Marcus Kjeldsen

Hi! I would love some guidance on the caching of HDA output in unity

The documentation for Houdini Engine for Unity outlines some considerations for version control:
https://www.sidefx.com/docs/unity20.5/_source_control.html#SourceControl_Files [www.sidefx.com]

As I understand it, the caching folder is supposedly a noisy folder where Houdini caches its output geometry and such, but when I work in Unity it seems every HDA is storing its output directly into the scene, or serializing the output in some way. The caching folder is always empty, apart from some folders named after some assets. This often lands us some challenges in terms of the github file limit of 100mb when working with work-in-progress-scenes heavily dependent on Houdini assets.

Is the caching folder still in use, or was the caching behavior an outdated approach? Or, just as likely, am I misunderstanding its function?

Any insight is greatly appreciated!

Is there a way to load the documentation for an HDA through the unity interface (running unity engine)?
I wouldn't be surprised if it is possible with the api, but I can't seem to find anything Perhaps with a python script attached to a button or something along those lines?

Any tips would be greatly appreciated!

Hi Ciscorp, this may not answer you question directly I think, but I'll chime in as best I can.

We're using/trying out multiple approaches, but for me it seems to depend on asset use case.

Things that are isolated assets (like furniture, single trees and such) we use simple fbx from Houdini proper. I use the path attribute and the fbx rop sop to control separation of geometry, which seems to work nicely and predictably.

When it is assets that are dependent on a unity scene, needs to fit with the surrounding geometry, or needs to change depending on level design, I sometimes create an actual HDA for unity. This also seems to work reliably, although it involves slightly more setup. For geometry separations in unity it depends on the use case, but usually it is one of the following approaches:

• Sop leve hdas with packed geometries where the number of sub-geometries are unknown or suitable for instancing.
  
• Sop level hdas with specific output nodes, when the number and type of outputs are predicable.
  
• Object level hdas with sops for each component, in cases where the output is predictable and the asset is more complex.
  

In cases where we do texture baking or texture compositing, we try to do it directly in the hda or use regular rops, but for our production so far we aren't doing much with textures, so there are probably better ways. Similarly, we haven't had need for LODs in our production, if that is what you are creating, so I don't have any good pointers there unfortunately.

In any case, I'm also hoping for some more indepth tutorials on the unity workflow from SideFX, and best of luck

Best,
Marcus