Spline Library for a Multitude of Programming Languages

TinySpline is a small, yet powerful library for interpolating, transforming, and querying arbitrary NURBS, B-Splines, and B├ęzier curves. The library is implemented in ANSI C (C89) and provides a wrapper for C++ along with auto-generated bindings for C#, D, Go, Java, Lua, Octave, PHP, Python, R, and Ruby.


MIT License - see the LICENSE file in the source distribution.

Getting Started

The following listing uses the ANSI C interface:

#include "tinyspline.h"
#include <stdlib.h>
#include <stdio.h>
int main(int argc, char **argv)
tsStatus status;
tsBSpline spline;
tsReal *ctrlp;
tsBSpline beziers;
tsReal *result;
/* ------------------------------------------------------------------------- */
/* TinySpline includes a powerful, system-independent, and thread-safe error
* handling system in the form of easy-to-use macros. All you need to do is to
* embed your code into TS_TRY/TS_END_TRY and use TS_CALL when calling a
* TinySpline function. Likewise, you can use any of the TS_THROW macros to
* raise an error if an external function (e.g. malloc) failed.
* Errors can be handled in TS_CATCH. TS_FINALLY contains code that is executed
* in any case, therefore being perfectly suitable for cleaning up resources.
* Yet, error handling is entirely optional. You may omit TS_TRY, TS_CALL, and
* TS_THROW and pass NULL instead of a pointer to a tsStatus object. */
spline = ts_bspline_init();
beziers = ts_bspline_init();
ctrlp = result = NULL;
TS_TRY(try, status.code, &status)
/* Create a spline... */
TS_CALL(try, status.code, ts_bspline_new(
7, /* ... consisting of 7 control points... */
2, /* ... in 2D... */
3, /* ... of degree 3... */
TS_CLAMPED, /* ... using a clamped knot vector. */
&spline, &status))
/* Setup control points of `spline`. */
TS_CALL(try, status.code, ts_bspline_control_points(
&spline, &ctrlp, &status))
ctrlp[0] = -1.75f; /* x0 */
ctrlp[1] = -1.0f; /* y0 */
ctrlp[2] = -1.5f; /* x1 */
ctrlp[3] = -0.5f; /* y1 */
ctrlp[4] = -1.5f; /* x2 */
ctrlp[5] = 0.0f; /* y2 */
ctrlp[6] = -1.25f; /* x3 */
ctrlp[7] = 0.5f; /* y3 */
ctrlp[8] = -0.75f; /* x4 */
ctrlp[9] = 0.75f; /* y4 */
ctrlp[10] = 0.0f; /* x5 */
ctrlp[11] = 0.5f; /* y5 */
ctrlp[12] = 0.5f; /* x6 */
ctrlp[13] = 0.0f; /* y6 */
TS_CALL(try, status.code, ts_bspline_set_control_points(
&spline, ctrlp, &status))
/* Evaluate `spline` at u = 0.4. */
TS_CALL(try, status.code, ts_bspline_eval(
&spline, 0.4f, &net, &status))
TS_CALL(try, status.code, ts_deboornet_result(
&net, &result, &status))
printf("x = %f, y = %f\n", result[0], result[1]);
/* Derive `spline` ... */
TS_CALL(try, status.code, ts_bspline_derive(
&spline, 1, &beziers, &status))
/* ... and subdivide it into a sequence of Bezier curves. */
TS_CALL(try, status.code, ts_bspline_to_beziers(
&beziers, &beziers, &status))
/* Evaluate `beziers` at u = 0.3. */
TS_CALL(try, status.code, ts_bspline_eval(
&beziers, 0.3f, &net, &status))
TS_CALL(try, status.code, ts_deboornet_result(
&net, &result, &status))
printf("x = %f, y = %f\n", result[0], result[1]);
if (ctrlp)
if (result)
return status.code? 1 : 0;

The same example using the C++ interface:

