如何编写此查询而不在GROUP BY中包含所有列

Question

The following query

SELECT child.* 
FROM [nested-set] AS parent
JOIN (
  SELECT [lft] AS parentLft, [rgt] AS parentRgt 
  FROM [nested-set] WHERE [id] = 3
) AS parent2 ON 1=1
JOIN [nested-set] AS child ON child.[lft] BETWEEN parent.[lft] AND parent.[rgt] 
WHERE parent.[lft] >= parentLft AND parent.[rgt] < parentRgt AND parent.[id] > 1 
GROUP BY child.[id] 
HAVING COUNT(child.[id]) = 1 
ORDER BY child.[lft] ASC

that gets children of a node in nested set model that runs perfectly fine on MySQL and SQLite (of course with different column quoting) but in order to get it to work on MS SQL Server I have to add all the columns in the GROUP BY clause. Since this query is supposed to run on different tables that only have id, lft and rgt columns in common having to specify all columns each time would be an overkill. Is there a solution for this?

Answer 1

You can just use a subquery that returns the Id's you need, and then link the main table to that.

That way you don't need to specify the columns in the outer query.

SELECT t.* 
FROM(
    SELECT child.id
    FROM [nested-set] headparent
    JOIN [nested-set] parent ON (parent.lft >= headparent.lft AND parent.rgt < headparent.rgt)
    JOIN [nested-set] child ON (child.lft BETWEEN parent.lft AND parent.rgt)
    WHERE headparent.id = 3 
      AND parent.id > 1
    GROUP BY child.id
    HAVING count(*) = 1
) q
JOIN [nested-set] t ON t.id = q.id
ORDER BY t.lft ASC;

And since this is pretty much "Old Standard SQL" (fe doesn't use window functions), this should work also on MySql or SQLite.

Post scriptum:

I was wondering if it would be possible to use EXISTS for this.
And from the experiment below, it does seem so. If you don't use a GROUP BY.
But imho, the first method is a more straight forward way to pull it off.

declare @NestedSet table (id int, lft int, rgt int, colA int);

insert into @NestedSet (id,lft,rgt,colA) values
(1,3,5,100),
(2,4,5,200),
(3,3,6,300),
(4,3,5,400),
(5,4,6,500);

select *
from @NestedSet t
where exists (
    select 1
    from @NestedSet headparent
    join @NestedSet parent on (parent.lft >= headparent.lft and parent.rgt < headparent.rgt and parent.id > 1)
    where headparent.id = 3
    and t.lft between parent.lft and parent.rgt
    having count(*) = 1
);