SQL書寫規(guī)范與各關鍵字詳解1、LIMIT語句2、隱式轉換3、關聯更新、刪除4、混合排序5、EXISTS語句6、條件下推7、提前縮小范圍8、中間結果集下推總結sql語句的執(zhí)行順序：FROM<left_table>ON<join_condition><join_type>JOIN<right_table>WHERE<where_condition>GROUPBY<group_by_list>HAVING<having_condition>SELECTDISTINCT<select_list>ORDERBY<order_by_condition>LIMIT<limit_number>1、LIMIT語句分頁查詢是最常用的場景之一，但也通常也是最容易出問題的地方。比如對于下面簡單的語句，一般DBA想到的辦法是在type,name,create_time字段上加組合索引。這樣條件排序都能有效的利用到索引，性能迅速提升。SELECT*FROMoperationWHEREtype='SQLStats'ANDname='SlowLog'ORDERBYcreate_timeLIMIT1000,10;好吧，可能90%以上的DBA解決該問題就到此為止。但當LIMIT子句變成“LIMIT1000000,10”時，程序員仍然會抱怨：我只取10條記錄為什么還是慢？要知道數據庫也并不知道第1000000條記錄從什么地方開始，即使有索引也需要從頭計算一次。出現這種性能問題，多數情形下是程序員偷懶了。在前端數據瀏覽翻頁，或者大數據分批導出等場景下，是可以將上一頁的最大值當成參數作為查詢條件的。SQL重新設計如下：SELECT*FROMoperationWHEREtype='SQLStats'ANDname='SlowLog'ANDcreate_time>'2017-03-1614:00:00'ORDERBYcreate_timelimit10;在新設計下查詢時間基本固定，不會隨著數據量的增長而發(fā)生變化。2、隱式轉換SQL語句中查詢變量和字段定義類型不匹配是另一個常見的錯誤。比如下面的語句：mysql>explainextendedSELECT*>FROMmy_balanceb>WHEREb.bpngt;ANDb.isverifiedISNULL;mysql>showwarnings;|Warning|1739|Cannotuserefaccessonindex'bpn'duetotypeorcollationconversiononfield'bpn'其中字段bpn的定義為varchar(20)，MySQL的策略是將字符串轉換為數字之后再比較。函數作用于表字段，索引失效。上述情況可能是應用程序框架自動填入的參數，而不是程序員的原意。現在應用框架很多很繁雜，使用方便的同時也小心它可能給自己挖坑。3、關聯更新、刪除雖然MySQL5.6引入了物化特性，但需要特別注意它目前僅僅針對查詢語句的優(yōu)化。對于更新或刪除需要手工重寫成JOIN。比如下面UPDATE語句，MySQL實際執(zhí)行的是循環(huán)/嵌套子查詢（DEPENDENTSUBQUERY)，其執(zhí)行時間可想而知。UPDATEoperationoSETstatus='applying'WHEREo.idIN(SELECTidFROM(SELECTo.id,o.statusFROMoperationoWHEREo.group=123ANDo.statusNOTIN('done')ORDERBYo.parent,o.idLIMIT1)t);執(zhí)行計劃：+----+--------------------+-------+-------+---------------+---------+---------+-------+------+-----------------------------------------------------+|id|select_type|table|type|possible_keys|key|key_len|ref|rows|Extra|+----+--------------------+-------+-------+---------------+---------+---------+-------+------+-----------------------------------------------------+|1|PRIMARY|o|index||PRIMARY|8||24|Usingwhere;Usingtemporary||2|DEPENDENTSUBQUERY||||||||ImpossibleWHEREnoticedafterreadingconsttables||3|DERIVED|o|ref|idx_2,idx_5|idx_5|8|const|1|Usingwhere;Usingfilesort|+----+--------------------+-------+-------+---------------+---------+---------+-------+------+-----------------------------------------------------+重寫為JOIN之后，子查詢的選擇模式從DEPENDENTSUBQUERY變成DERIVED，執(zhí)行速度大大加快，從7秒降低到2毫秒。UPDATEoperationoJOIN(SELECTo.id,o.statusFROMoperationoWHEREo.group=123ANDo.statusNOTIN('done')ORDERBYo.parent,o.idLIMIT1)tONo.id=t.idSETstatus='applying'執(zhí)行計劃簡化為：+----+-------------+-------+------+---------------+-------+---------+-------+------+-----------------------------------------------------+|id|select_type|table|type|possible_keys|key|key_len|ref|rows|Extra|+----+-------------+-------+------+---------------+-------+---------+-------+------+-----------------------------------------------------+|1|PRIMARY||||||||ImpossibleWHEREnoticedafterreadingconsttables||2|DERIVED|o|ref|idx_2,idx_5|idx_5|8|const|1|Usingwhere;Usingfilesort|+----+-------------+-------+------+---------------+-------+---------+-------+------+-----------------------------------------------------+4、混合排序MySQL不能利用索引進行混合排序。但在某些場景，還是有機會使用特殊方法提升性能的。