与聚合和分析功能不同

Question

我只是好奇，有没有办法在 Postgres 中做到这一点？

SUM(DISTINCT column_a) OVER(PARTITION BY column_b, column_c)

使用 DISTINCT 会导致错误：DISTINCT 未针对窗口函数实现

Answer 1

这应该可以解决问题：

SELECT column_a,
       column_b,
       column_c,
       sum(column_a) FILTER (WHERE is_new) OVER w
FROM (SELECT column_a,
             column_b,
             column_c,
             column_a IS DISTINCT FROM lag(column_a) OVER w AS is_new
      FROM atable
      WINDOW w AS (PARTITION BY column_b, column_c ORDER BY column_a)
     ) AS q
WINDOW w AS (PARTITION BY column_b, column_c ORDER BY column_a);

在内部查询中，column_a 的所有重复项都会得到is_new = FALSE，因此这些重复项不计入外部查询。