在 BigQuery 中将列转换为数组

Question

I would like to take a table such as this:

key	col1	col2	col3
"a"	0	2	3
"b"	1	1	3
"c"	5	0	2
"d"	0	0	1

and turn it into this:

key	col
"a"	[0, 2, 3]
"b"	[1, 1, 3]
"c"	[5, 0, 2]
"d"	[0, 0, 1]

Basically, I want to merge columns into an array. In real life I have 60 columns, all containing integers. I'll take a struct too if that's easier. Basically, I don't want to have to deal with 60 column names dynamically.

I can't seem to find an answer for this anywhere, which baffles me - this must be a common use case? It must be late for me...

Answer 1

How about using array ?

select key, array[col1, col2, col3] as cols
from t;

Answer 2

Consider below approach

select key, 
  (
    select array_agg(cast(value as int64))
    from unnest(split(trim(format('%t', (select as struct * except(key) from unnest([t]))), '()'))) value
  ) col
from `project.dataset.table` t  

if applied to sample data in your question - output is