Re: [Pbwebgui-development] Nested Set module

[Christian H. <ch...@ng...>]

On 2004-07-02, at 12.40, Martin Kamerbeek wrote:

[...]

> The mesurements for the Nested  Set module are not consistent with the=20=

> timings i posted earlier on. Please not that these values of the=20
> nested set module are taken without using the get_root method.=20
> Measuring with using get_root brings the timeing for the small tree on=20=

> about 3 miliseconds. Thats still a factor 2 difference, but it's only=20=

> off by 3 ms.
>
> The big tree with using get_root, however, has a runtime of 0.050=20
> which still differs an order of magnitude. Someone knows why this is=20=

> this way?

Try running timethese( 100, ...) or any other number. I'm not sure if=20
that will fix it but it=B4s worth a try.

Please also try the attached benchmark and see if you get the same=20
inconsistency.

[...]

>
>> If you don't have the time, I could whip up something by the weekend.
>
> If something is flawed with my method, please do so.

It looks that you are pulling the hole tree, but i could be wrong.

[...]

>> Is there a reason to keep the the "forrest" model if we are going for=20=

>> the Nested Set model?
>
> The way pages work in WebGUI force you to use a forrest model. Every=20=

> page-root is actually a seperate tree. Those are bound together with a=20=

> (non-exeistent) node with pageId=3D0. If you want to drop the forrest=20=

> model, you'll end up having one (existent) 'master' page (ie. only one=20=

> page root) under which alle the other pages are hung.

Yes, but same result could be achieved with one root. It will also make=20=

things easier when moving nodes. =46rom a performance view, the forrest=20=

model is better as it will lower the number of affected rows when=20
updating.

>>>>
>>>> Depth also comes for free, just count nodes lft between ancestors=20=

>>>> lft and rgt.
>>>
>>>
>>> Which would take a traversal of some kind.
>>
>>
>> Yes.
>>
> So it doesn't come for free, which it does when you calc the depth  a=20=

> creation/modification time and put it in the db as a property.

It does, try this:

   SELECT node.*,
          ( COUNT(node.lft) - 1 ) AS depth
     FROM tree AS node,
          tree AS ancestor
    WHERE node.lft BETWEEN ancestor.lft AND ancestor.rgt
GROUP BY node.lft
ORDER BY node.lft


--=20

Christian Hansen
nGmedia
+46 40 660 17 50