Rank order grouped data in MySQL without session variables? - mysql

Question

My data looks like this:
Table Name = sales_orders
Customer_id| Order_id| Item_Id
-------------------------------
1 | 1 | 10
1 | 1 | 24
1 | 1 | 37
1 | 2 | 11
1 | 2 | 15
1 | 3 | 28
2 | 4 | 37
4 | 6 | 10
2 | 7 | 10
However, I need it to look like this:
Customer_id| Order_id| Item_Id |Order_rank
------------------------------------------
1 | 1 | 10 | 1
1 | 1 | 24 | 1
1 | 1 | 37 | 1
1 | 2 | 11 | 2
1 | 2 | 15 | 2
1 | 3 | 28 | 3
2 | 4 | 37 | 1
4 | 6 | 10 | 1
2 | 7 | 10 | 2
Customer_Id is a unique person
Order_id is a unique order
item_id is the product code
To further explain, the first three lines are from Customer #1's first order (order_id = 1) where this person ordered 3 different items (10,24, and 37). They then purchased another order (order_id =2) with two other products. Person with customer_id =2 has 2 unique orders (4 and 6), while customer with ID '4' has one unique order (order_id =6)
Essentially, what I need to do is rank these orders by customer_id and order Id, so that I can say "Order_id = 7 is the second order for customer_id = 2, because Order_rank = 2"
The challenge here is that I can't use session variables (e.g. #grp := customer_id ) in the MySQL query
For example, a query such as this is NOT allowed:
SELECT
customer_id,
order_id,
#ss := CASE WHEN #grp = customer_id THEN #ss + 1 ELSE 1 END AS
order_rank,
#grp := customer_id
FROM
(
SELECT
customer_id,
order_id
FROM sales_orders
GROUP BY customer_id, order_id
ORDER BY customer_id, order_id ASC
) AS t_1
CROSS JOIN (SELECT #ss := 0, #grp = NULL)ss
ORDER BY customer_id asc
Thanks for the help!

Answer 1

In a Correlated Subquery, we can Count(..) the unique and previous order_id values for a specific row's customer_id and order_id to determine the rank.
We need to count unique values because you have multiple rows per order (due to multiple items).
Query
SELECT
t1.Customer_id,
t1.Order_id,
t1.Item_Id,
(SELECT COUNT(DISTINCT t2.Order_id)
FROM sales_orders t2
WHERE t2.Customer_id = t1.Customer_id AND
t2.Order_id <= t1.Order_id
) AS Order_rank
FROM sales_orders AS t1;
Result
| Customer_id | Order_id | Item_Id | Order_rank |
| ----------- | -------- | ------- | ---------- |
| 1 | 1 | 10 | 1 |
| 1 | 1 | 24 | 1 |
| 1 | 1 | 37 | 1 |
| 1 | 2 | 11 | 2 |
| 1 | 2 | 15 | 2 |
| 1 | 3 | 28 | 3 |
| 2 | 4 | 37 | 1 |
| 4 | 6 | 10 | 1 |
| 2 | 7 | 10 | 2 |
View on DB Fiddle

Answer 2

You can use a correlated subquery:
select so.*,
(select count(*)
from sales_orders so2
where so2.Customer_id = so.Customer_id and
so2.order_id <= so.order_id
) as rank_order
from sales_orders so;
Or in MySQL 8+:
select so.*,
dense_rank() over (partition by Customer_Id order by Order_Id) as rank_order
from sales_orders so;