在结构数组中找到最大数量

Question

I currently have a table that contains an organization name, org_name , and an array of structs called types that contains structs with properties name and count . I am attempting to use bigquery to figure out the "dominant" type by finding the type with the highest count and appending that types name to the organizations row. My code to get the organization name and the array of structs is as follows:

CREATE TEMP FUNCTION GetNamesAndCounts(elements ARRAY<STRING>) AS (
  ARRAY(
    SELECT AS STRUCT elem AS name, COUNT(*) AS count
    FROM UNNEST(elements) AS elem
    GROUP BY elem
    ORDER BY count
  )
);

select org_name, GetNamesAndCounts(types_of_professionals) as types from table

This is a picture of the results from that query. For context, I would like there to be another column dominant type that displays the name of the type with the highest count .

Answer 1

As I can see, you order your structs ascending by column count , so the last element is what you want (if there is no other name with same count ). So you can just GetNamesAndCounts(types_of_professionals)[ordinal(array_length(GetNamesAndCounts(types_of_professionals)))]

or here is full script

select
    org_name,

    array(
        select as struct
            elem as name,
            count(*) as count
        from
            unnest(elements) as elem
        group by
            elem
        order by
            count
    ) as types,

    (   select
            elem
        from
            unnest(elements) as elem
        group by
            elem
        order by
            count(*) desc
        limit
            1
    ) as dominant_type,

    array(
        select
            elem
        from
            (   select
                    elem,
                    count(*) as count,
                    rank() over(partition by
                                    elem
                                order by
                                    count(*) desc) as rank
                from
                    unnest(elements) as elem
                group by
                    elem
            )
        where
            rank = 1
    ) as all_dominant_types
from
    table