SQL与自我结合

Question

I have a table with ID numbers of people and then items of food they've ordered:

table "food_id"

food     id
ham      1
cheese   2
turkey   2
ham      3
ham      4
bread    5
cheese   6
turkey   6
cheese   7

And I'd like to use SQL to figure out, for each id, the total number of other IDs who ordered at least one of the same food items. For the above example, the answer should be:

"result_table"

count    id
3        1
3        2
3        3
3        4
1        5
3        6
3        7

The challenge is to avoid double counting here. For example, person number 2 got both cheese and turkey, so we want his final count to be 3 because person # 2, 6, and 7 got cheese, and person # 2 and 6 got turkey, and there are 3 unique IDs in this list of (2,6,7,2,6).

My initial thoughts were to first get a table with food items to distinct ID numbers, and then to join this table with the original table and then group by ID number and get a count of the distinct number of IDs. However, I'm a beginner to SQL and can't figure out how to implement the code correctly.

Any direction would be greatly appreciated.

Answer 1

To avoid the problem with the double counting you can concat both ids from the join and count only distinct combinations. I add a separator to make the combination unique with greater id values:

SELECT 
  COUNT(DISTINCT CONCAT(f1.id, ',', f2.id)) as count,
  f1.id 
FROM
    food_id f1
INNER JOIN
    food_id f2
ON
    f1.food = f2.food
GROUP BY f1.id;

See demo

Answer 2

Like you said, you can do a self join. You can join by food , and count the number of distinct matching ids.

select
  a.id,  -- Person you're investigating 
  count(distinct b.id) as samefoodcount -- number of people sharing the same food
from
  food_id a
  inner join food_id b on b.food = a.food
group by 
  a.id

Here you can see the query in action: http://sqlfiddle.com/#!2/c53884/1

Answer 3

You can run the following query to get the desired output:

select MAX(T.total),id
from table_name, (select count(*) as total,food from table_name group by food) T
where table_name.food=T.food
group by id

Check the DEMO