Question

我有这个问题。

select transaction, bbk, sbk, obk, ibk
from
(
  select 
    transaction, 
    case when "'BBK'" = 1 then country end bbk,
    case when "'SBK'" = 1 then country end sbk,
    case when "'OBK'" = 1 then country end obk,
    case when "'IBK'" = 1 then country end ibk  
  from (
    select 
    regexp_substr("col_a", '[^~]+', 1, 1) as transaction,
    regexp_substr("col_a", '[^~]+', 1, 2) as code,
    regexp_substr("col_a", '[^~]+', 1, 3) as country
    from Table1 t)
  pivot 
  ( 
    --case when count(code) = 1 then 
    count(code)
    for code in ('BBK','SBK','OBK','IBK')
  )  
)
group by transaction, bbk, sbk, obk, ibk
order by transaction

结果，如this fiddle

中所示

00004719    US  US  (null)  (null)
00004719    (null)  (null)  GB  (null)
00004719    (null)  (null)  (null)  DE

每笔交易显示多行。我想改变查询，以便每个事务只发生一行。

基本上垂直合并其他记录中的空值以实现：

00004719    US  US  GB  DE

怎么可能这样做？

Answer 1

您可以从查询结果开始，再次在事务列上分组：

with x as (
select transaction, bbk, sbk, obk, ibk
from
(
      select 
        transaction, 
        case when "'BBK'" = 1 then country end bbk,
        case when "'SBK'" = 1 then country end sbk,
        case when "'OBK'" = 1 then country end obk,
        case when "'IBK'" = 1 then country end ibk  
      from (
        select 
        regexp_substr("col_a", '[^~]+', 1, 1) as transaction,
        regexp_substr("col_a", '[^~]+', 1, 2) as code,
        regexp_substr("col_a", '[^~]+', 1, 3) as country
        from Table1 t)
      pivot 
      ( 
        --case when count(code) = 1 then 
        count(code)
        for code in ('BBK','SBK','OBK','IBK')
      )  
  )
group by transaction, bbk, sbk, obk, ibk
) select
transaction, max(bbk), max(sbk), max(obk), max(ibk)
from x
group by transaction
order by transaction;

Example SQLFiddle

编辑：

或等效地，只是在原始查询中按事务分组：

select 
  transaction, 
  max(case when "'BBK'" = 1 then country end) bbk,
  max(case when "'SBK'" = 1 then country end) sbk,
  max(case when "'OBK'" = 1 then country end) obk,
  max(case when "'IBK'" = 1 then country end) ibk  
from (
  select 
  regexp_substr("col_a", '[^~]+', 1, 1) as transaction,
  regexp_substr("col_a", '[^~]+', 1, 2) as code,
  regexp_substr("col_a", '[^~]+', 1, 3) as country
  from Table1 t)
pivot ( 
  count(code)
  for code in ('BBK','SBK','OBK','IBK')
)  
group by 
  transaction
order by 
  transaction;

Example SQLFiddle

Answer 2

这正是pivot的用途：

select 
  transaction, 
  "'BBK'",
  "'SBK'",
  "'OBK'",
  "'IBK'"
from (
  select 
  regexp_substr("col_a", '[^~]+', 1, 1) as transaction,
  regexp_substr("col_a", '[^~]+', 1, 2) as code,
  regexp_substr("col_a", '[^~]+', 1, 3) as country
  from Table1 t)
pivot 
( 
  MAX(country) for code in ('BBK','SBK','OBK','IBK')
);

分组创建垂直合并

2 个答案: