HYDRA-1057 : GeomSubsets wireframe highlighting #148
Conversation
…ht mirror use counters
debloip-adsk
force-pushed
the
debloip/HYDRA-1057/geomsubset-highlighting
branch
from
ad05cf6 to
2ccb0b0
Compare
| const HdRetainedContainerDataSourceHandle sRefinedWireOnSurfaceDisplayStyleDataSource | ||
| = HdRetainedContainerDataSource::New( | ||
| HdLegacyDisplayStyleSchemaTokens->displayStyle, | ||
| HdRetainedContainerDataSource::New( | ||
| HdLegacyDisplayStyleSchemaTokens->reprSelector, | ||
| HdRetainedTypedSampledDataSource<VtArray<TfToken>>::New( | ||
| { HdReprTokens->refinedWireOnSurf, TfToken(), TfToken() }))); |
There was a problem hiding this comment.
Everything now uses separate prims with refinedWire only
| // Computes the mask to use for an instancer's selection highlight mirror | ||
| // based on the instancer's topology and its selections. This allows | ||
| // highlighting only specific instances in the case of instance selections. | ||
| VtBoolArray _GetSelectionHighlightMask(const HdInstancerTopologySchema& originalInstancerTopology, const HdSelectionsSchema& selections) |
There was a problem hiding this comment.
Moved to a method
|
|
||
| // Returns the overall data source for an instancer's selection highlight mirror. | ||
| // This replaces the mask data source and blocks the selections data source. | ||
| HdContainerDataSourceHandle _GetSelectionHighlightInstancerDataSource(const HdContainerDataSourceHandle& originalDataSource) |
There was a problem hiding this comment.
Moved to a method
| return VtBoolArray(nbInstances, false); | ||
| }(); | ||
|
|
||
| if (!selections.IsDefined()) { |
There was a problem hiding this comment.
This is the newly handled case and why this function and _GetSelectionHighlightInstancerDataSource was moved to a method, in order to call GetSelectionHighlightPath
| // There already exists an explicit selection highlight mirror for this child (e.g. point instance prototypes), | ||
| // so don't create a duplicate implicit one. |
There was a problem hiding this comment.
This is just to avoid overdraw, as without the check everything still works, but a highlight could be drawn more than once
| affectedOriginalPrimPaths.insert(affectedOriginalPrimPaths.end(), instancingRelatedPaths.begin(), instancingRelatedPaths.end()); | ||
| return false; // We have found an instancer, don't process its children (nested instancers will be processed through the instancing-related paths). | ||
| if (prim.primType == HdPrimTypeTokens->instancer || prim.primType == HdPrimTypeTokens->mesh) { | ||
| if (direction & SelectionHighlightsCollectionDirection::Prototypes) { |
There was a problem hiding this comment.
Splitting the instancing-related paths in two prevents us from always traversing the entire instancing graph, and inadvertently creating selection highlights for non-selected prims. Without this, selecting an instanced mesh would create a highlight mirror for the mesh, then a highlight mirror for the instancer, but in creating the highlight mirror for the instancer and collecting its instancing paths, we would collect its other prototype paths, and then create highlight mirrors for those as well, even if they weren't actually selected.
| @@ -1894,6 +1894,7 @@ void MtohRenderOverride::_PickByRegion( | |||
| pickParams.viewMatrix.Set(viewMatrix.matrix); | |||
| pickParams.projectionMatrix.Set(adjustedProjMatrix.matrix); | |||
| pickParams.collection = _renderCollection; | |||
| pickParams.collection.SetExcludePaths(_wireframeSelectionHighlightSceneIndex->GetSelectionHighlightMirrorPaths()); | |||
There was a problem hiding this comment.
We need to exclude selection highlight prims. If we don't, selecting the highlight mirror of a mesh that is trimmed by a GeomSubset selection ends up selecting the original mesh, instead of selecting the GeomSubset.
| // This method takes in a path to a prim in a selection highlight hierarchy and ensures that | ||
| // the selection highlight graph is structured properly, and that the leaf mesh prims have | ||
| // the proper display style. | ||
| void assertSelectionHighlightCorrectness( |
There was a problem hiding this comment.
Extracted and slightly tweaked from testPointInstancingWireframeHighlight.cpp
There was a problem hiding this comment.
Can't say I delved very deeply into the logic, but the tests pass, so that works for me.
As discussed, agree that the approach is the right one, and is the way to get the draw flexibility we need, but concerned about z-fighting and resulting decrease in quality. Can be addressed at a later time.
debloip-adsk
added
the
ready-for-merge
This PR adds selection highlighting for GeomSubsets.
A brief overview of the thought process behind the implementation : previously, non-instanced meshes were highlighted through the
refinedWireOnSurfdisplay style. However, this does not allow us to create a wireframe for a subset of the mesh. So the idea is to instead use the existing architecture that maintains a selection highlight mirror hierarchy, and have a mirror prim of the mesh with arefinedWiredisplay style. We can then edit this mirror mesh to only contain the topology related to the selected GeomSubsets. However, to do so, it is much simpler and more consistent to move away fromrefinedWireOnSurfentirely, and just have mesh selections also use mirror prims : this way, we can use the selection highlight mirror hierarchy for everything, and we only have one consistent (though somewhat complex) way of handling all highlights.