如何为 BigQuery 标准 SQL case 语句创建多个 'THEN' 子句？

Question

我在 BigQuery 上使用标准 SQL 根据现有 table 中的某些条件创建新的 table。我有多个 WHEN 子句来支持这一点（因为我正在检查几个不同的条件）。我现在想要做的是在这些 WHEN 语句中有多个 THEN 子句，因为我打算添加多个列。

具体来说，我想将两个现有文本字段的串联添加为一个字段，然后将三个现有字段的聚合数组添加为一个字段：

CASE WHEN
    # all three match
    one_x1 = two_x1 = three_x1 THEN CONCAT( object1_name, ", ", object2_name, ", ", object3_name ) AND ARRAY_AGG(STRUCT(score_one, score_two, score_three))
    # one and two match
      WHEN one_x1 = two_x1 THEN CONCAT( object1_name, ", ", object2_name ) AND ARRAY_AGG(STRUCT(score_one, score_two))
    # one and three match
      WHEN one_x1 = three_x1 THEN CONCAT( object1_name, ", ", object3_name ) AND ARRAY_AGG(STRUCT(score_one, score_three))
    # two and three match
      WHEN two_x1 = three_x1 THEN CONCAT( object2_name, ", ", object3_name ) AND ARRAY_AGG(STRUCT(score_two, score_three))
   ELSE
    NULL
 END

是'AND ARRAY_AGG(STRUCT(xxxxx))'部分不行，我也试过用逗号分隔THEN子句。

重复相同的 case 语句来单独分隔 THEN 子句是唯一的选择吗？

示例数据： sample_data 示例数据第 1 行的预期结果： here

Answer 1

如果您希望在结果集中生成两个单独的列，则需要创建两个 CASE 语句，例如：

CASE 
    WHEN one_x1 = two_x1 = three_x1 THEN CONCAT( object1_name, ", ", object2_name, ", ", object3_name )
    WHEN one_x1 = two_x1 THEN CONCAT( object1_name, ", ", object2_name )
    WHEN one_x1 = three_x1 THEN CONCAT( object1_name, ", ", object3_name )
    WHEN two_x1 = three_x1 THEN CONCAT( object2_name, ", ", object3_name )
    ELSE NULL
END AS field1,
CASE
    WHEN one_x1 = two_x1 = three_x1 THEN ARRAY_AGG(STRUCT(score_one, score_two, score_three))
    WHEN one_x1 = two_x1 THEN ARRAY_AGG(STRUCT(score_one, score_two))
    WHEN one_x1 = three_x1 THEN ARRAY_AGG(STRUCT(score_one, score_three))
    WHEN two_x1 = three_x1 THEN ARRAY_AGG(STRUCT(score_two, score_three))
    ELSE NULL
 END AS field2

Answer 2

以下适用于 BigQuery 标准 SQL

首先，让我们更正您的初始查询，使其实际产生预期结果

#standardSQL
SELECT id, 
CASE 
    WHEN one_x1 = two_x1 AND one_x1 = three_x1 THEN CONCAT( object1_name, ", ", object2_name, ", ", object3_name )
    WHEN one_x1 = two_x1 THEN CONCAT( object1_name, ", ", object2_name )
    WHEN one_x1 = three_x1 THEN CONCAT( object1_name, ", ", object3_name )
    WHEN two_x1 = three_x1 THEN CONCAT( object2_name, ", ", object3_name )
    ELSE NULL
END AS field1,
CASE 
    WHEN one_x1 = two_x1 AND one_x1 = three_x1 THEN [score_one, score_two, score_three]
    WHEN one_x1 = two_x1 THEN [score_one, score_two]
    WHEN one_x1 = three_x1 THEN [score_one, score_three]
    WHEN two_x1 = three_x1 THEN [score_two, score_three]
    ELSE NULL
END AS field2
FROM `project.dataset.table`

如果应用于您问题中的示例数据 - 结果是

Row id  field1                  field2   
1   1   Dog, Animal             0.82     
                                0.72     
2   2   Horse, Animal, Bird     0.76     
                                0.73     
                                0.9  
3   3   Dog, Animal, Chicken    0.67     
                                0.75     
                                0.65     
4   4   Bird, Chicken           0.87     
                                0.86

接下来，据我了解，您希望避免在您的 CASE 中一次又一次地重复相同的条件 - 为此 - 您可以使用以下技巧

#standardSQL

SELECT id, fields.* FROM (
  SELECT id, 
  CASE 
      WHEN one_x1 = two_x1 AND one_x1 = three_x1 THEN 
        STRUCT(CONCAT( object1_name, ", ", object2_name, ", ", object3_name) AS field1, [score_one, score_two, score_three] AS field2)
      WHEN one_x1 = two_x1 THEN 
        STRUCT(CONCAT( object1_name, ", ", object2_name ) AS field1, [score_one, score_two] AS field2)
      WHEN one_x1 = three_x1 THEN 
        STRUCT(CONCAT( object1_name, ", ", object3_name ) AS field1, [score_one, score_three] AS field2)
      WHEN two_x1 = three_x1 THEN 
        STRUCT(CONCAT( object2_name, ", ", object3_name ) AS field1, [score_two, score_three] AS field2)
      ELSE NULL
  END AS fields
  FROM `project.dataset.table`
)

显然具有相同的输出...

最后，作为您的另一种选择 - 您可以使用以下方法消除所有这些 case/when/then

#standardSQL
SELECT id, 
  (SELECT STRING_AGG(object) FROM UNNEST(objects) object WITH OFFSET
    JOIN UNNEST(pos) OFFSET USING(OFFSET)
  ) field1,
    (SELECT ARRAY_AGG(score) FROM UNNEST(scores) score WITH OFFSET
    JOIN UNNEST(pos) OFFSET USING(OFFSET)
  ) field2
FROM (
  SELECT id, 
    [object1_name, object2_name, object3_name] objects,
    [score_one, score_two, score_three] scores,
    (SELECT ARRAY_AGG(OFFSET) 
      FROM UNNEST([one_x1, two_x1, three_x1]) x WITH OFFSET 
      GROUP BY x HAVING COUNT(1) > 1
    ) pos
  FROM `project.dataset.table`
)

再次使用相同的输出

如何为 BigQuery 标准 SQL case 语句创建多个 'THEN' 子句？

How to create multiple 'THEN' clauses for a BigQuery standard SQL case statement?

sql

case-statement

conditional-statements

google-bigquery

bigquery-standard-sql