Possible Bug - Weird Model Behaviour

[matteo.b (Matteo)]

Hi,

I have one model acting weird - which means it sometimes does this:

when really it should do this:

This two screens are from the same exact iFrame, before and after browser refreshing. Here is what I figured out so far:

• If the model loads weird - changing parameters may fix it, or not (it appears to be random)
• If the model loads correctly - changing parameters just work as expected
• It is not limited to visualization - the downloaded output is affected
• first time this was noticed it was yesterday
• I can deduce that the problem is -most likely- happening here
  
  
• However, I am absolutely unable to replicate the problem in grasshopper. The definition just works fine.
• ShapeDiver this is the incriminated model. Other models that use the same logic seams to be fine.

If this can help, I am happy to share the file I am using as input privately.

Cheers

[pavol]

It looks like your definition might hit the outputs limit with some parameter combinations. Make sure you manage your data tree to stay within the limits, more in this Documentation article.

https://support.shapediver.com/hc/en-us/articles/360009130740-Limitations-regarding-definition-outputs

Also, check that the imported curves/polylines are not too heavy and have reasonable number of control points.

Instead of copying the curves, you can try attaching transformations to improve the performance of your model, see the video tutorial below.

Optimizing Rhino Grasshopper Definitions | Pt. 3: Send Data To Online Apps!

[matteo.b (Matteo)]

Hi Pavol,

thank you for the quick response - I did previously study limitation regarding outputs, but went through it again twice.

I am almost sure this is not the problem.

In fact, the geometry passed to the viewer (via ‘preview on’ of the relevant component) is one single polyline (which is heavy -made up of several thousand small segments- by necessity). If my understanding is correct, it should count as one single file part.

Let me try to better explain what is happening:

• The geometry passed to the viewer is ‘correct’ - in fact, the ‘faulty polyline’ (see first screenshot) is both visualized in the viewer AND reflected in the downloaded output (i.e. a text file including all the coordinates of the polyline vertex)
• What is ‘weird’ is that the very first part of the ‘faulty polyline’, should be culled by ‘cull pattern’ component at the very beginning of my definition (see definition screenshot).

The model behavior is quite weird too: either it loads funny, and then acts funny only on certain parameters configurations, OR it loads correctly and then acts correctly on all parameter configuration I could randomly try within 10 minutes.

Hopefully the picture is now clear, let me know what you think.

Cheers
Matteo

[pavol]

Polyline culled form the list might be an issue related to list order, see following forum posts:

Geometry on Shapediver is different from the Rhino

This is a common problem where a list order varies from computer to computer or between Grasshopper version as the tolerances and other factors may have affected the order of your lists. In order to keep the order stable and predictable, you have to create a geometrical or mathematical check and sort the list accordingly. More in this tutorial or video below:

Hole in model becomes solid in Shapediver

Things can go wrong if you use Deconstruct Brep and similar components as the resulting element list order may vary between computers as explained in this blog post or tutorial below: This video can also give you some ideas how to build your model in more reliable and efficient way:

If this is not the case, please share a minimal version of your definition replicating this and we will have a look.

When it comes to model loading and acting funny, do you get any errors in the browser console? Please share screenshots if you do.

[matteo.b (Matteo)]

Thank you Pavol, I checked the posts.
I was already aware of those list order issues within grasshopper.

In fact, I am performing a geometry check that culls all the objects on the wrong side of the plane.
What I do not understand is why it fails - and only sometimes!

I am trying with the minimal definition, but I am unable to replicate the problem in anyway.
In fact - I have been using the same exact logic across 10+ models to cull or dispatch curves located on different sides of a plane. Only the model I linked above has been faulty.

Since I am reasonably sure the problem must be happening at geometry check - I tried replicate the problem here (you can use the attached dxf as input):
https://app.shapediver.com/m/troubleshooting-model
troubleShooting_Shapediver.gh (12.2 KB)
myCustomShape_wall (1).dxf (1.2 MB)

But after 15 minutes of refreshing and reuploading, the problem never showed up.

Finally - it doesnt matter if the model loads funny or not, my browser console look the same either ways (same as all my other models)

Console with bug happening:

Console with bug NOT happening:

I am starting to think re-uploading the model may fix it -and i need to update some other part of the definition anyway- but it would be interesting to understand where the problem comes from.

[pavol]

@matteo.b thank you for your detailed response. I couldn’t replicate the issue with your files and at this point, I don’t have any answer what went wrong in your case but our dev team will investigate further. Let us know if you experience any unexpected behaviour.