How can I summarise several columns in MySQL - mysql

Question

I have a table with a couple of thousand rows: from this I need to extract a total for column WTE for each value in column band, including those where the total is 0. I also need each total to be in a column of its own, so that I can easily update a summary table.
The code I have at present returns the values from the relevant rows:
SELECT
IF(band="E",WTE,0) AS `Band6_WTE`
FROM `orthoptists` AS o
LEFT JOIN `instances` AS i
ON o.instance_FK = i.id
WHERE i.region = 14
But when I add SUM(), the return is incorrect (zero, when it should be several thousands):
SELECT
IF(band="E",SUM(WTE),0) AS `Band6_WTE`
FROM `orthoptists` AS o
LEFT JOIN `instances` AS i
ON o.instance_FK = i.id
WHERE i.region = 14
I have looked at http://en.wikibooks.org/wiki/MySQL/Pivot_table, but I do not understand how that approach should be applied to my problem.
What should I do?

Answer 1

you must sum if:
SELECT
SUM(IF(band="E",WTE,0)) AS `Band6_WTE`
FROM `orthoptists` AS o
LEFT JOIN `instances` AS i
ON o.instance_FK = i.id
WHERE i.region = 14

Answer 2

In this particular case, wouldn't it be easier to use WHERE?
SELECT SUM(WTE) AS `Band6_WTE`
FROM `orthoptists` AS o
LEFT JOIN `instances` AS i
ON o.instance_FK = i.id
WHERE i.region = 14
AND band = "E"
For the general case, you could use GROUP BY since you say you need the result for each band (each value in column band):
SELECT band, SUM(WTE) AS `WTE`
FROM `orthoptists` AS o
LEFT JOIN `instances` AS i
ON o.instance_FK = i.id
WHERE i.region = 14
GROUP BY band
This will give either NULL or 0, you can instead use IFNULL to convert NULL to 0 if you like:
IFNULL(SUM(WTE), 0)
Edit: as you pointed out in the comments, you'd like multiple columns for the different bands rather than multiple rows. Generally speaking, you should not do that from SQL (use the second query and perform a pivot operation from your code), but there are exceptions, cases where it's significantly more complicated to do that outside of SQL, so here's how you could do it:
SELECT
(SELECT SUM(WTE)
FROM `orthoptists` AS o
LEFT JOIN `instances` AS i
ON o.instance_FK = i.id
WHERE i.region = 14
AND band = "E") AS `Band6_WTE`,
(SELECT SUM(WTE)
FROM `orthoptists` AS o
LEFT JOIN `instances` AS i
ON o.instance_FK = i.id
WHERE i.region = 14
AND band = "F") AS `Band7_WTE`,
(SELECT SUM(WTE)
FROM `orthoptists` AS o
LEFT JOIN `instances` AS i
ON o.instance_FK = i.id
WHERE i.region = 14
AND band = "G") AS `Band8_WTE`
The precise syntax might need a little bit of tweaking (some databases require each SELECT to include a FROM clause, some may require a name for each column in a subselect, I don't think MySQL does but I can't check right now), but the principle should be applicable regardless.

Answer 3

Noody has explained why your original query give you the incorrect result.
The original query has an aggregation function in the SELECT clause. This tells MySQL that this is an aggregation. There is no GROUP BY clause, so it returns one row, treating all rows as a single group.
Now, what happens to band in this case? In the SQL standard or any other dialect of SQL, the original query would return an error, saying something like band is not aggregated.
MySQL has a (mis)feature called Hidden Columns, which allows this syntax. It takes a random value of band from all the rows for the comparison. The value might have the value "E", in which case the sum of all WTE is returned. Or, it might have another value, in which case 0 is returned.
In general, you should avoid using Hidden Columns. Any "bare" column in your SELECT statement should also be in the GROUP BY clause, when you have an aggregation query.