MySQL multidimensional select from views - mysql

Question

I would like to display data re-arranged year by year and one of the possible solution is using views and select from them. The data matrix is something like (of course it's a ficticious demo dataset):
USA 2005 22 156
CAN 2005 14 101
MEX 2005 5 32
USA 2006 24 160
CAN 2006 16 103
USA 2007 26 163
MEX 2007 8 35
The SQL code to create and populate the table is:
DROP TABLE IF EXISTS `tab1`;<br>
CREATE TABLE `tab1` ( <br>
`id1` int(4) unsigned NOT NULL AUTO_INCREMENT,
`iso3` char(3) NOT NULL,
`year` int(4) unsigned NOT NULL,
`aaa` int(10) DEFAULT NULL,
`bbb` int(10) DEFAULT NULL,
PRIMARY KEY (`id1`)
)
INSERT INTO `tab1` VALUES
('1', 'USA', '2005', '22', '156'),
('2', 'CAN', '2005', '14', '101'),
('3', 'MEX', '2005', '5', '32'),
('4', 'USA', '2006', '24', '160'),
('5', 'CAN', '2006', '16', '103'),
('6', 'USA', '2007', '26', '163'),
('7', 'MEX', '2007', '8', '35');
COMMIT;
And now I would like to obtain for parameter 'aaa' a 2D table like this:
country 2005 2006 2007
USA 22 24 26
CAN 14 16
MEX 5 8
However the following SQL code is omitting all the lines with missing data, be it one single value and I only get one line
USA 22 24 26
The SQL code is:
SELECT view2005.Country, view2005.2005, view2006.2006, view2007.2007
FROM view2005, view2006, view2007
WHERE view2005.country = view2006.country
AND view2005.country = view2007.country
Any idea how to do it including lines with missing data? Thanks in advance.

Answer 1

Use left joins, and a view (or table, or inner select like below) which has all distinct countries:
SELECT c.country, view2005.2005, view2006.2006, view2007.2007
FROM (SELECT DISTINCT country FROM tab1) as c
LEFT JOIN view2005 ON view2005.country = c.country
LEFT JOIN view2006 ON view2006.country = c.country
LEFT JOIN view2007 ON view2007.country = c.country
GROUP BY c.country
EDIT:
In a more general context, what you are asking here is to create a pivot of this table, which is a common problem that has common solutions. Here is a nice "How To": http://www.artfulsoftware.com/infotree/queries.php?&bw=1339#78

Answer 2

It's better to use JOIN than implicit JOIN with WHERE. Additional advanatge is that you can convert it to a LEFT JOIN so data for 2005 that don't have a 2006 related row (and are not matched) will still be shown.
Use Galz's solution or search as correctly advised for how to create PIVOT queries.
One such logic to create a pivot query would be:
SELECT iso3 AS Country
, SUM(IF(year=2005, aaa, 0)) AS 2005
, SUM(IF(year=2006, aaa, 0)) AS 2006
, SUM(IF(year=2007, aaa, 0)) AS 2007
FROM tab1 AS t
GROUP BY iso3
If there are years without any data, you will get NULL in that column.
You can use COALESCE() function if you want 0 to be shown and not NULL:
SELECT iso3 AS Country
, COALESCE( SUM( IF(year=2004, aaa, 0) ) , 0) AS "2004"
, COALESCE( SUM( IF(year=2005, aaa, 0) ) , 0) AS "2005"
, COALESCE( SUM( IF(year=2006, aaa, 0) ) , 0) AS "2006"
, COALESCE( SUM( IF(year=2007, aaa, 0) ) , 0) AS "2007"
FROM tab1 AS t
GROUP BY iso3

Answer 3

Thank you Galz for the link to pivots and thank you ypercube for the SQL. It worked after enclosing the years into quotes to make them CHAR.
I was further intrigued by the question what happens if I add a row with no values at all or a row out of range of the years so I have added
INSERT INTO `tab1` VALUES
('7', 'ATA', '2004', '', '')
The result was that I got a mix of NULL and INT zero values. This is not good because zero is a valid number and legitimate data. So I have modified the query to get exactly the result I need:
SELECT iso3 AS countryб
SUM( IF(year=2004, aaa, NULL) ) AS "2004",
SUM( IF(year=2005, aaa, NULL) ) AS "2005",
SUM( IF(year=2006, aaa, NULL) ) AS "2006",
SUM( IF(year=2007, aaa, NULL) ) AS "2007"
FROM tab1
GROUP BY iso3