将 MySQL 查询转换为 BigQuery 查询

Question

我无法将 MySQL 查询转换为 Google Bigquery 查询。这是我的 MySQL 查询

SELECT id
FROM office_details
GROUP BY address
HAVING max(value)
ORDER BY id

此查询在 phpMyAdmin 和我的 php 脚本上完美运行。但是当我将其转换为 bigquery 时

SELECT id
FROM Office_db.office_details
GROUP BY address
HAVING max(value)
ORDER BY id

它说列 id 既不在 group by 也不聚合。

我需要的是唯一address 的ids，其中value 是最大值。例如

+-------------------------+
| id  |  address |  value |
+-------------------------+
| 1   |    a     |   4    |
| 2   |    a     |   3    |
| 3   |    b     |   2    |
| 4   |    b     |   2    |
+-------------------------+

我需要

+----+
| id |
+----+
| 1  |
| 3  |
+----+

Answer 1

#standardSQL
SELECT id FROM (
  SELECT 
    id, address, 
    ROW_NUMBER() OVER(PARTITION BY address ORDER BY value DESC, id) AS flag
  FROM office_details
)
WHERE flag = 1

Answer 2

试试这个：

#standardSQL
SELECT ARRAY_AGG(id ORDER BY value DESC, id LIMIT 1)[OFFSET(0)] AS id
FROM office_details
GROUP BY address;

与使用RANK 的解决方案相比，它更不容易耗尽内存（并且可能更快），因为它不需要在分区内计算排名时缓冲所有行。作为一个工作示例：

#standardSQL
WITH office_details AS (
  SELECT 1 AS id, 'a' AS address, 4 AS value UNION ALL
  SELECT 2, 'a', 3 UNION ALL
  SELECT 3, 'b', 2 UNION ALL
  SELECT 4, 'b', 2
)
SELECT
  address,
  ARRAY_AGG(id ORDER BY value DESC, id LIMIT 1)[OFFSET(0)] AS id
FROM office_details
GROUP BY address
ORDER BY address;

这给出了结果：

address | id
------------
a       | 1
b       | 3

Answer 3

一个有效的查询可能如下所示：

SELECT MIN(x.id) id
  FROM office_details x
  JOIN
     ( SELECT address
            , MAX(value) value
         FROM officer_details
        GROUP 
           BY address
     ) y
    ON y.address = x.address
   AND y.value = x.value
 GROUP 
    BY address
     , value