find all the employees under one manager (was "Need Help with Query")
I have an employee table with manager id and employee ids , i need to find all the employee ids for a manager id . Each employee can be a manager in turn . So I need to find all the employees under one manager and if any of the employee is in turn a manager , i need to find the employees under him as well .
The table structure is defined and i cannot edit it .
Please let me know if we could have a single query to do this .
Well, the code won't be identical on the various platforms anyway. CTEs are supported on SQL Server 2005 and DB2, and perhaps the latest version of Sybase as well. It is a part of the SQL Standard since SQL:1999, so I would suggest using it where possible. For the rest you could create a temporary table, and use pretty much the same idea as in the CTE:
1. Insert all without a parent (top level)
2. Insert all whose having those in 1 as parent
3. Recursively insert all whose having their parent inserted, but are not inserted themselves yet.
I know, it's not very pretty code, but it works. Hope you got my idea
roac approach would work i guess , but i was looking for a solution where in a single query would do the trick , but i guess its not going to be the case . I needed this solution to improve performance .
no single query is ever going to work the same in sql server 2000, sql server 2005, oracle, db2, and sybase
if you are writing an application that "abstracts" the database layer, you are never going to achieve your result by trying to abstract the sql
instead, you need to abstract the information request, and write specific sql modules for each database -- in this case, CONNECT BY for oracle, CTEs for those that support it (it's part of the sql standard), a recursive call for other databases, etc.
Thanks to all for helping me. I think i will do the recurvise calls using Java .
You would get excellent performance using the algorithm suggested by ROAC. I'd bet it would beat any recursive algorithm hands down. Recursion requires a separate call for each item in the hierarchy, while ROAC's suggestion only requires one call for each level in the hierarchy.
If it's not practically useful, then it's practically useless.
Without editing the layout of the table you're going to have difficulty getting a work around. If you know the max number of levels then you could hack it with a single query (based on rudy's example) :
e1.name AS 'Lvl1',
e2.name AS 'Lvl2',
e3.name AS 'Lvl3',
e4.name AS 'Lvl4',
e5.name AS 'Lvl5',
FROM employee AS e1
LEFT OUTER JOIN employee AS e2 ON e2.employee_number = e1.manager_number
LEFT OUTER JOIN employee AS e3 ON e3.employee_number = e2.manager_number
LEFT OUTER JOIN employee AS e4 ON e4.employee_number = e3.manager_number
LEFT OUTER JOIN employee AS e5 ON e5.employee_number = e4.manager_number
even if there is no maximum number of levels, you could still use the 4-way self-join to show just 4 levels of subtree, with links at each bottom level node if it has any further levels below it
and of course clicking on one of the links returns up 4 lower levels below that node, so in effect it "recurses" down the tree but it is the user driving the process 4 levels at a time
usually, any tree that is large enough to have "unlimited" levels is at the same time also too large to allow the complete display of the entire tree, so recursion would be neither practical nor desirable anyway