Sql 分组并计算每个不同值的百分比
带两个表的Postgresql 9.6:Sql 分组并计算每个不同值的百分比,sql,postgresql,Sql,Postgresql,带两个表的Postgresql 9.6: id | name | -------+-------+ 1 | Mars | 2 | Pluto | 3 | Moon | 4 | Venus | id | p_id | action | -------+-------+-----+ 1 | 1 | LANDED | 2 | 1 | UNSEEN |
id | name |
-------+-------+
1 | Mars |
2 | Pluto |
3 | Moon |
4 | Venus |
id | p_id | action |
-------+-------+-----+
1 | 1 | LANDED |
2 | 1 | UNSEEN |
3 | 1 | SEEN |
4 | 1 | SEEN |
5 | 2 | LANDED |
6 | 3 | SEEN |
7 | 3 | SEEN |
8 | 3 | UNSEEN |
9 | 3 | LANDED |
10 | 3 | LANDED |
11 | 3 | LANDED |
我无法找到一个查询来获得显示每个操作百分比的表,例如:
p_id | name | SEEN | UNSEEN | LANDED |
--------+-------+------+--------+--------+
1 | Mars | 10% | 30% | 60% |
2 | Pluto | 0% | 0% | 100% |
3 | Moon | 25% | 35% | 30% |
4 | Venus | 0% | 0% | 0% |
任何帮助都将不胜感激。谢谢,
Peraz您可以使用带有过滤器的
count(*)
计算每个类别的实例数:
select
n.id, name,
count(*) filter (where action = 'SEEN') as seen,
count(*) filter (where action = 'UNSEEN') as unseen,
count(*) filter (where action = 'LANDED') as landed,
count(*)::dec as total
from names n
left join actions a on a.p_id = n.id
group by n.id
order by n.id;
id | name | seen | unseen | landed | total
----+-------+------+--------+--------+-------
1 | Mars | 2 | 1 | 1 | 4
2 | Pluto | 0 | 0 | 1 | 1
3 | Moon | 2 | 1 | 3 | 6
4 | Venus | 0 | 0 | 0 | 1
(4 rows)
维纳斯的Total
不正确(1),因为左连接。事实上,这没关系,因为我们可以在下一步中避免除以0
在派生表(或CTE)中使用上述查询计算百分比:
select
id, name,
round(seen/ total* 100, 0) as seen,
round(unseen/ total* 100, 0) as unseen,
round(landed/ total* 100, 0) as landed
from (
select
n.id, name,
count(*) filter (where action = 'SEEN') as seen,
count(*) filter (where action = 'UNSEEN') as unseen,
count(*) filter (where action = 'LANDED') as landed,
count(*)::dec as total
from names n
left join actions a on a.p_id = n.id
group by n.id
) s
order by id;
id | name | seen | unseen | landed
----+-------+------+--------+--------
1 | Mars | 50 | 25 | 25
2 | Pluto | 0 | 0 | 100
3 | Moon | 33 | 17 | 50
4 | Venus | 0 | 0 | 0
(4 rows)
我将使用
avg()
执行此操作:
这将生成作为比率的值。使用
%
将它们格式化为字符串似乎更适合应用程序层。您可以使用窗口函数计算百分比,并使用交叉表将行移动到列。请核对:
SQL错误:错误:列“n.name”必须出现在GROUP BY子句中或在聚合函数中使用。addedgroupby n.id,name
有效。通常id
是主键,当不需要将其他列放在groupby
子句中时,也请参见@klin:return what-which-which-which-which-which-which-which-which-which-which-which-which-which-which-which-whtable@Ivan布卢斯基:有趣的方法,但是在一个巨大的table@GordonLinoff对我来说是最简单易懂的方法,在一个大数据集上受苦感谢你们所有人的宝贵回复!
select n.id, n.name,
avg( (action = 'SEEN')::int ) as seen,
avg( (action = 'UNSEEN')::int ) as unseen,
avg( (action = 'LANDED')::int ) as landed
from names n left join
actions a
on a.p_id = n.id
group by n.id, n.name;
create table body(id int, name varchar(256));
insert into body(id, name) values ( 1, 'Mars' ), ( 2, 'Pluto' ), (3, 'Moon' ), ( 4, 'Venus');
create table actions(id int, p_id int, action varchar(256));
insert into actions (id, p_id, action) values
( 1, 1, 'LANDED'),
( 2, 1, 'UNSEEN'),
( 3, 1, 'SEEN'),
( 4, 1, 'SEEN'),
( 5, 2, 'LANDED'),
( 6, 3, 'SEEN'),
( 7, 3, 'SEEN'),
( 8, 3, 'UNSEEN'),
( 9, 3, 'LANDED'),
(10, 3, 'LANDED'),
(11, 3, 'LANDED');
SELECT *
FROM crosstab($$
select p_id as id, action, ((times / sum(times) over (partition by p_id)) * 100)::float as percentage
from (
select action, p_id, count(*) as times
from actions
group by p_id, action
)x
order by 1, 2
$$
) as percentage ("id" int, "LANDED" float, "SEEN" float, "UNSEEN" float);