SELECT*FROMmy_orderoINNERJOINmy_appraiseaONa.orderid=o.idORDERBYa.is_replyASC,a.appraise_timeDESCLIMIT0,20執(zhí)行計劃顯示為全表掃描：+----+-------------+-------+--------+-------------+---------+---------+---------------+---------+-+|id|select_type|table|type|possible_keys|key|key_len|ref|rows|Extra+----+-------------+-------+--------+-------------+---------+---------+---------------+---------+-+|1|SIMPLE|a|ALL|idx_orderid|NULL|NULL|NULL|1967647|Usingfilesort||1|SIMPLE|o|eq_ref|PRIMARY|PRIMARY|122|a.orderid|1|NULL|+----+-------------+-------+--------+---------+---------+---------+-----------------+---------+-+由于is_reply只有0和1兩種狀態(tài)，我們按照下面的方法重寫后，執(zhí)行時間從1.58秒降低到2毫秒。SELECT*FROM((SELECT*FROMmy_orderoINNERJOINmy_appraiseaONa.orderid=o.idANDis_reply=0ORDERBYappraise_timeDESCLIMIT0,20)UNIONALL(SELECT*FROMmy_orderoINNERJOINmy_appraiseaONa.orderid=o.idANDis_reply=1ORDERBYappraise_timeDESCLIMIT0,20))tORDERBYis_replyASC,appraisetimeDESCLIMIT20;5、EXISTS語句MySQL對待EXISTS子句時，仍然采用嵌套子查詢的執(zhí)行方式。如下面的SQL語句：SELECT*FROMmy_neighbornLEFTJOINmy_neighbor_applysraONn.id=sra.neighbor_idANDsra.user_id='xxx'WHEREn.topic_status<4ANDEXISTS(SELECT1FROMmessage_infomWHEREn.id=m.neighbor_idANDm.inuser='xxx')ANDn.topic_type<>5執(zhí)行計劃為：+----+--------------------+-------+------+-----+------------------------------------------+---------+-------+---------+-----+|id|select_type|table|type|possible_keys|key|key_len|ref|rows|Extra|+----+--------------------+-------+------+-----+------------------------------------------+---------+-------+---------+-----+|1|PRIMARY|n|ALL||NULL|NULL|NULL|1086041|Usingwhere||1|PRIMARY|sra|ref||idx_user_id|123|const|1|Usingwhere||2|DEPENDENTSUBQUERY|m|ref||idx_message_info|122|const|1|Usingindexcondition;Usingwhere|+----+--------------------+-------+------+-----+------------------------------------------+---------+-------+---------+-----+去掉exists更改為join，能夠避免嵌套子查詢，將執(zhí)行時間從1.93秒降低為1毫秒。SELECT*FROMmy_neighbornINNERJOINmessage_infomONn.id=m.neighbor_idANDm.inuser='xxx'LEFTJOINmy_neighbor_applysraONn.id=sra.neighbor_idANDsra.user_id='xxx'WHEREn.topic_status<4ANDn.topic_type<>5新的執(zhí)行計劃：+----+-------------+-------+--------+-----+------------------------------------------+---------+-----+------+-----+|id|select_type|table|type|possible_keys|key|key_len|ref|rows|Extra|+----+-------------+-------+--------+-----+------------------------------------------+---------+-----+------+-----+|1|SIMPLE|m|ref||idx_message_info|122|const|1|Usingindexcondition||1|SIMPLE|n|eq_ref||PRIMARY|122|ighbor_id|1|Usingwhere||1|SIMPLE|sra|ref||idx_user_id|123|const|1|Usingwhere|+----+-------------+-------+--------+-----+------------------------------------------+---------+-----+------+-----+6、條件下推外部查詢條件不能夠下推到復雜的視圖或子查詢的情況有：1、聚合子查詢；2、含有LIMIT的子查詢；3、UNION或UNIONALL子查詢；4、輸出字段中的子查詢；如下面的語句，從執(zhí)行計劃可以看出其條件作用于聚合子查詢之后：SELECT*FROM(SELECTtarget,Count(*)FROMoperationGROUPBYtarget)tWHEREtarget='rm-xxxx'+----+-------------+------------+-------+---------------+-------------+---------+-------+------+-------------+|id|select_type|table|type|possible_keys|key|key_len|ref|rows|Extra|+----+-------------+------------+-------+---------------+-------------+---------+-------+------+-------------+|1|PRIMARY|<derived2>|ref|<auto_key0>|<auto_key0>|514|const|2|Usingwhere||2|DERIVED|operation|index|idx_4|idx_4|519|NULL|20|Usingindex|+----+-------------+------------+-------+---------------+-------------+---------+-------+------+-------------+確定從語義上查詢條件可以直接下推后，重寫如下：SELECTtarget,Count(*)FROMoperationWHEREtarget='rm-xxxx'GROUPBYtarget執(zhí)行計劃變?yōu)椋?