Question

Lets say I have this data set:

id   city   result fruit
1    boston good   apple
2    boston bad    banana
3    ny     good   peach
4    ny     good   banana
5    ny     good   apple
6    ny     bad    apple
...

So imagine that this data set goes on and has results for any number of fruit and any number of cities.
The problem is, I don’t know how to perform a query to give me results like this:

city   fruit   good   bad
ny     bananna 3      2
ny     apple   1      3
ny     peach   3      2
boston banana  2      5
boston peach   2      1
...

The idea is that we are grouping by two things, the city and the fruit, but I want to get only the count of fruit within each city. How can I do this?

Answer 1

This technique is generally called a "pivot":

select
    city,
    fruit,
    sum(result = 'good') as good,
    sum(result = 'bad') as bad
from mytable
group by 1,2

Notice the elegant use of sum(condition) rather than the mundane sum(case...)

Edited

As usual, people can’t help but make critical/irrelevant comments, so for the record…

The sum(condition) approach works for any database that treats true as 1 and false as 0. Mysql is one such database, thus this is an appropriate answer for a mysql tagged question. Mileage may vary with other databases.