# Installation There is no `pip install` yet (that comes with the packaging work, WP1 in the migration plan). For now, build DiffuseCode with its Python module and point Python at the build. ## Build On Debian or Ubuntu, install the build dependencies: ```bash sudo apt-get install -y gfortran cmake libfftw3-dev libhdf5-dev libreadline-dev \ libpng-dev libx11-dev zlib1g-dev pgplot5 python3-dev python3-numpy python3-matplotlib ``` Then, from the repository root: ```bash FC=gfortran CC=gcc CXX=g++ cmake -S . -B build -DCMAKE_BUILD_TYPE=Release \ -DDIFFUSE_PYTHON=ON -DPGPLOT_DIR=/usr/lib/pgplot5 cmake --build build -j ``` This builds `discus_suite` and the Python module `build/python/suite_python*.so`. ## Use Two settings tell Python where things are: * `DIFFUSE_PYTHON_PATH`: the folder with the compiled module (`build/python`). * the `python/` folder of the repository on the Python path, for the `diffuse` package. ```bash export DIFFUSE_PYTHON_PATH=$PWD/build/python export PYTHONPATH=$PWD/python python3 python/examples/powder_demo.py ``` In a Jupyter notebook, set them before `import diffuse`: ```python import os, sys os.environ["DIFFUSE_PYTHON_PATH"] = "/path/to/DiffuseCode/build/python" sys.path.insert(0, "/path/to/DiffuseCode/python") import diffuse as dc ``` ## How it works, and its limits This version drives DISCUS through the f2py module `suite_python`: Python writes the commands to a small macro, DISCUS runs it in the same process, and the results come back through files or KUPLOT data sets. * DISCUS holds **one structure at a time** per process. A {class}`~diffuse.Crystal` remembers how it was made, and {func}`~diffuse.powder` reads it in again if another structure was loaded since. * DISCUS output is captured, not printed. Set `DIFFUSE_ECHO=1`, or call `dc.session(echo=True)`, to see it; the last output is in `dc.session().last_output`. * A failed DISCUS command raises {class}`~diffuse.MacroError` with DISCUS's own error lines. * {func}`~diffuse.run_macro` runs any DISCUS commands that are not wrapped yet.