SQL — segment by groups [closed] - mysql

Question

Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.
Questions must demonstrate a minimal understanding of the problem being solved. Tell us what you've tried to do, why it didn't work, and how it should work. See also: Stack Overflow question checklist
Closed 9 years ago.
Improve this question
In a MySQL DB, I have a purchases table that has these columns:
USERID PURCHASE_AMOUNT
3 20
9 30
3 5
4 5
1 10
1 5
I would like to generate a report like this
SUM_OF_PURCHASES_RANGE NUM_OF_USERS
0-1 0
1-5 1
5-20 1
20-30 2
Where it means: there are 0 users who bought up to 1(SUM of purchases) (inclusive), there are 1 users who bought between 1 to 5 etc...
What query should I use to generate it?

Answer 1

You can create the range using a UNION, and just LEFT JOIN to that to get all categories; (edited for your change in the desired result)
SELECT CONCAT(base.lower,'-',base.upper) PURCHASE_RANGE, COUNT(userid) NUM_OF_USERS
FROM (
SELECT 0 lower, 1 upper UNION SELECT 2, 5 UNION SELECT 6,20 UNION SELECT 21,30
) base
LEFT JOIN (
SELECT userid, SUM(purchase_amount) pa FROM purchases GROUP BY userid
) p
ON p.pa >= base.lower AND p.pa <= base.upper
GROUP BY base.upper
An SQLfiddle to test with.

Answer 2

More easier syntax :
SELECT PURCHASE_RANGE , COUNT(*) as NUM_OF_USERS
FROM
(
SELECT
CASE
WHEN PURCHASE_AMOUNT <= 1 THEN 1
WHEN PURCHASE_AMOUNT > 1 AND PURCHASE_AMOUNT <= 5 THEN 5
WHEN PURCHASE_AMOUNT > 5 AND PURCHASE_AMOUNT <= 10 THEN 10
WHEN PURCHASE_AMOUNT > 10 AND PURCHASE_AMOUNT <= 20 THEN 20
WHEN PURCHASE_AMOUNT > 20 AND PURCHASE_AMOUNT <= 30 THEN 30 END AS PURCHASE_RANGE
FROM Table1
) AS A
GROUP BY PURCHASE_RANGE
ORDER BY PURCHASE_RANGE
SqlFiddle

Answer 3

try this
select PURCHASE_RANGE , NUM_OF_USERS
from (
select 1 as PURCHASE_RANGE ,count(*) as NUM_OF_USERS from table1 where PURCHASE_AMOUNT between 0 and 1
union all
select 5 ,count(*) from table1 where PURCHASE_AMOUNT between 1 and 5
union all
select 20 ,count(*) from table1 where PURCHASE_AMOUNT between 6 and 20
union all
select 30 ,count(*) from table1 where PURCHASE_AMOUNT between 21 and 30
)t
DEMO HERE

Answer 4

There are faster ways to do this if you need the performance (this will do a full table scan), but try this:
SELECT
SUM(CASE WHEN purchase_amount BETWEEN 0 AND 1 THEN 1 ELSE 0) bucket_0_to_1,
SUM(CASE WHEN purchase_amount BETWEEN 1 AND 5 THEN 1 ELSE 0) bucket_1_to_5,
SUM(CASE WHEN purchase_amount BETWEEN 5 AND 20 THEN 1 ELSE 0) bucket_5_to_20,
SUM(CASE WHEN purchase_amount BETWEEN 20 AND 30 THEN 1 ELSE 0) bucket_20_to_30,
SUM(CASE WHEN purchase_amount > 30 THEN 1 ELSE 0) bucket_over_30, FROM my_table LIMIT 1;

Answer 5

To get the values you want in rows, you need to start with a driver table that has all the values you are interested in, and then left outer join to the data:
select driver.mina, coalesce(sum(cnt), 0) as Num_Of_Users
from (select 1 as mina, 5 as maxa union all
select 5, 10 union all
select 10, 20 union all
select 20, 30 union all
select 30, NULL
) driver left outer join
(select purchase_amount, count(*) as cnt
from purchases
group by purchase_amount
) pa
on driver.mina >= pa.purchase_amount and
(pa.purchase_amount < driver.maxa or driver.maxa is null)
group by driver.mina
order by driver.mina
You can actually do this without the inner group by. That is likely to reduce the size of the data significantly (especially in your example) before join.
I would encourage you to include both the lower and upper bounds of the range on each row.

Answer 6

This might be easier if the ranges will ever change.
with ranges(rstart, rfinish) as (
select 0, 1 union all
select 2, 5 union all
select 6, 20 union all
select 21, 30
), purchases(amount) as (
select sum(PURCHASE_AMOUNT)
from <purchases_basetable> -- <-- your tablename goes here
group by USERID
)
select
-- concat(case when r.rstart = 0 then 0 else r.rstart-1 end, '-', r.rfinish) as SUM_OF_PURCHASES_RANGE /* op's name for the group */,
concat(r.rstart, '-', r.rfinish) as SUM_OF_PURCHASES_RANGE /* better name for the group */,
count(*) as NUM_OF_USERS
from
purchases as p inner join
ranges as r
on p.amount between r.start and r.finish
group by r.rstart, r.rfinish
order by r.rstart, r.rfinish
I don't know what the mysql query plan will look like. It's trivial to change the query to use derived tables rather than table expressions. (But I include it below anyway.)
You might also find the UNPIVOT operation to be useful on a platform that supports it.
select
-- concat(case when r.rstart = 0 then 0 else r.rstart-1 end, '-', r.rfinish) as SUM_OF_PURCHASES_RANGE /* op's name for the group */,
concat(r.rstart, '-', r.rfinish) as SUM_OF_PURCHASES_RANGE /* better name for the group */,
count(*) as NUM_OF_USERS
from
(
select sum(PURCHASE_AMOUNT) as amount
from <purchases_basetable> -- <-- your tablename goes here
group by USERID
) as p inner join
(
select 0 as rstart, 1 as rfinish union all
select 2, 5 union all
select 6, 20 union all
select 21, 30
) as r
on p.amount between r.start and r.finish
group by r.rstart, r.rfinish
order by r.rstart, r.rfinish