SymCXX is an experimental Python package (with its core implemented in C++) for symbolic manipulation with limited scope (as in minimal). It tries to stay compatible with the API of SymPy. Its main goal is to explore designs related to automatic differentiation and numerical evaluation similar to SymEngine's Lambdify functionality.
The capabilities include:
- Differentiation (including taking the jacobian of a matrix)
- Numerical evaluation
The above capabilities are exactly what is needed by pyodesys and pyneqsys.
Note that integers in symcxx may overflow, i.e. if correctness is important please use SymPy/SymEngine.
Currently users may refer to SymPy's / SymEngine's documentation.
Simplest way to install SymCXX and its (optional) dependencies is to use the conda package manager:
$ conda install -c bjodah symcxx pytest $ python -m pytest --pyargs symcxx
or pip (requires a C++11 compliant compiler):
$ pip install --user symcxx pytest $ python -m pytest --pyargs symcxx
There are no requirements outside the Python standard library.
Source distribution is available here: https://pypi.python.org/pypi/symcxx
Differentiation
>>> from symcxx import NameSpace
>>> ns = NameSpace()
>>> x, y = map(ns.Symbol, 'x y'.split())
>>> expr = x*y**2 - ns.tan(2*x)
>>> print(expr.diff(x))
Sub(Pow(Symbol(y), Integer(2)), Mul2(Integer(2), Add2(Integer(1), Pow(Tan(Mul2(Symbol(x), Integer(2))), Integer(2)))))Lambdify
>>> lmb = ns.Lambdify([x, y], expr)
>>> import numpy as np
>>> np.all(np.abs(lmb([1.3, 0.2]) - 0.65359661) < 1e-6)
TrueThe source code is Open Source and is released under the simplified 2-clause BSD license. See LICENSE for further details.
Contributors are welcome to suggest improvements at https://github.com/bjodah/symcxx
Björn I. Dahlgren, contact:
- gmail address: bjodah