为多个 window 函数重复使用相同的移动 window

Question

我在 Postgres 数据库中有 table 个数字，我正在尝试为每一行计算前 3 行和后 3 行的范围（最大-最小）。给定一个测试 table:

CREATE TABLE test_table AS (
SELECT
    generate_series(1,100) id,
    random()*generate_series(1,100) val
);

这可以通过以下查询完成：

SELECT
    id,
    MAX(val) OVER (
        ORDER BY id
        ROWS
            BETWEEN 3 PRECEDING
            AND 3 FOLLOWING
    ) - 
    MIN(val) OVER (
        ORDER BY id
        ROWS
            BETWEEN 3 PRECEDING
            AND 3 FOLLOWING
    ) val_range
FROM test_table;

但是 OVER 子句重复了。有没有办法改进查询并删除重复项？重复会影响性能吗？如果我想为每个id计算其他聚合，我应该如何处理移动window？

我提出了以下查询来重用 window，但我不确定这是否是最佳方法。

WITH
windowed AS (
SELECT
    id,
    UNNEST(ARRAY_AGG(val) OVER (
        ORDER BY id
        ROWS
            BETWEEN 3 PRECEDING
            AND 3 FOLLOWING
    )) val
FROM test_table
)
SELECT
    id,
    MAX(val)-MIN(val) val_range,
    AVG(val) val_mean,
    STDDEV(val) val_stddev
FROM windowed
GROUP BY id
ORDER BY id;

Answer 1

您可以使用 WINDOW clause 来缩短代码：

SELECT id
     , MAX(val) OVER w - 
       MIN(val) OVER w AS val_range
FROM   test_table
WINDOW w AS (ORDER BY id ROWS BETWEEN 3 PRECEDING AND 3 FOLLOWING);

但这对性能几乎没有影响。如果您也重复拼写，Postgres 将重用 window 框架。 The manual:

When multiple window functions are used, all the window functions having syntactically equivalent PARTITION BY and ORDER BY clauses in their window definitions are guaranteed to be evaluated in a single pass over the data.

为多个 window 函数重复使用相同的移动 window

Reusing the same moving window for multiple window functions

sql

postgresql

window-functions