ldrawloader
These files implement a LDraw™ (.ldr/.mpd) file loader in C/C++ and provide some functionality to enhance the 3D mesh for rendering.
You need to have a LDraw™ Parts Library installed and pass the path at creation time of the loader.
Features
Note: API and featureset are still work in progress in this personal project.
- LdrModel and LdrPart:
- basic ldraw part/model definitions
- parts can be "fixed" (re-triangulate t-junctions, split edges on non-manifold joints, map edges to triangle topology etc.)
- you can express ldraw's primitives as procedural shapes as well (e.g various flavors of
stud.dat)
- LdrRenderModel and LdrRenderPart
- meant for rendering, handles per-vertex normal splitting, and provides resolved per-triangle material lists if required.
- optional chamferring for hard edges.
- operates on "fixed parts" (implicit or explicit)
- rudimentary multi-threading support to defer loading of parts across threads
Example
// The C api is used here, however there is also ldr::Loader which pretty much
// works the same.
// initialize library
LdrLoaderCreateInfo createInfo = {};
// while parts are not directly fixed, we will implicitly create a fixed version
// for renderparts
createInfo.partFixMode = LDR_PART_FIX_NONE;
createInfo.renderpartBuildMode = LDR_RENDERPART_BUILD_ONLOAD;
// required for chamfering
createInfo.partFixTjunctions = LDR_TRUE;
// optionally look for higher subdivided ldraw primitives
createInfo.partHiResPrimitives = LDR_FALSE;
// leave 0 to disable
createInfo.renderpartChamfer = 0.35f;
// installation path of the LDraw Part Library
createInfo.basePath = m_ldrawPath.c_str();
LdrResult result = ldrCreateLoader(&createInfo, &m_loader);
assert(result == LDR_SUCCESS);
if (singleThreaded) {
// load a model, either in the simple (typical singleThreaded) "autoResolve" mode
result = ldrCreateModel(m_loader, m_modelFilename.c_str(), LDR_TRUE, &m_scene.model);
assert(result == LDR_SUCCESS || result == LDR_WARNING_PART_NOT_FOUND);
} else {
// or by passing FALSE to autoResolve allow multi-threaded loading
result = ldrCreateModel(m_loader, m_modelFilename.c_str(), LDR_FALSE, &m_scene.model);
assert(result == LDR_SUCCESS || result == LDR_WARNING_PART_NOT_FOUND);
uint32_t numParts = ldrGetNumRegisteredParts(m_loader);
uint32_t numThreads = std::thread::hardware_concurrency();
uint32_t perThread = (numParts + numThreads - 1) / numThreads;
std::vector<LdrPartID> partIds(numParts);
std::vector<LdrResult> partResults(numParts);
std::vector<std::thread> threads(numThreads);
for (uint32_t i = 0; i < numThreads; i++)
{
threads[i] = std::thread( [&](uint32_t idx){
uint32_t offset = idx * perThread;
uint32_t numLocal = offset > numParts ? 0 : std::min(perThread, numParts - idx * perThread);
if (numLocal){
for (uint32_t p = 0; p < numLocal; p++){
partIds[offset + p] = (LdrPartID)(offset + p);
}
ldrLoadDeferredParts(m_loader, numLocal, &partIds[offset], sizeof(LdrPartID), &partResults[offset]);
}
}, i);
}
for (uint32_t i = 0; i < numThreads; i++){
threads[i].join();
}
// must do manual resolve after parts are loaded
ldrResolveModel(m_loader, m_scene.model);
}
// access the model and part details directly
for (uint32_t i = 0; i < model->num_instances; i++){
const LdrInstance& instance = model->instances[i];
const LdrPart* part = ldrGetPart(m_loader, instance.part);
const LdrRenderPart* rpart = ldrGetRenderPart(m_loader, instance.part);
...
}
// the rendermodel is self-contained
result = ldrCreateRenderModel(m_loader, m_scene.model, LDR_TRUE, &m_scene.renderModel);
// so you can delete the regular model afterwards
ldrDestroyModel(m_loader, m_scene.model);
Debugging / OpenGL Viewer
ldrawloader_viewer is a supplementary project that is used to debug this library.
Not a "Complete" Solution
LDraw™ has many features and enhancements of various complexity (animation, grouping, deformable parts etc.). Some of those features may be added over time, however the goal of the project is mostly to provide easy-to-embed loader for basic use-cases.
Furthermore, the ldraw parts are modelled with varying complexity, and are typically not watertight, contain self-intersections and often have "floating" edges not matching the triangle or quad topology. Therefore the chamfering or per-vertex normal computations may not always yield desired results. Hence there are some "hacks" with magic values in place to workaround some pathologies found in parts so far. A "proper" solution would be deriving the solids through remeshing the parts completely.
Building
The easiest is to embed the files directly, but you could also configure it as library if needed. Furthermore a C interface exists to aid binding to other languages. C++ 11 is required.
Acknowledgements
- The chamfering algorithm is derived from the Master's Thesis of Diana Algma
- The vertex merging is a modified line-sweep, with sorted vertices along a single vector and was inspired by the algorithm in http://www.assimp.org/
History
- 11/3/2024 Version 0.4:
- major overhaul
- revised non-manifold handling completely, more resililent to non manifold edges
- improve renderpart building for better vertex splitting / smoothing and improve chamfer
- chamfer no longer relies on fixing t-junctions
- 2/4/2024 Version 0.3: improve material handling and support color codes
- 1/5/2020 Version 0.2: various fixes
- 8/18/2019 Version 0.1: initial release
Other Related Projects
In case this is not what you were looking for, here are some other related projects to working with ldr files.
- Blender Importer https://github.com/TobyLobster/ImportLDraw
- three.js Loader https://threejs.org/examples/webgl_loader_ldraw.html
- weldr ldr to glTF 2.0 converter https://github.com/djeedai/weldr
- LDD to LDR conversion, for example:
LDraw™ is a trademark owned and licensed by the Estate of James Jessiman, which does not sponsor, endorse, or authorize this project.
LEGO® and the LEGO logo are registered trademarks of The Lego Group, which does not sponsor, endorse, or authorize this site. You can find the official LEGO website at http://www.lego.com.