如何在不使用 unnest 的情况下在 BigQuery 中取消数组的嵌套？

Question

我试图在查询中选择hits.page.searchKeyword，但出现经典的嵌套错误。除了解除嵌套之外，我还有其他选择吗？当我从点击（GA3 - UA）中取消嵌套数据时，所有其他选择都会重复它们的值，并且修复起来很混乱。我只需要将搜索词添加到我的查询中，我已在方案中将其标识为 hist.page.searchKeyword，然后就完成了。但我需要在没有取消嵌套函数的情况下取消嵌套命中数组。请有人帮忙。

SELECT 
  parse_date('%Y%m%d', date) AS date,
  channelGrouping,trafficSource.campaign, 
  hits.page.searchKeyword as searchterm,
  SUM(totals.visits) AS sessions, 
  count(distinct clientId),
  SUM(totals.transactions) AS transactions,
  SUM(totals.totalTransactionRevenue) AS revenue1mil
FROM `vdxl-prod-data-reporting-01.200759185.ga_sessions_*`
WHERE _TABLE_SUFFIX BETWEEN '20230101' AND '20231102'
GROUP BY 1, channelGrouping, trafficSource.campaign, hits.page.searchKeyword
ORDER BY 1 DESC

Cannot access field page on a value with type ARRAY<STRUCT<hitNumber INT64, time INT64, hour INT64, ...>> at [4:8]

Answer 1

UNNEST()

只是将数组转换为关系。它不会自动将该关系与主表连接起来。仅当您执行类似

FROM `vdxl-prod-data-reporting-01.200759185.ga_sessions_*` t cross join t.hits

之类的操作时才会发生这种情况（有时

cross join

会被其别名

,

替换）

hits.page.searchKeyword

是内部搜索，意味着每个会话可以有多个搜索 - 您希望如何将它们聚合到会话范围？

一种方法可能是获取按这些关键字排序的不同列表：

SELECT 
  parse_date('%Y%m%d', date) AS date,
  channelGrouping,
  trafficSource.campaign, 
  (select string_agg(distinct page.searchKeyword order by page.searchKeyword) from unnest(hits)) as searchterm,
  SUM(totals.visits) AS sessions, 
  count(distinct clientId),
  SUM(totals.transactions) AS transactions,
  SUM(totals.totalTransactionRevenue) AS revenue1mil
FROM `vdxl-prod-data-reporting-01.200759185.ga_sessions_*`
WHERE _TABLE_SUFFIX BETWEEN '20230101' AND '20231102'
GROUP BY 1, 2, 3, 4
ORDER BY 1 DESC

如果您想要每行一个关键字，则需要交叉联接 - 当然，它会为该行中的每个命中重复会话范围的行...这就是联接的作用。然后，您需要计算点击范围的指标，并且不能再使用会话聚合/总计：

SELECT 
  parse_date('%Y%m%d', date) AS date,
  channelGrouping,
  trafficSource.campaign, 
  h.page.searchKeyword,
  count(distinct fullvisitorid || visitstarttime) AS sessions, 
  count(distinct clientId) as clients,
  countif(h.ecommerceaction.action_type='6') AS transactions,
  SUM(h.transaction.transactionrevenue /1000000) AS revenue
FROM `vdxl-prod-data-reporting-01.200759185.ga_sessions_*` as t
  CROSS JOIN t.hits as h
WHERE 
  _TABLE_SUFFIX BETWEEN '20230101' AND '20231102'
GROUP BY 
  1, 2, 3, 4
ORDER BY 
  1 DESC, 2, 3, 4