views:

65

answers:

2

Given a table that represents a hierarchical tree structure and has three columns

  1. ID (Primary Key, not-autoincrementing)
  2. ParentGroupID
  3. SomeValue

I know the lowest most node of that branch, and I want to copy that to a new branch with the same number of parents that also need to be cloned.

I am trying to write a single SQL INSERT INTO statement that will make a copy of every row that is of the same main has is part one GroupID into a new GroupID.

Example beginning table:

ID | ParentGroupID | SomeValue
------------------------
1  |      -1       |    a
2  |       1       |    b
3  |       2       |    c

Goal after I run a simple INSERT INTO statement:

ID | ParentGroupID | SomeValue
------------------------
1  |      -1       |    a
2  |       1       |    b
3  |       2       |    c
4  |      -1       |    a-cloned
5  |       4       |    b-cloned
6  |       5       |    c-cloned

Final tree structure

+--a (1)
|  +--b (2)
|     +--c (3)
|
+--a-cloned (4)
|  +--b-cloned (5)
|     +--c-cloned (6)

The IDs aren't always nicely spaced out as this demo data is showing, so I can't always assume that the Parent's ID is 1 less than the current ID for rows that have parents.

Also, I am trying to do this in T-SQL (for Microsoft SQL Server 2005 and greater).

This feels like a classic exercise that should have a pure-SQL answer, but I'm too used to programming that my mind doesn't think in relational SQL.

+3  A: 

Try this, based on a query from Quassnoi's article Adjacency List vs Nested Sets: SQL Server:

WITH q AS
(
    SELECT  h.*, 1 AS level
    FROM    Table1 h
    WHERE   id = 3
    UNION ALL
    SELECT  hp.*, level + 1
    FROM    q
    JOIN    Table1 hp
    ON      hp.id = q.ParentGroupID
), q2 AS (
    SELECT
        ID,
        ParentGroupID,
        SomeValue,
        (SELECT MAX(level) FROM q) - level AS level
    FROM q
)
INSERT INTO table1
SELECT
    (SELECT MAX(ID) FROM Table1) + level + 1 AS ID,
    CASE WHEN level = 0 THEN -1
         ELSE (SELECT MAX(ID) FROM Table1) + level
    END AS ParentGroupID,
    SomeValue + '-cloned'
FROM    q2

Result when run on your test data:

ID  ParentGroupID  SomeValue  
1   -1             a          
2   1              b          
3   2              c          
4   -1             a-cloned
5   4              b-cloned
6   5              c-cloned
Mark Byers
I'm going to try that out! Thanks again Mark! I need to start catching up on posts from you and Quassnoi, since you guys always seem to have an answer!
AmoebaMan17
+2  A: 

Assming that ID is an identity column, with auto assigned values. I'm doing this off the cuff, so appologies for any syntax errors. I hope the comments make the intent clear.

/* Find all ancestors for a given ID */
WITH Ancestors(ChildID, AncestorID) AS
(
  SELECT id AS ChildID, id As AncestorID
     FROM YourTable WHERE ParentGroupID=--1
  UNION ALL
  SELECT a.ChildID, d.ParentGroupID
     FROM  Ancestors AS a INNER JOIN
       YourTable d ON d.ID = a.AncestorID
     WHERE (b1.ParentGroupID <> -1))
),
/* Insert new rows for all ancestors of a given ID and save the results, so we have access to the new ID. we also have a column for the old ID. */
IDMap(ID, ParentGroupID, SomeValue, OldID) AS
{
    // insert, using identity column assigned ID, and save the old ID
   INSERT INTO YourTable SELECT d.ParentGroupID, d.SomeValue+"-cloned", d.ID FROM YourTable d
      INNER JOIN Ancestors a ON a.ChildID = d.ID 
      WHERE a.AncestorID=<the ID to clone>
}
/* Now update the parentID in the inserted data to the new ID */
UPDATE YourTable
SET ParentGroupID = map.ID
FROM YouTable t INNER JOIN (SELECT * FROM IDMap) map
  ON t.ParentGroupID=map.OldID
mdma
I don't see any calls to `@@identity`, where do you get the it
Omu
I agree, I don't see it either. I was expecting an OUTPUT clause on the IDMap CTE. Maybe it got chomped by the late night brain fog. I'll update the query.
mdma