pyparaspace: a Python wrapper for the ParaSpace timelines planner -- a simple, flexible and extensible solver for timeline-based planning problems using the Z3 Theorem Prover.

Installing

Install using Pip.

pip install pyparaspace

See the file testPyParaspace.py for example usage.

Building locally

Requirements: Rust, Cargo, Clang/LLVM/LibClang, CMake.

• Create a virtual environment

python3 -m venv env
source env/bin/activate

• Install maturin

pip install maturin

• Build package

maturin develop

Building and releasing

This section is intended for package maintainers. The pyparaspace package is released on PyPi with Python wheel packages that make it convenient to use paraspace without needing to set up Rust and C++ compilers and tools. Through the z3-sys package's static link option, we get the whole planner, including the Z3 solver, statically linked. This greatly increases the convenience for users of the library.

Windows and Manylinux platforms are currently supported.

Windows

If building and installing the local package works, then using maturin build --release should also correctly build a wheel package, which can be uploaded to PyPi using maturin publish.

Because the z3-rs feature "static-link-z3" is used, the Z3 library is built during the building of paraspace. To make sure that C++ compiler paths are correctly set up, use the "x64 Native Tools Command Prompt" that is installed by Visual Studio. If you get compilation errors after trying to build in a different environment, the z3-rs package doesn't always correctly rebuild the right parts, so it might be necessary to explicitly clean the build cache (cargo clean or git clean).

Note that building paraspace with the statically linked Z3 can take some time, typically 10-20 minutes.

When in the correct environment, publishing a new release should be possible with:

maturin publish --skip-existing

Manylinux

paraspace requires an Rust version 1.60 and Clang version 3.5 (to compile the Z3 solver), which makes it requires a bit of setup to correcly build the manylinux wheel. There is a Dockerfile available that can be used to build a Docker image with an up-to-date Rust version and version 7 of the LLVM/Clang toolchain.

The builds should work using the following commands.

docker build -t mybuild .
docker run --rm -v $(pwd):/io mybuild publish --skip-existing --compatibility manylinux2014 -i python3.10