ON_MeshTessellationCache Class Reference

#include <opennurbs_plus_meshtessellationcache.h>

Public Member Functions
	ON_MeshTessellationCache ()=default

Public Attributes
std::mutex	m_build_mutex
	Guards the one-time fill only; see the note above. More...
std::atomic< bool >	m_built { false }
ON_SimpleArray< unsigned int >	m_F_to_T
std::atomic< bool >	m_naked_built { false }
ON_SimpleArray< ON_2udex >	m_naked_edges
ON_SimpleArray< unsigned int >	m_T_to_F

Detailed Description

Internal header - NOT part of the shipped SDK. Included by opennurbs_plus_mtree.cpp, which owns the cache's lifetime, and by opennurbs_plus_meshbooleans.cpp, which owns its contents. ON_Mesh refers to this type only as an incomplete type (opennurbs_mesh.h), exactly as it does for ON_MeshTreeCache, so nothing here is visible to a plug-in and no public layout depends on it. Cached triangulation of one mesh's faces: which diagonal splits each quad, and the mesh's naked edge list. Both are pure functions of the mesh, and both cost O(face count) to derive, so the intersection routines share one copy per mesh rather than rebuilding it per call.

Held by ON_MeshTreeCache, so it is discarded exactly when the cached trees are. That is the correct trigger: the diagonal choice compares the two diagonals' lengths, so it depends on vertex POSITIONS as well as on m_F, and ON_Mesh::Transform drops the tree cache.

Created empty. The first consumer fills it under m_build_mutex and sets m_built; every later reader only reads. The encoding of m_F_to_T / m_T_to_F belongs to that consumer.

Constructor & Destructor Documentation

ON_MeshTessellationCache()

ON_MeshTessellationCache::ON_MeshTessellationCache ( )

default

Member Data Documentation

m_build_mutex

std::mutex ON_MeshTessellationCache::m_build_mutex

Guards the one-time fill only; see the note above.

NOT cached here: the ngon map. ON_Mesh::NgonMap() returns a pointer INTO the mesh's own m_NgonMap, and ON_Mesh::RemoveNgonMap() frees that array without going through the tree cache, so a copy kept across calls could dangle. It stays owned per Tessellation, exactly as it was before this cache existed.

m_built

std::atomic<bool> ON_MeshTessellationCache::m_built { false }

Stored with release once the arrays below are fully written, loaded with acquire on the unlocked fast path, so a reader either sees nothing or sees a completed build.

m_F_to_T

ON_SimpleArray<unsigned int> ON_MeshTessellationCache::m_F_to_T

m_naked_built

std::atomic<bool> ON_MeshTessellationCache::m_naked_built { false }

Derived only once some caller asks for naked edges, then kept. Tracked separately from m_built so a caller that did not ask still sees an empty list, as it did before caching.

m_naked_edges

ON_SimpleArray<ON_2udex> ON_MeshTessellationCache::m_naked_edges

m_T_to_F

ON_SimpleArray<unsigned int> ON_MeshTessellationCache::m_T_to_F