All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Groups Pages
Common Idioms and Examples

Traversing a Stage

The Usd API provides a number of methods for fast, configurable traversal of a stage or sub-trees of prims on a stage. The two categories of traversal are recursion and iteration. Common to both is the ability to specify a "predicate" that governs which prims will actually be yielded by the API during a traversal, and which will be skipped. Predicates are conjunctions, disjunctions, and negations of a set of tests of core properties of UsdPrim that are cached for speedy access during stage population. See Prim Flags for more detail and examples.

Traversing with Recursion

UsdPrim provides a number of methods for accessing its direct children via a "range", with which one can construct very efficient recursive functions. The standard UsdPrim::GetChildren() uses the "canonical traversal predicate," which implies all active, loaded, defined, non-abstract children. UsdPrim::GetFilteredChildren() allows specification of a custom predicate.

Traversing with Iteration

UsdPrim::GetDescendants() returns a range for all of a prim's "canonical" decendants, and UsdPrim::GetFilteredDescendants() allows the range's predicate to be specified.

UsdPrimRange enables highly customized iterations, adding the ability to perform pre-and-post-order traversals, and to prune subtrees. The convenience method UsdStage::Traverse() returns a UsdPrimRange that visits all descendants of the pseudo-root, but not the pseudo-root itself.

Working With Schema Classes

Later we'll talk about how to create new schema classes...

Create vs Get Property Methods

For a UsdAttribute named "points" defined in a schema class, you will find two access methods in the schema class in C++ and python:

  • GetPointsAttr()
  • CreatePointsAttr()

What is the difference between these two methods, and when should you use each? GetPointsAttr() simply returns a UsdAttribute object, without authoring any scene description. It is therefore threadsafe, and it should be your choice in any code that may be executed in a multithreaded, "multiple readers" section. Of course, the UsdAttribute it returns has an API that includes mutating methods that will author scene description, and those methods should be avoided in multi-threaded sections. However, there is no guarantee that the UsdAttribute returned by GetPointsAttr() will be valid for use in authoring scene description, because there may not be any definition available for the attribute in the Stage's scene description, which means (for example) the Set() method might not know what the attribute's datatype is. CreatePointsAttr(), by contrast, will actually create a typed definition for the attribute in the current UsdEditTarget if no definition for the attribute yet exists - it follows that it is legal and safe to call CreatePointsAttr() even if the attribute is already defined in the current UsdEditTarget().

In general, you should use CreateXXXAttr() when authoring scene description, and GetXXXAttr() when reading/importing scene description. CreateXXXAttr() also allows you to specify a default value to assign to the attribute, which you can specify to author sparsely, which means that we will refrain from authoring the provided value iff:

  • there is currently no authored value for the attribute on the composed prim
  • the attribute has a fallback value on the composed prim, and the fallback value matches the provided value. If the prim has no composed typeName (i.e. in the authoring context, the prim is a pure override), then no attribute on the prim will possess a fallback, since fallbacks apply via the prim's typeName.

The same pattern holds for relationships, except that CreateXXXRel() does not allow specification of a default target, because relationships are list-edited, and it would be unclear whether to add or set the target when there are already authored target opinions.

USDeeper: Schema Registry of IsA Schemas. IsA schemas are registered in the UsdSchemaRegistry, which provides automatic (also directly queryable) access to the typed definition of each property listed in a schema's definition in its originating schema.usda file. The properties defined in each IsA schema are called "builtins", which means they are considered present on prims of that type even when unauthored, and builtin attributes can be provided a fallback value that will be returned by UsdAttribute::Get() when the attribute possesses no authored scene description. Therefore, when authoring builtin properties on a defined, typed prim, one is technically not required to use the CreateXXXAttr() method, because the core consults the schema registry using the prim typeName and the property name to define the property, if no definition has yet been authored. Note also that when querying the typeName of a builtin, UsdAttribute will always consult only the property definition in the schema registry, even if one has attempted to change the attribute's type in a stringer layer. However, HasA schemas record no typename on prims, nor are they present in the schema registry. So one must use CreateXXXAttr() when authoring scene description using HasA schemas. CreateXXXAttr() is also useful with IsA schemas for authoring onto an untyped, "pure override" prim that the author knows will later compose onto a typed prim of a compatible schema type (though of course, care should be exercised). However, code that consistently uses CreateXXXAttr() for authoring will be easier to maintain since it will help distinguish reader sections from writer sections.

Bool Return Values and Safe Operator Bool

Error Reporting Policy and Control