自连接因 where 子句而失败

Question

我正在使用 apache drill 使用自引用连接在 CSV 文件中对 YTD（年初至今）求和。（缩短的）查询是

select
  ... fields from table a ...
  a.PeriodAmount,
  sum(cast(b.PeriodAmount as dec(18,3))) as YTDAmount
from dfs.`/home/foo/data/a.csv` a
  left join dfs.`/home/foo/data/a.csv` b
on
  ... join-conditions ...
*** where a.Year = '2018' ***
group by
  ... group-conditions ...
order by
  ... order-conditions ...
;

查询在没有 where 子句的情况下有效。当 where 子句在同一数据集上 included 时，我收到以下错误：

Error: UNSUPPORTED_OPERATION ERROR: This query cannot be planned possibly due to either a cartesian join or an inequality join
[Error Id: b62e6b63-eda7-4a52-8f95-2499a1f5c278 on foo:31010] (state=,code=0)

我可以通过删除 where 子句并改为执行子查询来避免错误：

from (select * from dfs.`/home/foo/data/a.csv` where Year = '2017') a
  from (select * from dfs.`/home/foo/data/a.csv` where Year = '2017') b

但我不确定这是正确的方法。它使查询更容易出错，因为必须将相同的条件应用于多个子查询，而不是将其作为自然所属的 where 子句。

是否可以重写此自连接以便保留 where 子句？

这是在 ubuntu 16.04 上，在 win10 上使用 WSL，apache drill 是 ver. 1.13.

完成（正在练习）查询：

select
  a.Dep_id,
  a.Dep,
  substr(a.Post_id, 1, 4) as Kap,
  a.Post_id,
  substr(a.Post_id, 5, 2) as Post,
  a.Art_id,
  a.Art,
  a.V_id,
  a.Reg,
  a.Dep_V_id,
  a.Dep_V,
  concat(substr(a.Periode, 1, 4), '-', substr(a.Periode, 5, 2), '-15') as PeriodDate,
  a.Period,
  a.Year,
  a.PeriodAmount,
  sum(cast(b.PeriodAmount as dec(18,3))) as YTDAmount
from dfs.`/home/foo/data/a.csv` a
  left join dfs.`/home/foo/data/a.csv` b
on
  a.Dep_id = b.Dep_id
  and a.Post_id = b.Post_id
  and a.Post_id is not null
  and a.Art_id = b.Art_id
  and a.V_id = b.V_id
  and a.Reg = b.Reg
  and a.Dep_V_id = b.Dep_V_id
  and a.Dep_id = b.Dep_id
  and b.Period <= a.Period
  and a.Year = b.Year
  and a.Post_id = b.Post_id
  and a.Art_id = b.Art_id
where a.Year in ('2018') and b.Year in (a.Year)
group by
  a.Dep_id,
  a.Dep,
  a.Post_id,
  a.Art_id,
  a.Art,
  a.V_id,
  a.Reg,
  a.Dep_V_id,
  a.Dep_V,
  a.Dep_id,
  a.Period,
  a.Year,
  a.PeriodAmount
order by
  a.Year,
  a.Dep_id,
  a.Post_id,
  a.Art_id,
  a.V_id,
  a.Reg,
  a.Dep_V_id,
  a.Dep_id,
  a.Period,
  a.PeriodAmount
;

Answer 1

我没有像这样查询过 csv 文件，所以这更像是一个尝试的建议。

像这样完成 a 和 b 的 where 子句以帮助编译器怎么样

WHERE a.Year = ‘2018’ AND b.Year = ‘2018’

或

WHERE a.Year = ‘2018’ AND b.Year = a.Year

自连接因 where 子句而失败

Self-join fails with where-clause

sql

csv

self-join

windows-subsystem-for-linux

apache-drill