BigQuery LEFT JOIN的值增加了一倍

Question

我正在尝试按日期和市场（美国/英国）合并两个数据集-一个销售目标和另一个实际销售。

为此，我使用的是第三张表，该表使用GENERATE_DATE_ARRAY创建要报告的日期的主列表-这样，我就不会出现没有设置目标且没有报告销售的差距。

我发现我要对销售进行两次计数，因此将数据和查询减少到可重复的状态：

#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

得到以下结果：

结果显示，尽管数据为15，但7月6日（第3行）的报告销售数量为30。

当targets数据中有两个针对该日期和市场的行时，就会发生这种情况，但是我不知道该如何编码。

谢谢你的帮助！

Answer 1

下面应该工作。 这个想法是预先汇总销售表和目标表，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