Group By specified numbers of ordered rows

I have such table in my MySQL database:

---------------------------
|fid | price | date       |
---------------------------
|  1 | 1.23  | 2开发者_JAVA技巧011-08-11 |
|  1 | 1.43  | 2011-08-12 |
|  1 | 1.54  | 2011-08-13 |
|  1 | 1.29  | 2011-08-14 |
|  1 | 1.60  | 2011-08-15 |
|  1 | 1.80  | 2011-08-16 |

fid - this is the product id

price - this is the price of the product in specified day

I want to calculate average price of the product fid=1. I want to calculate the average price of first n=3 rows ordered by date for specified fid, and then calculate average price for another 3 rows ordered by date.

How can I group first 3 rows and calculate avg and then group next 3 rows and calculate avg. Before calculation I need to sort the rows by date and then group n rows.

If n=3 this should return such result:

--------------
|fid | price |
--------------
|  1 | 1.40  | 2011-08-11 -> 2011-08-13 - average price for 3 days
|  1 | 1.56  | 2011-08-14 -> 2011-08-16 - average price for 3 days

How can I create SQL Query to do such calculations?

Thanks in advance.

---

Unluckily mysql doesn't offer analytic functions like oracle,mssql and postgres do. So you have to play with variables to reach your goal.

create table mytest (
id int not null auto_increment primary key,
fid int,
price decimal(4,2),
fdate date
) engine = myisam;

insert into mytest (fid,price,fdate)
values 
(1,1.23,'2011-08-11'),
(1,1.43,'2011-08-12'),
(1,1.54,'2011-08-13'),
(1,1.29,'2011-08-14'),
(1,1.60,'2011-08-15'),
(1,1.80,'2011-08-16');


select 
concat_ws('/',min(fdate),max(fdate)) as rng,
format(avg(price),2) as average from (
select *,@riga:=@riga+1 as riga
    from mytest,(select @riga:=0) as r order by fdate
     ) as t
group by ceil(riga/3);


+-----------------------+---------+
| rng                   | average |
+-----------------------+---------+
| 2011-08-11/2011-08-13 | 1.40    |
| 2011-08-14/2011-08-16 | 1.56    |
+-----------------------+---------+
2 rows in set (0.02 sec)

---

maybe you could use

GROUP BY FLOOR(UNIX_TIMESTAMP(date)/(60*60*24*3))

= convert to secounds, divide by secounds for 3 days, and round down

---

SELECT AVG( price ) FROM my_table
    GROUP BY ( date - ( SELECT MIN( date ) FROM my_table WHERE fid = 1 ) ) DIV 3
    WHERE fid = 1

---

select fid, avg(price), min(date), max(date)
from
    (select floor((@rownum := @rownum + 1)/3) as `rank`,  prices.*
    from prices, (SELECT @rownum:=-1) as r
    order by date) as t
group by rank