如何将旧版 SQL BigQuery 转换为标准 SQL？

Posted 2023-03-24

【中文标题】如何将旧版 SQL BigQuery 转换为标准 SQL？

【英文标题】：How to convert legacy SQL BigQuery to standard SQL?

【发布时间】：2018-09-27 12:48:31

【问题描述】：

我一直在尝试将旧版 SQL BigQuery 代码转换为标准 SQL，但我不断收到大量错误。

这是原始的旧版 SQL：

    SELECT t.page_path,
        t.second_page_path,
        t.third_page_path,
        t.fourth_page_path,
        CONCAT(t.page_path,IF(t.second_page_path IS NULL,"","-"),
        IFNULL(t.second_page_path,""),IF(t.third_page_path IS NULL,"","-"),
        IFNULL(t.third_page_path,""),IF(t.fourth_page_path IS NULL,"","-"),
        IFNULL(t.fourth_page_path,"")) AS full_page_journey,
        count(sessionId) AS total_sessions

FROM (

SELECT
     CONCAT(fullVisitorId,"-",STRING(visitStartTime)) AS sessionId,
     hits.hitNumber,
     hits.page.pagePath AS page_path,
     LEAD(hits.page.pagePath) OVER (PARTITION BY fullVisitorId, visitStartTime ORDER BY hits.hitNumber) AS second_page_path,
     LEAD(hits.page.pagePath,2) OVER (PARTITION BY fullVisitorId, visitStartTime ORDER BY hits.hitNumber) AS third_page_path,
     LEAD(hits.page.pagePath,3) OVER (PARTITION BY fullVisitorId, visitStartTime ORDER BY hits.hitNumber) AS fourth_page_path
   FROM
    TABLE_DATE_RANGE( [xxxxxxx:xxxxxxx.ga_sessions_],
TIMESTAMP('2017-01-01'), TIMESTAMP('2017-01-02') )
   WHERE
     hits.type="PAGE"

     ) t
     WHERE t.hits.hitNumber=1
     GROUP BY t.page_path,
              t.second_page_path,
              t.third_page_path,
              t.fourth_page_path,
              full_page_journey
     ORDER BY total_sessions DESC

已更新（已编辑）：这是迄今为止我能够做到的：

    SELECT t.page_path,
        t.second_page_path,
        t.third_page_path,
        t.fourth_page_path,
        CONCAT(t.page_path,IF(t.second_page_path IS NULL,"","-"),
        IFNULL(t.second_page_path,""),IF(t.third_page_path IS NULL,"","-"),
        IFNULL(t.third_page_path,""),IF(t.fourth_page_path IS NULL,"","-"),
        IFNULL(t.fourth_page_path,"")) AS full_page_journey,
        count(sessionId) AS total_sessions

FROM (

SELECT
     CONCAT(fullVisitorId,"-",cast(visitStartTime as string)) AS sessionId,
     hits.hitNumber,
     hits.page.pagePath AS page_path,
     LEAD(hits.page.pagePath) OVER (PARTITION BY fullVisitorId, visitStartTime ORDER BY hits.hitNumber) AS second_page_path,
     LEAD(hits.page.pagePath,2) OVER (PARTITION BY fullVisitorId, visitStartTime ORDER BY hits.hitNumber) AS third_page_path,
     LEAD(hits.page.pagePath,3) OVER (PARTITION BY fullVisitorId, visitStartTime ORDER BY hits.hitNumber) AS fourth_page_path
   FROM
       `xxxxxxxxxxx.xxxxxxx.ga_sessions_*`,
        UNNEST(hits) AS hits
     WHERE
          _TABLE_SUFFIX BETWEEN 
          FORMAT_DATE('%Y%m%d', DATE_ADD(CURRENT_DATE(), INTERVAL -16 DAY))AND 
          FORMAT_DATE('%Y%m%d', DATE_ADD(CURRENT_DATE(), INTERVAL -1 DAY))AND
          hits.type = 'PAGE' ) AS t
          WHERE t.hits.hitNumber = 1
     GROUP BY t.page_path,
              t.second_page_path,
              t.third_page_path,
              t.fourth_page_path,
              full_page_journey
     ORDER BY total_sessions DESC

如果有人能帮助找出语法错误，那就太好了。

得到的一些错误包括：

无法访问类型为 ARRAY 的值上的字段 hitNumber

我读到的“_TABLE_SUFFIX”问题与通配符有关。

【参考方案1】：

作为起点，DATE_ADD 需要一个日期，但您要给它一个时间戳，而 _TABLE_SUFFIX 需要一个字符串，但您要给它一个日期（有点）。

尝试围绕现有语法使用 CURRENT_DATE() 和 FORMAT_DATE：

 FORMAT_DATE('%Y%m%d', DATE_ADD(CURRENT_DATE(), INTERVAL -16 DAY))

这个问题可能对 hitNumber 错误有用：

query-hits-and-custom-dimensions-in-the-bigquery

尝试使用 CTE 而不是子查询，因为它使事情更清晰，更易于调试。

WITH CTE AS 
(SELECT
  CONCAT(fullVisitorId,"-",cast(visitStartTime as string)) AS sessionId,
  hits.hitNumber as hitNumber,
  hits.page.pagePath AS page_path,
  LEAD(hits.page.pagePath) OVER (PARTITION BY fullVisitorId, visitStartTime     
ORDER BY hits.hitNumber) AS second_page_path, 
LEAD(hits.page.pagePath,2) OVER (PARTITION BY fullVisitorId, visitStartTime
ORDER BY hits.hitNumber) AS third_page_path,
  LEAD(hits.page.pagePath,3) OVER (PARTITION BY fullVisitorId,
  visitStartTime ORDER BY hits.hitNumber) AS fourth_page_path
FROM
 `xxxxxxxxxxx.xxxxxxx.ga_sessions_*`,
  UNNEST(hits) AS hits
WHERE
  _TABLE_SUFFIX BETWEEN 
  FORMAT_DATE('%Y%m%d', DATE_ADD(CURRENT_DATE(), INTERVAL -16 DAY))AND 
  FORMAT_DATE('%Y%m%d', DATE_ADD(CURRENT_DATE(), INTERVAL -1 DAY))AND
  hits.type = 'PAGE' )

SELECT page_path,
  second_page_path,
  third_page_path,
  fourth_page_path,
  CONCAT(page_path,IF(second_page_path IS NULL,"","-"),
  IFNULL(second_page_path,""),IF(third_page_path IS NULL,"","-"),
  IFNULL(third_page_path,""),IF(fourth_page_path IS NULL,"","-"),
  IFNULL(fourth_page_path,"")) AS full_page_journey,
  count(sessionId) AS total_sessions
FROM CTE
WHERE hitNumber = 1
GROUP BY page_path,
    second_page_path,
    third_page_path,
    fourth_page_path,
    full_page_journey
ORDER BY total_sessions DESC