Question

我有下表：

PersonID    CW_MilesRun    PW_MilesRun    CM_MilesRun    PM_MilesRun
1           15             25             35             45         
2           10             20             30             40         
3           5              10             15             20         
...

我需要将此表拆分为一个垂直表，每个字段都有一个 id（即 CD_MilesRun =1、CW_MilesRun = 2 等），以便我的表看起来像这样：

PersonID    TimeID    Description    C_MilesRun    P_MilesRun
1           1         Week           15            25
1           2         Month          35            45
2           1         Week           10            20
2           2         Month          30            40
3           1         Week           5             10
3           2         Month          15            20

在 postgres 中，我会使用类似于：

SELECT
    PersonID
  , unnest(array[1,2]) AS TimeID
  , unnest(array['Week','Month']) AS "Description"
  , unnest(array["CW_MilesRun","CM_MilesRun"]) C_MilesRun
  , unnest(array["PW_MilesRun","PM_MilesRun"]) P_MilesRun
FROM myTableHere
;

但是，我无法在雪花中使用类似的功能。有什么想法吗？

Answer 1

您可以将 FLATTEN() 与 LATERAL 结合使用以获得您想要的结果，尽管查询完全不同。

with tbl as (select $1 PersonID, $2   CW_MilesRun, $3    PW_MilesRun, $4   CM_MilesRun, $5    PM_MilesRun from values (1, 15, 25, 35, 45),(2, 10, 20, 30, 40),(3, 5, 10, 15, 20))

select
    PersonID,
    t.value[0] TimeID,
    t.value[1] Description,
    iff(t.index=0,CW_MilesRun,CM_MilesRun) C_MilesRun,
    iff(t.index=1,PW_MilesRun,PM_MilesRun) P_MilesRun
from tbl, lateral flatten(parse_json('[[1, "Week"],[2, "Month"]]')) t;

PERSONID    TIMEID  DESCRIPTION C_MILESRUN  P_MILESRUN
1   1   "Week"  15  25
1   2   "Month" 35  45
2   1   "Week"  10  20
2   2   "Month" 30  40
3   1   "Week"  5   10
3   2   "Month" 15  20

附言使用 t.* 查看展平后可用的内容（也许很明显。）

Answer 2

您也可以使用 UNPIVOT 和 NATURAL JOIN。

上面的答案很好......就像考虑其他的做事方式一样......你永远不知道它什么时候可能适合你的需求 - 并且让你接触到一些新的很酷的功能。

with cte as (
select
    1 PersonID,
    15 CW_MilesRun,
    25 PW_MilesRun,
    35 CM_MilesRun,
    45 PM_MilesRun
union
select
    2 PersonID,
    10 CW_MilesRun,
    20 PW_MilesRun,
    30 CM_MilesRun,
    40 PM_MilesRun
union
select
    3 PersonID,
    5 CW_MilesRun,
    10 PW_MilesRun,
    15 CM_MilesRun,
    20 PM_MilesRun
)
select * from
(select
   PersonID,
   CW_MilesRun weekly,
   CM_MilesRun monthly
 from
   cte
) unpivot (C_MilesRun for description in (weekly, monthly)) 
natural join 
(select * from
    (select
       PersonID,
       PW_MilesRun weekly,
       PM_MilesRun monthly
     from
       cte
    ) unpivot (P_MilesRun for description in (weekly, monthly))) f

从雪花中的表中取消嵌套

2 个答案: