How to calculate a moving average in MYSQL - mysql

Question

I have an application that stores stock quotes into my MySQL database.
I have a table called stock_history:
mysql> desc stock_history;
+-------------------+---------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+-------------------+---------------+------+-----+---------+----------------+
| id | int(11) | NO | PRI | NULL | auto_increment |
| date | date | NO | MUL | NULL | |
| close | decimal(12,5) | NO | MUL | NULL | |
| dmal_3 | decimal(12,5) | YES | MUL | NULL | |
+-------------------+---------------+------+-----+---------+----------------+
5 rows in set (0.01 sec)
These are all the values in this table:
mysql> select date, close, dmal_3 from stock_history order by date asc;
+------------+----------+----------+
| date | close | dmal_3 |
+------------+----------+----------+-
| 2000-01-03 | 2.00000 | NULL |
| 2000-01-04 | 4.00000 | NULL |
| 2000-01-05 | 6.00000 | NULL |
| 2000-01-06 | 8.00000 | NULL |
| 2000-01-07 | 10.00000 | NULL |
| 2000-01-10 | 12.00000 | NULL |
| 2000-01-11 | 14.00000 | NULL |
| 2000-01-12 | 16.00000 | NULL |
| 2000-01-13 | 18.00000 | NULL |
| 2000-01-14 | 20.00000 | NULL |
+------------+----------+----------+-
10 rows in set (0.01 sec)
I am guaranteed that there will be 0 or 1 record for each date.
Can I write a single query that will insert the three-day moving average (ie: the average closing prices of that day and the two previous trading days before it) into the dmal_3 field? How?
When the query is done, I want the table to look like this:
mysql> select date, close, dmal_3 from stock_history order by date asc;
+------------+----------+----------+
| date | close | dmal_3 |
+------------+----------+----------+
| 2000-01-03 | 2.00000 | NULL |
| 2000-01-04 | 4.00000 | NULL |
| 2000-01-05 | 6.00000 | 4.00000 |
| 2000-01-06 | 8.00000 | 6.00000 |
| 2000-01-07 | 10.00000 | 8.00000 |
| 2000-01-10 | 12.00000 | 10.00000 |
| 2000-01-11 | 14.00000 | 12.00000 |
| 2000-01-12 | 16.00000 | 14.00000 |
| 2000-01-13 | 18.00000 | 16.00000 |
| 2000-01-14 | 20.00000 | 18.00000 |
+------------+----------+----------+
10 rows in set (0.01 sec)

Answer 1

That is what I call a good challenge. My solution first creates a counter for the values and uses it as a table. From it I select everything and join with the same query as a subquery checking the position of the counter on both. Once the query works it just need an inner join with the actual table to do the update. Here it is my solution:
update stock_history tb1
inner join
(
select a.id,
case when a.step < 3 then null
else
(select avg(b.close)
from (
select hh.*,
#stp:=#stp+1 stp
from stock_history hh,
(select #sum:=0, #stp:=0) x
order by hh.dt
limit 17823232
) b
where b.stp >= a.step-2 and b.stp <= a.step
)
end dmal_3
from (select h1.*,
#step:=#step+1 step
from stock_history h1,
(select #sum:=0, #step:=0) x
order by h1.dt
limit 17823232
) a
) x on tb1.id = x.id
set tb1.dmal_3 = x.dmal_3;
I changed some columns names for easiness of my test. Here it is the working SQLFiddle: http://sqlfiddle.com/#!9/e7dc00/1
If you have any doubt, let me know so I can clarify!
Edit
The limit 17823232 clause was added there in the subqueries because I don't know which version of MySql you are in. Depending on it (>= 5.7, not sure exactly) the database optimizer will ignore the internal order by clauses making it not work the way it should. I just chose a random big number usually you can use the maximum allowed.
The only column with different colunm name between your table and mine is the date one which I named dt because date is a reserved word and you should use backticks ( ` ) to use such columns, therefore I will left it as dt in above query.