问题
I feel like I'm missing something obvious. I am trying to test out the distribution of random()
. Here is the table:
create table test (
id int,
random_float float,
random_int int
);
Here is what I want to do:
truncate table test;
insert into test (id)
values (generate_series(1,1000));
update test
set
random_float = random() * 10 + 1;
update test
set
random_int = trunc(random_float);
select
random_int,
count(random_int) as Count,
cast( count(random_int) / max(id) as float) as Percent
from test
group by random_int
order by random_int;
However, the "Percent" column returns zero for every record. I tried casting it as float, as decimal, I tried changing the random_int
column to decimal instead of integer, always same result.
Here is a fiddle.
Any insight as to what I am doing wrong?
回答1:
You should cast before you divide, but also you were missing a subquery to get the total count from the table. Here's the sample.
select
random_int,
count(random_int) as Count,
cast(count(random_int) as decimal(7,2)) / cast((select count(random_int) from test) as decimal(7,2)) as Percent
from test
group by random_int
order by random_int;
回答2:
Try this query instead:
select
random_int,
count(random_int) as Count,
cast( count(random_int) / max(id) as float) as Percent,
(100.0 * count(random_int) / max(id))::numeric(5,2) as pct
from test
group by random_int
order by random_int;
PostgreSQL has a strong types system. In your case, type is implied by count()
function, which returns bigint
(or int8
) and id
column, which is integer
.
I would recommend using 100.0
as initial multiplier, that'll cause whole expression to be calculated as numeric and will also provide real percents. You might also want to cast to numeric(5,2)
at the end to get rid of too big number.
来源:https://stackoverflow.com/questions/26537140/division-of-integers-returns-0