MYSQL解释Rows_examined不匹配

Question

我在mysql_slow_queries日志中有如下查询：

Query_time: 4.642323 Lock_time: 1.821996 Rows_sent: 14 Rows_examined: 27099

SET TIMESTAMP=1356068688;
SELECT gw.id website_id, gw.name,gw.url,gw.language,gw.title,gw.nickname, gd.id, 
        gd.deal_title, gd.cdeal_title, gd.deal_details, gd.cdeal_details, 
        gd.discount_price, gd.original_price, gd.savings, gd.expiry, gd.shop, 
        gd.location, gd.clocation, gd.limited_offer, gd.contact, gd.url website, 
        gd.affiliate_url, gd.tags, gd.pic_url, gd.featured, gd.top_pos, 
        gd.sub_pos, gd.appeal, gd.redeem_until, gd.noofpurchased 
FROM groupon_deals gd 
INNER JOIN groupon_websites gw ON gw.id=gd.groupon_websites_id 
WHERE gd.tags LIKE '%technology-and-gadgets%' AND gd.pubDate >= SYSDATE() - INTERVAL 24 HOUR AND 
        gd.hidden = 0 AND gd.pubDate < SYSDATE() AND 
        gd.id NOT IN (1,30079,30090,30116,30118,30070,30136,30137,30138,30156,30103,30157,30038,30044,30084,30025,30013,30111,30030,30020,30059,30087,30026,30016,30112,30031,30021,30005,30092,30027,30017,30113,30049,30032,30023,30006,30096,30040,30028,30018,30120,30033,30024,30008,30110,30029,30019,30128,30131,30129,30100,30004,29995,30076,30126,30069,30078,30071,30034,30080,30065,30073,30082,29987,30074,30117,30068,29981,30098,30102,30088,30119,30135,30155,30107,29997,30041,30046,30077,30003,29992,30058,30097,30014,29999,30066,30127,30009,30081,29993,30060,30015,30114,30000,29985,30099,30010,30083,29994,30061,30022,30115,30001,30072,29986,30011,30086,30062,30123,30002,30075,29990,30054,30160,30094,30012,29998,30064,30125,30039,30130,30134,29982,30159,30048,30047,30158,30043,30101,30104,30106,30122,30056,30057,30063,30161,30053,29984,30132,30109,30036,30108,30037,30121,30045,30124) AND 
        (gw.language = 'C' OR gw.language = 'B') 
ORDER BY gd.sub_pos,gd.noofpurchased DESC

现在，当我转到phpMyAdmin并使用EXPLAIN运行相同的查询时，我在这里得到了输出：http://algaryeung.com/temp/explain-output.jpg

我有两个问题：

1)为什么日志中的rows_examined是27099，而解释37,756中的rows_examined是不同的？我是否需要在“解释”中多个两个值才能检查真实的行？

2)我知道这是一种开放的方式，但是如何改进现有的查询呢？我已经对字段groupon_deals.groupon_websites_id进行了索引，我认为可能有一些方法可以改进查询中的NOT部分。这里没有一个完整的答案，但是知道从哪里开始挖掘/学习吗？

Answer 1

MySQL EXPLAIN提供了对每一步返回的行数的估计和预测。

EXPLAIN真正提供给您的是执行计划，即将要使用的访问路径、操作顺序以及将使用的索引。它实际上并不是为了获得准确的行计数而处理语句，它只是根据它所掌握的关于表中行数的信息以及列中值的基数和分布来预测将检索多少行。

根据您提供的解释输出，查询正在对groupon_websites表进行全面扫描。对于检索到的每个id值(不被谓词消除)，MySQL在groupon_deals表的groupon_websites_id列上执行索引查找。

对于这个查询，使用索引可以提高一些性能。

... ON groupon_deals (groupon_websites_id, hidden, pubDate, id)

我认为从“挖掘”开始的一个好地方是理解EXPLAIN语句。

如果您对MySQL实际如何处理一个MySQL语句、哪些“操作”MySQL可以执行，以及哪些“操作”可以使用合适的索引有一定的了解，那么这是理解来自EXPLAIN的输出的基础。

我建议从这里开始，在MySQL文档中：理解查询执行计划