Python wrappers to the C++ library SymEngine, a fast C++ symbolic manipulation library.

See License section for information about wheels

pip install symengine --user

conda install python-symengine -c conda-forge

Install prerequisites.

CMake       >= 3.21
Python3     >= 3.11
SymEngine   >= 0.15.0
pip
setuptools_scm           # will be automatically downloaded by pip
scikit-build-core        # will be automatically downloaded by pip
cython      >= 0.29.24   # will be automatically downloaded by pip
cython-cmake             # will be automatically downloaded by pip

For SymEngine, only a specific commit/tag (see symengine_version.txt) is supported. The latest git master branch may not work as there may be breaking changes in SymEngine.

Python wrappers can be installed by,

pip install . -vv

If you are building SymEngine for development, you should implement the dependencies automatically installed by pip and do,

pip install -e . -vv --no-build-isolation

A few additional options to scikit-build-core are mentioned here:

pip install -e . -vv --no-build-isolation
    -Ccmake.build-type=Release|Debug                                # Set build-type for multi-configuration generators like MSVC
    -Ccmake.define.SymEngine_DIR=/path/to/symengine/install/dir     # Path to SymEngine install directory or build directory

If you intend to evaluate floating-point expressions (using lambdify), you should consider linking against LLVM. Many users might also benefit from linking against FLINT, as it is now LGPL-licensed.

In general, sudo is only required if you are installing to the default prefix (/usr/local). We recommend specifying a custom prefix (--prefix=$HOME/.local) to avoid requiring administrative privileges, which most users can do without using sudo.

If you're uncomfortable specifying the prefix manually, we suggest using Conda or installing the pre-built wheels via pip instead of building from source.

You can verify the installation of SymEngine by using the provided code snippet in this README. This snippet ensures that the installation works as expected and that basic functionality is available.

from symengine import var
x, y, z = var('x y z')
e = (x + y + z)**2
expanded_e = e.expand()
print(expanded_e)

This will output:

x**2 + y**2 + z**2 + 2*x*y + 2*x*z + 2*y*z

Note: The verification code provided above checks the functionality of SymEngine. For additional verification specific to SymEngine, please refer to the official SymEngine Python bindings repository for further tests and examples.

symengine.py is MIT licensed and uses several LGPL, BSD-3, and MIT licensed libraries.

Licenses for the dependencies of pip wheels are as follows:

• pip wheels on Unix use GMP (LGPL-3.0-or-later), MPFR (LGPL-3.0-or-later), MPC (LGPL-3.0-or-later), LLVM (Apache-2.0), zlib (Zlib), libxml2 (MIT), zstd (BSD-3-Clause), and symengine (MIT AND BSD-3-Clause).
• pip wheels on Windows use MPIR (LGPL-3.0-or-later) instead of GMP above and pthreads-win32 (LGPL-3.0-or-later) additionally.
• NumPy (BSD-3-Clause) and SymPy (BSD-3-Clause) are optional dependencies.
• Sources for these binary dependencies can be found on symengine-wheels.