如何 select 并计算 PostgreSQL 中 5 列的对或三元组

Question

我有一个 postgreSQL table，我在其中存储了五个不同的列（比如 col1、col2、col3、col4、col5）数字。每行中存储的数字彼此不同。

我想制作一些 select 来给出存在的对以及它们出现的次数（计算它们在行中出现的次数）。

示例：

col1	col2	col3	col4	col5
1	5	10	20	100
5	20	30	40	100

结果（大致是这样的）：

pair	total
1,5	1
1,10	1
1,20	1
1,100	1
5,10	1
5,20	2
5,100	2
10,20	1
10,100	1
20,100	2
5,30	1
5,40	1
20,30	1
20,40	1

我可以做一些SQL来执行特定列中的代码：

SELECT count (*) as total, col1, col2
FROM numbers
group by col1, col2;

但这不会给我所有的组合，而且我不知道如何获得其余的组合。此外，这在性能方面可能效率低下。

任何帮助将不胜感激。

此致，米格尔.

Answer 1

您可以使用 recursive 查询来生成您期望的输出

Demo

-- You can change "select 3" to "select 2" or any number you want

with recursive col_join as (select 3),
               numbers_row as (
                   select *, row_number() over () as row
                   from numbers
               ),
               cte_r as (
                   select col1 as value, row, 1 as col
                   from numbers_row
                   union all
                   select col2 as value, row, 2 as col
                   from numbers_row
                   union all
                   select col3 as value, row, 3 as col
                   from numbers_row
                   union all
                   select col4 as value, row, 4 as col
                   from numbers_row
                   union all
                   select col5 as value, row, 5 as col
                   from numbers_row
               ),
               cte as (
                   select array_agg(value order by col) as value,
                          row,
                          col
                   from cte_r
                   group by row, col
                   union all
                   select c.value || cr1.value as value,
                          cr1.row,
                          cr1.col
                   from cte c,
                        cte_r cr1
                   where c.row = cr1.row
                     and not c.value @> array [cr1.value]
                     and c.col < cr1.col
                     and array_length(c.value || cr1.value, 1) <= (select * from col_join)
               )
select array_to_string(value, ','), count(*)
from cte
where array_length(value, 1) = (select * from col_join)
group by 1
order by 1

**加入双列的旧方案

Demo

with recursive 
  numbers_row as (
    select *, row_number() over () as row from numbers
  ),
  cte_r as (
    select col1 as value, row, 1 as col from numbers_row
    union all
    select col2 as value, row, 2 as col from numbers_row
    union all
    select col3 as value, row, 3 as col from numbers_row
    union all
    select col4 as value, row, 4 as col from numbers_row
    union all
    select col5 as value, row, 5 as col from numbers_row
  ),
  cte as (
    select 
      value as val1, 
      value as val2, 
      row, 
      col
    from cte_r
    union all
    select 
      c.val1 as val1,
      cr.value as val2,
      cr.row,
      cr.col
    from 
      cte c,
      cte_r cr 
    where c.row = cr.row and c.col = cr.col - 1
  )
select val1 || ',' || val2, count(*)
from cte
where val1 <> val2
group by val1, val2, val1 || ',' || val2
order by val1, val2

如何 select 并计算 PostgreSQL 中 5 列的对或三元组

How to select and count pairs or triplets from 5 columns in PostgreSQL

postgresql

combinations

combinatorics