Link Search Menu Expand Document
PyCSP3
Models & Data GitHub XCSP3 About

PyCSP3


PyCSP3 is a Python library for developping models of combinatorial constrained problems in a declarative manner; you can write models of constraint satisfaction (CSP) and constraint optimization (COP) problems. In this website, you will find all that you need to know about Constraint Programming (CP) and PyCSP3, with more than 60 Jupyter Notebooks:

You can also:

  • read this well-documented guide
  • explore a GitHub repository containing more than 340 models at pycsp3-models

As an illustration, below, you can find different examples of PyCSP3 models:


Last News

December 10, 2023
New GitHub repository for Models and Data; see pycsp3-models and this page.
December 10, 2023
PyCSP3 2.2, with new (control) structures 'If' and 'Match', new derivated constraint forms (Hamming, Exist, NotExist, ExactlyOne, AtLeastOne, AtMostOne and AllHold), new functions 'both' and 'either', auto-adjustment of array indexing, and the predefined named tuple 'Task'. See other extensions in Changelog Chapter of the complete guide.
August 30, 2023
2023 XCSP3 Competition, held with CP 2023; see the results.
November 10, 2022
PyCSP3 2.1, with new versions of embedded solvers (ACE and Choco) and additional constraints (Precedence, BinPacking, Knapsack, MinimumArg, MaximumArg).
November 10, 2022
Specifications of XCSP3, Version 3.1
November 10, 2022
ACE 2.1 , Version 2.1 of the generic constraint solver ACE (written in Java).
August 3, 2022
2022 XCSP3 Competition, held with FLoC 2022 (Olympic Games); see the results.
December 15, 2021
PyCSP3 2.0 , Version 2 of the Python library PyCSP3.
December 15, 2021
ACE 2.0 , Version 2 of the generic constraint solver ACE (written in Java).
January 16, 2021
Specifications of XCSP3, Version 3.0.7

Requirements

  • For building and compiling models, you need:
    • Python 3.10 (at least) to be installed.
    • Python package PyCSP3 to be installed, for example from PyPi: pip install pycsp3
  • For running the embedded solvers (ACE and Choco), you need Java 11 to be installed

Importantly, there is a complete separation between the modeling and solving phases: you write a model, you compile it (while providing some data) in order to generate an XCSP3 instance (file), and you solve that problem instance by means of a constraint solver. You can also directly pilot the solving procedure in PyCSP3 possibly conducting an incremental solving strategy.

In a nutshell, the main ingredients of the complete tool chain we propose for handling combinatorial constrained problems are:

  • PyCSP3: a Python library for modelling constrained problems, which is described in this document
  • XCSP3: an intermediate format used to represent problem instances while preserving structure of models

A shown in the figure below, the user who wishes to solve a combinatorial constrained problem has to:

  • write a model using the Python library PyCSP3 (i.e., write a Python file)
  • provide a data file (in JSON format) for a specific problem instance to be solved
  • compile both files (model and data) so as to generate an XCSP3 instance (file)
  • solve the XCSP3 file (problem instance under format XCSP3) by using a constraint solver as, e.g., ACE, Choco, OscaR or PicatSAT

modeling

This approach has many advantages:

  • Python, JSON, and XML are robust mainstream technologies
  • using JSON for data permits to have a unified notation, easy to read for both humans and computers
  • using Python 3 for modeling allows the user to avoid learning again a new programming language
  • using a coarse-grained XML structure permits to have compact and readable problem instances
  • Note that using JSON instead of XML for representing instances would have been possible but has some drawbacks, as explained in an appendix of XCSP3 Specifications.