如何将 SQL PIVOT 用于多个聚合？

Question

我有一个如下所示的源表：

╔════════════════════╦════════════════╦══════════════╦══════════════════╗
║       Topic        ║     Person     ║ PersonsReply ║  PersonsComment  ║
╠════════════════════╬════════════════╬══════════════╬══════════════════╣
║ Is the earth flat? ║ This Person    ║ Yes          ║ It's flat.       ║
║ Is the earth flat? ║ That Person    ║ No           ║ It's round       ║
║ Is the earth flat? ║ Another Person ║ Maybe        ║ Don't care.      ║
╚════════════════════╩════════════════╩══════════════╩══════════════════╝

但从我在网上看到的示例来看，我似乎无法将我的数据转换为如下表格表示：

╔════════════════════╦══════════════════╦══════════════════╦═════════════════════╦════════════════════╦════════════════════╦═══════════════════════╗
║       Topic        ║ ThisPersonsReply ║ ThatPersonsReply ║ AnotherPersonsReply ║ ThisPersonsComment ║ ThatPersonsComment ║ AnotherPersonsComment ║
╠════════════════════╬══════════════════╬══════════════════╬═════════════════════╬════════════════════╬════════════════════╬═══════════════════════╣
║ Is the earth flat? ║ Yes              ║ No               ║ Maybe               ║ It's flat          ║ It's round         ║ Don't care            ║
╚════════════════════╩══════════════════╩══════════════════╩═════════════════════╩════════════════════╩════════════════════╩═══════════════════════╝

如何使用 SQL 的 PIVOT 函数来实现我想要实现的目标？这是我的沙盒：http://sqlfiddle.com/#!18/e198d/1

现在我得到：

topic   ThisReply   ThatReply   AnotherReply
Is the earth flat?  (null)  (null)  (null)

Answer 1

你想要条件聚合：

select topic,
       max(case when Person = 'This' then reply end) as ThisPersonsReply,
       max(case when Person = 'That' then reply  end) as ThatPersonsReply,
       max(case when Person = 'Another' then reply  end) as AnotherPersonsReply,
       max(case when Person = 'This' then comment end) as ThisPersonsComment,
       max(case when Person = 'That' then comment end) as ThatPersonsComment,
       max(case when Person = 'Another' then comment end) as AnotherPersonsComment
from ptest pt
group by topic;

这里是DB 小提琴。

Answer 2

如果要使用非动态PIVOT，用CROSS APPLY就成了小事

示例

SELECT  *
FROM (
    SELECT topic
          ,B.*
    FROM ptest
    Cross Apply (values (person+'Reply',reply)
                       ,(person+'Comment',comment)

                )B(Item,Value)
) AS source
PIVOT ( max(value) FOR item IN ([ThisReply], [ThatReply], [AnotherReply],[ThisComment], [ThatComment], [AnotherComment]) ) AS pivotTable

退货

topic               ThisReply   ThatReply   AnotherReply    ThisComment ThatComment AnotherComment
Is the earth flat?  Yes         No          Maybe           Its flat    Its round   Dont care

Answer 3

我建议使用条件聚合来透视您的数据集：

• 这是一个跨数据库解决方案，因此您只需学习一次该技术，然后就可以在任何地方使用它

• 它的性能通常至少与供应商特定的实现一样好

• 另一个优点是您可以处理比使用 pivot 时更复杂的聚合表达式（您的问题不是这种情况，但最终可能会出现）

查询：

select 
    topic,
    max(case when person = 'This person' then personsReply end) ThisPersonsReply,
    max(case when person = 'That person' then personsReply end) ThatPersonsReply,
    max(case when person = 'Another person' then personsReply end) AnotherPersonsReply,
    max(case when person = 'This person' then personsComment end) ThisPersonsComment,
    max(case when person = 'That person' then personsComment end) ThatPersonsComment,
    max(case when person = 'Another person' then personsComment end) AnotherPersonsComment
from ptest 
group by topic

Demo on DB Fiddle：

话题 |这个人回复 |那个人回复 |另一个人回复 |这个人评论 |那个人评论 |另一个人评论 :----------------- | :--------------- | :--------------- | :----------------- | :----------------- | :----------------- | :-------------------- 地球是平的吗？ |是 |没有 |也许 |它的平|它的圆|不在乎