----+-------------+-----------+------+---------------+-------+---------+-------+------+--------------------+|id|select_type|table|type|possible_keys|key|key_len|ref|rows|Extra|+----+-------------+-----------+------+---------------+-------+---------+-------+------+--------------------+|1|SIMPLE|operation|ref|idx_4|idx_4|514|const|1|Usingwhere;Usingindex|+----+-------------+-----------+------+---------------+-------+---------+-------+------+--------------------+關于MySQL外部條件不能下推的詳細解釋說明請參考以前文章：MySQL·性能優(yōu)化·條件下推到物化表/monthly/2016/07/087、提前縮小范圍先上初始SQL語句：SELECT*FROMmy_orderoLEFTJOINmy_userinfouONo.uid=u.uidLEFTJOINmy_productinfopONo.pid=p.pidWHERE(o.display=0)AND(o.ostaus=1)ORDERBYo.selltimeDESCLIMIT0,15該SQL語句原意是：先做一系列的左連接，然后排序取前15條記錄。從執(zhí)行計劃也可以看出，最后一步估算排序記錄數為90萬，時間消耗為12秒。+----+-------------+-------+--------+---------------+---------+---------+-----------------+--------+----------------------------------------------------+|id|select_type|table|type|possible_keys|key|key_len|ref|rows|Extra|+----+-------------+-------+--------+---------------+---------+---------+-----------------+--------+----------------------------------------------------+|1|SIMPLE|o|ALL|NULL|NULL|NULL|NULL|909119|Usingwhere;Usingtemporary;Usingfilesort||1|SIMPLE|u|eq_ref|PRIMARY|PRIMARY|4|o.uid|1|NULL||1|SIMPLE|p|ALL|PRIMARY|NULL|NULL|NULL|6|Usingwhere;Usingjoinbuffer(BlockNestedLoop)|+----+-------------+-------+--------+---------------+---------+---------+-----------------+--------+----------------------------------------------------+由于最后WHERE條件以及排序均針對最左主表，因此可以先對my_order排序提前縮小數據量再做左連接。SQL重寫后如下，執(zhí)行時間縮小為1毫秒左右。SELECT*FROM(SELECT*FROMmy_orderoWHERE(o.display=0)AND(o.ostaus=1)ORDERBYo.selltimeDESCLIMIT0,15)oLEFTJOINmy_userinfouONo.uid=u.uidLEFTJOINmy_productinfopONo.pid=p.pidORDERBYo.selltimeDESClimit0,15再檢查執(zhí)行計劃：子查詢物化后（select_type=DERIVED)參與JOIN。雖然估算行掃描仍然為90萬，但是利用了索引以及LIMIT子句后，實際執(zhí)行時間變得很小。+----+-------------+------------+--------+---------------+---------+---------+-------+--------+----------------------------------------------------+|id|select_type|table|type|possible_keys|key|key_len|ref|rows|Extra|+----+-------------+------------+--------+---------------+---------+---------+-------+--------+----------------------------------------------------+|1|PRIMARY|<derived2>|ALL|NULL|NULL|NULL|NULL|15|Usingtemporary;Usingfilesort||1|PRIMARY|u|eq_ref|PRIMARY|PRIMARY|4|o.uid|1|NULL||1|PRIMARY|p|ALL|PRIMARY|NULL|NULL|NULL|6|Usingwhere;Usingjoinbuffer(BlockNestedLoop)||2|DERIVED|o|index|NULL|idx_1|5|NULL|909112|Usingwhere|+----+-------------+------------+--------+---------------+---------+---------+-------+--------+----------------------------------------------------+8、中間結果集下推再來看下面這個已經初步優(yōu)化過的例子(左連接中的主表優(yōu)先作用查詢條件)：SELECTa.*,c.allocatedFROM(SELECTresourceidFROMmy_distributedWHEREisdelete=0ANDcusmanagercode='1234567'ORDERBYsalecodelimit20)aLEFTJOIN(SELECTresourcesid，sum(ifnull(allocation,0)*12345)allocatedFROMmy_resourcesGROUPBYresourcesid)cONa.resourceid=c.resourcesid那么該語句還存在其它問題嗎？不難看出子查詢c是全表聚合查詢，在表數量特別大的情況下會導致整個語句的性能下降。其實對于子查詢c，左連接最后結果集只關心能和主表resourceid能匹配的數據。因此我們可以重寫語句如下，執(zhí)行時間從原來的2秒下降到2毫秒。SELECTa.*,c.allocatedFROM(SELECTresourceidFROMmy_distributedWHEREisdelete=0ANDcusman