where clause sql and display NULL values - mysql

Question

I have a database with information about several servers :
mysql> select DATE_FORMAT(timestamp,'%M %Y') as 'Mois',AVG(value_perf) as 'moy' from info WHERE server = 'pi4' GROUP BY MONTH(timestamp);
+----------------+-----------+
| Mois | moy |
+----------------+-----------+
| June 2014 | 98.465500 |
| July 2014 | 98.854516 |
| August 2014 | 98.227097 |
| September 2014 | 95.008667 |
| October 2014 | 77.880000 |
+----------------+-----------+
5 rows in set (0.00 sec)
However, each pi server has different starting date ( pi4 stats in june, but pi1 stats in january)
Example :
mysql> select DISTINCT DATE_FORMAT(timestamp,'%M %Y') as 'Mois' from info ORDER BY timestamp;
+----------------+
| Mois |
+----------------+
| January 2014 |
| February 2014 |
| March 2014 |
| April 2014 |
| May 2014 |
| June 2014 |
| July 2014 |
| August 2014 |
| September 2014 |
| October 2014 |
+----------------+
And i would like a query to get the average score per server, BUT with the most large view, including NULL.
Example for pi4 :
+----------------+-----------+
| Mois | moy |
+----------------+-----------+
| January 2014 | NULL |
| February 2014 | NULL |
| March 2014 | NULL |
| April 2014 | NULL |
| May 2014 | NULL |
| June 2014 | 98.465500 |
| July 2014 | 98.854516 |
| August 2014 | 98.227097 |
| September 2014 | 95.008667 |
| October 2014 | 77.880000 |
+----------------+-----------+
Ps : Here is an example of the database structure
mysql> select * from info limit 10;
+----+------------+-----------+------------+-------------+
| id | timestamp | server | value_perf | value_avail |
+----+------------+-----------+------------+-------------+
| 45 | 2014-06-11 | pi4 | 98.33 | 99.91 |
| 46 | 2014-06-12 | pi4 | 97.92 | 100.00 |
| 52 | 2014-06-18 | pi4 | 98.15 | 99.97 |
| 54 | 2014-06-20 | pi | 98.33 | 99.94 |
+----+------------+-----------+------------+-------------+
and i want the average per month, with NULL value for unavailable months.
How can i get this ?
Thanks

Answer 1

The simplest way to achieve what you want is to create a date lookup table. Something like
CREATE TABLE `month_lookup` (
`month_year` varchar(30),
`st` datetime,
`en` datetime,
PRIMARY KEY (`month_year`),
KEY `st` (`st`)
KEY `end` (`end`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
Then insert data for each month's start and end date. Go ahead and populate it for 1970 to 2070 or something. This table can come in handy. An example row would look like
month-year st en
Jan-2014 2014-01-01 00:00:00 2014-01-31 23:59:59
Then what you can do is something like this:
SELECT a.month_year, avg(b.value_perf) AS moy
FROM month_lookup AS a
LEFT JOIN info AS b
ON b.timestamp BETWEEN a.st AND b.en
GROUP BY a.month_year, b.server;
The indices on the month_lookup table might be better if you did st_end (st, en) instead of st and en separately, but you'd have to test that... I don't recall off the top of my head if a compound index would help you with with a join ... between clause or not
Either way, create a month_lookup table and use it in a left join... probably your best option. Otherwise you'll be doing unions ond subqueries and it'll just get hard to manage.