对于存储在Google BigQuery中的Firebase事件，请在一行中选择多个事件参数

Question

我正在尝试对存储在Google BigQuery中的Firebase事件执行一个非常简单的查询，但我无法找到方法来执行此操作。

在Android应用中，我记录了这样的事件：

Bundle params = new Bundle();
params.putInt("productID", productId);
params.putInt(FirebaseAnalytics.Param.VALUE, value);
firebaseAnalytics.logEvent("productEvent", params);

所以，在BigQuery中我有这样的东西：

 ___________________ _______________________ ____________________________ 
| event_dim.name    | event_dim.params.key  | event_dim.params.int_value | 
|___________________|_______________________|____________________________|
| productEvent      | productID             | 25                         | 
|                   |_______________________|____________________________| 
|                   | value                 | 1253                       |
|___________________|_______________________|____________________________| 

当我从这个表中获取数据时，我得到两行：

 ___________________ _______________________ ____________________________
|event_dim.name     | event_dim.params.key  | event_dim.params.int_value |
|___________________|_______________________|____________________________|
| productEvent      | productID             | 25                         |
| productEvent      | value                 | 12353                      |

但我真正需要的是此表中的SELECT子句以获取如下数据：

 ___________________ _____________ _________
|   name            | productID   | value   |
|___________________|_____________|_________|
| productEvent      | 25          | 12353   |

有什么想法或建议吗？

Answer 1

您可以将值转换为这样的列

SELECT 
  event_dim.name as name,
  MAX(IF(event_dim.params.key = "productID", event_dim.params.int_value, NULL)) WITHIN RECORD productID,
  MAX(IF(event_dim.params.key = "value", event_dim.params.int_value, NULL)) WITHIN RECORD value,
FROM [events] 

如果要使用SQL生成此命令，请参阅此解决方案：Pivot Repeated fields in BigQuery

Answer 2

使用standard SQL（取消选中UI中“显示选项”下的“使用旧版SQL”），可以将查询表达为：

SELECT 
  event_dim.name as name,
  (SELECT value.int_value FROM UNNEST(event_dim.params)
   WHERE key = "productID") AS productID,
  (SELECT value.int_value FROM UNNEST(event_dim.params)
   WHERE key = "value") AS value
FROM `dataset.mytable` AS t,
  t.event_dim AS event_dim;

编辑：更新示例，根据下面的评论将int_value作为value的一部分。这是一个独立的示例，演示了该方法：

WITH T AS (
  SELECT ARRAY_AGG(event_dim) AS event_dim
  FROM (
    SELECT STRUCT(
        "foo" AS name,
        ARRAY<STRUCT<key STRING, value STRUCT<int_value INT64, string_value STRING>>>[
          ("productID", (10, NULL)), ("value", (5, NULL))
        ] AS params) AS event_dim
    UNION ALL
    SELECT STRUCT(
        "bar" AS name,
        ARRAY<STRUCT<key STRING, value STRUCT<int_value INT64, string_value STRING>>>[
          ("productID", (13, NULL)), ("value", (42, NULL))
        ] AS params) AS event_dim
  )
)
SELECT 
  event_dim.name as name,
  (SELECT value.int_value FROM UNNEST(event_dim.params)
   WHERE key = "productID") AS productID,
  (SELECT value.int_value FROM UNNEST(event_dim.params)
   WHERE key = "value") AS value
FROM T AS t,
  t.event_dim AS event_dim;