Question

我正在尝试优化以下查询：

SELECT idnum AS ID, 
   (SELECT stddev(close) 
    from hist.eurusd1 
    where idnum between ID-7 AND ID) AS Sdev 
FROM hist.eurusd1 
where idnum between 1001 and 2000;

大约1000行需要 1800秒，我有大约100万行需要处理。

简单STDDEV查询的执行速度要快得多：

SELECT stddev(close)
from hist.eurusd1
where idnum between 1003 AND 1010;

对于获取部分，

首次为0.047秒，因此1000次查询将为47秒而不是1800秒。

我认为子查询中的某些内容会使整个第一个查询变慢，但不知道如何优化它。

Answer 1

我意识到这不是一个答案，因为我无法解释为什么给定的子查询如此缓慢。但我确实知道如何在大约30秒内计算百万行，所以我想发布代码：

SELECT a.idnum as id, stddev(b.close)
from eurusd1 as a
JOIN (
   select idnum+7 as ix, close from eurusd1
   UNION ALL
   select idnum+6 as ix, close from eurusd1
   UNION ALL
   select idnum+5 as ix, close from eurusd1
   UNION ALL
   select idnum+4 as ix, close from eurusd1
   UNION ALL
   select idnum+3 as ix, close from eurusd1
   UNION ALL
   select idnum+2 as ix, close from eurusd1
   UNION ALL
   select idnum+1 as ix, close from eurusd1
   UNION ALL
   select idnum as ix, close from eurusd1
    ) as b on a.idnum = b.ix
group by a.idnum

Answer 2

可以接近MariaDB 10.2 Window functions。

我没有看到STD，但AVG会像：

AVG(close) OVER(ORDER BY idnum
                RANGE BETWEEN 6 PRECEDING
                          AND CURRENT ROW)
     AS MovingAvg

使用stddev为外汇数据历史优化子查询

2 个答案: