在左连接中使用子查询时如何优化查询

Question

表格： 请在此处查看表格。 How to query counting specific wins of team and find the winner of the series

问题：

• 如何让查询更优化？
• 如何减少查询冗余？
• 如何让这个查询更快？

总结

正如您在示例查询中看到的，这部分被多次使用。

WHERE leagueid = 2096
AND start_time >= 1415938900
AND ((matches.radiant_team_id= 1848158 AND matches.dire_team_id= 15)
OR (matches.radiant_team_id= 15 AND matches.dire_team_id= 1848158))

---

SELECT matches.radiant_team_id,
       matches.dire_team_id,
       matches.radiant_name,
       matches.dire_name,
       TA.Count AS teamA,
       TB.Count AS teamB,
       TA.Count + TB.Count AS total_matches,
       SUM(TA.wins),
       SUM(TB.wins),
       (CASE
            WHEN series_type = 0 THEN 1
            WHEN series_type = 1 THEN 2
            WHEN series_type = 2 THEN 3
        END) AS wins_goal
FROM matches
LEFT JOIN
  (SELECT radiant_team_id,
          COUNT(id) AS COUNT,
          CASE
              WHEN matches.radiant_team_id = radiant_team_id && radiant_win = 1 THEN 1
          END AS wins
   FROM matches
   WHERE leagueid = 2096
     AND start_time >= 1415938900
     AND ((matches.radiant_team_id= 1848158
           AND matches.dire_team_id= 15)
          OR (matches.radiant_team_id= 15
              AND matches.dire_team_id= 1848158))
   GROUP BY radiant_team_id) AS TA ON TA.radiant_team_id = matches.radiant_team_id
LEFT JOIN
  (SELECT dire_team_id,
          COUNT(id) AS COUNT,
          CASE
              WHEN matches.dire_team_id = dire_team_id && radiant_win = 0 THEN 1
          END AS wins
   FROM matches
   WHERE leagueid = 2096
     AND start_time >= 1415938900
     AND ((matches.radiant_team_id= 1848158
           AND matches.dire_team_id= 15)
          OR (matches.radiant_team_id= 15
              AND matches.dire_team_id= 1848158))
   GROUP BY dire_team_id) AS TB ON TB.dire_team_id = matches.dire_team_id
WHERE leagueid = 2096
  AND start_time >= 1415938900
  AND ((matches.radiant_team_id= 1848158
        AND matches.dire_team_id= 15)
       OR (matches.radiant_team_id= 15
           AND matches.dire_team_id= 1848158))
GROUP BY series_id

预定比赛

ID| leagueid| team_a_id| team_b_id| starttime
 1|     2096|   1848158|        15| 1415938900

Answer 1

我相信不用子查询也可以做到。

我做了下面的比赛表

并使用以下查询对结果进行分组，每个系列一行

SELECT 
      matches.leagueid,
      matches.series_id,
      matches.series_type,
      COUNT(id) as matches,
      IF(radiant_team_id=1848158,radiant_name, dire_name) AS teamA,
      IF(radiant_team_id=1848158,dire_name, radiant_name) AS teamB,
      SUM(CASE
           WHEN radiant_team_id=1848158 AND radiant_win=1  THEN 1
           WHEN dire_team_id=1848158 AND radiant_win=0 THEN 1
           ELSE 0 END) AS teamAwin,
      SUM(CASE
           WHEN radiant_team_id=15 AND radiant_win=1  THEN 1
           WHEN dire_team_id=15 AND radiant_win=0 THEN 1
           ELSE 0 END) AS teamBwin

FROM `matches` 
WHERE leagueid = 2096
     AND start_time >= 1415938900
AND dire_team_id IN (15, 1848158)
AND radiant_team_id IN  (15, 1848158)
group by leagueid,series_id,series_type,teamA,teamB

产生以下结果

请注意，在对一个系列的结果进行分组时，并没有光辉队或可怕队之类的东西。在同一个系列中，光辉和可怕的角色可能会多次互换，所以我只称呼队伍为teamA和teamB。

现在，看看您之前的问题，我发现您需要根据系列赛类型和每支球队的胜利来确定系列赛的获胜者。这需要包装前一个查询并将其用作子查询，例如

SELECT matchresults.*,
      CASE series_type
      WHEN 0 then IF(teamAwin>=1, teamA,teamB)
      WHEN 1 then IF(teamAwin>=2, teamA,teamB)
      ELSE IF(teamAwin>=3, teamA,teamB)
      END as winner

from ( THE_MAIN_QUERY) as matchresults

Answer 2

可能有更有效的方法来获得您想要的结果。但是，为了使这个查询更有效，您可以添加索引。这是重复的where 子句：

WHERE leagueid = 2096 AND
      start_time >= 1415938900 AND
      ((matches.radiant_team_id= 1848158 AND matches.dire_team_id= 15) OR
       (matches.radiant_team_id= 15 AND matches.dire_team_id= 1848158))

or 的条件对于优化器来说很难。以下索引会有所帮助：matches(leagueid, start_time)。覆盖索引（至少对于where 条件）是matches(leagueid, start_time, radiant_team_id, dire_team_id)。我将从后一个索引开始，看看它是否能充分提高性能以达到您的目的。