It is one way to do it. With the approach you suggested, it means going through this process:

1. fetch an entry
2. find any children
3. for each child, call step 1

And with an additional table, the steps are?

1. fetch an entry
2. fetch all descendants of that entry (gives a collection)
3. build a tree doing ... what exactly? Probably a select to get its immediate parent.

There. :-)

update - having read the Nested Set Model of Trees - well, that's one way to do it. You have to jump through hoops still to get your data into a tree structure suitable to be displayed in a TT template. But anyways, trading simplicity and additional database calls for just one call (which I doubt would satisfy the OP's requirements) and more code complexity is something I would do only if I had proven the fact that multiple queries are the bottleneck in that application.

update 2: Thinking about it, building a tree given the parent_id in each record would probably be pretty simple, so there's no point discussing your improvement.

cheers,

_($_=" "x(1<<5)."?\n".q·/)Oo.  G°\        /
                              /\_¯/(q    /
----------------------------  \__(m.====·.(_("always off the crowd"))."·
");sub _{s./.($e="'Itrs `mnsgdq Gdbj O`qkdq")=~y/"-y/#-z/;$e.e && print}

Good point :)

However, by selecting all the descendants at once, you do have all the information to build a tree within your program. I'm guessing here, but I would think that doing it in memory would still be faster than doing it with multiple selects.

I would concede to those with more experience though :)

Clint

Update I'm assuming that by selecting all of the objects, you are also receiving the parent_id of each object, which would give you the information to build the tree.