Question

你能帮我解决一下Bigquery的标准语法问题吗？我试图理解为什么（以及如何修复它）这个standardql查询返回2行，而这个legacysql只返回1（我希望只有一行）。

StandardSQL

  SELECT 
  hits2.transaction.transactionId as transactionId
  FROM `ga-export-TTTT.1234567890.ga_sessions_*`
  ,UNNEST (hits) as hits2

WHERE 
hits2.transaction.transactionId = '03971163'

LegacySQL

select
hits.transaction.transactionId
FROM
 TABLE_DATE_RANGE([ga-export-TTTT:1234567890.ga_sessions_], TIMESTAMP('2016-09-01'), TIMESTAMP('2017-02-14'))
WHERE 
hits.transaction.transactionId = '03971163'

在阅读帮助后，我还尝试了 StandardSQL ，其结果中包含相同的两行：

select 
title
from
(
  SELECT
    ARRAY(SELECT transaction.transactionId FROM UNNEST(hits)
          WHERE transaction.transactionId = '03971163') AS title
  FROM `ga-export-TTTTT.1234567890.ga_sessions_*`
  )
WHERE ARRAY_LENGTH(title) > 0;

非常感谢您的帮助。

Answer 1

请改为尝试：

#standardSQL
SELECT 
  ARRAY(SELECT transaction.transactionId
        FROM UNNEST(hits)
        WHERE transaction.transactionId = '03971163')
    AS transactionIds
FROM `ga-export-TTTT.1234567890.ga_sessions_*`;

对数组使用CROSS JOIN时，每个数组元素的结果都会得到一行。如果您希望与表中的行一一对应，则可以使用数组子查询（如上面的查询中所示）在应用过滤器后“重新打包”数组元素。另外一个例子，您可以使用ARRAY子查询以及SELECT AS STRUCT：

#standardSQL
SELECT 
  ARRAY(SELECT AS STRUCT transaction.*
        FROM UNNEST(hits)
        WHERE transaction.transactionId = '03971163')
    AS transactions
FROM `ga-export-TTTT.1234567890.ga_sessions_*`;

对于匹配transaction条件的匹配，这将返回transaction.transactionId = '03971163'内所有字段的数组。如果您只想要数组的单个元素，则可以在选择列表中使用带有LIMIT的子查询：

#standardSQL
SELECT 
 (SELECT transaction.transactionId
  FROM UNNEST(hits)
  WHERE transaction.transactionId = '03971163'
  LIMIT 1)
    AS transactionId
FROM `ga-export-TTTT.1234567890.ga_sessions_*`;

或者：

#standardSQL
SELECT 
 (SELECT transaction
  FROM UNNEST(hits)
  WHERE transaction.transactionId = '03971163'
  LIMIT 1)
    AS transaction
FROM `ga-export-TTTT.1234567890.ga_sessions_*`;

Answer 2

我看到的两个查询（对于Legacy和Standard SQL版本）的唯一区别是你可能会查询不同的表集！

在旧版SQL版本中 - 您将时间段的表格列表从TIMESTAMP（＆＃39; 2016-09-01＆＃39;）到TIMESTAMP（＆＃39; 2017-02-14＆＃39;），而在标准中SQL查询ga-export-TTTT.1234567890数据集

中的所有ga_sessions_表

请尝试使用以下内容将表格过滤到与旧版SQL中相同的列表

#standardSQL
SELECT hits2.transaction.transactionId as transactionId
FROM `ga-export-TTTT.1234567890.ga_sessions_*`, UNNEST (hits) as hits2
WHERE hits2.transaction.transactionId = '03971163'
AND _TABLE_SUFFIX BETWEEN '20160901' AND '20170214'

为什么Bigquery的standardsql返回多行而legacysql只返回一行？

2 个答案: