使用 BigQuery SQL 计算同一 ID 的所有列值的模式

Question

Let's say I have a Bigquery table with columns id , species , genre , and level . There are cases that for the same id , species , and genre my table could have different level values in multiple rows.

Finally, I want to have 1 row per id with level value as mode of all the level values present in the original table for that id .

Example

#standardSQL
with `project.dataset.table` as (
  select '123' id, 'dog' species, 'suspense' genre, 3 level  union all 
  select '124', 'cat', 'love', 4 union all 
  select '123', 'dog', 'suspense', 5 union all
  select '123', 'dog', 'suspense', 5 
)
select *
from `project.dataset.table`

Expected Outcome: Same dataset with one row for each id. For eg. in the above example, for id 123 level will be 5 (which occurred the most number of times)

How could I achieve this?

[Update] The above data is just an example. I have 20 million rows in my actual dataset with more than 4 columns.

Answer 1

Try this:

with `project.dataset.table` as (
  select '123' id, 'dog' species, 'suspense' genre, 3 level  union all 
  select '124', 'cat', 'love', 4 union all 
  select '123', 'dog', 'suspense', 5 union all
  select '123', 'dog', 'suspense', 5 
)
select id, array_agg(level order by cnt desc limit 1)[offset(0)] as mode
from (
  select id, level, count(level) as cnt
  from `project.dataset.table`
  group by id, level
)
group by id