表的每个字段的值计数

Question

在 PostgresSQL 中，我试图获取每个字段和值对的值计数。 为了说明，我有 2 列，但我确实有 100 列。 所以，我需要一个程序化的解决方案。

对于一张桌子，

| id | name  | gender |
-----------------------
| 1  | john  | male   |
| 2  | susan | female |
| 3  | john  | male   |
| 4  | john  | male   |
| 5  | susan | female |
| 6  | john  | male   |
| 7  | susan | female |
| 8  | julie | female |
| 9  | julie | female |

我知道

select name, count(*) as value_count
from table_name
group by name

给了我我想要的一栏：

| name  | value_count |
-----------------------
| john  | 4           |
| susan | 3           |
| julie | 2           |

但是我如何以编程方式遍历所有列并生成它，例如：

| column_name | value_name | value_count |
------------------------------------------
| name        | john       | 4           |
| name        | susan      | 3           |
| name        | julie      | 2           |
| gender      | male       | 4           |
| gender      | female     | 5           |

Answer 1

使用grouping sets ：

select name, gender, count(*) as value_count
from my_table
group by grouping sets ((name), (gender))

 name  | gender | value_count 
-------+--------+-------------
 john  |        |           4
 julie |        |           2
 susan |        |           3
       | female |           5
       | male   |           4
(5 rows)

如果您确定要以上述格式获得结果：

select 
    case when gender is null then 'name' else 'gender' end as column_name,
    coalesce(gender, name) as value_name,
    value_count
from (
    select name, gender, count(*) as value_count
    from my_table
    group by grouping sets ((name), (gender))
    ) s

 column_name | value_name | value_count 
-------------+------------+-------------
 name        | john       |           4
 name        | julie      |           2
 name        | susan      |           3
 gender      | female     |           5
 gender      | male       |           4
(5 rows)

Answer 2

你可以在这些上做一个联合

select 'name' as column_name, name as value_name, count(distinct id) as value_count
from table_name
group by name
union
select 'gender' as column_name, gender as value_name, count(distinct id) as value_count
from table_name
group by gender

Answer 3

如果值兼容（即所有字符串），您只能做您想做的事情，所以让我假设是这种情况。

然后，您可以使用横向连接：

select col, val, count(*)
from t cross join lateral
     (values ('name', name::text),
             ('gender', gender::text),
             . . .
     ) v(col, val)
group by col, val
order by col, count(*) desc;

您需要填写values()语句。 实际上，您可以使用INFORMATION_SCHEMA.COLUMNS上的select语句生成代码。