The fastest and most intuitive library to manipulate STL files (stereolithography) for C++ and Python, header-only.
pip install openstl
Discover the staggering performance of OpenSTL in comparison to numpy-stl, meshio and stl-reader, thanks to its powerful C++ backend. See benchmark.py. Benchmark performed on an Intel i5-9600KF CPU @ 3.70GHz.
Performance gains over numpy-stl, meshio and stl-reader
#openstl vs numpy-stl
Write: OpenSTL is 1.262 to 5.998 X faster than numpy-stl
Read: OpenSTL is 2.131 to 11.144 X faster than numpy-stl
Rotate: OpenSTL is 0.971 to 13.873 X faster than numpy-stl
Rotate: OpenSTL + PyTorch is 0.022 to 100.25 X faster than numpy-stl
#openstl vs meshio
Write: OpenSTL is 4.289 to 80.714 X faster than meshio
Read: OpenSTL is 15.915 to 311.365 X faster than meshio
#openstl vs stl_reader
Read: OpenSTL is 0.719 to 2.2 X faster than stl_reader
Note: meshio has no specific way of rotating vertices, so it was not benchmarked.
pip install openstl or pip install -U git+https://github.com/Innoptech/OpenSTL@main
import openstl
import numpy as np
# Define an array of triangles
# Following the STL standard, each triangle is defined with : normal, v0, v1, v2
quad = np.array([
# normal, vertices 0, vertices 1, vertices 2
[[0.0, 0.0, 1.0], [0.0, 0.0, 0.0], [1.0, 0.0, 0.0], [1.0, 1.0, 0.0]], # Triangle 1
[[0.0, 0.0, 1.0], [0.0, 0.0, 0.0], [0.0, 1.0, 0.0], [1.0, 1.0, 0.0]], # Triangle 2
])
# Serialize the triangles to a file
success = openstl.write("quad.stl", quad, openstl.format.binary) # Or openstl.format.ascii (slower but human readable)
if not success:
raise Exception("Error: Failed to write to the specified file.")
# Deserialize triangles from a file
deserialized_quad = openstl.read("quad.stl")
# Print the deserialized triangles
print("Deserialized Triangles:", deserialized_quad)import openstl
import numpy as np
quad = openstl.read("quad.stl")
# Rotating
rotation_matrix = np.array([
[0,-1, 0],
[1, 0, 0],
[0, 0, 1]
])
rotated_quad = np.matmul(rotation_matrix, quad.reshape(-1,3).T).T.reshape(-1,4,3)
# Translating
translation_vector = np.array([1,1,1])
quad[:,1:4,:] += translation_vector # Avoid translating normals
# Scaling
scale = 1000.0
quad[:,1:4,:] *= scale # Avoid scaling normalsimport openstl
# Define an array of triangles
triangles = [
# normal, vertices 0, vertices 1, vertices 2
[[0.0, 0.0, 1.0], [0.0, 0.0, 0.0], [1.0, 0.0, 0.0], [1.0, 1.0, 0.0]], # Triangle 1
[[0.0, 0.0, 1.0], [0.0, 0.0, 0.0], [0.0, 1.0, 0.0], [1.0, 1.0, 0.0]], # Triangle 2
]
# Convert triangles to vertices and faces
vertices, faces = openstl.convert.verticesandfaces(triangles)import openstl
# Define vertices and faces
vertices = [
[0.0, 0.0, 0.0],
[1.0, 1.0, 1.0],
[2.0, 2.0, 2.0],
[3.0, 3.0, 3.0],
]
faces = [
[0, 1, 2], # Face 1
[1, 3, 2] # Face 2
]
# Convert vertices and faces to triangles
triangles = openstl.convert.triangles(vertices, faces)import openstl
# Deserialize triangles from a file
triangles = openstl.read("disjoint_solids.stl")
# Convert triangles to vertices and faces
vertices, faces = openstl.convert.verticesandfaces(triangles)
# Identify connected components of faces
connected_components = openstl.topology.find_connected_components(vertices, faces)
# Print the result
print(f"Number of connected components: {len(connected_components)}")
for i, component in enumerate(connected_components):
print(f"Faces of component {i + 1}: {component}")import openstl
import torch
quad = torch.Tensor(openstl.read("quad.stl")).to('cuda')
# Rotating
rotation_matrix = torch.Tensor([
[0,-1, 0],
[1, 0, 0],
[0, 0, 1]
]).to('cuda')
rotated_quad = torch.matmul(rotation_matrix, quad.reshape(-1,3).T).T.reshape(-1,4,3)
# Translating
translation_vector = torch.Tensor([1,1,1]).to('cuda')
quad[:,1:4,:] += translation_vector # Avoid translating normals
# Scaling
scale = 1000.0
quad[:,1:4,:] *= scale # Avoid scaling normalsTo read STL file with a large triangle count > 1 000 000, the openstl buffer overflow safety must be unactivated with
openstl.set_activate_overflow_safety(False) after import. Deactivating overflow safety may expose the application
to a potential buffer overflow attack vector since the stl standard is not backed by a checksum.
