Ballooning Stability — Multidimensional Slicer

Interactive exploration tool for tokamak ideal-ballooning stability data. Built with Panel + Plotly.

Quick Start

# Create virtual environment and install deps
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Launch the app (uses data/IdealBallooningSamples.csv by default)
panel serve gui/app.py --show --autoreload

# Or point at a different CSV
SLICER_CSV=data/kappa_delta_5k_narrow.csv panel serve gui/app.py --show --autoreload

The browser will open at http://localhost:5006/app.

Features

Slicer Tab

• 2D / 3D scatter: pick X, Y, and optional Z axes from all numeric columns
• Colour & size mapping: map any numeric column to point colour or size
• Boolean filters: toggle isapar / isbpar (Any / True / False)
• Discrete selectors: multi-select for low-cardinality columns like psi_n
• Range sliders: auto-generated for every continuous numeric column
• Axis log scale: per-axis toggle
• CSV reload: re-read data without restarting the server
• Reset controls: restore all widgets to defaults
• Download: export filtered data as CSV

Analysis Tab

• Pearson correlation ranking: horizontal bar chart showing correlation of each variable with IBMgr, updated reactively as filters change
• 1D marginal histograms: strip of histograms for all varying columns
• 2D pairplot: NxN grid of Histogram2d (off-diagonal) and Histogram (diagonal) for selected columns, with selectable columns via checkbox (defaults to top 6 by correlation)

IBMgr Generator (CLI)

• ibm/generate_ibmgr.py: grid-based geometry scans wrapping pyrokinetics
• ibm/generate_kappa_delta_scan.py: random kappa-delta scans with marginal kinetic sampling and --nice CPU throttling
• Parallel execution: --workers N flag for multiprocessing (spawn context)
• Incremental saves: partial results written to CSV during long runs; safe to interrupt with Ctrl-C
• Tested: 5,000-point κ-δ scan completed with 0 failures

Running Tests

source .venv/bin/activate
python -m pytest tests/ -v

Tests also run automatically on push/PR via GitHub Actions.

Project Structure

gui/                    # Slicer application
  app.py                  Panel app — slicer + analysis tabs
  data_utils.py           Data loading, cleaning, column classification

ibm/                    # IBMgr generation tools
  ibm_generator.py        Core library (pyrokinetics wrapper)
  generate_ibmgr.py       Grid-based scan CLI
  generate_kappa_delta_scan.py  Random scan CLI
  validate_ibmgr.py       Cross-validation script

data/                   # Datasets (committed)
  IdealBallooningSamples.csv    Original 10k dataset
  kappa_delta_5k_narrow.csv     5000 samples, δ ∈ [0.3, 0.8]
  input2.cgyro                  CGYRO template
  fixture.csv                   100-row test fixture

tests/                  # Test suite (53 tests)
  test_app.py             App tests
  test_data_utils.py      Data layer tests
  test_ibm_generator.py   Generator tests (pyrokinetics mocked)
  conftest.py             Test fixture configuration