使用 WITH 子句的语法错误

Question

我正在尝试设置一个包含一系列季度日期的 CTE 表。
查询返回 [42601] ERROR: syntax error at or near "values" Position: 38

with q(qqyy, firstday, lastday) as (
values
('Q4_10', '09-30-2010', '12-31-2010'),
('Q1_11', '12-31-2010', '03-31-2011'),
('Q2_11', '03-31-2011', '06/30/2011'),
('Q3_11', '06/30/2011', '09/30/2011'),
('Q4_11', '09/30/2011', '12/31/2011'),
('Q1_12', '12/31/2011', '03/31/2012'),
('Q2_12', '03/31/2012', '06/30/2012'),
('Q3_12', '06/30/2012', '09/30/2012'),
('Q4_12', '09/30/2012', '12/31/2012'),
('Q1_13', '12/31/2012', '03/31/2013'),
('Q2_13', '03/31/2013', '06/30/2013'),
('Q3_13', '06/30/2013', '09/30/2013'),
('Q4_13', '09/30/2013', '12/31/2013'),
('Q1_14', '12/31/2013', '03/31/2014'),
('Q2_14', '03/31/2014', '06/30/2014'),
('Q3_14', '06/30/2014', '09/30/2014'),
('Q4_14', '09/30/2014', '12/31/2014'),
('Q1_15', '12/31/2014', '03/31/2015'),
('Q2_15', '03/31/2015', '06/30/2015'),
('Q3_15', '06/30/2015', '09/30/2015'),
('Q4_15', '09/30/2015', '12/31/2015'),
('Q1_16', '12/31/2015', '03/31/2016'),
('Q2_16', '03/31/2016', '06/30/2016'),
('Q3_16', '06/30/2016', '09/30/2016'),
('Q4_16', '09/30/2016', '12/31/2016')
  )

SELECT q.qqyy, cobrand_id, sum(calc)
into   temp_08.cmg_calc
from temp_08.cmg s 
join q on 
s.transaction_date >= q.firstday 
and s.transaction_date <= q.lastday
GROUP BY q.qqyy, cobrand_id;

由于 Redshift 使用旧版本的 postgresql (http://docs.aws.amazon.com/redshift/latest/dg/c_unsupported-postgresql-features.html)，上述查询似乎卡在“值”上。但由于某种原因，以下也使用“值”的查询工作正常。知道如何使用 redshift 使上述查询工作吗？

 create table temp_08.cmgquarters (
quarter_col text
  , date_from   date
  , date_to     date
);


 insert into temp_08.cmgquarters
   values
('Q4_10', '09-30-2010', '12-31-2010'),
('Q1_11', '12-31-2010', '03-31-2011'),
('Q2_11', '03-31-2011', '06/30/2011'),
('Q3_11', '06/30/2011', '09/30/2011'),
('Q4_11', '09/30/2011', '12/31/2011'),
('Q1_12', '12/31/2011', '03/31/2012'),
('Q2_12', '03/31/2012', '06/30/2012'),
('Q3_12', '06/30/2012', '09/30/2012'),
('Q4_12', '09/30/2012', '12/31/2012'),
('Q1_13', '12/31/2012', '03/31/2013'),
('Q2_13', '03/31/2013', '06/30/2013'),
('Q3_13', '06/30/2013', '09/30/2013'),
('Q4_13', '09/30/2013', '12/31/2013'),
('Q1_14', '12/31/2013', '03/31/2014'),
('Q2_14', '03/31/2014', '06/30/2014'),
('Q3_14', '06/30/2014', '09/30/2014'),
('Q4_14', '09/30/2014', '12/31/2014'),
('Q1_15', '12/31/2014', '03/31/2015'),
('Q2_15', '03/31/2015', '06/30/2015'),
('Q3_15', '06/30/2015', '09/30/2015'),
('Q4_15', '09/30/2015', '12/31/2015'),
('Q1_16', '12/31/2015', '03/31/2016'),
('Q2_16', '03/31/2016', '06/30/2016'),
('Q3_16', '06/30/2016', '09/30/2016'),
('Q4_16', '09/30/2016', '12/31/2016');

Answer 1

由于 Redshift 不支持将 values() 作为“表替换”，您需要将其重写为 union：

with q(qqyy, firstday, lastday) as (
  select 'Q4_10', '09-30-2010', '12-31-2010' union all
  select 'Q1_11', '12-31-2010', '03-31-2011' union all
  ....
)
SELECT ...;

但是，您应该使用正确的 DATE 文字：

with q(qqyy, firstday, lastday) as (
  select 'Q4_10', DATE '2010-09-30', DATE '2010-12-31' union all
  select 'Q1_11', DATE '2010-12-31', DATE '2011-03-31' union all
  ....
)
SELECT ...;

Answer 2

我不太了解 Postgres，但是对于 SQL-Server，您不能像直接使用表一样使用 VALUES。您必须使用括号括起来并提供带有列名的表别名来定义派生表。

这将是这样的：

with q as (
select * from
(
    values
    ('Q4_10', '09-30-2010', '12-31-2010'),
    ('Q1_11', '12-31-2010', '03-31-2011'),
    ('Q2_11', '03-31-2011', '06/30/2011'),
    ('Q3_11', '06/30/2011', '09/30/2011'),
    ('Q4_11', '09/30/2011', '12/31/2011'),
    ('Q1_12', '12/31/2011', '03/31/2012'),
    ('Q2_12', '03/31/2012', '06/30/2012'),
    ('Q3_12', '06/30/2012', '09/30/2012'),
    ('Q4_12', '09/30/2012', '12/31/2012'),
    ('Q1_13', '12/31/2012', '03/31/2013'),
    ('Q2_13', '03/31/2013', '06/30/2013'),
    ('Q3_13', '06/30/2013', '09/30/2013'),
    ('Q4_13', '09/30/2013', '12/31/2013'),
    ('Q1_14', '12/31/2013', '03/31/2014'),
    ('Q2_14', '03/31/2014', '06/30/2014'),
    ('Q3_14', '06/30/2014', '09/30/2014'),
    ('Q4_14', '09/30/2014', '12/31/2014'),
    ('Q1_15', '12/31/2014', '03/31/2015'),
    ('Q2_15', '03/31/2015', '06/30/2015'),
    ('Q3_15', '06/30/2015', '09/30/2015'),
    ('Q4_15', '09/30/2015', '12/31/2015'),
    ('Q1_16', '12/31/2015', '03/31/2016'),
    ('Q2_16', '03/31/2016', '06/30/2016'),
    ('Q3_16', '06/30/2016', '09/30/2016'),
    ('Q4_16', '09/30/2016', '12/31/2016')
  ) AS tbl(qqyy, firstday, lastday)
)
SELECT *
from q

注意

你处于高度危险之中！

• 您正在使用文化相关的日期格式。这可能适用于您的系统，但在另一个系统上会中断......
• 此外，您甚至不一致！
• 您的VALUES 将您的日期值作为字符串提供。

在 SQL-Server 中，我建议使用未分隔的 ISO8601 或 - 我最喜欢的 - ODBC。但我确信 Postgres 中的文字日期也有与文化无关的格式。

我建议让 CTE 以类型值返回或使用带有类型列的临时表。