背景

近期在做用戶標(biāo)簽項目，目前標(biāo)簽的存儲是用戶id，標(biāo)簽ids（每個標(biāo)簽以,形式存儲） 的形式，但是如果想統(tǒng)計前后兩天標(biāo)簽的變化，使用find_in_set函數(shù)，一方面查詢速度慢（因為不能使用索引），另一方面目前標(biāo)簽已有將近300多個，以后還會更多，一個標(biāo)簽一個標(biāo)簽的寫，使得sql特別長。
目前采取的策略是：將用戶標(biāo)簽表拆分成 用戶id，標(biāo)簽id的形式。這樣將前后兩天的表關(guān)聯(lián)，就能查出昨天到今天有那些標(biāo)簽離開，哪些標(biāo)簽進(jìn)來。
那么采用上述策略就需要研究怎么將用戶id，標(biāo)簽ids的形式轉(zhuǎn)換成用戶id，標(biāo)簽id的形式。

探索

我們知道像 1,2,3,4,5,212 這種字符串，如果要分別取到1 2 3 4 5 212，用編程的思想就是先將該字符串用","分隔成一個數(shù)據(jù)，然后遍歷取到數(shù)組里的每一個值，但是在mysql里并沒有數(shù)組的概念，但是我們可以用各種方法求得字符串的長度，以及求得使用“,”分隔后有多少個值。也可以用mysql可以采取的字符串截取的形式去獲得相應(yīng)位置的數(shù)值。下面就讓我們看一下吧~

實現(xiàn)

• 相關(guān)表結(jié)構(gòu)

CREATE TABLE `tagids_label` (
  `userid` int(11) NOT NULL COMMENT '用戶id',
  `label` int(11) NOT NULL COMMENT '標(biāo)記，暫時 保留三天的數(shù)據(jù)，day%3 ',
  `day` int(11) NOT NULL COMMENT '對應(yīng)的統(tǒng)計日期的天',
  `tagids` text NOT NULL COMMENT '標(biāo)簽id，以，（英文）分隔',
  `createTime` datetime NOT NULL COMMENT '創(chuàng)建時間',
  `updateTime` datetime NOT NULL COMMENT '更新時間',
  PRIMARY KEY (`userid`,`label`),
  KEY `index_day` (`day`),
  KEY `index_label` (`label`),
  KEY `index_label_userid` (`userid`,`label`),
  KEY `index_createTime_userid` (`userid`,`createTime`),
  KEY `index_userid` (`userid`),
  KEY `index_createtime` (`createTime`) USING BTREE,
  FULLTEXT KEY `index_tagids` (`tagids`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COMMENT='用戶標(biāo)簽結(jié)果表'

CREATE TABLE `sequence` (
  `seq` int(3) DEFAULT NULL
) ENGINE=InnoDB DEFAULT CHARSET=utf8

ps:sequence表表示標(biāo)簽的個數(shù)，從1到最大個數(shù)

• 獲得固定分隔符分隔后元素個數(shù)

  • 原數(shù)據(jù)

  SELECT * FROM `tagids_label` WHERE `userid` =2
  

  blockchain

  171,172,173,174,175,184,187,189,191,192,49,52,55,90,96,101,104,110,7,9,253,270,277,280,129,131,134,136,138,139,231,241,58,63,66,70,72,75,77,79,84,149,150,159,163,165,166,193,195,256,225,236,246,248,197,200,207,221,210,278,227
  

  • 分隔符分隔后元素個數(shù)

