如何结合这两个mysql查询来保持它的速度？

Question

如何结合这两个mysql查询来保持它的速度？

我有两个MySQL查询,一个接一个,运行速度非常快:

查询1

SELECT Ads.AdId FROM Ads, AdsGeometry WHERE 
      AdsGeometry.AdId = Ads.AdId AND
      (ST_CONTAINS(GeomFromText('Polygon((
         -4.9783515930176 36.627100703563, 
         -5.0075340270996 36.61222072018, 
         -4.9896812438965 36.57638676015, 
         -4.965991973877 36.579419508882, 
         -4.955005645752 36.617732160006, 
         -4.9783515930176 36.627100703563
      ))'), AdsGeometry.GeomPoint)) 
GROUP BY Ads.AdId

Run Code Online (Sandbox Code Playgroud)

此查询以0.0013秒运行,并返回4行.

查询2

 SELECT Ads.AdId FROM Ads, AdsHierarchy WHERE 
      Ads.AdId = AdsHierarchy.ads_AdId AND  
      AdsHierarchy.locations_LocationId = 148022797 
 GROUP BY Ads.AdId

Run Code Online (Sandbox Code Playgroud)

此查询以0.0094秒运行,并返回67行(其中3行与上述查询相同).

我试图将这两个查询合并为一个查询,因为稍后,两个查询的结果集应该一起排序,我想使用MySQL进行排序.这是我尝试过的,在它下面,你会发现它也是解释:

SELECT Ads.AdId FROM Ads, AdsHierarchy, AdsGeometry WHERE 
      Ads.AdId = AdsHierarchy.ads_AdId AND 
      AdsGeometry.AdId = Ads.AdId AND ( 
          ST_CONTAINS(GeomFromText('Polygon((
             -4.9783515930176 36.627100703563, 
             -5.0075340270996 36.61222072018, 
             -4.9896812438965 36.57638676015, 
             -4.965991973877 36.579419508882, 
             -4.955005645752 36.617732160006, 
             -4.9783515930176 36.627100703563
          ))'), AdsGeometry.GeomPoint) OR 
          AdsHierarchy.locations_LocationId = 148022797
      ) 
GROUP BY Ads.AdId

id  select_type     table           type                  possible_keys                                 key                key_len  ref                      rows       Extra
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1   SIMPLE          AdsGeometry     ALL                   PRIMARY,GeomPoint,sx_adsgeometry_geompoint    NULL               NULL     NULL                     682848     Using temporary; Using filesort
1   SIMPLE          Ads             eq_ref                PRIMARY                                       PRIMARY            4        dbname.AdsGeometry.AdId  1          Using where; Using index
1   SIMPLE          AdsHierarchy    ref                   Ads_AdsHierarchy,locations_LocationId         Ads_AdsHierarchy   4        dbname.Ads.AdId          1          Using where

Run Code Online (Sandbox Code Playgroud)

虽然此查询返回正确的结果集(68行),但运行需要6.5937秒.如果我理解正确,AdsHierarchy表不使用它的索引,也不使用它AdsGeometry.

有没有办法将两个查询(或可能更多的位置,或这些基于多边形的查询)合并在一起,并保持合理的运行速度？

谢谢!

编辑:一些信息,关于3个表的索引

AdsGeometry表是MyISAM,主键是AdId.

结果SHOW INDEXES FROM AdsGeometry是:

Table           Non_unique  Key_name                    Seq_in_index    Column_name     Collation   Cardinality     Sub_part    Packed      Null    Index_type    Comment   Index_comment
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
AdsGeometry     0           PRIMARY                     1               AdId            A           682848          NULL        NULL        BTREE       
AdsGeometry     1           Latitude                    1               Latitude        A           NULL            NULL        NULL        BTREE       
AdsGeometry     1           Longitude                   1               Longitude       A           NULL            NULL        NULL        BTREE       
AdsGeometry     1           GeomPoint                   1               GeomPoint       A           NULL            32          NULL        SPATIAL         
AdsGeometry     1           sx_adsgeometry_geompoint    1               GeomPoint       A           NULL            32          NULL        SPATIAL         
AdsGeometry     1           Latitude_2                  1               Latitude        A           NULL            NULL        NULL        BTREE       
AdsGeometry     1           Latitude_2                  2               Longitude       A           NULL            NULL        NULL        BTREE

Run Code Online (Sandbox Code Playgroud)

AdsHierarchy表类型是InnoDB,主键是AdsHierarchyId.

结果SHOW INDEXES FROM AdsHierarchy是:

Table           Non_unique  Key_name                    Seq_in_index    Column_name           Collation     Cardinality     Sub_part    Packed      Null    Index_type    Comment   Index_comment
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
AdsHierarchy    0           PRIMARY                     1               AdsHierarchyId        A             2479044         NULL        NULL                BTREE       
AdsHierarchy    1           Ads_AdsHierarchy            1               ads_AdId              A             2479044         NULL        NULL                BTREE       
AdsHierarchy    1           locations_LocationId        1               locations_LocationId  A             123952          NULL        NULL                BTREE

Run Code Online (Sandbox Code Playgroud)

Ads表类型是InnoDB,主键是AdId.

