Every example here can be run two ways, and they are the same problem:

JSON and bin/PeriDEM: ./run.sh, or bin/PeriDEM -i input.json.
Python: ./run.py, which calls the problem.py next to it. That file builds the deck through the peridem interface: geometry, mesh, material, contact, boundary conditions and particle placement. Requires -DEnable_Python=ON.

Some examples also have a C++ driver (-DEnable_Examples=ON) that builds the same deck in C++. Those are the examples whose decks are compared key by key.

What each folder holds

File	Role
`problem.py`	the problem, set up in Python (`build_deck()` plus a CLI)
`run.py`	runs `problem.py` in the calling process
`run.sh`	runs `bin/PeriDEM` on the JSON deck
`input*.json`	the same problem as a deck, for `bin/PeriDEM`
`main.cpp`	C++ driver, where one exists
`runs/`	output (gitignored)

The examples

Path	Problem	C++ driver
`PeriDEM/twop_circ_contact/`	Two circles, one dropped on the other	yes
`PeriDEM/ellipse_triangle/`	Hollow ellipse split on a tip-up triangle	yes
`PeriDEM/compressive/n12/`	4×3 grain pack compressed by a plate	yes (`jha2021_comp_n50`)
`PeriDEM/compressive/n500/`	Paper N≈502 pack, two-stage settle → compress	—
`PeriDEM/attrition/sim1_rotating_cylinder/`	Thick rotating drum with a protrusion	—
`PeriDEM/attrition/sim2_thin_container/`	Thin drum, offset spin axis	—
`PeriDEM/silling_kw/`	Silling 2003 Kalthoff–Winkler impact, 2D and 3D	yes (`notched_impact_inbuilt`)
`Peridynamics/circle/`	Single disc, fixed patch and linear pull	—
`Peridynamics/rectangle/`	Single plate on a uniform in-process grid	—

Running them

cmake -S . -B build -DEnable_Python=ON -DEnable_Examples=ON -DEnable_Tests=ON
cmake --build build --target peridem_core PeriDEM -j
export PYTHONPATH="$PWD/build/python:$PYTHONPATH"
 
cd examples/Peridynamics/rectangle
./run.py                             # Python, mesh built in memory
./problem.py --help                  # per-example options
./problem.py --write-deck /tmp/input.json
../../../build/bin/PeriDEM -i /tmp/input.json -nThreads 4

run.sh still honours DECK / NP / NTHREADS for the JSON path.

Meshes

Where the mesh does not need to be a specific one, problem.py has Gmsh generate it in the calling process and writes no file. That is what --in-process-mesh and --mesh-size select.

Where a specific mesh is needed, which is the case for the attrition packs and for runs compared against archived output, problem.py reads the committed .msh. Those files, the particle-location CSVs and the setup scripts are what is checked in. runs/ and generated input*.json are gitignored.

Comparing the Python and C++ decks

python/tests/test_example_parity.py rebuilds each example's deck in Python, compares it against the deck its C++ driver writes, then runs that deck through bin/PeriDEM and through the interface and compares the nodes. See python/README.md.