三个表的 BigQuery 连接答案

【问题标题】：BigQuery join of three tables三个表的 BigQuery 连接
【发布时间】：2016-09-20 15:17:00
【问题描述】：

我正在尝试在 BigQuery 中加入三个表；表 1 有一个事件的记录（即每一行是一个记录），表 2 有第二个事件的记录，表 3 有类别名称。

我想按类别和设备平台生成一个包含表 1 和表 2 计数的最终表。但是，每次我运行它时，我都会收到一条错误消息，提示 joined.t3.category 不是连接中任一表的字段。

这是我当前的代码：

Select count(distinct joined.t1.Id) as t1_events, count(distinct t2.Id) as t2_events, joined.t1.Origin as platform, joined.t3.category as category

from 

(

SELECT 
        Id,
        Origin,
        CatId

    FROM [testing.table_1] as t1

JOIN (SELECT category,
            CategoryID

FROM [testing.table_3]) as t3

on t1.CatId = t3.CategoryID

) AS joined

JOIN (SELECT Id,
            CategoryId

FROM [testing.table_2]) as t2

ON (joined.t1.CatId = t2.CategoryId)    

Group by platform,category;

作为参考，下面是表 1 和表 2 之间的一个更简单的连接，效果很好：

Select count(distinct t1.Id) as t1_event, count(distinct t2.Id) as t2_events, t1.Origin as platform

from testing.table_1 as t1

JOIN testing.table_2 as t2

on t1.CatId = t2.CategoryId

Group by platform;

【问题讨论】：

标签： sql google-bigquery

【解决方案1】：

简单的解决方法是在第一个内部SELECT 中添加category 字段 - 否则最外面的SELECT 看不到它 - 因此出现错误！这就是问题所在！

此外，在 BigQuery Legacy SQL 中，您可以使用 EXACT_COUNT_DISTINCT 否则您会得到统计近似值 - 请参阅 COUNT([DISTINCT]) 中的更多信息

因此，对于 Legacy SQL，您的查询可能如下所示：

SELECT
  EXACT_COUNT_DISTINCT(joined.t1.Id) AS t1_events,
  EXACT_COUNT_DISTINCT(t2.Id) AS t2_events,
  joined.t1.Origin AS platform,
  joined.t3.category AS category
FROM (
  SELECT
    Id, Origin, CatId, category
  FROM [testing.table_1] AS t1
  JOIN (SELECT category, CategoryID FROM [testing.table_3]) AS t3
  ON t1.CatId = t3.CategoryID 
) AS joined
JOIN (SELECT Id, CategoryId FROM [testing.table_2]) AS t2
ON joined.t1.CatId = t2.CategoryId
GROUP BY platform, category

此外，我觉得您可以进一步简化它（假设不会有歧义的字段）

SELECT
  EXACT_COUNT_DISTINCT(joined.t1.Id) AS t1_events,
  EXACT_COUNT_DISTINCT(t2.Id) AS t2_events,
  joined.t1.Origin AS platform,
  joined.t3.category AS category
FROM (
  SELECT
    Id, Origin, CatId, category
  FROM [testing.table_1] AS t1
  JOIN [testing.table_3] AS t3
  ON t1.CatId = t3.CategoryID 
) AS joined
JOIN [testing.table_2] AS t2
ON joined.t1.CatId = t2.CategoryId
GROUP BY platform, category

当然，如果您使用标准 SQL 版本，您将需要进行相同的修复（正如 Elliott 建议的那样：

SELECT
  COUNT(DISTINCT joined.t1.Id) AS t1_events,
  COUNT(DISTINCT t2.Id) AS t2_events,
  joined.t1.Origin AS platform,
  joined.t3.category AS category
FROM (
  SELECT 
    Id, Origin, CatId, category
  FROM `testing.table_1` AS t1
  JOIN `testing.table_3` AS t3
  ON t1.CatId = t3.CategoryID
) AS joined
JOIN `testing.table_2` AS t2
ON joined.t1.CatId = t2.CategoryId
GROUP BY platform, category

【讨论】：

【解决方案2】：

您可以尝试使用standard SQL 代替您的查询吗？它对别名有更好的处理，COUNT(DISTINCT ...) 会给你一个准确的结果，而不是像旧版 SQL 中的近似值。如果有帮助，您应该对查询进行的唯一更改是使用反引号来转义表名而不是括号。例如：

SELECT
  COUNT(DISTINCT joined.t1.Id) as t1_events,
  COUNT(DISTINCT t2.Id) as t2_events,
  joined.t1.Origin as platform,
  joined.t3.category as category
FROM (
  SELECT 
    Id,
    Origin,
    CatId
  FROM `testing.table_1` AS t1
  JOIN (
    SELECT
      category,
      CategoryID
    FROM `testing.table_3`
  ) AS t3
  ON t1.CatId = t3.CategoryID
) AS joined
JOIN (
  SELECT
    Id,
    CategoryId
  FROM `testing.table_2`
) AS t2
ON joined.t1.CatId = t2.CategoryId
GROUP BY platform, category;

【讨论】：

【解决方案3】：

我不知道 google-bigquery，但我的 SQL 知识告诉我，在列名之前有两个别名会导致问题。尝试在joined 之后删除t-aliases，例如使用joined.category 而不是joined.t3.category。

【讨论】：