优化复杂的 mysql 以减少查询时间

Question

我有一个非常复杂的查询，我已经优化了很多，但是我找不到更好的方法来编写更优化的查询并减少查询时间。让我分享详细信息，以便您更好地理解它。

下面是我的查询。

SELECT a.msisdn, GROUP_CONCAT( COALESCE(a.answer, 'skip') order by a.question_id separator ',') as answer,

(SELECT GROUP_CONCAT( COALESCE(answer, 'skip') order by question_id separator ',') as answer FROM  `campaign_survey_responses`
 WHERE campaign_id = 11559 and question_id=14751 and msisdn=a.msisdn) as a1,
 
 (SELECT GROUP_CONCAT( COALESCE(s.answer, 'skip') order by s.question_id separator ',') as answer 
FROM  `campaign_survey_responses` s left join campaign_survey_questions q on q.id = s.question_id 
WHERE s.campaign_id = 11559 and q.parent_id=5128 and q.sort_order = 0 and s.msisdn=a.msisdn) as sur1,
 
(SELECT GROUP_CONCAT( COALESCE(answer, 'skip') order by question_id separator ',') as answer FROM  `campaign_survey_responses`
WHERE campaign_id = 11559 and question_id=14768 and msisdn=a.msisdn) as a2,

(SELECT GROUP_CONCAT( COALESCE(s.answer, 'skip') order by s.question_id separator ',') as answer 
FROM  `campaign_survey_responses` s left join campaign_survey_questions q on q.id = s.question_id 
WHERE s.campaign_id = 11559 and q.parent_id=5108 and q.sort_order = 0 and s.msisdn=a.msisdn) as sur2,

(SELECT GROUP_CONCAT( COALESCE(answer, 'skip') order by question_id separator ',') as answer FROM  `campaign_survey_responses`
WHERE campaign_id = 11559 and question_id=14785 and msisdn=a.msisdn) as a3,

(SELECT GROUP_CONCAT( COALESCE(s.answer, 'skip') order by s.question_id separator ',') as answer 
FROM  `campaign_survey_responses` s left join campaign_survey_questions q on q.id = s.question_id 
WHERE s.campaign_id = 11559 and q.parent_id=5148 and q.sort_order = 0 and s.msisdn=a.msisdn) as sur3

FROM  `campaign_survey_responses` a
WHERE a.campaign_id = 11559 and a.question_id=14750 group by msisdn limit 500;

上面的查询结合了 100 万行，只有大约 50,000 行，现在唯一的目的是导出为 csv。但是上面的查询运行起来太耗时了

我已经从该查询应用了 500 行的限制，500 行限制大约需要 78.844856977463 秒，但是 50K 行呢？它将关闭服务器。

任何优化查询的更好方法，我正在使用 Mysql 5.7，谢谢

它returns数据如下，限制500，会有500条记录，下面仅以2条记录为例。

0 => 
    array (size=8)
      'msisdn' => string '3003932957' (length=10)
      'answer' => string '1' (length=1)
      'answer1' => string '4' (length=1)
      'survey1' => null
      'answer2' => null
      'survey2' => null
      'answer3' => null
      'survey3' => null
  1 => 
    array (size=8)
      'msisdn' => string '3013555354' (length=10)
      'answer' => string '1' (length=1)
      'answer1' => string '3' (length=1)
      'survey1' => string '2,1,1' (length=5)
      'answer2' => null
      'survey2' => null
      'answer3' => null
      'survey3' => null
  2 =>

Table结构：

CREATE TABLE `campaign_survey_responses` (
 `id` int unsigned NOT NULL AUTO_INCREMENT,
 `campaign_id` int unsigned NOT NULL,
 `question_id` int unsigned NOT NULL,
 `answer` varchar(20) DEFAULT NULL,
 `msisdn` int unsigned NOT NULL,
 `campaign_date` date DEFAULT NULL,
 PRIMARY KEY (`id`),
 KEY `campaign_id` (`campaign_id`),
 KEY `question_id` (`question_id`)
) ENGINE=MyISAM AUTO_INCREMENT=9016122 DEFAULT CHARSET=utf8

添加了

Table 结构，基本上是系统生成的调用，基于用户输入，它得到像 1、2、3 等的响应被添加到上面 table。例如，它是一种多层次调查。有一个家长问题，有3个选项，如果用户选择选项2，系统会播放与选项2相关的问题。在该问题之后，如果用户选择任何选项，将出现一系列与选项2相对应的10个问题。

与选择 1 和 3 类似。

Parent Question ( it has 3 sub questions )
Each 3 of subquestions has 10 questions each.

如果用户选择子题1，系统会播放与子题1相关的10道题，不会播放其他选择2,3的相关题。这取决于用户，他想玩哪个问题。

Answer 1

SELECT中的相关子查询可真是拖后腿了。你有很多这样的。如果这是不可避免的，我会将其应用于结果的子集而不是整个结果。即我会先创建内联视图

SELECT 
   -- your correlated subqueries as fields here
FROM
   ( -- start inline view
       SELECT -- your columns
       FROM  `campaign_survey_responses` a
       WHERE a.campaign_id = 11559 and a.question_id=14750 
       GROUP BY msisdn 
       LIMIT 500;
   ) T1

这样关联子查询只涉及500条记录

Answer 2

campaign_survey_responses 上的这些复合索引将帮助很多：

INDEX(campaign_id, question_id,  msisdn, answer)
INDEX(campaign_id, msisdn, question_id,  answer)

（第一个可能会帮助 a；第二个可能会帮助 s。）

优化复杂的 mysql 以减少查询时间

Optimizing a complex mysql to reduce query time

mysql

query-optimization