palatini-pt-spurion

Reproducible codebase for Palatini × PT-even + spurion.
This repo accompanies the manuscript and reproduces all C1/C2/C3 results and figures end-to-end.

Highlights

• ✅ Analytic claims (C1/C2/C3) organized as paper-checkable identities
• ✅ One-command figure regeneration (make paper)
• ✅ Full test suite (44 passed) and CI-ready
• ✅ Notebooks for demonstrations + Colab viewing
• ✅ Snakemake DAG of the figure pipeline (tex/snakemake_dag.pdf)
• ✅ App. D: 2×2 mixing matrix artifact + locking check（make mixing）

---

TL;DR — One command

# Option A: conda (recommended)
conda env create -f environment.yml
conda activate palpt

# Option B: venv/pip
python -m venv .venv && source .venv/bin/activate
pip install -U pip
pip install -e .[dev]

# Reproduce all paper figures & data
make paper

# Run tests exactly as used for the paper
make paper-test

Expected output (abridged):

Generated:
  [pdfs] figs/pdf/fig1_c1_pure_trace.pdf
  ...
  [pdfs] figs/pdf/fig9_flux_ratio.pdf
  [data] figs/data/*.csv, *.npz
44 passed in ~5–6s

---

Installation

# conda
conda env create -f environment.yml
conda activate palpt

# or pip/venv
python -m venv .venv && source .venv/bin/activate
pip install -U pip
pip install -e .[dev]

Check tooling:

pre-commit install && pre-commit run -a
pytest -q
palpt --help

---

Reproducing the paper artifacts

Task	Command	Output
Generate all figures & data (paper grids)	make paper	figs/pdf/fig*.pdf, figs/data/*
Run the exact paper test suite	make paper-test	44 passed
App. D mixing matrix + locking check	make mixing	figs/data/mixing_matrix.csv、figs/data/mixing_matrix_meta.json（含 .md5），並打印鎖定比 w*_ROD:w*_CM
Sync notebooks (Jupytext)	make nb	notebooks/*.py updated
Execute notebooks headless	make nb-test	notebooks executed in-place
Build the pipeline DAG	make dag	tex/snakemake_dag.pdf

Note (DAG dependencies). make dag 需要：

• Python 套件：snakemake, graphviz（py binding 已在 .[dev]）；

• 系統層的 dot 可執行檔：如 sudo apt-get install graphviz 或 brew install graphviz。

---

Google Colab (view/run the notebooks)

• 00 — sanity checks
  

• 10 — C2 symbolic demo (three-chain equivalence at quadratic order)
  

• 20 — Flux-ratio visualization (finite domain)
  

• 30 — NLO offsets and δcT2(k)\delta c_T^2(k) toy forecast
  

Colab 安裝小抄（每本筆記首格可用）

# Option 1: 安裝已公開的 GitHub 套件（最簡）
!pip -q install "git+https://github.com/ice91/palatini-pt-spurion.git"

# Option 2: clone + 本地 editable 安裝（便於開發）
!git clone https://github.com/ice91/palatini-pt-spurion.git
%cd palatini-pt-spurion
!pip -q install -e .

筆記僅依賴 numpy/sympy/matplotlib 等常見庫，Colab 原生可用；不需要 graphviz/snakemake。
在 matplotlib 標題/標籤含 LaTeX 字串時請使用 raw string，例如：r"$\\mathcal R_{X/Y}$"。

---

Command-line interface

palpt --help
palpt figs --which all --config configs/paper_grids.yaml   # 生成論文圖
palpt figs --which smoke                                    # 快速 smoke

---

Project layout

palatini-pt-spurion/
├─ palatini_pt/           # 核心邏輯（algebra / palatini / equivalence / gw / spurion / io）
├─ scripts/               # 產圖腳本（fig1–fig9 + make_all_figs + mixing_matrix_extract.py）
├─ configs/               # 掃描與係數設定（含 paper_grids.yaml，coeffs/*）
├─ figs/                  # 產出 PDF/PNG 與中間數據（含 .md5；含 mixing_matrix.csv）
├─ notebooks/             # Colab 友善展示 notebook
├─ tests/                 # 單元與整合測試（**44 tests**，含 test_mixing_matrix_extract.py）
├─ tex/                   # Snakemake DAG、TikZ 原檔
├─ Snakefile              # 圖片產製規則
├─ Makefile               # 一鍵：paper / tests / nb / dag / mixing
├─ environment.yml, pyproject.toml, CITATION.cff, LICENSE

---

Snakemake DAG

產製流程圖（需 snakemake 與系統 dot）：

make dag
# 輸出：tex/snakemake_dag.pdf

---

Reproducibility checklist (for reviewers)

• Figures regenerate with make paper under the pinned environment.

• Tests pass with make paper-test (44 tests).

• App. D: make mixing 生成 mixing_matrix.csv 與 meta，並檢查兩條混合條件的比例一致（C3 鎖定）。

• Static artifacts provided in figs/ with .md5 sidecars.

• No shell-escape required in LaTeX; figures are external PDFs.

• DAG of the pipeline: tex/snakemake_dag.pdf.

---

How to cite

See CITATION.cff. If you use this repository, please cite both the companion paper and this software snapshot.

---

License

MIT (see LICENSE).

---

Acknowledgments

This repository separates paper-checkable identities (C1/C2/C3) from computational artifacts. We thank the community for reproducibility best practices that inspired this layout.