Too many rows when Joining to same table twice - mysql

Question

I have two tables, Product and Benchmark
A benchmark is linked to only one product. There can only be one benchmark per year per product.
I would like to retrieve every products' name for a set of years, and count how many benchmark there are for each product.
SELECT p.name,
p.id,
COUNT(p.id) AS nb_benchmark
FROM product p
INNER JOIN benchmark b0 ON b0.product_id = p.id
INNER JOIN benchmark b1 ON b1.product_id = p.id
WHERE p.owner = "MyCompany"
AND b0.year = 2011
AND b1.year = 2012
GROUP BY p.id
ORDER BY nb_trials DESC
But the count is wrong, it's way to high, it even gives me more results than there actually are in the database. I guess it's because of the JOINs, but I don't know how to build the query.

Answer 1

Remember that the basis of SQL joining is the cartesian product of rows in the referenced tables, which are then eliminated by filters and join conditions. Because you are joining TWICE to table benchmark, which from the nature of your query, we can assume has many benchmark rows per product per benchmark year.
e.g. 1 Product with 3 Benchmark rows each for 2011 and 2012
FROM product p -- 1 Product Row
INNER JOIN benchmark b0 ON b0.product_id = p.id -- 1 x 3 = 3
INNER JOIN benchmark b1 ON b1.product_id = p.id -- 1 x 3 x 3 = 9
So the multiple joins to benchmark introduces duplicate rows for product, which are then counted.
You can use COUNT(DISTINCT xx) to count distinct values, so your query should be of the form:
SELECT p.name,
p.id,
COUNT(DISTINCT p.id) AS distinct_products,
COUNT(DISTINCT b.name) AS distinct_benchmark_names
-- etc
FROM ...
Other Notes
for correctness sake you should GROUP BY both p.id and p.name. Although MySql allows this, other RDBMS are more strict.

Answer 2

Try this:
SELECT p.name,
p.id,
COUNT(b0.id) AS nb_benchmark
FROM product p
INNER JOIN benchmark b0 ON b0.product_id = p.id
WHERE p.owner = "MyCompany"
AND b0.year IN (2011, 2012)
GROUP BY p.name, p.id
ORDER BY nb_trials DESC

Answer 3

I have found a way to achieve what I wanted
SELECT p.name, p.id, COUNT(DISTINCT(b0.id)) + COUNT(DISTINCT(b1.id)) as nb_benchmark
FROM product p
INNER JOIN benchamrk b0 ON b0.product_id = p.id AND b0.year = 2011
INNER JOIN benchamrk b1 ON b1.product_id = p.id AND b1.year = 2012
WHERE
p.owner = "myCompany"
GROUP BY p.id
ORDER BY nb_benchmark DESC

Answer 4

Try this.
SELECT p.id, p.name, b.nb_benchmark
FROM product p
JOIN (
/* number of benchpark per product for years 2011 and 2012 */
SELECT product_id, COUNT(*) AS nb_benchmark
FROM benchmark
WHERE year = 2011 OR year = 2012
GROUP BY product_id
) b ON p.id = b.product_id
WHERE p.owner = "MyCompany"
ORDER BY nb_benchmark DESC