需要在 Impala 中创建一个 Un-Pivots、Pivo​​ts 和 Union 的视图

Posted 2023-04-12

【中文标题】需要在 Impala 中创建一个 Un-Pivots、Pivo​​ts 和 Union 的视图

【英文标题】：Need to create a view in Impala which Un-Pivots, Pivots and Union's it altogether

【发布时间】：2020-11-24 12:02:54

【问题描述】：

我第一次尝试在 SQL 中做一些适当的事情，除了复制一些我可以通过 Alteryx 轻松完成的事情时遇到问题。

本质上，需要一些基本的数据整理来创建我已经在 Impala/Hive 中创建的表的摘要。需要将基本表分解为较小的表（未透视表和已透视表），然后将它们联合在一起以创建聚合表。

表格如下：

Run_Code | ID | ColB | ColC | ColD | ColE | ColF | ColG | TaxExpense | RetainedExpense | IncomeExpense | Year
-----------------------------------------------------------------------------------------------------------------
run1     | 21 | 1234 | 1234 | 1234 | 1234 | 1234 | 1234 | 1234.56789 | 1234.56789      |  1234.56789   | Year1 
run1     | 22 | 1234 | 1234 | 1234 | 1234 | 1234 | 1234 | 1234.56789 | 1234.56789      |  1234.56789   | Year2
run1     | 23 | 1234 | 1234 | 1234 | 1234 | 1234 | 1234 | 1234.56789 | 1234.56789      |  1234.56789   | Year3
run1     | 24 | 1234 | 1234 | 1234 | 1234 | 1234 | 1234 | 1234.56789 | 1234.56789      |  1234.56789   | Year4

目前在 Alteryx 中完成了以下工作；即只选择了 TaxExpense，然后将其取消透视，然后作为一列返回。

Run_Code | ID | ColB | ColC | ColD |     Name    | Year1 | Year2 | Year3 | Year4
-----------------------------------------------------------------------------------------------------------------
run1     | 21 | 1234 | 1234 | 1234 | Tax Expense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | Tax Expense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | Tax Expense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | Tax Expense | (sum) | (sum) | (sum) | (sum)

保留费用和收入费用也是如此。

Run_Code | ID | ColB | ColC | ColD |      Name      | Year1 | Year2 | Year3 | Year4
-----------------------------------------------------------------------------------------------------------------
run1     | 21 | 1234 | 1234 | 1234 | RetainedExpense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | RetainedExpense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | RetainedExpense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | RetainedExpense | (sum) | (sum) | (sum) | (sum)

最终结果如下，期望的结果如下：

Run_Code | ID | ColB | ColC | ColD |      Name       | Year1 | Year2 | Year3 | Year4
-----------------------------------------------------------------------------------------------------------------
run1     | 21 | 1234 | 1234 | 1234 | TaxExpense      | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | TaxExpense      | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | TaxExpense      | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | TaxExpense      | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | RetainedExpense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | RetainedExpense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | RetainedExpense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | RetainedExpense | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | IncomeExpense   | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | IncomeExpense   | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | IncomeExpense   | (sum) | (sum) | (sum) | (sum)
run1     | 21 | 1234 | 1234 | 1234 | IncomeExpense   | (sum) | (sum) | (sum) | (sum)

感谢任何帮助创建解决上述问题的 SQL。

【参考方案1】：

嗯。 . .如果我理解正确，您可以取消透视并重新聚合：

select Run_Code, ID, ColB, ColC, ColD,
       sum(case when year = 'year1' then expense end) as year_1,
       sum(case when year = 'year2' then expense end) as year_2,
       sum(case when year = 'year3' then expense end) as year_3,
       sum(case when year = 'year4' then expense end) as year_4
from ((select Run_Code, ID, ColB, ColC, ColD, 'TaxExpense' as name, TaxExpense as expense, year
       from t
      ) union all
      (select Run_Code, ID, ColB, ColC, ColD, 'RetainedExpense' as name, RetainedExpense, year
       from t
      ) union all
      (select Run_Code, ID, ColB, ColC, ColD, 'IncomeExpense' as name, IncomeExpense, year
       from t
      )
     ) t
group by Run_Code, ID, ColB, ColC, ColD, name

【讨论】：

我收到以下错误：AnalysisException：SUM 需要一个数字参数：sum(CASE WHEN year = 'YEAR1' THEN Name END)。当我尝试使用 MAX 而不是 Sum 时，名称字段不会出现，而是填充在 Year_1、Year_2 等年份下。