组连接查询性能答案

【问题标题】：Group concat Query Performance组连接查询性能
【发布时间】：2021-12-16 21:04:31
【问题描述】：

我有一个查询几乎需要 22 秒来填充数据。

SELECT 
    TP.*, 
    GROUP_CONCAT(DISTINCT TS.`subject_name`) AS subjects, 
    GROUP_CONCAT(DISTINCT TC.`class_name`) AS classes, 
    GROUP_CONCAT(DISTINCT TT.`tution_name`) AS tution_type, 
    GROUP_CONCAT(DISTINCT TL.`name`) AS locations 
FROM `tutor_profile` TP 
LEFT JOIN `tutor_to_subject` TTS ON TP.`tutor_id`=TTS.`tutor` 
LEFT JOIN `tutor_subjects` TS ON TS.`subject_id`=TTS.`subject` 
LEFT JOIN `tutor_to_class` TTC ON TP.`tutor_id`=TTC.`tutor` 
LEFT JOIN `tutor_classes` TC ON TC.`class_id`=TTC.`class` 
LEFT JOIN `tutor_to_tution_type` TTTT ON TP.`tutor_id`=TTTT.`tutor` 
LEFT JOIN `tution_types` TT ON TT.`tution_id`=TTTT.`tution_type` 
LEFT JOIN `tutor_to_locality` TTL ON TP.`tutor_id`=TTL.`tutor` 
LEFT JOIN `tutor_locality` TL ON TL.`id`=TTL.`locality` 
WHERE 1=1 AND TP.`status` = 1 
GROUP BY TP.`tutor_id` 
ORDER BY TP.`date_added` DESC LIMIT 0 , 25

有没有办法提高它的性能？

【问题讨论】：

标签： mysql performance

【解决方案1】：

您可以尝试在输出列表中使用相关子查询，而不是 JOIN 和 GROUP BY：

SELECT 
    TP.*, 
    ( SELECT GROUP_CONCAT(DISTINCT TS.`subject_name`)
      FROM `tutor_to_subject` TTS 
      LEFT JOIN `tutor_subjects` TS ON TS.`subject_id`=TTS.`subject`
      WHERE TP.`tutor_id`=TTS.`tutor` ) AS subjects, 
-- the same for another output columns
FROM `tutor_profile` TP 
WHERE 1=1 AND TP.`status` = 1 
ORDER BY TP.`date_added` DESC LIMIT 0 , 25

【讨论】：

【解决方案2】：

替换这些

GROUP_CONCAT(DISTINCT TS.`subject_name`) AS subjects, 

LEFT JOIN `tutor_to_subject` TTS ON TP.`tutor_id`=TTS.`tutor` 
LEFT JOIN `tutor_subjects` TS ON TS.`subject_id`=TTS.`subject`

与

( SELECT GROUP_CONCAT(DISTINCT TS.`subject_name`
    FROM `tutor_to_subject` TTS
    JOIN `tutor_subjects` TS ON TS.`subject_id` = TTS.`subject`
    WHERE  TP.`tutor_id` = TTS.`tutor`  
) AS subjects,

然后为TS 抛出JOIN。对其他 3 个 group_concat + Left join 执行同样的操作。

对于多对多表，改进索引：

CREATE TABLE tutor_to_subject
    tutor_id ...,
    subject_id ...,
    PRIMARY KEY(tutor_id, subject_id),
    INDEX(subject_id)
) ENGINE=InnoDB;

更多讨论：Many:many mapping

另外，去掉GROUP BY，因为它不再需要了。

并添加

INDEX(status, date_added)

【讨论】：