mysql - 如何优化我的复杂 MySQL 查询？

Question

桌子

每行代表在特定日期特定时间播出的视频。每天大约有 1600 个视频。

CREATE TABLE `air_video` (
    `id` INT(10) UNSIGNED NOT NULL AUTO_INCREMENT,
    `date` DATE NOT NULL,
    `time` TIME NOT NULL,
    `duration` TIME NOT NULL,
    `asset_id` INT(10) UNSIGNED NOT NULL,
    `name` VARCHAR(100) NOT NULL,
    `status` VARCHAR(100) NULL DEFAULT NULL,
    `updated` TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
    PRIMARY KEY (`id`),
    UNIQUE INDEX `date_2` (`date`, `time`),
    INDEX `date` (`date`),
    INDEX `status` (`status`),
    INDEX `asset_id` (`asset_id`)
)
ENGINE=InnoDB

任务

有两个条件。

1. 每个视频每天必须播放不超过 24 次。
2. 每个视频的轮播时间不得超过 72 小时。

轮播意味着从拳头到最后一次视频播出之间的时间跨度。

所以我需要选择所有违反这些条件的视频，给定用户指定的日期范围。

结果必须按天和asset_id（视频ID）分组。例如：

      date      asset_id          name           dailyCount      rotationSpan
   2012-04-27      123      whatever_the_name         35              76
   2012-04-27      134      whatever_the_name2        39              20
   2012-04-28      125      whatever_the_name3        26              43

询问

现在我已经写了这个查询：

SELECT
    t1.date, t1.asset_id, t1.name,
    (SELECT 
        COUNT(t3.asset_id) 
        FROM air_video AS t3 
        WHERE t2.asset_id = t3.asset_id AND t3.date = t1.date
    ) AS 'dailyCount',
    MIN(CONCAT(t2.date, ' ', t2.time)) AS 'firstAir',
    MAX(CONCAT(t2.date, ' ', t2.time)) AS 'lastAir',
    ROUND(TIMESTAMPDIFF(
        MINUTE,
        MIN(CONCAT(t2.date, ' ', t2.time)),
        MAX(CONCAT(t2.date, ' ', t2.time)) 
    ) / 60) as 'rotationSpan'
FROM
    air_video AS t1
INNER JOIN
    air_video AS t2 ON
        t1.asset_id = t2.asset_id 
WHERE
    t1.status NOT IN ('bumpers', 'clock', 'weather')
    AND t1.date BETWEEN '2012-04-01' AND '2012-04-30'
GROUP BY
    t1.asset_id, t1.date
HAVING
    `rotationSpan` > 72
    OR `dailyCount` > 24
ORDER BY
    `date` ASC,
    `rotationSpan` DESC,
    `dailyCount` DESC    

问题

1. 用户指定天数之间的范围越大 - 完成查询所需的时间越长（一个月范围内大约需要 9 秒）
2. lastAir时间戳不是视频在特定日期播出的最晚时间，而是它完全播出的最晚时间。

Answer 1

如果您需要加快查询速度，您需要删除第 3 行的 select 子查询。要保持该计数，您可以在 from 子句中使用您最初使用的确切参数再次将其内部加入。它应该是这样的：

SELECT
    t1.date, t1.asset_id, t1.name,
    COUNT(t3.asset_id) AS 'dailyCount',
    MIN(CONCAT(t2.date, ' ', t2.time)) AS 'firstAir',
    MAX(CONCAT(t2.date, ' ', t2.time)) AS 'lastAir',
    ROUND(TIMESTAMPDIFF(
        MINUTE,
        MIN(CONCAT(t2.date, ' ', t2.time)),
        MAX(CONCAT(t2.date, ' ', t2.time)) 
    ) / 60) as 'rotationSpan'
FROM
    air_video AS t1
INNER JOIN
    air_video AS t2 ON
        (t1.asset_id = t2.asset_id)
INNER JOIN
    air_video AS t3 
        ON (t2.asset_id = t3.asset_id AND t3.date = t1.date)
WHERE
    t1.status NOT IN ('bumpers', 'clock', 'weather')
    AND t1.date BETWEEN '2012-04-01' AND '2012-04-30'
GROUP BY
    t1.asset_id, t1.date
HAVING
    `rotationSpan` > 72
    OR `dailyCount` > 24
ORDER BY
    `date` ASC,
    `rotationSpan` DESC,
    `dailyCount` DESC

由于 t2 不受日期约束，因此您显然正在查看整个表格，而不是日期范围。

编辑： 由于很多日期绑定，查询仍然运行得太慢。然后我采取了不同的方法。我创建了 3 个视图（您显然可以将它们组合成一个没有视图的普通查询，但我更喜欢最终结果查询）

--T1--

CREATE VIEW t1 AS select date,asset_id,name from air_video where (status not in ('bumpers','clock','weather')) group by asset_id,date order by date;

--T2--

CREATE VIEW t2 AS select t1.date,t1.asset_id,t1.name,min(concat(t2.date,' ',t2.time)) AS 'firstAir',max(concat(t2.date,' ',t2.time)) AS 'lastAir',round((timestampdiff(MINUTE,min(concat(t2.date,' ',t2.time)),max(concat(t2.date,' ',t2.time))) / 60),0) AS 'rotationSpan' from (t1 join air_video t2 on((t1.asset_id = t2.asset_id))) group by t1.asset_id,t1.date;

--T3--

CREATE VIEW t3 AS select t2.date,t2.asset_id,t2.name,count(t3.asset_id) AS 'dailyCount',t2.firstAir,t2.lastAir,t2.rotationSpan AS rotationSpan from (t2 join air_video t3 on(((t2.asset_id = t3.asset_id) and (t3.date = t2.date)))) group by t2.asset_id,t2.date;

然后，您可以从那里运行以下查询：

SELECT
    date,
    asset_id, 
    name,
    dailyCount,
    firstAir,
    lastAir,
    rotationSpan
FROM
    t3
WHERE
    date BETWEEN '2012-04-01' AND '2012-04-30'
    AND (
        rotationSpan > 72
        OR 
        dailyCount > 24
    )
ORDER BY
    date ASC,
    rotationSpan DESC,
    dailyCount DESC