Component to split a list based on data type?

Replies to This Discussion

Permalink Reply by Nathan Miller on June 6, 2012 at 12:52pm

I have a deserialization component in Slingshot! that does basically this under the hood.  

I'll see about getting a more general purpose version up that will just sort objects as you describe....

-Nate

Permalink Reply by Andrew Heumann on June 6, 2012 at 3:01pm

While we wait for Nathan to do the job properly :) here's my take... one easily extensible GH version and a scripted version for the 5 types you mentioned. 

Attachments:

• Sort_By_Type.gh, 14 KB

Permalink Reply by Marc Syp on June 6, 2012 at 4:28pm

I had a feeling you'd whip something up.  Glad you took the bait.  :)

Incidentally, I had the hacky way of doing it working, but thought a scripted version might be more efficient.  Looks like it actually uses more computational cycles.  Any idea why that might be?

Marc

Permalink Reply by Andrew Heumann on June 6, 2012 at 4:46pm

Oh no... you've figured out that I can't resist a little scripting problem :-p

It's weird though... on my machine the scripted version is markedly faster, with the "Contains" function as the bottleneck of the GH-only approach. Not sure why you'd be seeing something different. 

Permalink Reply by Marc Syp on June 7, 2012 at 11:04am

I think there must be some kind of one-time start-up penalty for the scripting components.  The first time it ran I was getting 187ms on the scripting component, but I can't repeat that now.  Anyway, after playing with it for a bit, it turns out that the C# script maintains complex data tree structures while the hacky way does not (without some serious gymnastics).

So anyway, I am now trying to add text objects to this list... no success thus far.  My modifications below (not working):

List<Rhino.Geometry.Point> Pts = new List<Rhino.Geometry.Point>();
List<Curve> Crvs = new List<Curve>();
List<Mesh> Meshes = new List<Mesh>();
List<Brep> Breps = new List<Brep>();
List<Brep> Srfs = new List<Brep>();
List<String> Txts = new List<String>();

foreach(GeometryBase x in geo){
Rhino.DocObjects.ObjectType objectType = x.ObjectType;
switch(objectType){
case Rhino.DocObjects.ObjectType.TextEntity: Txts.Add(x as String);
break;
case Rhino.DocObjects.ObjectType.Point: Pts.Add(x as Rhino.Geometry.Point);
break;
case Rhino.DocObjects.ObjectType.Curve: Crvs.Add(x as Curve);
break;
case Rhino.DocObjects.ObjectType.Mesh: Meshes.Add(x as Mesh);
break;
case Rhino.DocObjects.ObjectType.Brep:
Brep brepObj = x as Brep;
if(brepObj.IsSurface){
Srfs.Add(brepObj);
} else {
Breps.Add(brepObj);
}
break;

}
}
Pt = Pts;
Crv = Crvs;
Mesh = Meshes;
Brep = Breps;
Srf = Srfs;
Txt = Txts;

Any suggestions?  Is it perhaps because TextEntity is not a member of GeometryBase?  Also, I wouldn't mind having an "Other" category to pick up things like shaders/materials, etc.  I had that working in the GH environment but again, the scripted version is better for a number of reasons.

Thanks,

Marc

Permalink Reply by David Rutten on June 7, 2012 at 12:03pm

The first time a new script runs it will need to be compiled. This is quite a performance hit. But, it only happens the first time. It should run much faster after that.

There is also a penalty the first time you run a regular component as the JIT compiler needs to consume the MSIL code and generate machine code. If you want to see more realistic times on the profiler then you should press F5 to recompute the entire file without any startup overheads.

--

David Rutten

david@mcneel.com

Poprad, Slovakia