记一次解决oracle sql性能瓶颈的问题

版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/qq_17011423/article/details/79533487

先上sql：

SELECT
        (SELECT M.ALBUM_ID FROM ALBUM_R_MUSIC AM,ALBUM M WHERE AM.MUSIC_ID = M.MUSIC_ID AND AM.ALBUM_ID = M.ALBUM_ID  AND ROWNUM = 1) ALBUM_ID,
        (SELECT M.ALBUM_NAME FROM ALBUM_R_MUSIC AM,ALBUM M WHERE AM.MUSIC_ID = M.MUSIC_ID AND AM.ALBUM_ID = M.ALBUM_ID AND ROWNUM = 1) ALBUM_NAME,
        (SELECT TO_CHAR(WM_CONCAT(T.ARTIST_ID || ':' ||
        T.ARTIST_NAME))
        FROM MUSIC_R_ARTIST MR, ARTIST T
        WHERE M.MUSIC_ID = MR.MUSIC_ID(+)
        AND MR.ARTIST_ID = T.ARTIST_ID
        AND MR.ARTIST_TYPE = 1) ARTIST_ID,
        M.MUSIC_ID,
        M.LENGTH,
        M.MUSIC_NAME,
        M.SINGER_NAME,
        M.RANK_NUM,
        M.PUBLISH_YEAR,
        M.STATUS
        FROM
        <if test="tagId != null">
            TAG TA,
            OBJECT_R_TAG O,
        </if>
        MUSIC_R_ARTIST T,
        ARTIST_R_TYPE Y,
        MUSIC M
        <where>
        T.ARTIST_ID = Y.ARTIST_ID
        AND Y.TYPE_ID = '1'
        AND T.MUSIC_ID = M.MUSIC_ID
        AND M.STATUS = '1'
        AND T.ARTIST_TYPE = 1
        <if test="tagId != null">
            AND M.MUSIC_ID = O.OBJECT_ID(+)
            AND TA.TAG_ID = O.TAG_ID
            AND TA.TAG_ID = #{tagId}
            AND O.OBJECT_TYPE = 1
        </if>
        </where>
</php>

这个sql有九百多万条数据，分页采用的mybatis框架的RowBounds类。会把整个结果集扫描一次，很耗时间，大概分一次页要5S左右，到后面几页可能需要几分钟。。。

解决办法是将要关联的表写成存储过程，将MUSIC_R_ARTIST T,ARTIST_R_TYPE Y,MUSIC M这三张需要关联的表融合成一张大的表，并且将rownum作为一列，加上索引。存储过程会提前把需要的数据跑出来，这样我们在查询的时候去操作musicianas，如下：

CREATE OR REPLACE 
procedure migu_musician_songs
AS
begin
execute immediate 'drop table temp_musicianas';
execute   immediate 'create table temp_musicianas As SELECT m.*,rownum rn
  from MUSIC M
 where exists
 (select 1
          from (select music_id
                  from (select artist_id from ARTIST_R_TYPE where type_id = 1) Y,
                       MUSIC_R_ARTIST T
                 where y.artist_id = t.artist_id) t
         where t.music_id = m.music_id
           and m.status = 1)';
execute   immediate 'rename musicianas to temp_musicianas1';
execute   immediate 'rename temp_musicianas to musicianas';
execute   immediate 'create index IND_musicianas_rn on musicianas(rn)';
EXCEPTION
   WHEN OTHERS THEN
   ROLLBACK;
END;

这样我们可以通过行号，也就是rownum去做分页，取得从startIndex到endIndex的数据，就是我们需要的这一页的数据。sql如下：

SELECT
        (SELECT M.ALBUM_ID FROM ALBUM_R_MUSIC AM,ALBUM M WHERE AM.MUSIC_ID = M.MUSIC_ID AND AM.ALBUM_ID = M.ALBUM_ID  AND ROWNUM = 1) ALBUM_ID,
        (SELECT M.ALBUM_NAME FROM ALBUM_R_MUSIC AM,ALBUM M WHERE AM.MUSIC_ID = M.MUSIC_ID AND AM.ALBUM_ID = M.ALBUM_ID AND ROWNUM = 1) ALBUM_NAME,
        (SELECT TO_CHAR(WM_CONCAT(T.ARTIST_ID || ':' ||
        T.ARTIST_NAME))
        FROM MUSIC_R_ARTIST MR, ARTIST T
        WHERE M.MUSIC_ID = MR.MUSIC_ID(+)
        AND MR.ARTIST_ID = T.ARTIST_ID
        AND MR.ARTIST_TYPE = 1) ARTIST_ID,
        M.MUSIC_ID,
        M.LENGTH,
        M.MUSIC_NAME,
        M.SINGER_NAME,
        M.RANK_NUM,
        M.PUBLISH_YEAR,
        M.STATUS
        FROM
        <if test="tagId != null">
            OBJECT_R_TAG O,
        </if>
        musicianas M
        <where> 1= 1
        <if test="tagId != null">
            AND M.MUSIC_ID = O.OBJECT_ID(+)
            AND o.TAG_ID = #{tagId}
            AND O.OBJECT_TYPE = 1
        </if>
            and m.rn between ${startIndex} and ${endIndex}
        </where>

极大提升了效率：