Question

这是我拥有的数据集（约10 TB）的示例

+----+------------+----------+----------------+--------------+
| id | date       | campaign | campaign_start | campaign_end |
+----+------------+----------+----------------+--------------+
| 1  | 2018-01-01 | 1        | 2018-01-01     | 2018-02-03   |
+----+------------+----------+----------------+--------------+
| 1  | 2018-02-01 | 2        | 2018-02-01     | 2018-02-03   |
+----+------------+----------+----------------+--------------+
| 1  | 2018-02-02 | 2        | 2018-02-01     | 2018-02-03   |
+----+------------+----------+----------------+--------------+
| 1  | 2018-02-03 | 2        | 2018-02-01     | 2018-02-03   |
+----+------------+----------+----------------+--------------+
| 2  | 2018-01-23 | 1        | 2018-01-01     | 2018-02-03   |
+----+------------+----------+----------------+--------------+
| 2  | 2018-02-03 | 2        | 2018-02-01     | 2018-02-03   |
+----+------------+----------+----------------+--------------+

我要：

对于每个唯一的ID +广告系列：

获取特定广告系列在一段时间内ID出现的频率
获取在广告系列开始前的可变回溯期内（例如3个月）内ID出现的频率。说“> = campaign_start + 3个月”
在该窗口中获取最早（第一）和最新（最后）日期

我想要的输出是：

+----+----------+--------------------+--------------------------+----------------+--------------+------------+------------+
| id | campaign | campaign_frequency | total_lookback_frequency | campaign_start | campaign_end | first_date | last_date  |
+----+----------+--------------------+--------------------------+----------------+--------------+------------+------------+
| 1  | 1        | 1                  | 1                        | 2018-01-01     | 2018-02-03   | 2018-01-01 | 2018-01-01 |
+----+----------+--------------------+--------------------------+----------------+--------------+------------+------------+
| 1  | 2        | 3                  | 4                        | 2018-02-01     | 2018-02-03   | 2018-01-01 | 2018-02-03 |
+----+----------+--------------------+--------------------------+----------------+--------------+------------+------------+
| 2  | 1        | 1                  | 1                        | 2018-01-01     | 2018-02-03   | 2018-01-23 | 2018-01-23 |
+----+----------+--------------------+--------------------------+----------------+--------------+------------+------------+
| 2  | 2        | 1                  | 2                        | 2018-02-01     | 2018-02-03   | 2018-01-23 | 2018-02-03 |
+----+----------+--------------------+--------------------------+----------------+--------------+------------+------------+

我一直遇到的问题是我无法使total_lookback_frequency正常工作，它总是返回与campaign_frequency（这只是id（广告系列）的count（id）组）相同的结果。

以下是我所拥有的（不起作用）：

SELECT  
  id,
  campaign,
  min(date) as first_date,
  max(date) as end_date,
  count(id) as total_lookback_frequency,
WHERE
  date >= sub(date, INTERVAL 730 hour)
GROUP BY
  id,
  campaign,
  date

您能在这里帮忙吗？

谢谢！

Answer 1

以下是用于BigQuery标准SQL

#standardSQL
SELECT 
  id,
  campaign,
  COUNT(1) campaign_frequency,
  (
    SELECT COUNT(1) 
    FROM `project.dataset.table` 
    WHERE id = t.id
    AND dt BETWEEN  DATE_SUB(t.campaign_start, INTERVAL 3 MONTH) AND DATE_SUB(t.campaign_start, INTERVAL 1 DAY)
  ) total_lookback_frequency,
  campaign_start,
  campaign_end,
  MIN(dt) AS first_date,
  MAX(dt) AS end_date
FROM `project.dataset.table` t
GROUP BY id, campaign, campaign_start, campaign_end

您可以使用下面的问题中的虚拟数据进行测试，操作

#standardSQL
WITH `project.dataset.table` AS (
  SELECT 1 id, DATE '2018-01-01' dt, 1 campaign, DATE '2018-01-01' campaign_start, DATE '2018-02-03' campaign_end UNION ALL
  SELECT 1, '2018-02-01', 2, '2018-02-01', '2018-02-03' UNION ALL
  SELECT 1, '2018-02-02', 2, '2018-02-01', '2018-02-03' UNION ALL
  SELECT 1, '2018-02-03', 2, '2018-02-01', '2018-02-03' UNION ALL
  SELECT 2, '2018-01-23', 1, '2018-01-01', '2018-02-03' UNION ALL
  SELECT 2, '2018-02-03', 2, '2018-02-01', '2018-02-03' 
)
SELECT 
  id,
  campaign,
  COUNT(1) campaign_frequency,
  (
    SELECT COUNT(1) 
    FROM `project.dataset.table` 
    WHERE id = t.id
    AND dt BETWEEN  DATE_SUB(t.campaign_start, INTERVAL 3 MONTH) AND DATE_SUB(t.campaign_start, INTERVAL 1 DAY)
  ) total_lookback_frequency,
  campaign_start,
  campaign_end,
  MIN(dt) AS first_date,
  MAX(dt) AS end_date
FROM `project.dataset.table` t
GROUP BY id, campaign, campaign_start, campaign_end
-- ORDER BY id, campaign

如何在回溯期内返回多个字段唯一的计数？

1 个答案: