EDGAR — Experimental Design Generator and Randomiser

A modern Python implementation of EDGAR, originally developed as Excel/VBA workbooks by the Biometrics team at Rothamsted Research. This package replaces legacy macro-driven workbooks with a deterministic, reproducible, fully tested Python system.

Supported Designs

Installation

pip install edgar_design

Quick Start

Command-line

# List all design types
edgar-design list-designs

# Generate an RCB design
edgar-design generate rcb --treatment-count 8 --block-count 4 --seed 42 --format csv

# Generate an alpha design as XLSX
edgar-design generate alpha --treatment-count 30 --reps 3 --blocks-per-replicate 6 --seed 1 --format xlsx -o design.xlsx

Web interface

uvicorn edgar_design.api.main:app --reload --host 0.0.0.0 --port 8000
# Open http://localhost:8000

A read-only static preview is available at https://rotsl.github.io/edgar/.

Python API

from edgar_design.designs import generate_design

result = generate_design("rcb", treatment_count=4, block_count=3, seed=42)
for row in result.rows:
    print(row)

Key Features

• Deterministic — random.Random(seed) ensures same seed always gives same design
• 9 design types — covers the most common agricultural and industrial experiment structures
• Three interfaces — web UI, CLI (edgar-design), and Python API
• Multiple export formats — CSV, JSON, XLSX (with metadata, matching legacy workbook layout)
• 229 tests, 94–100% coverage on core modules
• No Excel required — pure Python, runs on any OS

Links

• Repository: https://github.com/rotsl/edgar
• Full documentation: https://github.com/rotsl/edgar/blob/main/README.md
• Issue tracker: https://github.com/rotsl/edgar/issues
• Original Excel workbooks: https://www.edgarweb.org.uk

License

Copyright © 2026 rotsl. MIT License.