FMeshElementIndexer

Navigation

API > API/Runtime > API/Runtime/MeshDescription

This is an efficient container for holding back references to mesh elements from attributes. Typically this will be a one-to-many relationship, where a given key has many different references.

Naively this would be implemented as an array of arrays, but we want to minimise allocations and fragmentation as much as possible. This is achieved in two ways:

1) Arrays representing the list of references are consolidated into chunks of a user-specified size. This reduces the number of individual allocations which need to be made. By default, 256 keys are consolidated into a single chunk.

2) The arrays are lazily updated. The most expensive aspect of consolidating adjacent keys' values into a single chunk of data is having to insert new space into the middle of the chunk. If a new value can't be immediately added into a chunk without inserting, the key is marked as stale and the referencer is remembered. If a stale key is queried, the indexer will be updated first, building all stale keys by getting back references from the remembered referencers.

There is also a way to fully rebuild an indexer, which will perform optimally few allocations, and collect all back references at once. It may sometimes be desirable to suspend an indexer, build a mesh description with indexing suspended, and then perform an indexer rebuild for additional speed and allocation efficiency.

Some element relationships don't lend themselves to chunking, for example polygon groups to triangles. In this case, there are generally few polygon groups compared to many triangles. Chunking can be turned off for such indexers, resulting in a more simple array representation which can be arbitrarily grown without requiring elements to be inserted.

Syntax

class FMeshElementIndexer  
Copy full snippet

class FMeshElementIndexer

Constructors

Structs

Variables

Protected

Functions

Public

Operators