在 SQL 中将表转换为矩阵

Question

我有这张最喜欢的食物表（称为“food_table”） - 请注意，没有（汉堡，汉堡）的信息：

    food_1    food_2 number_of_people
     pizza     pizza                3
 chocolate     pizza                3
     tacos     pizza               10
    burger     pizza                2
     pizza chocolate                6
 chocolate chocolate                5
     tacos chocolate                4
    burger chocolate                6
     pizza     tacos                9
 chocolate     tacos               10
     tacos     tacos                5
    burger     tacos                3
     pizza    burger                9
 chocolate    burger                9
     tacos    burger                9

我正在尝试制作一个 4x4 矩阵，显示喜欢每种组合的人的相对百分比：

# step 1: counts (food1 , food2)

               pizza chocolate tacos burger

 pizza         3         6     9      9
 chocolate     3         5    10      9
 tacos        10         4     5      9
 burger      NULL        6     3      3

# step 2 :precents (each row should add to 100)
(food1 , food2)
e.g. (pizza,pizza) = 3/(3+6+9+9), (pizza,chocolate) = 6/(3+6+9+9), (pizza,tacos) = 9/(3+6+9+9), (pizza,burger) = 9/(3+6+9+9)

             pizza chocolate    tacos   burger
pizza     11.11111  22.22222 33.33333 33.33333
chocolate 11.11111  18.51852 37.03704 33.33333
tacos     35.71429  14.28571 17.85714 32.14286
burger    NULL       50       25       25

我尝试过这样的：

with step1 as (
select
 food_1, 
food_2,
 number_of_people * 100/ sum(number_of_people) over (partition by food_1) as percent
from food_table
group by 
food_1, 
food_2),

step2 as(
select food_1 as "food1/food2",
max(case when food_2 = 'pizza' then percent end) as pizza,
max(case when food_2 = 'tacos' then percent end) as tacos,
max(case when food_2 = 'burger' then percent end) as burger,
max(case when food_2 = 'chocolate' then percent end) as chocolate
from step1
group by "food1/food2")

select * from step2;

但我认为这是不正确的 - 代码的结果与我的手算不符。

我该如何解决这个问题？

Answer 1

首先，您显示的表格不正确 - 汉堡和披萨不能为 NULL，因为上面列出了 2。

其次，步骤 1 中的 SQL 使用 OLAP 函数，不需要 GROUP BY。通过.

尝试这个修改后的 SQL

 with step1 as (
select
 food_1, 
food_2,
 number_of_people * 100/ sum(number_of_people) over (partition by food_1) as percent
from food
)
, step2 as(
select food_1 as "food1/food2",
max(case when food_2 = 'pizza' then percent end) as pizza,
max(case when food_2 = 'tacos' then percent end) as tacos,
max(case when food_2 = 'burger' then percent end) as burger,
max(case when food_2 = 'chocolate' then percent end) as chocolate
from step1
group by food_1)

select * from step2;

由于四舍五入，它不会加到 100 - 因此您可能需要将数据类型更改为十进制。