Hierarchical Data Storage

[Trevor DeVore]

On Feb 12, 2007, at 11:26 AM, Ruslan Zasukhin wrote:

>> I thought about using double values as well.  Looking at how updates
>> are performed in the modified preorder tree traversal algorithm it
>> seems to me that adding child nodes can potentially invalidate data
>> in many records that aren't part of the current hierarchy.  If I am
>> understanding the concept correctly, this seems like a bad thing if
>> you have lots of clients working on parts of the tree.
>>
>> Using doubles to at least isolate the tree connected to each root
>> node seems like it could solve that issue for some projects.
>
> I think should exists only one root.
>
> May be I have not think deeply yet, but it seems to me if AddNode()  
> touch
> not ALL right nodes but only nearest nodes, then this improve  
> concurrency.

Let's say I have a lessons and steps table in my database.  A lesson  
is made up of one or more steps.  A step with no children means it's  
content is displayed to the user, a step with children is used to  
create a hierarchal lesson structure.

Lesson 1
	Step 1
		Step 1.1
		Step 1.2
		Step 1.3
	Step 2
		Step 2.1
		Step 2.2
Lesson 2
	Step 1
		Step 1.1
		Step 1.2

So I was picturing Leson 1.Step 1, Leson 1.Step 2, Leson 2.Step 1 as  
"root" nodes and using the algorithm on the steps table.  When I  
modify the order or number of steps in Lesson 1 I wouldn't want to  
modify relationship data in Lesson 2.  That is where the thought of  
using doubles came from.  For steps, the number to the left of the  
decimal represented the lesson and the number to the right  
represented the numbers used left left/right values.

This is the first I've really thought about the modified preorder  
tree traversal algorithm before and perhaps the above scenario isn't  
how it would be used.  Or maybe the tables would have to be designed  
differently.

Make any sense?

-- 
Trevor DeVore
Blue Mango Learning Systems - www.bluemangolearning.com
trevor at bluemangolearning.com