Dantzig is a lightweight and concise linear programming solver suitable for small and large-scale problems alike.
Dantzig is implemented in both Rust and Python, meaning you get the expressiveness and flexibility of a Python frontend plus the raw computing speed of a Rust backend.
Dantzig supports
- A solver featuring a parametric self-dual algorithm
- Arbitrarily restricted variables, including completely unrestricted free variables
==,<=, and>=constraints- Both minimization and maximization problems
- A numerically stable LU factorization with partial pivoting routine for robust linear algebra operations
- Memory-efficient sparse matrix representations
- Modern Python type-checking
Dantzig supports Python 3.10+ and can be installed with pip.
pip install dantzig Dantzig prides itself on being both lightweight (zero-dependency) and concise. The API is designed to be extremely expressive and terse, saving you keystrokes without sacrificing clarity. To this end, Dantzig provides several short aliases for common classes and methods.
A few examples are listed below,
Var == VariableMin == MinimizeMax == MaximizeVar.free() == Variable(lb=0.0, ub=0.0)Var.nn() == Var.nonneg() == Variable(lb=0.0, ub=None)Var.np() == Var.nonpos() == Variable(lb=None, ub=0.0)
import dantzig as dz
x = dz.Variable(lb=0.0, ub=None)
y = dz.Variable(lb=0.0, ub=None)
z = dz.Variable(lb=0.0, ub=None)
soln = dz.Minimize(x + y - z).subject_to(x + y + z == 1).solve()
assert soln.objective_value == -1.0
assert soln[x] == 0.0
assert soln[y] == 0.0
assert soln[z] == 1.0Using aliases, the previous example can alternately be written
from dantzig import Min, Var
x = Var.nn()
y = Var.nn()
z = Var.nn()
soln = Min(x + y - z).st(x + y + z == 1)- Mixed integer linear programing (MILP)
- SIMD-accelerated linear algebra operations
- General optimizations to make the library competitive with
ortools - Improved documentation