结果SHOW INDEXES FROM Ads是:

Table           Non_unique  Key_name                    Seq_in_index    Column_name           Collation     Cardinality     Sub_part    Packed      Null    Index_type    Comment   Index_comment
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Ads             0           PRIMARY                     1               AdId                   A            705411          NULL        NULL                BTREE       
Ads             1           Accounts_Ads                1               accounts_AccountId     A            2               NULL        NULL                BTREE       
Ads             1           Ads_Locations               1               locations_LocationId   A            88176           NULL        NULL                BTREE       
Ads             1           Categories_Ads              1               categories_CategoryId  A            16              NULL        NULL                BTREE       
Ads             1           Currencies_Ads              1               currencies_Currency    A            2               NULL        NULL                BTREE       
Ads             1           countries_CountryId         1               countries_CountryId    A            204             NULL        NULL                BTREE       
Ads             1           ExternalId                  1               ExternalId             A            705411          NULL        NULL                BTREE       
Ads             1           ExternalId                  2               accounts_AccountId     A            705411          NULL        NULL                BTREE       
Ads             1           xml_XMLId                   1               xml_XMLId              A            4               NULL        NULL                BTREE       
Ads             1           streets_StreetId            1               streets_StreetId       A            2               NULL        NULL        YES     BTREE

Run Code Online (Sandbox Code Playgroud)

编辑2:使用隐式连接重写查询,并解释:

这是查询,重写为使用隐式连接,但它仍然运行得非常慢(5.503秒)

 SELECT a.AdId FROM Ads AS a 
   JOIN AdsHierarchy AS ah ON a.AdId = ah.ads_AdId   
   JOIN AdsGeometry AS ag ON a.AdId = ag.AdId 
   WHERE 
      ST_CONTAINS(GeomFromText('Polygon((
          -4.9783515930176 36.627100703563, 
          -5.0075340270996 36.61222072018, 
          -4.9896812438965 36.57638676015, 
          -4.965991973877 36.579419508882, 
          -4.955005645752 36.617732160006, 
          -4.9783515930176 36.627100703563
      ))'), ag.GeomPoint) 
      OR ah.locations_LocationId = 148022797
   GROUP BY a.AdId

id  select_type     table           type                  possible_keys                                 key                key_len  ref                      rows       Extra
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1   SIMPLE          a               index                 PRIMARY                                       PRIMARY            4        NULL                     627853     Using index
1   SIMPLE          ag              eq_ref                PRIMARY,GeomPoint,sx_adsgeometry_geompoint    PRIMARY            8        micasa_dev.a.AdId        1          Using index condition
1   SIMPLE          ah              ref                   Ads_AdsHierarchy,locations_LocationId         Ads_AdsHierarchy   4        micasa_dev.a.AdId        1          Using where

Run Code Online (Sandbox Code Playgroud)

编辑3:尝试联合两个查询

还尝试了UNION@RobertKoch提供的方法.

以下UNION查询运行速度非常快(0.06秒)

SELECT Ads.AdId FROM Ads, AdsGeometry 
WHERE 
    AdsGeometry.AdId = Ads.AdId AND 
    ST_CONTAINS(GeomFromText('Polygon(( 
         -4.9783515930176 36.627100703563, 
         -5.0075340270996 36.61222072018, 
         -4.9896812438965 36.57638676015, 
         -4.965991973877 36.579419508882, 
         -4.955005645752 36.617732160006, 
         -4.9783515930176 36.627100703563 
   ))'), AdsGeometry.GeomPoint) 
GROUP BY Ads.AdId 
UNION 
SELECT Ads.AdId FROM Ads, AdsHierarchy WHERE 
   Ads.AdId = AdsHierarchy.ads_AdId AND  
   AdsHierarchy.locations_LocationId = 148022797 
GROUP BY Ads.AdId

Run Code Online (Sandbox Code Playgroud)

我仍然无法使用此方法,因为稍后我需要根据Ads表对两个查询的合并得到的结果集进行排序.

