Grouping items between 2 numbers - mysql

Question

I have a query that looks like this:
select
price,
item_id,
sum(price),
count(item_id)
from transactions
group by
(price <= 20),
(price between 21 and 30),
(price between 31 and 40),
(price between 41 and 50),
(price > 50)
I have never done a group like this before when I wrote it I was just guessing to see if the query was even valid, and it was. But my question is, is it really getting me what I want?
I want all transactions grouped by:
Items that cost less than or equal to $20
Items that cost between $21 and $30
Items that cost between $31 and $40
Items that cost between $41 and $50
Items that cost more than $50
So, is that query doing what I am asking?

Answer 1

The way to do this in standard SQL (and MySQL) is to use the case statement. Also, I put the definition in a subquery like this:
select pricegrp, sum(price), count(item_id)
from (select t.*,
(case when price <= 20 then '00-20'
when price between 21 and 30 then '21-30'
when price between 31 and 40 then '31-40'
when price between 41 and 50 then '41-50'
when price > 50 then '50+'
end) as pricegrp
from transactions t
) t
group by pricegrp
Also, do you want to group by item_id as well? Or are you just trying to return one arbitrary item? Based on what you want, I'm removing the item_id from the select clause. It doesn't seem necessary.
Your query actually does work in MySQL, in the sense that it runs. It is going to produce one row for each group that you want, so in that sense it "works". However, within each group, it is going to choose an arbitrary price and item_id. These are not explicitly mentioned in the group by clause, so you are using a MySQL (mis)feature called Hidden Columns. Different runs of the query or slight changes to the data or slight changes to the query can change the values of price and item_id returned for each group.
I strongly suggest that you actually name the group. This makes the query and the output much clearer.
Also, I recommend that you get in the habit of putting all columns in the select in the group by clause. There are a few cases where hidden columns are actually useful, but I think, in general, you should depend on them sparingly.
If the price is not stored as an integer, then correct logic is:
select pricegrp, sum(price), count(item_id)
from (select t.*,
(case when price <= 20 then '00-20'
when price <= 30 then '21-30'
when price <= 40 then '31-40'
when price <= 50 then '41-50'
when price > 50 then '50+'
end) as pricegrp
from transactions t
) t
group by pricegrp

Answer 2

SELECT
price,
item_id,
sum(price),
count(item_id),
IF(price<=20,0,IF(price<=30,1,IF(price<=40,2,IF(price<=50,3,4)))) AS pricegroup
FROM transactions
GROUP BY pricegroup
or even
SELECT
price,
item_id,
sum(price),
count(item_id)
FROM transactions
GROUP BY
IF(price<=20,0,IF(price<=30,1,IF(price<=40,2,IF(price<=50,3,4))))

Answer 3

SELECT price,
item_id,
SUM(CASE WHEN price <= 20 THEN price ELSE 0 END) `(price <= 20) SUM`,
SUM(CASE WHEN price <= 20 THEN 1 ELSE 0 END) `(price <= 20) COUNT`,
SUM(CASE WHEN price between 21 and 30 THEN price ELSE 0 END) `(price <= 20) SUM`,
SUM(CASE WHEN price between 21 and 30 THEN 1 ELSE 0 END) `(price <= 20) COUNT`,
SUM(CASE WHEN price between 31 and 40 THEN price ELSE 0 END) `price between 31 and 40 SUM`,
SUM(CASE WHEN price between 31 and 40 THEN 1 ELSE 0 END) `price between 31 and 40 COUNT`,
SUM(CASE WHEN price between 41 and 50 THEN price ELSE 0 END) `price between 41 and 50 SUM`,
SUM(CASE WHEN price between 41 and 50 THEN 1 ELSE 0 END) `price between 41 and 50 COUNT`,
SUM(CASE WHEN price > 50 THEN price ELSE 0 END) `price > 50 SUM`,
SUM(CASE WHEN price > 50 THEN 1 ELSE 0 END) `price > 50 COUNT`
FROM transactions
GROUP BY price, item_id