2

我正在使用高效的（ish）BigQuery SQL 查询来解决以下问题：

我有一张看起来像这样的桌子：

    
    Row | Col_A | Col_B |
    ---------------------
     1  |   2   |   3   |
     2  |   1   |   4   |
     3  |   5   |   7   |
     4  |   2   |   3   |
     5  |   6   |   1   |

    ...and so on (>million rows)

每列的值是一个范围为 [1..7] 的 ID。

查询应产生以下内容，即对每一列的每个代码求和：

    
    Code | Total Col_A | Total Col_B
    --------------------------------
      1  |      1      |      0  
      2  |      2      |      0  
      3  |      0      |      2  
      4  |      0      |      1  
      5  |      1      |      0  
      6  |      1      |      0  
      7  |      0      |      1

任何人都知道在 BigQuery 中不使用多个 SELECT 的方法吗？

干杯。

4

2 回答 2

2

你能用你的样本数据创建一个公共数据集吗？这将使编写适用于您的数据的查询和验证结果变得更加容易。

一个起始查询：

SELECT Code, COUNT(Col_A) count_column_x, COUNT(Col_B) count_column_y
FROM [your:list.of_codes] a
LEFT JOIN EACH [your:sample.table] b
ON a.Code=b.Col_A
GROUP BY 1

（这并不完美，如果您共享一张桌子，会更进一步）

于 2013-10-14T17:46:26.807 回答
1

任何人都知道在 BigQuery 中不使用多个 SELECT 的方法吗？

一个带有标准 SQL 的 SELECT

#standardSQL
WITH logs AS (
  SELECT 2 AS Col_A, 3 AS Col_B UNION ALL
  SELECT 1 AS Col_A, 4 AS Col_B UNION ALL
  SELECT 5 AS Col_A, 7 AS Col_B UNION ALL
  SELECT 2 AS Col_A, 3 AS Col_B UNION ALL
  SELECT 6 AS Col_A, 1 AS Col_B   
)
SELECT 
  id, 
  SUM(CAST(id = Col_A AS INT64)) AS Total_Col_A, 
  SUM(CAST(id = Col_B AS INT64)) AS Total_Col_B
FROM logs, UNNEST(GENERATE_ARRAY(1,7)) AS id
GROUP BY id
ORDER BY id

或与COUNTIF() 

#standardSQL
WITH logs AS (
  SELECT 2 AS Col_A, 3 AS Col_B UNION ALL
  SELECT 1 AS Col_A, 4 AS Col_B UNION ALL
  SELECT 5 AS Col_A, 7 AS Col_B UNION ALL
  SELECT 2 AS Col_A, 3 AS Col_B UNION ALL
  SELECT 6 AS Col_A, 1 AS Col_B   
)
SELECT 
  id, 
  COUNTIF(id = Col_A) AS Total_Col_A, 
  COUNTIF(id = Col_B) AS Total_Col_B
FROM logs, UNNEST(GENERATE_ARRAY(1,7)) AS id
GROUP BY id
ORDER BY id
于 2017-01-13T22:16:33.637 回答