大查询：查询不允许我们获取结构的最小值

Question

I recently moved from using apache hive to GCP big query as my SQL database.

I have a set of data as follows,

ID	GROUP	GROUP_TYPE	OFFSET	OFFSET_IDENTIFIER
1	100	A	120	1
2	100	A	140	2
3	500	B	120	1
4	500	B	140	2
5	5000	C	300	1
6	5000	C	250	2
7	10000	D	24	1
8	10000	D	46	2
9	10000	D	99	3
10	10000	D	11	4

In hive I used the following query to get the min of offset by group_type and returned the min offset and corresponding offset identifier,

select 
  ID,
  GROUP,
  GROUP_TYPE,
  min(struct(OFFSET,OFFSET_IDENTIFIER)).col1 as min_offset,
  min(struct(OFFSET,OFFSET_IDENTIFIER)).col2 as offset_identifier from eng.offsets 
group by ID,GROUP,GROUP_TYPE;

The output I get from this query is as follows,

ID	GROUP	GROUP_TYPE	OFFSET	OFFSET_IDENTIFIER
1	100	A	120	1
3	500	B	120	1
6	5000	C	250	2
10	10000	D	11	4

But when I execute the same query in BQ,

select 
  ID,
  GROUP,
  GROUP_TYPE,
  min(struct(OFFSET,OFFSET_IDENTIFIER)).OFFSET as min_offset,
  min(struct(OFFSET,OFFSET_IDENTIFIER)).OFFSET_IDENTIFIER as offset_identifier 
from eng.offsets 
  group by ID,GROUP,GROUP_TYPE;

I get the following error: MIN is not defined for arguments of type STRUCT

May I check if there is any solution to this issue/a way to get the min of struct?

Thank you for your inputs!

Answer 1

Consider query below:

with sample_data as (
  select 1 as id, 100 as groupp, 'A' as group_type, 120 as offsett, 1 as offset_identifier
  union all select 2 as id, 100 as groupp, 'A' as group_type, 140 as offsett, 2 as offset_identifier
  union all select 3 as id, 500 as groupp, 'B' as group_type, 120 as offsett, 1 as offset_identifier
  union all select 4 as id, 500 as groupp, 'B' as group_type, 140 as offsett, 2 as offset_identifier
  union all select 5 as id, 5000 as groupp, 'C' as group_type, 300 as offsett, 1 as offset_identifier
  union all select 6 as id, 5000 as groupp, 'C' as group_type, 250 as offsett, 2 as offset_identifier
  union all select 7 as id, 10000 as groupp, 'D' as group_type, 24 as offsett, 1 as offset_identifier
  union all select 8 as id, 10000 as groupp, 'D' as group_type, 46 as offsett, 2 as offset_identifier
  union all select 9 as id, 10000 as groupp, 'D' as group_type, 99 as offsett, 3 as offset_identifier
  union all select 10 as id, 10000 as groupp, 'D' as group_type, 11 as offsett, 4 as offset_identifier

),
add_min_offset as(
select 
  id,
  groupp,
  group_type,
  offsett,
  min(offsett) over (partition by group_type) as min_offset,
  offset_identifier
from sample_data

)
select * except(offsett) from add_min_offset
where offsett=min_offset

Output: