MySQL SET rank over First Column and Second Column - mysql

Question

I'm trying to make a query setting rank column by First and Second column. Like Rank over Partition which doesn't exist MySQL
For example,
From
+----+-------+--------+------+
| id | First | Second | Rank |
+----+-------+--------+------+
| 1 | a | 10 | |
| 2 | a | 9 | |
| 3 | b | 10 | |
| 4 | b | 7 | |
| 5 | a | 1 | |
| 6 | b | 1 | |
+----+-------+--------+------+
To
+----+-------+--------+------+
| id | First | Second | Rank |
+----+-------+--------+------+
| 1 | a | 10 | 3 |
| 2 | a | 9 | 2 |
| 3 | b | 10 | 3 |
| 4 | b | 7 | 2 |
| 5 | a | 1 | 1 |
| 6 | b | 1 | 1 |
+----+-------+--------+------+
The Rank doesn't continue. It starts from 1 again when it reaches the last value of 'a' of 'First' column.
And it's gotta be SET not SELECT.
I wouldn't mind using SELECT but my point is I'm not trying to retrieve data from Database but setting values.
Cheers in advance mates.

Answer 1

DROP TABLE IF EXISTS my_table;
CREATE TABLE my_table
(id INT NOT NULL AUTO_INCREMENT PRIMARY KEY
,first CHAR(1) NOT NULL
,second INT NOT NULL
);
INSERT INTO my_table VALUES
(1,'a',10),
(2,'a',9),
(3,'b',10),
(4,'b',7),
(5,'a',1),
(6,'b',1);
SELECT id
, first
, second
, rank
FROM
( SELECT x.*
, CASE WHEN #prev = first THEN #i:=#i+1 ELSE #i:=1 END rank
, #prev:=first
FROM my_table x
, (SELECT #prev:=null,#i:=0) vars
ORDER
BY first
, second
, id
) a
ORDER
BY id;
+----+-------+--------+------+
| id | first | second | rank |
+----+-------+--------+------+
| 1 | a | 10 | 3 |
| 2 | a | 9 | 2 |
| 3 | b | 10 | 3 |
| 4 | b | 7 | 2 |
| 5 | a | 1 | 1 |
| 6 | b | 1 | 1 |
+----+-------+--------+------+
6 rows in set (0.00 sec)

Answer 2

Came up with a solution which I was looking for.
I'm not sure if these queries are completely safe but so far no harms.
SET #rank = 0, #First = ''
UPDATE 'Table' SET
rank = IF(#First = First, #rank:= #rank +1, #rank := 1 AND #First := First)
ORDER BY First ASC, Second;

Answer 3

One method is a correlated subquery. For rank() you can do:
select t.*,
(select count(*) + 1
from t t2
where t2.first = t.first and t2.second < t.second
) as rank
from t;
Ranks are tricky to handle with variables (dense_rank() and row_number() are simpler).
EDIT:
This is easy to turn into an update:
update t join
(select t.*,
(select count(*) + 1
from t t2
where t2.first = t.first and t2.second < t.second
) as new_rank
from t
) tt
on t.id = tt.id
set t.rank = tt.new_rank;