尝试通过 visitNumber 使用 STRING_AGG 函数获取唯一的来源、媒介、活动

Question

我正在尝试使用 STRING_AGG 函数按 fullVisitorId 和 ORDER BY visitNumber 分区来获得唯一的来源、媒介和活动。 但是，我在使用 STRING_AGG 的来源、媒体和广告系列中获得了很多重复字段。

所以我想到了使用 DISTINCT，但它不适用于 STRING_AGG。

任何建议表示赞赏。

代码：

SELECT
  CAST(concat(SUBSTR(date,0,4), '-', SUBSTR(date,5,2),'-',SUBSTR(date,7,2)) as date ) as date,
  fullVisitorId,
  trafficSource.keyword,
  trafficSource.adContent,

STRING_AGG(channelGrouping), ' > ') OVER (PARTITION BY fullVisitorId ORDER BY visitNumber) AS channelPath,
STRING_AGG(trafficSource.medium), ' > ') OVER (PARTITION BY fullVisitorId ORDER BY visitNumber ) AS medium_Path,
STRING_AGG(trafficSource.source), ' > ') OVER (PARTITION BY fullVisitorId ORDER BY visitNumber) AS source_Path,
STRING_AGG(trafficSource.campaign), ' > ') OVER (PARTITION BY fullVisitorId ORDER BY visitNumber) AS campaign_Path

FROM
  `tfa-big-query.74006564.ga_sessions_*`,UNNEST(hits) as hit
WHERE _table_suffix BETWEEN '20180801' and '201808015'

STRING_AGG列中的结果（channelgrouping、medium和source是重复字段的出现，可以看出如下：

实际结果：

自然搜索>自然搜索>自然搜索

预期输出：

自然搜索

Answer 1

以下是 BigQuery 标准 SQL

#standardSQL
CREATE TEMP FUNCTION DeDup(arr ANY TYPE) AS ((
  SELECT STRING_AGG(str, ' > ') FROM (
    SELECT CONCAT(ANY_VALUE(x), ' (', CAST(COUNT(1) AS STRING), ')') str FROM (
      SELECT x, COUNTIF(flag) OVER(ORDER BY OFFSET) grp FROM (
        SELECT *,x!=IFNULL(LAG(x) OVER(ORDER BY OFFSET), x) flag FROM UNNEST(arr) x WITH OFFSET
      ))
    GROUP BY grp
)));
SELECT 
  DATE,
  fullVisitorId,
  keyword,
  adContent,
  DeDup(channelPath) AS channelPath,
  DeDup(medium_Path) AS medium_Path,
  DeDup(source_Path) AS source_Path,
  DeDup(campaign_Path) AS campaign_Path
FROM (
  SELECT
    CAST(CONCAT(SUBSTR(DATE,0,4), '-', SUBSTR(DATE,5,2),'-',SUBSTR(DATE,7,2)) AS DATE ) AS DATE,
    fullVisitorId,
    trafficSource.keyword,
    trafficSource.adContent,
    ARRAY_AGG(channelGrouping) OVER(PARTITION BY fullVisitorId ORDER BY visitNumber) AS channelPath,
    ARRAY_AGG(trafficSource.medium) OVER(PARTITION BY fullVisitorId ORDER BY visitNumber ) AS medium_Path,
    ARRAY_AGG(trafficSource.source) OVER(PARTITION BY fullVisitorId ORDER BY visitNumber) AS source_Path,
    ARRAY_AGG(trafficSource.campaign) OVER(PARTITION BY fullVisitorId ORDER BY visitNumber) AS campaign_Path
  FROM `tfa-big-query.74006564.ga_sessions_*`,UNNEST(hits) as hit
  WHERE _table_suffix BETWEEN '20180801' and '201808015'
)

上面会返回类似

的东西

自然搜索 (3) > 其他 (2) > 直接 (1) > 自有 (1) > 电子邮件 (2) > (none) (2)

这是您预期结果的一点扩展Organic Search < Other < Direct < Owned < email < (None)

Answer 2

我认为 STRING_AGG 与 DISTINCT 一起工作

从 bigquery 文档中尝试这个 sql

SELECT STRING_AGG(DISTINCT fruit, " & ") AS string_agg
FROM UNNEST(["apple", NULL, "pear", "banana", "pear"]) AS fruit;