  SELECT length(`tagids`) ,length(REPLACE (`tagids`,',','')),length(`tagids`)-length(REPLACE (`tagids`,',',''))+1  FROM `tagids_label` WHERE `userid` =2
  

  image.png
  • 注：length(tagids)計算字符串長度，以字節(jié)為單位，每個數(shù)字、英文標(biāo)點符號是一個字節(jié)，每個中文、中文標(biāo)點符號是3個字節(jié)。length(tagids)表示tagids含有多少個數(shù)字和標(biāo)點符號
    replace(tagids,',','')將字符串tagids里的","用空字符來代替，length(replace(tagids,',',''))就表示tagids含有多少個數(shù)字。
    length(tagids)-length(replace(tagids,',','')) 表示tagsid含有多少個標(biāo)點符號，那標(biāo)點符號+1就表示tagids用","分隔符分隔后含有多少個元素，即標(biāo)簽個數(shù)。

• substring_index 截取字符串
  • 根據(jù)關(guān)鍵字","截取字符串

  SELECT substring_index('171,172,173,174,175,184,187,189,191,192,49,52,55,90,96,101,104,110,7,9,253,270,277,280,129,131,134,136,138,139,231,241,58,63,66,70,72,75,77,79,84,149,150,159,163,165,166,193,195,256,225,236,246,248,197,200,207,221,210,278,227', ',',1)
  UNION ALL 
  SELECT substring_index('171,172,173,174,175,184,187,189,191,192,49,52,55,90,96,101,104,110,7,9,253,270,277,280,129,131,134,136,138,139,231,241,58,63,66,70,72,75,77,79,84,149,150,159,163,165,166,193,195,256,225,236,246,248,197,200,207,221,210,278,227', ',',2)
  UNION ALL 
  SELECT substring_index('171,172,173,174,175,184,187,189,191,192,49,52,55,90,96,101,104,110,7,9,253,270,277,280,129,131,134,136,138,139,231,241,58,63,66,70,72,75,77,79,84,149,150,159,163,165,166,193,195,256,225,236,246,248,197,200,207,221,210,278,227', ',',3)
  UNION ALL 
  SELECT substring_index('171,172,173,174,175,184,187,189,191,192,49,52,55,90,96,101,104,110,7,9,253,270,277,280,129,131,134,136,138,139,231,241,58,63,66,70,72,75,77,79,84,149,150,159,163,165,166,193,195,256,225,236,246,248,197,200,207,221,210,278,227', ',',4)
  UNION ALL 
  ......
  UNION ALL
  SELECT substring_index('171,172,173,174,175,184,187,189,191,192,49,52,55,90,96,101,104,110,7,9,253,270,277,280,129,131,134,136,138,139,231,241,58,63,66,70,72,75,77,79,84,149,150,159,163,165,166,193,195,256,225,236,246,248,197,200,207,221,210,278,227',',',61)
  

  結(jié)果：

  171
  171,172
  171,172,173
  171,172,173,174
  ......
  171,172,173,174,175,184,187,189,191,192,49,52,55,90,96,101,104,110,7,9,253,270,277,280,129,131,134,136,138,139,231,241,58,63,66,70,72,75,77,79,84,149,150,159,163,165,166,193,195,256,225,236,246,248,197,200,207,221,210,278,227
  

  • 注：substring_index（str,delim,count） 說明：substring_index（被截取字段，關(guān)鍵字，關(guān)鍵字出現(xiàn)的次數(shù)），如果count=-1我們就可以截取到倒數(shù)第一個被關(guān)鍵字分隔的元素。只要在上面查詢結(jié)果中再使用一次substring_index即可獲得每個被關(guān)鍵字分隔的元素。

• 最終實現(xiàn)

SELECT
    userid,
    SUBSTRING_INDEX(
        SUBSTRING_INDEX(tagids, ',', seq),
        ',' ,- 1
    ) sub_id,
    seq
FROM sequence
 JOIN (SELECT * FROM `tagids_label` WHERE  userid = 2)b
WHERE
    seq BETWEEN 1
AND (
    SELECT
        1 + LENGTH(tagids) - LENGTH(replace(tagids, ',', ''))
)
ORDER BY
    userid,
    tagids;

• 結(jié)果：

  
  
  image.png
  
  
  image.png