Skip to content

Latest commit

 

History

History
219 lines (153 loc) · 7.44 KB

README.rst

File metadata and controls

219 lines (153 loc) · 7.44 KB
Raw

CyLP

CyLP is a Python interface to COIN-OR’s Linear and mixed-integer program solvers (CLP, CBC, and CGL). CyLP’s unique feature is that you can use it to alter the solution process of the solvers from within Python. For example, you may define cut generators, branch-and-bound strategies, and primal/dual Simplex pivot rules completely in Python.

You may read your LP from an mps file or use the CyLP’s easy modeling facility. Please find examples in the documentation.

Docker

If you're comfortable with Docker, you can get started right away with the container available on Dockerhub that comes with CyLP pre-installed.

https://hub.docker.com/repository/docker/coinor/cylp

Otherwise, read on.

Prerequisites and installation

On Windows: Installation as a binary wheel

On Windows, a binary wheel is available and it is not necessary to install Cbc. Just do:

$ python -m pip install cylp

On Linux/macOS: Installation as a binary wheel

Binary wheels are available for Linux and some versions of OS X for some versions of Python. To see if there is a wheel available for your platform, you can browse

https://pypi.org/project/cylp/#files

or just try:

$ python -m pip install cylp

In case this fails, it is most likely that there is no wheel for your platform. In particular, there are no wheels for MacOS running on Apple Silicon. If you are on Linux, this can probably be addressed by switching to a supported Python version with, e.g., conda:

$ conda create -n cylp python=3.9
$ conda activate cylp

If all else fails, it is easy to install from source, but Cbc must be installed first, as detailed below. The easiest route for this is to use conda.

On Linux/macOS with conda: Installation from source

To install from source, you will need to install binaries for Cbc or also build Cbc from source. The version should be 2.10 (recommended) or earlier (current master branch of Cbc will not work with this version of CyLP).

The following commands will create and activate a new conda environment with all these prerequisites installed:

$ conda create -n cylp coin-or-cbc cython numpy pkg-config scipy -c conda-forge
$ conda activate cylp

Now you can install CyLP from PyPI:

$ pip install --no-build-isolation cylp

(The option --no-build-isolation ensures that cylp uses the Python packages installed by conda in the build phase.)

Alternatively, if you have cloned CyLP from GitHub:

$ pip install --no-build-isolation .

On Linux/macOS with pip: Installation from source

You will need to install binaries for Cbc. The version should be 2.10 (recommended) or earlier (current master branch of Cbc will not work with this version of CyLP). You can install Cbc by either by installing with your system's package manager, by downloading pre-built binaries, or by building yourself from source using coinbrew.

  1. To install Cbc in Linux, the easiest way is to use a package manager. Install coinor-libcbc-dev on Ubuntu/Debian or coin-or-Cbc-devel on Fedora, or the corresponding package on your distribution.

  2. On macOS, it is easiest to install Cbc with homebrew:

    $ brew install cbc pkg-config

You should no longer need to build Cbc from source on any platform unless for some reason, none of the above recipes applies to you. If you do need to build from source, please go to the Cbc project page and follow the instructions there. After building and installing, make sure to either set the COIN_INSTALL_DIR variable to point to the installation or set PKG_CONFIG_PATH to point to the directory where the .pc files are installed. You may also need to set either LD_LIBRARY_PATH (Linux) or DYLD_LIBRARY_PATH (macOS).

Next, build and install CyLP:

$ python -m pip install cylp

This will build CyLP install the runtime dependencies (install-requires), NumPy and SciPy <https://scipy.org> and build and install CyLP.

Testing your installation

Optional step:

If you want to run the doctests (i.e. make doctest in the doc directory) you should also define:

$ export CYLP_SOURCE_DIR=/Path/to/cylp

Now you can use CyLP in your python code. For example:

>>> from cylp.cy import CyClpSimplex
>>> s = CyClpSimplex()
>>> s.readMps('../input/netlib/adlittle.mps')
0
>>> s.initialSolve()
'optimal'
>>> round(s.objectiveValue, 3)
225494.963

Or simply go to CyLP and run:

$ python -m unittest discover

to run all CyLP unit tests (this is currently broken).

Modeling Example

Here is an example of how to model with CyLP's modeling facility:

import numpy as np
from cylp.cy import CyClpSimplex
from cylp.py.modeling.CyLPModel import CyLPArray

s = CyClpSimplex()

# Add variables
x = s.addVariable('x', 3)
y = s.addVariable('y', 2)

# Create coefficients and bounds
A = np.matrix([[1., 2., 0],[1., 0, 1.]])
B = np.matrix([[1., 0, 0], [0, 0, 1.]])
D = np.matrix([[1., 2.],[0, 1]])
a = CyLPArray([5, 2.5])
b = CyLPArray([4.2, 3])
x_u= CyLPArray([2., 3.5])

# Add constraints
s += A * x <= a
s += 2 <= B * x + D * y <= b
s += y >= 0
s += 1.1 <= x[1:3] <= x_u

# Set the objective function
c = CyLPArray([1., -2., 3.])
s.objective = c * x + 2 * y.sum()

# Solve using primal Simplex
s.primal()
print(s.primalVariableSolution['x'])

This is the expected output:

Clp0006I 0  Obj 1.1 Primal inf 2.8999998 (2) Dual inf 5.01e+10 (5) w.o. free dual inf (4)
Clp0006I 5  Obj 1.3
Clp0000I Optimal - objective value 1.3
[ 0.2  2.   1.1]

Documentation

You may access CyLP's documentation:

  1. Online : Please visit http://coin-or.github.io/CyLP/
  2. Offline : To install CyLP's documentation in your repository, you need Sphinx (https://www.sphinx-doc.org/). You can generate the documentation by going to cylp/doc and run make html or make latex and access the documentation under cylp/doc/build. You can also run make doctest to perform all the doctest.

Who uses CyLP

The following software packages make use of CyLP:

  1. CVXPY, a Python-embedded modeling language for convex optimization problems, uses CyLP for interfacing to CBC, which is one of the supported mixed-integer solvers.

CyLP has been used in a wide range of practical and research fields. Some of the users include:

  1. PyArt, The Python ARM Radar Toolkit, used by Atmospheric Radiation Measurement (U.S. Department of energy).
  2. Meteorological Institute University of Bonn.
  3. Sherbrooke university hospital (Centre hospitalier universitaire de Sherbrooke): CyLP is used for nurse scheduling.
  4. Maisonneuve-Rosemont hospital (L'hôpital HMR): CyLP is used for physician scheduling with preferences.
  5. Lehigh University: CyLP is used to teach mixed-integer cuts.
  6. IBM T. J. Watson research center
  7. Saarland University, Germany