使用 postgres tablefunc crosstab() 计算错误答案

Question

我有一个视图（我们称之为 'fruit'），其中有一列正确答案和相关的多项选择题错误答案，我想计算哪些错误答案被选得最频繁（容易混淆的）。视图看起来像这样：

correct_answer | wrong_response
-------------------------------
apple          |   pear
apple          |   pear
apple          |   banana
banana         |   apple
banana         |   pear
banana         |   pear
banana         |   pear
pear           |   apple

我想要的是一个枢轴 table，它计算相对于正确答案的错误响应，这样列代表正确答案，行代表错误答案的计数。

wrong_response | apple | banana | pear
---------------------------------------
apple          | 0     | 1      | 1
banana         | 1     | 0      | 0
pear           | 2     | 3      | 0

我一直在这里使用此功能，但当时我并没有尝试计算东西。任何帮助将不胜感激！

编辑：对于未来的人们，这两种解决方案都有效！然而，条件聚合更灵活。交叉表解决方案仅在您完全具有查询中的所有可能性时才有效。例如，如果您排除 pear（或添加 kiwi），则交叉表解决方案 returns 会出错。无论您是否排除（或添加当前不存在的）记录，条件聚合 returns 都会产生结果。谢谢大家的帮助。

Answer 1

如果您知道可以使用条件聚合的列：

select wrong_response,
       count(*) filter (where correct_answer = 'apple') as apple,
       count(*) filter (where correct_answer = 'pear') as pear,
       count(*) filter (where correct_answer = 'banana') as banana
from t
group by wrong_response;

Answer 2

假设您已经完成：CREATE EXTENSION tablefunc;

那么通过crosstab()函数实现你想要的就是：

SELECT *
FROM crosstab('SELECT wrong_response,
                      correct_answer,
                      count(*)
               FROM fruit
               GROUP BY wrong_response, correct_answer 
               ORDER BY wrong_response',

              'SELECT correct_answer
               FROM fruit
               GROUP BY correct_answer
               ORDER BY correct_answer')

AS (wrong_answer varchar(20),
    apple bigint,
    banana bigint,
    pear bigint);

上面的代码会给你下面的结果，这就是你想要的：

注意这里的0输出为null，为了得到你想要的，你只需要稍微修改一下select：

SELECT
    wrong_answer,
    coalesce(apple, 0) as apple,
    coalesce(banana, 0) as banana,
    coalesce(pear, 0) as pear
FROM crosstab('SELECT wrong_response,
                      correct_answer,
                      count(*)
               FROM fruit
               GROUP BY wrong_response, correct_answer 
               ORDER BY wrong_response',

              'SELECT correct_answer
               FROM fruit
               GROUP BY correct_answer
               ORDER BY correct_answer')

AS (wrong_answer varchar(20),
    apple bigint,
    banana bigint,
    pear bigint)

上面的内容会让你得到你想要的：

使用 postgres tablefunc crosstab() 计算错误答案

Using postgres tablefunc crosstab() to count incorrect answers

sql

postgresql

crosstab