Question

使用Big Query，我正在尝试查询

count（distinct col ），count（distinct col ）over（partition by col2 ）
< / p>

但是我收到了一个意想不到的错误。

以下是我试图执行的查询模板：

SELECT
          country,
          partner,
          segment_id,
          COUNT(DISTINCT pv_id) pvs,
          COUNT(DISTINCT pv_id) over(PARTITION BY country) country_total_pvs
FROM (...)
GROUP BY
          country,
          partner,
          segment_id

我不断得到的错误：

错误：表达式＆＃39; pv_id＆＃39;在GROUP BY列表中不存在

没有第5列（分析计数），查询执行时没有任何错误。

思想？

非常感谢！

Answer 1

以下情况应该有效，但我猜不到你想做的事情：

SELECT country, partner, segment_id,
       COUNT(DISTINCT pv_id) pvs,
       SUM(COUNT(DISTINCT pv_id)) OVER (PARTITION BY country) as country_total_pvs
FROM (...) q
GROUP BY country, partner, segment_id;

相反：

SELECT country, partner, segment_id,
       COUNT(DISTINCT pv_id) pvs,
       country_total_pvs
FROM (SELECT q.*,
             COUNT(DISTINCT pv_id) OVER (PARTITION BY country) as country_total_pvs
      FROM (...) q
     ) q
GROUP BY country, partner, segment_id, country_total_pvs;

Answer 2

尝试以下

SELECT
  a.country AS country,
  partner,
  segment_id,
  COUNT(DISTINCT pv_id) pvs,
  country_total_pvs
FROM youTable AS a
LEFT JOIN (
  SELECT country, COUNT(DISTINCT pv_id) country_total_pvs
  FROM youTable GROUP BY country
) AS b
ON a.country = b.country
GROUP BY
  country,
  partner,
  segment_id,
  country_total_pvs

请记住 - 以上将提供＆＃34;正确＆＃34; BigQuery StandardSQL的不同计数，但在BigQuery Legacy SQL中COUNT(DISTINCT)是一种统计近似值，并不保证是精确的。您可以使用EXACT_COUNT_DISTINCT代替

以下是略微优化的版本

SELECT
  a.country AS country,
  partner,
  segment_id,
  pvs,
  country_total_pvs
FROM (
  SELECT
    country,
    partner,
    segment_id,
    COUNT(DISTINCT pv_id) pvs
  FROM youTable 
  GROUP BY country, partner, segment_id
) AS a
LEFT JOIN (
  SELECT country,
    COUNT(DISTINCT pv_id) country_total_pvs
  FROM youTable GROUP BY country
) AS b
ON a.country = b.country

在同一个Big Query查询中查询聚合和分析聚合

2 个答案: