Siggraph Presentation

This guide will be officially introduced at Siggraph 2023 - Houdini Hive on Wednesday, 9. of August 2023 at 11:00 AM PST.

Plugin System

Usd has a plugin system over which individual components are loaded.

Usd Pipeline Plugins

In this guide we will cover how to create these minimal set of plugins needed to setup a production ready Usd pipeline:

  • Kinds Plugin (No compiling needed): For this plugin all you need is a simple .json file that adds custom kinds. Head over to our kind section to get started.
  • Asset Resolver Plugin (Compiling needed or use pre-packed resolvers): Head over to our asset resolver section to get started. The actual code and guide is hosted here as it is a big enough topic of its own.
  • Schema Plugin (Optional, compiling needed if you want Python C++/Bindings): A schema plugin allows you to create own prim types/API schemas. This is useful when you need to often create a standardized set of attributes on prims that are relevant for your pipeline. Head over to our schemas section to get going or to our schemas overview section to get an overview of what schemas are.
  • Metadata Plugin (Optional, No compiling needed): A metadata plugin allows you to create own metadata entries, so that you don't have to use the assetInfo/customData dict fields for transporting custom metadata. Schema plugins can also specify metadata, it is limited to the prim/applied schema though, so a standalone metadata plugin allows us to make metadata available on all prims/properties regardless of prim type/schema.

Compiling against USD

As listed above, some plugins need to be compiled. Most DCCs ship with a customized USD build, where most vendors adhere to the VFX Reference Platform and only change USD with major version software releases. They do backport important production patches though from time to time. That's why we recommend using the USD build from the DCC instead of trying to self compile and link it to the DCC, as this guarantees the most stability. This does mean though, that you have to compile all plugins against each (major version) releases of each individual DCC.

Typical plugins are:

  • Schemas
  • Metadata
  • Kinds
  • Asset Resolver
  • Hydra Delegates (Render Delegates)
  • File Format Plugins (.abc/.vdb)

You can inspect if whats plugins were registered by setting the TF_DEBUG variable as mentioned in the debugging section:

export TF_DEBUG=PLUG_REGISTRATION

If you want to check via Python, you have to know under what registry the plugin is installed. There are several (info shamelessly copied from the below linked USD-CookBook page ;)):

  • KindRegistry
  • PlugRegistry
  • Sdf_FileFormatRegistry
  • ShaderResourceRegistry
  • UsdImagingAdapterRegistry

Colin Kennedy's USD-Cookbook has an excellent overview on this topic: USD Cook-Book Plugins

Plugins are detected by looking at the PXR_PLUGINPATH_NAME environment variable for folders containing aplugInfo.json file.

To set it temporarily, you can run the following in a shell and then run your Usd application:

// Linux
export PXR_PLUGINPATH_NAME=/my/cool/plugin/resources:${PXR_PLUGINPATH_NAME}
// Windows
set PXR_PLUGINPATH_NAME=/my/cool/plugin/resources:${PXR_PLUGINPATH_NAME}

If you search you Usd installation, you'll find a few of these for different components of Usd. They are usually placed in a /resources folder as a common directory convention.

Via Python you can also partially search for plugins (depending on what registry they are in) and also print their plugInfo.json file content via the .metadata attribute.

from pxr import Plug
registry = Plug.Registry()
for plugin in registry.GetAllPlugins():
    print(plugin.name, plugin.path, plugin.isLoaded)
    # To print the plugInfo.json content run:
    # print(plugin.metadata)

We can also use the plugin registry to lookup from what plugin a specific type/class (in this case a schema) is registered by:

from pxr import Plug, Tf, Usd
registry = Plug.Registry()
print(">>>>>", "Typed Schemas")
for type_name in registry.GetAllDerivedTypes(Usd.Typed):
    print(type_name)
print(">>>>>", "API Schemas")
for type_name in registry.GetAllDerivedTypes(Usd.APISchemaBase):
    print(type_name)

# For example to lookup where the "Cube" type is registered from,
# we can run:
print(">>>>>", "Cube Schema Plugin Source")
plugin = registry.GetPluginForType(Tf.Type.FindByName("UsdGeomCube"))
print(plugin.name)
print(plugin.path)
print(plugin.resourcePath)
print(plugin.metadata)