#include <iostream>
#include "tinysplinecpp.h"
int main(int argc, char **argv)
// Create a cubic spline with 7 control points in 2D using
// a clamped knot vector. This call is equivalent to:
// tinyspline::BSpline spline(7, 2, 3, TS_CLAMPED);
// Setup control points.
std::vector<tinyspline::real> ctrlp = spline.controlPoints();
ctrlp[0] = -1.75; // x0
ctrlp[1] = -1.0; // y0
ctrlp[2] = -1.5; // x1
ctrlp[3] = -0.5; // y1
ctrlp[4] = -1.5; // x2
ctrlp[5] = 0.0; // y2
ctrlp[6] = -1.25; // x3
ctrlp[7] = 0.5; // y3
ctrlp[8] = -0.75; // x4
ctrlp[9] = 0.75; // y4
ctrlp[10] = 0.0; // x5
ctrlp[11] = 0.5; // y5
ctrlp[12] = 0.5; // x6
ctrlp[13] = 0.0; // y6
// Evaluate `spline` at u = 0.4 using 'eval'.
std::vector<tinyspline::real> result = spline.eval(0.4).result();
std::cout << "x = " << result[0] << ", y = " << result[1] << std::endl;
// Derive `spline` and subdivide it into a sequence of Bezier curves.
tinyspline::BSpline beziers = spline.derive().toBeziers();
// Evaluate `beziers` at u = 0.3 using '()' instead of 'eval'.
result = beziers(0.3).result();
std::cout << "x = " << result[0] << ", y = " << result[1] << std::endl;
return 0;


Prebuilt Binaries

Snapshot releases are available at releases.

Compiling From Source

TinySpline uses the CMake build system to compile and package its interfaces. The following compiler suites are tested: GCC, Clang, and MSVC. In order to create the bindings, Swig (3.0.1 or later) must be available. Each binding may have further dependencies to generate the source code of the target language. The following table gives an overview:

Language Dependencies to Generate Source (Relative) Output Directory
C# csharp
D - dlang
Golang - go
Java Java Development Kit org/tinyspline
Lua Lua headers lua
Octave Octave headers octave
PHP PHP (Zend) headers * php
Python Python headers python
R R headers and RCPP r
Ruby Ruby headers ruby

The following tools are required if you want to compile and package the the source code files of the corresponding binding:

Language Required Tool(s) Output File
C# Any of: csc, mcs, dmcs, gmcs TinySpline.dll
Java javac and jar (available in JDK) tinyspline.jar

Checkout the repository and cd into it:

git clone git@github.com:msteinbeck/tinyspline.git tinyspline
cd tinyspline

Afterwards, create a build directory and cd into it:

mkdir build
cd build

Finally, run CMake and build the project:

cmake ..
cmake --build .

If you want to build a specific binding, use -DTINYSPLINE_ENABLE_<LANGUAGE> when setting up cmake (<LANGUAGE> is interface you want to build) . For example:

cmake --build .

To enable all interfaces, use -DTINYSPLINE_ENABLE_ALL_INTERFACES:

cmake --build .

You will find the resultant libraries and packages in tinyspline/build/lib.

Python 2 vs. Python 3

While generating the Python binding, Swig needs to distinguish between Python 2 and Python 3. That is, Swig uses the command line parameter -py to generate Python 2 compatible code and -py3 to generate Python 3 compatible code. Accordingly, Swig is configured depending on the Python version found by CMake during initialization. On systems with multiple versions of Python installed, CMake usually chooses the more recent one. If you want to use a specific version of Python instead, set the environment variable 'TINYSPLINE_PYTHON_VERSION' to '2' or '3'.TINYSPLINE_VERSION

The following example shows how to force CMake to use Python 2 rather than Python 3:


Install the C and C++ Libraries

The following command installs TinySpline to your system:

cmake --build . --target install

This command also installs a set of CMake config scripts and pkg-config files (for the C and C++ interface respectively). The CMake config script of the C interface exports the following variables:

The CMake config script of the C++ interface exports the same variables except that they have prefix TINYSPLINECXX, e.g. TINYSPLINECXX_INCLUDE_DIRS.

Use the CMake commands find_package(tinyspline) (C) and find_package(tinysplinecxx) (C++) to include TinySpline into your project.

Install the Bindings

Depending on your configuration, binding-related distribution files are generated within the root of your build directory. For example, the file setup.py is generated if Python is enabled. Currently, the following build tools are supported: Setuptools (Python), Maven (Java), and Luarocks (Lua).

Theoretical Backgrounds

[1] is a very good starting point for B-Splines.

[2] explains De Boor's Algorithm and gives some pseudo code.

[3] provides a good overview of NURBS with some mathematical background.

[4] is useful if you want to use NURBS in TinySpline.