如果我尝试执行以下操作,查询将再次变得非常慢(3.7秒):

SELECT Ads.AdId FROM Ads WHERE Ads.AdId IN (
  SELECT Ads.AdId FROM Ads, AdsGeometry 
  WHERE 
      AdsGeometry.AdId = Ads.AdId AND 
      ST_CONTAINS(GeomFromText('Polygon(( 
         -4.9783515930176 36.627100703563, 
         -5.0075340270996 36.61222072018, 
         -4.9896812438965 36.57638676015, 
         -4.965991973877 36.579419508882, 
         -4.955005645752 36.617732160006, 
         -4.9783515930176 36.627100703563 
      ))'), AdsGeometry.GeomPoint) 
  GROUP BY Ads.AdId 
  UNION 
  SELECT Ads.AdId FROM Ads, AdsHierarchy WHERE 
     Ads.AdId = AdsHierarchy.ads_AdId AND  
     AdsHierarchy.locations_LocationId = 148022797 
  GROUP BY Ads.AdId
) WHERE Ads.AdId > 100000
ORDER BY Ads.ModifiedDate ASC

Run Code Online (Sandbox Code Playgroud)

编辑4:改变UNION所在的位置,似乎可以解决问题

如果我修改上面的UNION查询

SELECT Ads.AdId
FROM Ads,
(SELECT Ads.AdId   
    FROM Ads,
    AdsGeometry
   WHERE AdsGeometry.AdId = Ads.AdId
     AND ST_CONTAINS(GeomFromText('Polygon(( 
         -4.9783515930176 36.627100703563, 
         -5.0075340270996 36.61222072018, 
         -4.9896812438965 36.57638676015, 
         -4.965991973877 36.579419508882, 
         -4.955005645752 36.617732160006, 
         -4.9783515930176 36.627100703563 
      ))'), AdsGeometry.GeomPoint)
   GROUP BY Ads.AdId
   UNION SELECT Ads.AdId
   FROM Ads,
        AdsHierarchy
   WHERE Ads.AdId = AdsHierarchy.ads_AdId
     AND AdsHierarchy.locations_LocationId = 148022797
   GROUP BY Ads.AdId) AS nt
WHERE Ads.AdId = nt.AdId
  AND Ads.AdId > 1000000
ORDER BY Ads.ModifiedDate ASC

Run Code Online (Sandbox Code Playgroud)

然后查询再次快速运行(~0.0007秒).

如果没有解决方案还没有UNION,我愿意给赏金的人谁可以解释两者之间的区别UNION版本(这个和那个在,编辑3),并且向我解释,为什么在查询运行速度快,当它被写入按以下顺序排列,并按以上顺序写入时运行缓慢.

如果需要任何其他信息,请在评论中询问,我尝试提供它们!谢谢

*注意:*我已经为两个UNION查询添加了一个ORDER,以使其更清晰,虽然我只是AdId从表中选择,但我仍然需要表中的其他字段Ads.

编辑5:@bovko的请求

1   SIMPLE  Ads     index   NULL                countries_CountryId     2   NULL                    627853  Using index; Using temporary
1   SIMPLE  ag      eq_ref  PRIMARY             PRIMARY                 8   micasa_dev.Ads.AdId     1       Using where; Distinct  
1   SIMPLE  ah      ref     Ads_AdsHierarchy    Ads_AdsHierarchy        4   micasa_dev.Ads.AdId     1       Using where; Distinct

Run Code Online (Sandbox Code Playgroud)

Answer 1

Ric*_*mes 3

IN ( SELECT ... )通常效率很低。躲开它。

到目前为止，所有的答案都比他们需要的更加努力。在之前似乎JOINs都是不必要的。请参阅下面的更多注释。UNION

SELECT Ads.AdId FROM Ads, JOIN ( ( SELECT AdId FROM AdsGeometry WHERE ST_CONTAINS(GeomFromText('Polygon(( -4.9783515930176 36.627100703563, -5.0075340270996 36.61222072018, -4.9896812438965 36.57638676015, -4.965991973877 36.579419508882, -4.955005645752 36.617732160006, -4.9783515930176 36.627100703563 ))'), AdsGeometry.GeomPoint) AND AdId > 1000000 ) UNION DISTINCT ( SELECT ads_AdId AS AdId FROM AdsHierarchy WHERE locations_LocationId = 148022797 AND ads_AdId > 1000000 ) ) AS nt ON Ads.AdId = nt.AdId ORDER BY Ads.ModifiedDate ASC
Run Code Online (Sandbox Code Playgroud)
笔记：

和AdsGeometry都有AdsHierarchyadId（名称不同）；不需要JOIN在内部查询中执行，除非可能验证它是否存在于中Ads。这是一个问题吗？无论如何，我的查询将在外部处理SELECT's JOIN。

UNION DISTINCT是必需的，因为两者SELECTs可能获取相同的 id。

移动> 1000000内部以减少收集的值的数量UNION。

总是UNION（在较旧版本的 MySQL 中）或有时（在较新版本中）创建临时表。你被困住了。

IN ( SELECT ... )通常优化得非常好；躲开它。

我加了一些括号；可以（但目前没有必要）将等添加到;ORDER BY中。UNION括号清楚地表明了它属于什么。

外部查询的唯一原因是为了ModifiedDate排序。您可以通过删除该要求来加快速度。（UNION可能会创建一个临时表；这ORDER BY可能会创建另一个表。）

归档时间：	9 年，8 月前
查看次数：	204 次
最近记录：	9 年，8 月前