CODEWITHSUNDEEP
sqlsql-servercountgroup-bymultiple-columns
shaolinmonk
shaolinmonk

Reputation: 31

Multiple count columns with multiple conditions

I have 2 tables that looks like this:

People:

1 Andy        Relative
2 Jim         Friend
3 Anderson    Friend
4 Pamela      Relative

Likes:

1  Pizza
1  Soda
2  Pizza
3  Soda
4  Pizza
4  Soda

I need multiple counts with multiple conditions. If it were just the one string value I were looking for I could count(Case...) but I need to include this query that looks for TWO strings per id:

Code to count BOTH Pizza and Soda:

select relation, count(*) 
from People p
join likes l1 on l1.id = p.id and pizzavalue = 'Pizza'
join likes l2 on l2.id = p.id and pizzavalue = 'Soda'
group by relation

This gives me:

Relative 2
Friend   0

I then need to join additional columns also grouped by relation to count such as counting just soda and just pizza. This is the intended result:

Relation | Pizza and Soda | Pizza Only | Soda Only
Relative         2              2            2
Friend           0              1            1

Upvotes: 1

Views: 851

Answers (3)

TMNT2014
TMNT2014

Reputation: 2130

You might need to tweak the joins a bit to get the desired result but this should get you going in the right direction.

select Relationship, Count(l1.Likes) 'Pizza Only', Count(l2.likes) 'Soda Only', Count(PS.Id) 'Both' FROM @People p LEFT Outer join @likes l1 on l1.id = p.id and l1.Likes = 'Pizza' LEFT Outer join @likes l2 on l2.id = p.id and l2.Likes = 'Soda' LEFT Outer JOIN (select p2.Id, Count(distinct(Likes)) count FROM @People p2 INNER JOIN likes l3 ON p2.Id = l3.Id WHERE Likes in ( 'Pizza','Soda') Group by p2.Id Having Count(distinct(Likes))> 1) PS ON p.Id = pS.id GROUP BY RelationShip

Upvotes: 1

podiluska
podiluska

Reputation: 51494

You can do this with a pivot

select *
from
(
    select 
        distinct
        p.id,
        relation,
        (select likes + ' ' from @likes l2 where l2.id=l1.id order by likes for xml path('') ) as likes
    from @likes l1
        inner join @people p on l1.id = p.id
) s
pivot
    (count(id) for likes in ([pizza],[soda],[pizza soda])) p

Upvotes: 2

Richard Hansell
Richard Hansell

Reputation: 5403

This might help?

DECLARE @People TABLE (
    Id INT PRIMARY KEY,
    Name VARCHAR(50),
    Relationship VARCHAR(50));
INSERT INTO @People VALUES (1, 'Andy', 'Relative');
INSERT INTO @People VALUES (2, 'Jim', 'Friend');
INSERT INTO @People VALUES (3, 'Anderson', 'Friend');
INSERT INTO @People VALUES (4, 'Pamela', 'Relative');

DECLARE @Likes TABLE (
    Id INT,
    Likes VARCHAR(50));
INSERT INTO @Likes VALUES (1, 'Pizza');
INSERT INTO @Likes VALUES (1, 'Soda');
INSERT INTO @Likes VALUES (2, 'Pizza');
INSERT INTO @Likes VALUES (3, 'Soda');
INSERT INTO @Likes VALUES (4, 'Pizza');
INSERT INTO @Likes VALUES (4, 'Soda');

WITH Likes AS (
    SELECT
        l.Id,
        SUM(CASE WHEN l.Likes = 'Pizza' THEN 1 ELSE 0 END) AS Pizza,
        SUM(CASE WHEN l.Likes = 'Soda' THEN 1 ELSE 0 END) AS Soda
    FROM
        @Likes l
    GROUP BY
        l.Id)
SELECT
    p.Relationship,
    SUM(CASE WHEN l.Pizza = 1 AND l.Soda = 0 THEN 1 ELSE 0 END) AS PizzaOnly,
    SUM(CASE WHEN l.Pizza = 0 AND l.Soda = 1 THEN 1 ELSE 0 END) AS SodaOnly,
    SUM(CASE WHEN l.Pizza = 1 AND l.Soda = 1 THEN 1 ELSE 0 END) AS PizzaAndSoda
FROM
    @People p
    LEFT JOIN Likes l ON l.Id = p.Id
GROUP BY
    p.Relationship;

Results:

Relationship    PizzaOnly   SodaOnly    PizzaAndSoda
Friend          1           1           0
Relative        0           0           2

Upvotes: 0

Related Questions