PC-SAFT

pcsaft is a Cython/C++ implementation of the PC-SAFT equation of state with dipole, association, and electrolyte terms.

This repository now serves three roles at once:

• the installable pcsaft Python package
• the runtime parameter-data source for ePC-SAFT datasets
• the in-repo analysis and paper-validation workspace used to recreate figures and compare model behavior

Install

Standard installs:

pip install .

Editable installs for local development and analysis:

pip install -e . --no-build-isolation

The build requires a C++ toolchain, Cython, NumPy, and the vendored Eigen headers included in this repository.

Package Layout

• src/pcsaft/: installable runtime package, Cython/C++ sources, and packaged parameter datasets
• data/: workspace and analysis datasets that are not required for package runtime
• scripts/: validation, fitting, and paper-reproduction workflows
• tests/: core package regression tests
• docs/: user documentation and paper/reference material

Runtime Parameter Data

The runtime dataset loader lives in pcsaft.parameters, and the package bundles the dataset files needed at runtime.

Development Workflow

For local work inside this repo, use an editable install and rerun it when package metadata or compiled sources change:

python scripts/build_pcsaft.py
python run_pytest.py

python scripts/build_pcsaft.py now skips the reinstall when the editable build is already current. Use python scripts/build_pcsaft.py --force to force a fresh editable reinstall, or python run_pytest.py --force-build to force that rebuild before tests.

Analysis and validation scripts are expected to run from the repo-named PC-SAFT Conda environment with pcsaft installed editable. A source checkout by itself is not a supported package-import path.

The paper-validation and exploratory analysis scripts intentionally remain outside pytest; they are workspace tools, not package-unit tests.

Codex / Agent Workflow

For Codex agents and other automated local tooling on Windows, this repo overrides the generic "default to conda run" policy: prefer the repo-local environment wrapper first.

pwsh -NoProfile -ExecutionPolicy Bypass -File scripts/run_in_repo_env.ps1 scripts/codex_doctor.py
pwsh -NoProfile -ExecutionPolicy Bypass -File scripts/run_in_repo_env.ps1 run_pytest.py

The wrapper resolves the repo-named Conda environment's python.exe directly and only falls back to conda run if needed, which avoids some Windows temp-file issues seen with conda run.

Use raw conda run -n PC-SAFT ... only when you already know it is healthy in the current session. In that case, these are equivalent:

conda run -n PC-SAFT python scripts/codex_doctor.py
conda run -n PC-SAFT python run_pytest.py

scripts/codex_doctor.py reports which checkout pcsaft is importing from, whether that checkout is compatible with the current repo, and whether a rebuild is needed before testing.

scripts/build_pcsaft.py now treats a same-commit editable install from another worktree as compatible for testing, so agents do not need to reinstall the package just because the active environment still points at a sibling checkout with the same git HEAD.

When an editable reinstall is required, the build helper uses repo-local pip temp/cache directories under build/ instead of machine-global temp paths.

MIAC Data Workflow

Canonical MIAC experimental datasets live under data/MIAC.

• Canonical MIAC CSVs: data/MIAC/<solvent_system>/*.csv
• Canonical MIAC fit outputs: data/MIAC/<solvent_system>/miac_fits/*.png
• Canonical MIAC_m fit outputs: data/MIAC/<solvent_system>/miac_m_fits/*.png

Regenerate the MIAC figures with:

conda run -n PC-SAFT python scripts\\fits\\validate_miac_fits.py

The legacy data/MIAC_m, data/MIAC_combined, and data/MIAC/*/plot_fits layouts are no longer part of the supported workflow.