on sites like stackexchange and more there are python, tcl etc. samples of code to create quad-tree's.

 

Should be not too hard to execute in fme.

 

I don't think you need to clip it as the qaud (wich is a grid/tile anyway) has its identification. You could just select the data using a spatialrelator.

 

 

Quad tree is suppose to branch out if objects dont fit in prior level quads.

 

Also after a certain amount of levels, it becomes computational heavy. (why usualy a max level is set)

 

 

Oracle site "discourages" quad tree's as R-tree's are more efficient and faster.

 

(mind you, i hav'nt tried it out).

 

 

I was in the process of trying to implement a quad tree, haven't worked on it lately though.

 

 

I still wonder what fme uses, point relating on large sets can take a while if you dont tile them.

---

Thank you for your reply Gio,

 

 

I have been looking at r-tree's for other projects (it seems to be efficient with large sets) but for my project now, I already have the data quad-tree indexed.

 

 

My problem is on how to use a spatialrelator transformer with a quad-tree input. What kind of reader do I have to connect ?

 

 

Any idea ?

---

If assume the objects in a Quad have a attribute referring to this quad.

 

What is the format of your quad tree, just the relational tree or the actual qads with objects? Or objects with tree relations as attributes?

 

 

If you have quad geomtery u can

 

Clip existing polygons with the quads and relate the polygon data to the objects in the quads?

 

Then u could AreOnAre overly, Clipp, build Topology to relate or filter objects.

 

 

 

 

can post/show example maybe?

---