Question

我在PostgreSQL数据库中有这3个表：

艺术家： id，name
相册：ID，title，year，artist_id
歌曲： id，title，album_id

基本上每位艺术家都有多张专辑，每张专辑都有多首歌曲。

我的查询正确地返回了25个不同的artist.id，他们的歌曲的标题以“The”开头，以相册的年份排序：

SELECT id
FROM (

    -- Remove the duplicate artists
    SELECT DISTINCT ON (a.id) id, row
    FROM (

        -- Get all matching artists 
        -- This is slow because there is no limit
        SELECT
            artist.id,
            row_number() OVER(ORDER BY album.year DESC) as row
        FROM artist
        LEFT JOIN album ON album.artist_id = artist.id
        LEFT JOIN song ON song.album_id = album.id
        WHERE song.title ilike 'The %'
        ORDER BY album.year DESC

    ) as a

) as b
ORDER BY row
LIMIT 25

然而，它是缓慢且低效的，因为最里面的查询没有LIMIT，因此它将在整个表中搜索所有可能的匹配。理想情况下，当找到25个不同的artist.id时，它会停止搜索。

可以重写或优化此查询以更快地执行吗？

我认为window functions可以在这里加快速度，但我无法找到有用的东西。

谢谢！

Answer 1

select id, year
from (
    SELECT DISTINCT ON (artist.id) artist.id, album.year
    FROM artist
    inner JOIN album ON album.artist_id = artist.id
    inner JOIN song ON song.album_id = album.id
    WHERE song.title ilike 'The %'
    ORDER BY artist.id, album.year DESC
) s
order by year desc
LIMIT 25

song.title上的索引可以在ilike表达式以%

开头时提供帮助

Answer 2

试试这个，应该比当前查询更快地工作

SELECT
  artist.id,                        
  MAX( album.year ) as latest_album_date
FROM 
  artist
  JOIN album ON album.artist_id = artist.id  -- JOIN, not LEFT JOIN
  JOIN song ON song.album_id = album.id      -- Since you have song.title in WHERE clause, it makes no sense to make these as a LEFT JOIN
WHERE 
  song.title ilike 'The %'
GROUP BY
  artist.id
ORDER BY
  latest_album_date DESC
limit 25;

SQLFIDDLE

Answer 3

试试这个，

Select id, year
from (
    SELECT DISTINCT ON (artist.id) artist.id, album.year
    FROM artist
    rightJOIN album ON album.artist_id = artist.id
    left JOIN song ON song.album_id = album.id
    WHERE song.title ilike 'The %'
    ORDER BY artist.id, album.year DESC
) s
order by year desc
LIMIT 25

SQL窗口函数 - 按限制选择DISTINCT ORDER

3 个答案: