加入查询来计算记录数不正确

Question

I have a categories table that works sort of like the tags on StackOverflow. There are categories like JavaScript, which would be a top category, and maybe jQuery which would be a sub-category of JavaScript.

What I am trying to query is how many records (in my application they are "problems") there are under each category.

Here is the SQL I have so far. The problem_categories is just a joining table which has a problem_id and a category_id.

select problems.problem_id , categories.category_id , category_name , count(problems.problem_id) as num_problems , is_top
from problems 
    left join problem_categories on
    problems.problem_id = problem_categories.category_id
    left join categories on
    problem_categories.category_id = categories.category_id
    where is_top = 1;

This returns only one line. What I was hoping for is to have returned the number of records where is_top = 1 (That would mean that it is a top category).

How could I change my query to do that?

Thanks!!

Answer 1

Without guessing at the logic, what's wrong is that you need a GROUP BY to count:

SELECT problems.problem_id, categories.category_id, category_name, 
COUNT(problems.problem_id) as num_problems, is_top
FROM problems 
LEFT JOIN problem_categories 
ON problems.problem_id = problem_categories.category_id
LEFT JOIN categories 
ON problem_categories.category_id = categories.category_id
WHERE is_top = 1;
GROUP BY  problems.problem_id, categories.category_id, category_name

But if you want the number of problems per top-category, your logic probably goes like this instead:

SELECT category_name, categories.category_id, problems.problem_id, 
COUNT(problems.problem_id) as num_problems
FROM categories 
JOIN problem_categories 
ON problem_categories.category_id = categories.category_id
JOIN problems 
ON problems.problem_id = problem_categories.category_id
WHERE is_top = 1;
GROUP BY category_name, categories.category_id, problems.problem_id

Notice that:

• For each category, you get the problems, and you count those instead
• You use JOIN instead of LEFT JOIN, since you don't care about categories that have no problem(s) anyway.
• You can leave the is_top out of the select, since you put it in your WHERE clause anyway. Selecting something which is not in your GROUP BY will get any random value, but since all values are 1, you can safely do it, or just leave it.

Answer 2

"This returns only one line. What I was hoping for is to have returned the number of records where is_top = 1 (That would mean that it is a top category)." I suspect that that is what you are getting.

...count(problems.problem_id) as num_problems...
...where is_top = 1;

I think you need to introduce a GROUP BY clause on one or more of your fields

Answer 3

There are a few things that are amiss in your query:

1. As mentioned in other answers, the query is missing a GROUP BY clause

2. The first JOIN in your query is matching unrelated columns - "on problems.problem_id = problem_categories.category_id". As you can see, it is joining problem_id with category_id

3. Though this might not be the problem with the current data-set but to get counts per category, it makes more sense to keep the categories table on the extreme left in the LEFT JOIN
4. From optimization point of view, I do not think you need to put problems table in your query because there is nothing in that table that is needed in the output

Please below another version of your query:

SELECT `categories`.`category_id`, `categories`.`category_name`, COUNT(`problem_categories`.`problem_id`) AS `num_problems`, `categories`.`is_top`
FROM `categories`
LEFT JOIN `problem_categories` ON `problem_categories`.`category_id` = `categories`.`category_id`
WHERE `categories`.`is_top` = 1
GROUP BY `categories`.`category_id`;

Hope the above helps!