BigQuery LEFT JOIN 是加倍值

Posted 2023-03-24

【中文标题】BigQuery LEFT JOIN 是加倍值

【英文标题】：BigQuery LEFT JOIN is doubling-up values

【发布时间】：2018-07-10 19:29:14

【问题描述】：

我正在尝试合并两个数据集——一个是销售目标，另一个是实际销售额，按天和市场（美国/英国）。

为此，我使用了第三个表，该表使用GENERATE_DATE_ARRAY 创建要报告的日期的主列表 - 这样我就不会在没有设定目标和没有报告销售的情况下出现空白.

我发现我的销售额被计算了两次，因此已将我的数据和查询减少到可重现的状态：

#standardSQL

WITH dates AS (
  SELECT day FROM UNNEST(GENERATE_DATE_ARRAY(DATE '2018-07-05', '2018-07-09', INTERVAL 1 DAY)) AS day
),
targets AS (
  SELECT DATE '2018-07-06' AS day, 'UK' AS Market, NUMERIC '2.4' AS quantity
  UNION ALL SELECT '2018-07-06', "US", 8.4
  UNION ALL SELECT '2018-07-06', "US", 1.2
  UNION ALL SELECT '2018-07-08', "UK", 3.0
  UNION ALL SELECT '2018-07-08', "US", 10.9
),
sales AS (
  SELECT DATE '2018-07-08' AS day, 'UK' AS Market, 4 AS quantity
  UNION ALL SELECT '2018-07-06', 'US', 15
)

SELECT 
  dates.day AS day,
  targets.market AS market,
  SUM(targets.quantity) AS targetQuantity,
  SUM(sales.quantity) AS quantity
FROM dates
LEFT JOIN targets
  ON dates.day = CAST(targets.day AS DATE)
LEFT JOIN sales
  ON dates.day = CAST(sales.day AS DATE) AND targets.market = sales.market
GROUP BY day, market
ORDER BY day, market

这给出了以下结果：

结果显示，7 月 6 日（第 3 行）报告的销售量为 30，尽管数据中为 15。

当targets 数据中有两行该日期和市场时，就会发生这种情况，但我不知道如何为此编码。

感谢您的帮助！

【参考方案1】：

下面应该工作。这个想法是预先聚合销售和目标表以避免重复

#standardSQL
WITH dates AS (
  SELECT day FROM UNNEST(GENERATE_DATE_ARRAY(DATE '2018-07-05', '2018-07-09', INTERVAL 1 DAY)) AS day
), targets AS (
  SELECT DATE '2018-07-06' AS day, 'UK' AS Market, NUMERIC '2.4' AS quantity
  UNION ALL SELECT '2018-07-06', "US", 8.4
  UNION ALL SELECT '2018-07-06', "US", 1.2
  UNION ALL SELECT '2018-07-08', "UK", 3.0
  UNION ALL SELECT '2018-07-08', "US", 10.9
), sales AS (
  SELECT DATE '2018-07-08' AS day, 'UK' AS Market, 4 AS quantity
  UNION ALL SELECT '2018-07-06', 'US', 15
)
SELECT 
  dates.day AS day,
  t.market AS market,
  targetQuantity,
  quantity
FROM dates 
LEFT JOIN (SELECT day, market, SUM(quantity) AS targetQuantity FROM targets GROUP BY day, market) t
  ON dates.day = CAST(t.day AS DATE)
LEFT JOIN (SELECT day, market, SUM(quantity) AS quantity FROM sales GROUP BY day, market) s
  ON dates.day = CAST(s.day AS DATE) AND t.market = s.market
ORDER BY day, market