Appendices

Appendix B

Reproducing This Book

Definition

B.1 What is reproducible, and how exactly

Artifact Command (from the repository root) Reproducibility
theorem checks (small-case exhaustive) python3 experiments/check_theorems.py exact
research experiments EXP-001 … EXP-010 sh experiments/run_all.sh counted steps bit-for-bit; wall-clock fields vary
comparison with committed results python3 experiments/verify_reproduction.py exits non-zero if any counted field differs
SDK test suite cd sdk && python3 -m unittest discover -s tests exact
book examples python3 book/tools/check_examples.py every printed output must match exactly
laboratory–SDK agreement cd book && npm test exact (floating point to 10−1410^{-14})
site build with integrity checks cd book && npm install && npm run build deterministic HTML; fails on any integrity error
print edition (PDF) cd book && npm run pdf generated from the same sources via headless Chromium

Requirements: Python 3.10 or later (standard library only) and, for the book build, Node.js 18 or later with the pinned packages of book/package.json. All random inputs are generated from fixed seeds.

B.2 What reproducibility does not cover

Wall-clock times depend on the machine and are reported as secondary data only. The book’s arguments rest on counted steps under stated cost models, which are machine-independent. The environment in which the recorded results were produced is listed in REPRODUCIBILITY_REPORT.md in the repository, together with the outcome of each command.

Report an erratum or an unclear passage in this appendix