This can cause significant risks if openstl (and any other STL reader) is used as part of a service in a backend server for example. For
domestic usage, ignore this warning. OpenSTl is the only stl reader to provide such default safety feature.
#include <openstl/core/stl.h>
std::ifstream file(filename, std::ios::binary);
if (!file.is_open()) {
std::cerr << "Error: Unable to open file '" << filename << "'" << std::endl;
}
// Deserialize the triangles in either binary or ASCII format
std::vector<openstl::Triangle> triangles = openstl::deserializeStl(file);
file.close();std::ofstream file(filename, std::ios::binary);
if (!file.is_open()) {
std::cerr << "Error: Unable to open file '" << filename << "'" << std::endl;
}
std::vector<openstl::Triangle> originalTriangles{}; // User triangles
openstl::serialize(originalTriangles, file, openstl::StlFormat::Binary); // Or StlFormat::ASCII
if (file.fail()) {
std::cerr << "Error: Failed to write to file " << filename << std::endl;
} else {
std::cout << "File " << filename << " has been successfully written." << std::endl;
}
file.close();std::stringstream ss;
std::vector<openstl::Triangle> originalTriangles{}; // User triangles
openstl::serialize(originalTriangles, ss, openstl::StlFormat::Binary); // Or StlFormat::ASCIIusing namespace openstl
std::vector triangles = {
// normal, vertices 0, vertices 1, vertices 2
Triangle{{0.0f, 0.0f, 1.0f}, {1.0f, 1.0f, 1.0f}, {2.0f, 2.0f, 2.0f}, {3.0f, 3.0f, 3.0f}},
Triangle{{0.0f, 0.0f, 1.0f}, {2.0f, 2.0f, 2.0f}, {3.0f, 3.0f, 3.0f}, {4.0f, 4.0f, 4.0f}}
};
const auto& [vertices, faces] = convertToVerticesAndFaces(triangles);using namespace openstl
std::vector vertices = {
Vec3{0.0f, 0.0f, 0.0f}, Vec3{1.0f, 1.0f, 1.0f}, Vec3{2.0f, 2.0f, 2.0f}, Vec3{3.0f, 3.0f, 3.0f}
};
std::vector<Face> faces = {
{0, 1, 2}, {3, 1, 2}
};
const auto& triangles = convertToTriangles(vertices, faces);using namespace openstl;
// Convert to vertices and faces
const auto& [vertices, faces] = convertToVerticesAndFaces(triangles);
// Find connected components
const auto& connected_components = findConnectedComponents(vertices, faces);
std::cout << "Number of connected components: " << connected_components.size() << "\\n";
for (size_t i = 0; i < connected_components.size(); ++i) {
std::cout << "Component " << i + 1 << ":\\n";
for (const auto& face : connected_components[i]) {
std::cout << " {" << face[0] << ", " << face[1] << ", " << face[2] << "}\\n";
}
}Include this repository with CMAKE Fetchcontent and link your executable/library to openstl::core library.
Choose weither you want to fetch a specific branch or tag using GIT_TAG. Use the main branch to keep updated with the latest improvements.
include(FetchContent)
FetchContent_Declare(
openstl
GIT_REPOSITORY https://github.com/Innoptech/OpenSTL.git
GIT_TAG main
GIT_SHALLOW TRUE
GIT_PROGRESS TRUE
)
FetchContent_MakeAvailable(openstl)Simply add stl.h to your codebase.
git clone https://github.com/Innoptech/OpenSTL
mkdir OpenSTL/build && cd OpenSTL/build
cmake -DOPENSTL_BUILD_TESTS=ON .. && cmake --build .
ctest .C++17 or higher.
The STL format is simple and widely used, but that simplicity brings important limitations:
- No validation: STL files include no checksums, hashes, or structure verification, so corruption (e.g., truncated or malformed data) often goes undetected until parsing.
- Sensitive to corruption: Errors during download, storage, or editing may only fail at runtime, causing crashes or undefined behavior.
- Security concerns: Without built-in bounds checks, malformed STL files can potentially trigger buffer overflows, especially risky when handling untrusted input.
Because STL offers no internal protection, applications must implement their own validation and error-handling when loading these files.