) 
LEFT OUTER JOIN `CHART_DATA` AS Copen_earlier 
  ON DAY(C.`DTE`) = DAY(Copen_earlier.`DTE`) AND Copen.`DTE` < Copen_earlier.`DTE`
WHERE Cclose_later.`DTE` IS NULL AND Copen_earlier .`DTE` IS NULL 
GROUP BY trading_day;
于 2011-09-20T20:44:13.747 回答