添加 JOIN 语句时值增加

Question

这几天我在这个问题上遇到了一些困难，我无法弄清楚。

（运行这是 Bigquery）

对于营销客户，我试图获取每个广告客户的展示次数和点击次数。

p_impressions 包含所有展示数据，包括 IDs Campaign_ID、Advertiser_ID p_click 拥有所有的点击数据，包括 IDs Campaign_ID, Advertiser_ID match_table_campaigns 有广告系列数据：Campaign_ID、广告系列名称

（基本上我的简单目标是获取展示次数和点击次数并将其与 Campaign 表相匹配，这样我就可以看到广告系列的名称而不仅仅是 ID。

当我使用 UNION all 对 Impression 和 Click 表运行查询时，仅使用 IDS，所有数字都与我的平台匹配，一切看起来都很好。

但是，当我插入一个 JOIN（我已经尝试了所有这些）以包含广告系列的名称时，展示次数和点击次数的所有指标似乎都在增加，而且我似乎无法再获得正确的数字了。

with combinedDeliveryData AS (
SELECT
  FORMAT_DATETIME("%F",DATETIME(TIMESTAMP_MILLIS( CAST(SUBSTR(CAST( Event_Time AS String),0,LENGTH(CAST( Event_Time AS String)) - 3) AS INT64 )),"America/Toronto")) AS Date,
  Impression_ID,
  DBM_Advertiser_ID,
  Campaign_ID,
  Ad_ID,
  Rendering_ID,
  Event_Type,
  Advertiser_ID,
FROM
  CampaignManager.p_impression
UNION ALL
SELECT
  FORMAT_DATETIME("%F",DATETIME(TIMESTAMP_MILLIS( CAST(SUBSTR(CAST( Event_Time AS String),0,LENGTH(CAST( Event_Time AS String)) - 3) AS INT64 )),"America/Toronto")) AS Date,
  Impression_ID,
  DBM_Advertiser_ID,
  Campaign_ID,
  Ad_ID,
  Rendering_ID,
  Event_Type,
  Advertiser_ID,
FROM
  CampaignManager.p_click)
  
SELECT cpg.Campaign_ID, cpg.Campaign,data.Date,COUNT(case data.Event_Type when 'VIEW' then 1 else null end) AS Impressions
FROM match_table_campaigns AS cpg
LEFT JOIN combinedDeliveryData AS data ON cpg.Campaign_ID = data.Campaign_ID
WHERE cpg.Advertiser_ID = "4739279"
GROUP BY 3,1,2

任何帮助将不胜感激。我不确定我错过了什么。

请注意，这是没有 JOIN 的情况

SELECT
  Impression_ID,
  Event_Time,
  DBM_Advertiser_ID,
  Campaign_ID,
  Ad_ID,
  Rendering_ID,
  Event_Type,
  Advertiser_ID
FROM
  CampaignManager.p_impression
UNION ALL
SELECT
  Impression_ID,
  Event_Time,
  DBM_Advertiser_ID,
  Campaign_ID,
  Ad_ID,
  Rendering_ID,
  Event_Type,
  Advertiser_ID
FROM
  CampaignManager.p_click)
SELECT FORMAT_DATETIME("%F",DATETIME(TIMESTAMP_MILLIS( CAST(SUBSTR(CAST( clicks.Event_Time AS String),0,LENGTH(CAST( clicks.Event_Time AS String)) - 3) AS INT64 )),"America/Toronto")) AS exposureDate,clicks.Campaign_ID,COUNT(case Event_Type when 'VIEW' then 1 else null end) AS Impressions,COUNT(case Event_Type when 'CLICK' then 1 else null end) AS Clicks
FROM 
combinedDeliveryData AS clicks
WHERE clicks.Campaign_ID = "4739279"
GROUP BY 1,2

新问题： 我添加了以下代码，看起来很棒

  this_is_deduplicated AS (
    SELECT x.*
    FROM (
      SELECT Campaign_ID, ARRAY_AGG(a LIMIT 1)[OFFSET(0)] x
      FROM combinedDeliveryData a
      GROUP BY 1
    ) 
  )

LEFT JOIN this_is_deduplicated AS data ON cpg.Campaign_ID = data.Campaign_ID

得到我所有的价值观，它们是不准确的。我现在遇到了另一个问题，我必须加入决赛桌，但上面的解决方案不起作用。

新表中的值： 日期 广告系列_ID Activity_ID

我现在需要加入最后一个表，这样对于每个日期，我都会获得 Activity_ID 的 COUNT 作为 Conversiions。

我的决赛桌应该是：

广告商、广告系列、展示次数、点击次数、转化次数（如上）

Answer 1

我几乎可以确定原因是combinedDeliveryData 表中有多个相同的campaign_id。

在加入之前确保每个广告系列只有一个 ID。像这样的东西：

WITH [...], 
  this_is_deduplicated AS (
    SELECT x.*
    FROM (
      SELECT Campaign_ID, ARRAY_AGG(a LIMIT 1)[OFFSET(0)] x
      FROM combinedDeliveryData a
      GROUP BY 1
    ) 
  )

而JOIN 反对，如：

LEFT JOIN this_is_deduplicated AS data ON cpg.Campaign_ID = data.Campaign_ID

而不是

LEFT JOIN combinedDeliveryData AS data ON cpg.Campaign_ID = data.Campaign_ID

或更短的

LEFT JOIN this_is_deduplicated AS data USING(Campaign_ID)

Answer 2

您可能需要加入campaign 和campaign_id。提供以下示例查询

select
    t1.date,
    t1.campaign_id,
    t1.campaign,
    sum(case when t1.event = 'view' then 1 else 0 end) as impressions,
    sum(case when t2.event = 'click' then 1 else 0 end) as clicks
from table1 as t1
left join table2 as t2
on t1.campaign = t2.campaign
and on t1.campaign_id = t2.campaign_id

where advertiser_id = '12345'
group by
    t1.campaign_id,
    t1.campaign,
    t1.date