无法将窗口调用分解为组。错误：org.apache.hadoop.hive.ql.parse

Posted 2021-03-27

我正在尝试从两个表连接两列，以生成一列的唯一id.Max列值，其中包含另一个表的行号。

select (MAX(S.m_id))from MPPO S;
select row_number() OVER(ORDER BY G.a,G.r,G.f1,STG.filler2,G.n_p,G.fe,G.se) 
FROM mmp G
LEFT OUTER JOIN mppo S
ON TRIM(G.pc) = S.pc;

但是将这两个查询结合起来如下：

select (MAX(S.m_id))+ row_number() OVER(ORDER BY G.a,G.r,G.f1,STG.filler2,G.n_p,G.fe,G.se) 
FROM mmp G LEFT OUTER JOIN mppo S
ON TRIM(G.pc) = S.pc;

我收到以下错误：

SemanticException Failed to breakup Windowing invocations into Groups. At least 1 group
must only depend on input columns. Also check for circular dependencies. Underlying error:
org.apache.hadoop.hive.ql.parse.SemanticException

我究竟做错了什么？请帮忙

答案

分别从每个表中选择id然后加入输出：

select concat(t.id,'',t1.id) from (select MAX(S.m_id) as id from MPPO s) t join (
select row_number() OVER(ORDER BY G.a,G.r,G.f1,STG.filler2,G.n_p,G.fe,G.se)  as id
FROM mmp G
LEFT OUTER JOIN mppo S
ON TRIM(G.pc) = S.pc) t1 on 1=1

另一答案

我在Hive中遇到了类似的问题。我希望分享我的经验，以防有人遇到类似的问题

select语句有以下分区子句

COUNT () OVER (PARTITION BY mc.source_well_key, mc.report_dt order by mc.report_dt ) AS r_number

正确的语法是

replace count() with count(1) or count(*)