使用 GROUP BY 排序的前 N 行？

Question

我有以下 transaction table:

customer_id, category, product_id, score

我按 customer_id 和 category 分组以创建 product_id-score 地图对列表：

SELECT
    s.customer_id,    
    s.category,
    collect_list(s.pair) 
FROM
    (
        SELECT
            customer_id,
            category,
            map(product_id, score) AS pair
        FROM
            transaction
        WHERE
            score > {score_threshold}
    ) s 
GROUP BY
    s.customer_id,
    s.category

现在我想更进一步。对于每个组，我希望只保留顶部的 n 对，按 score（降序）排序。我尝试 OVER (PARTITION BY...ORDER BY) 我是运行问题。

注意：transaction table 被 category

分割

谢谢

Answer 1

试试这个：

SELECT
        s.customer_id,    
        s.category,
        collect_list(s.pair) 
    FROM
        (
            SELECT
                ROW_NUMBER() OVER (PARTITION BY customer_id, category ORDER BY score desc) as RowId 
                customer_id,
                category,
                map(product_id, score) AS pair
            FROM
                transaction
            WHERE
                score > {score_threshold}
        ) s 
        where s.RowId < n
    
    GROUP BY
        s.customer_id,
        s.category

使用 GROUP BY 排序的前 N 行？

Top N sorted rows with GROUP BY?

hive

apache-spark-sql

hiveql

top-n

使用 GROUP BY 排序的前 N ​​行？

Top N sorted rows with GROUP BY?

hive

apache-spark-sql

hiveql

top-n

使用 GROUP BY 排序的前 N 行？