在 bigquery 中聚合列并构建数组

Posted 2023-03-25

【中文标题】在 bigquery 中聚合列并构建数组

【英文标题】：Aggregate columns and build array in bigquery

【发布时间】：2020-07-28 20:07:12

【问题描述】：

从这样的表开始

WITH table AS (
SELECT 1001 as ID, 1 As Color_blue, 0  AS Color_red,  0 AS Color_black UNION ALL
SELECT 1002 as ID, 0 As Color_blue, 0  AS Color_red,  1 AS Color_black UNION ALL
SELECT 1003 as ID, 0 As Color_blue, 1  AS Color_red,  0 AS Color_black UNION ALL
SELECT 1004 as ID, 0 As Color_blue, 0  AS Color_red,  1 AS Color_black )

SELECT *
FROM table

我想聚合所有特征列并构建一个数组，如下所示：

select ID, array<float64>[Color_blue, Color_red, Color_black] features
from table

但是，我想动态地包含所有列名称，而无需在查询中对它们进行硬编码。输出应该保持不变。 假设我不知道列名（只有“ID”），如何创建具有所有功能的浮点数组？

【参考方案1】：

以下是 BigQuery 标准 SQL

#standardSQL
SELECT id, 
  ARRAY(
    SELECT CAST(val AS FLOAT64) 
    FROM UNNEST(SPLIT(TRIM(FORMAT('%t', t), '()'))) val 
    WHERE CAST(id AS STRING) != val
  ) AS features
FROM `project.dataset.table` t   

当应用于您问题的样本数据时 - 结果是

Row id      features     
1   1001    1.0  
            0.0  
            0.0  
2   1002    0.0  
            0.0  
            1.0  
3   1003    0.0  
            1.0  
            0.0  
4   1004    0.0  
            0.0  
            1.0