Installation

PyCCAPT requires Python >=3.9.

Recommended Conda Workflow

For most users, the smoothest setup is a conda environment with a pip install:

conda create -n pyccapt python=3.11
conda activate pyccapt
python -m pip install --upgrade pip
pip install "pyccapt[full]"

For a local checkout of this repository:

git clone https://github.com/mmonajem/pyccapt.git
cd pyccapt
conda activate pyccapt
pip install -e ".[full]"

PyPI Installation (Online)

Default installation (core dependencies only):

pip install pyccapt

Install optional dependency groups:

pip install "pyccapt[calibration]"
pip install "pyccapt[control]"
pip install "pyccapt[full]"

[full] installs both control and calibration dependency sets.

Local Development Installation

From the project root:

pip install -e ".[full]"

Module-specific editable installs:

pip install -e ".[control]"
pip install -e ".[calibration]"

Faster Local Installs

The editable (-e) step itself is fast; the time goes into resolving and downloading dependencies. A few of the scientific dependencies are heavy to resolve or build:

numba pulls in llvmlite (a large binary). If your Python is newer than the latest numba supports, pip backtracks through many versions or tries to build llvmlite from source.
tables (PyTables) and h5py depend on the HDF5 C library and build from source when no matching wheel is available.
The dependencies are unpinned, so pip downloads metadata for many candidate versions to find a set that satisfies the numpy ranges that numba, tables, and h5py each require.

Options to speed this up:

Use uv, which resolves and downloads in parallel:
```
uv pip install -e ".[full]"
```

Let conda provide the binaries, then install PyCCAPT without re-resolving them:

conda install -c conda-forge numpy scipy pandas h5py pytables numba matplotlib
pip install -e . --no-deps

Skip dependency resolution during iterative development, once the environment already has the dependencies:
```
pip install -e . --no-deps
```
Keep pip current (older resolvers backtrack more slowly):
```
python -m pip install --upgrade pip
```

Conda Installation

If PyCCAPT is available on your selected conda channel:

conda install -c conda-forge pyccapt

Build and install locally from this repository:

conda install -c conda-forge conda-build
conda build conda-recipe
conda install --use-local pyccapt

If you need both optional dependency sets in that environment:

pip install "pyccapt[full]"

You can also create predefined environments:

conda env create -f environment.yml
conda env create -f environment.full.yml

Running PyCCAPT Control

After installation:

pyccapt

Alternative entrypoint:

python -m pyccapt.control

Running Tutorials

Start JupyterLab:

jupyter lab

Then open notebooks under pyccapt/calibration/tutorials.

Testing

Run module-specific test suites:

pytest -q --run-calibration
pytest -q --run-control

Run all discoverable tests with currently installed optional dependencies:

pytest -q