Question

我有一个小应用程序，它在每日计划中针对Redshift数据库执行一系列SQL脚本，并使用聚合数据填充表，以供客户提取。脚本保存在文本文件中，可以轻松更新，SQL从文件中提取出来＆＃39; | businessday |＆＃39;替换为所需日期，例如＆＃39; 20180501＆＃39;。没有强大的逻辑来基于当前日历日期的日期。

客户的要求已更改，现在有两个脚本只需要在该月的最后一天填充表格。我可以更新脚本，以便谓词读取：

WHERE (SELECT businessday FROM bd) = LAST_DAY((SELECT businessday FROM bd))

其中bd是CTE，所以我可以将日期字符串转换为DATE。

虽然这没有正确地返回任何记录，但它只需要稍微少的时间来执行，就像我在整个月运行它一样 - 它需要超过一分钟才能返回0行。我希望它能够快速将此谓词识别为失败，并且几乎不会立即返回任何行。

有没有办法重新构建SQL以首先评估此谓词？

我的理解是你不能在Redshift中使用过程IF语句，所以我只能在SQL字符串中添加谓词。

我尝试添加第二个CTE，它不会在关键表的businessday列上返回任何谓词：

WITH bd as (SELECT CAST('20180425' as date) as businessday WHERE 
    (SELECT CAST('20180425' as date)) = LAST_DAY(( CAST('20180425' as date)))
...
WHERE ts.businessday in (select businessday from bd)

（这需要修改以获得我需要的东西，但原理似乎不起作用）

简化的SQL字符串（删除了几个表和列）：

with cte as (select storeid from ttl_store_processed where
        businessday = '20180425'),
    bd as (SELECT CAST('20180425' as date) as businessday 
        WHERE (SELECT CAST('20180425' as date)) = LAST_DAY(( CAST('20180425' as date))))
SELECT store.storenumber AS COST_CENTER,
     TO_CHAR(DATE(tii.BusinessDay), 'YYYYMM') AS YEAR_MONTH,
     ii.ItemCode AS MATERIAL_NUMBER,
     SUM(tii.Quantity) AS UNITS
FROM cte s
    inner join transactionsale ts 
        on s.storeid = ts.storeid
    inner join Store store 
        on ts.storeid = store.storeid
    inner join transactionsaleitem tsi 
        on ts.transactionsaleid = tsi.transactionsaleid
    inner join transactioninventoryitem tii 
        on tsi.transactionsaleitemid = tii.transactionsaleitemid
    inner join inventoryitem ii 
        on tii.inventoryitemid = ii.inventoryitemid
WHERE (SELECT businessday FROM bd) = LAST_DAY((SELECT businessday FROM bd))
    AND ts.storeid IN (SELECT storeid FROM cte)
    AND ts.businessday BETWEEN DATE_TRUNC('MONTH', (SELECT businessday FROM bd)) 
        AND LAST_DAY((SELECT businessday FROM bd))
GROUP BY 
     store.storenumber,
     TO_CHAR(DATE(tii.BusinessDay), 'YYYYMM'),
     ii.ItemCode;

cte目前返回~20家商店，但这将增加到可能180+。我试过应用逻辑，所以这个表是空的：

with cte as (select storeid from mcdonaldshk.ttl_store_processed 
    where businessday = '20180425' and (SELECT CAST('20180425' as date)) 
        = LAST_DAY(( CAST('20180425' as date))))

这似乎不起作用

Answer 1

所以，你基本上说当你(SELECT businessday FROM bd) = LAST_DAY((SELECT businessday FROM bd))为假时，你想让它真的快速运行，先让它进行评估吗？

您可以尝试将查询加入子查询：

JOIN (SELECT 'end of month'
      FROM bd
      WHERE businessday = LAST_DAY(businessday)
      ) lastday ON (true)

这样，如果它不是最后一天，则返回零行，因此没有要连接的行。如果首先评估此问题，则查询的其余部分将不会执行，因为没有要加入的行。

顺便说一句，您还可以简化一些代码：

WHERE (SELECT CAST('20180425' as date)) = LAST_DAY(( CAST('20180425' as date)))

可以简单地说：

WHERE ('20180425'::date) = LAST_DAY('20180425'::date)

另外，如果您向JOIN添加bd，则可以简化

ts.businessday BETWEEN 
    DATE_TRUNC('MONTH', (SELECT businessday FROM bd)) 
    AND LAST_DAY((SELECT businessday FROM bd))

成：

ts.businessday BETWEEN 
    DATE_TRUNC('MONTH', businessday) 
    AND LAST_DAY(businessday)

强制Redshift首先评估特定谓词

1 个答案: