#!/usr/bin/env python3 from __future__ import annotations import html import re import subprocess from pathlib import Path ROOT = Path(__file__).resolve().parent SOURCE = ROOT / "paper.md" ARCHITECTURE_FIGURE = r""" \begin{figure}[!htbp] \centering \includegraphics[width=\linewidth]{assets/architecture.pdf} \caption{Canvas-4B system architecture. A Qwen3.5-4B-class native multimodal backbone receives a mode-routed visual representation. A learned global compressor preserves page structure while protected high-resolution regions retain typography and small controls. Shared training builds the common visual--code substrate; lightweight mode adapters separate literal reconstruction from redesign and open-ended generation. Render feedback is used during training and optional quality-mode inference, while the public response remains a single HTML document.} \label{fig:architecture} \end{figure} """ DATA_FIGURE = r""" \begin{figure}[!htbp] \centering \includegraphics[width=\linewidth]{assets/data-flow.pdf} \caption{Experiment-first, provenance-preserving data engine. Public research corpora, current front ends, curated design galleries, and controlled synthesis enter through deterministic build-and-render workers, multi-view deduplication, and executable quality gates. The same canonical page can yield reconstruction, redesign, prompt-generation, preference-distilled, and visual-repair records. Human designers calibrate the uncertain boundary rather than label the bulk corpus.} \label{fig:data-flow} \end{figure} """ # One entry per Markdown table, in document order. Widths sum to at most 0.94 # to leave room for inter-column padding. TABLE_WIDTHS = [ [0.12, 0.24, 0.27, 0.27], [0.13, 0.20, 0.17, 0.18, 0.22], [0.20, 0.09, 0.11, 0.15, 0.25, 0.10], [0.18, 0.21, 0.23, 0.28], [0.15, 0.17, 0.17, 0.18, 0.23], [0.19, 0.16, 0.23, 0.32], [0.18, 0.22, 0.23, 0.27], [0.13, 0.18, 0.19, 0.19, 0.21], [0.16, 0.22, 0.22, 0.30], [0.22, 0.13, 0.16, 0.17, 0.22], [0.16, 0.21, 0.17, 0.18, 0.18], [0.20, 0.12, 0.28, 0.30], [0.20, 0.16, 0.16, 0.16, 0.22], [0.18, 0.18, 0.18, 0.18, 0.18], [0.19, 0.30, 0.18, 0.23], [0.22, 0.11, 0.11, 0.11, 0.35], [0.25, 0.12, 0.13, 0.40], [0.16, 0.18, 0.18, 0.18, 0.20], [0.20, 0.30, 0.40], [0.16, 0.18, 0.18, 0.18, 0.20], [0.18, 0.24, 0.20, 0.28], [0.24, 0.14, 0.18, 0.34], [0.24, 0.20, 0.20, 0.26], ] def clean_markdown(text: str) -> str: text = html.unescape(text) text = re.sub( r'
\s*]*architecture\.svg[^>]*>' r"\s*
.*?
\s*
", "\nCAPYARCHFIGUREMARKER\n", text, flags=re.S, ) text = re.sub( r'
\s*]*data-flow\.svg[^>]*>' r"\s*
.*?
\s*
", "\nCAPYDATAFIGUREMARKER\n", text, flags=re.S, ) text = re.sub(r'
', "", text) text = re.sub( r'
(.*?)
', lambda match: "\n> " + re.sub(r"<[^>]+>", "", match.group(1)).strip() + "\n", text, flags=re.S, ) text = re.sub(r"^##\s+\d+\.\s+", "# ", text, flags=re.M) text = re.sub(r"^###\s+\d+\.\d+\s+", "## ", text, flags=re.M) text = re.sub(r"^####\s+\d+\.\d+\.\d+\s+", "### ", text, flags=re.M) text = re.sub(r"^##\s+Appendix\s+[A-Z]\.\s+", "# ", text, flags=re.M) return text.strip() + "\n" def pandoc(source: Path, destination: Path) -> None: subprocess.run( [ "pandoc", str(source), "--from=markdown+pipe_tables+fenced_code_blocks+raw_html+raw_tex+citations", "--to=latex", "--top-level-division=section", "--listings", "--natbib", "--wrap=none", "-o", str(destination), ], check=True, ) def column_spec(widths: list[float]) -> str: return "@{}" + "".join( rf">{{\raggedright\arraybackslash}}p{{{width:.3f}\linewidth}}" for width in widths ) + "@{}" def format_tables(text: str) -> str: pattern = re.compile( r"\\begin\{longtable\}\[\]\{.*?@\{\}\}", re.S, ) matches = list(pattern.finditer(text)) if len(matches) != len(TABLE_WIDTHS): raise RuntimeError( f"expected {len(TABLE_WIDTHS)} tables, found {len(matches)}" ) pieces: list[str] = [] cursor = 0 for match, widths in zip(matches, TABLE_WIDTHS, strict=True): pieces.append(text[cursor : match.start()]) pieces.append("\\begin{longtable}[]{" + column_spec(widths) + "}") cursor = match.end() pieces.append(text[cursor:]) return "".join(pieces) def normalize_longtable_headers(text: str) -> str: pattern = re.compile( r"(\\toprule\\noalign\{\}\n.*?\\midrule\\noalign\{\}\n)" r"\\endhead", re.S, ) def replace(match: re.Match[str]) -> str: header = match.group(1).replace( r"\toprule\noalign{}" + "\n", r"\toprule\noalign{}" + "\n" + r"\rowcolor{capywash}" + "\n", 1, ) return header + "\\endfirsthead\n" + header + "\\endhead" return pattern.sub(replace, text) def normalize_symbols(text: str) -> str: replacements = { "≤": r"\(\leq\)", "≥": r"\(\geq\)", "×": r"\(\times\)", "±": r"\(\pm\)", "α": r"\(\alpha\)", "κ": r"\(\kappa\)", "τ": r"\(\tau\)", "ρ": r"\(\rho\)", "→": r"\(\rightarrow\)", "–": "--", "—": "---", "≈": r"\(\approx\)", } for source, target in replacements.items(): text = text.replace(source, target) return text def prepare() -> None: source = SOURCE.read_text() body_start = source.index("## 1. Scope and recommendation") references_start = source.index("## References") appendix_start = source.index("## Appendix A.") body_markdown = clean_markdown(source[body_start:references_start]) appendix_markdown = clean_markdown(source[appendix_start:]) body_md = ROOT / "body.md" appendix_md = ROOT / "appendix.md" body_md.write_text(body_markdown) appendix_md.write_text(appendix_markdown) pandoc(body_md, ROOT / "body.raw.tex") pandoc(appendix_md, ROOT / "appendix.raw.tex") body = (ROOT / "body.raw.tex").read_text() body = body.replace("CAPYARCHFIGUREMARKER", ARCHITECTURE_FIGURE) body = body.replace("CAPYDATAFIGUREMARKER", DATA_FIGURE) body = format_tables(body) body = normalize_longtable_headers(body) body = normalize_symbols(body) section_two = body.index("\\hypertarget{literature-review-2025}") (ROOT / "introduction.tex").write_text(body[:section_two]) (ROOT / "body.tex").write_text(body[section_two:]) appendix = (ROOT / "appendix.raw.tex").read_text() appendix = normalize_longtable_headers(appendix) appendix = normalize_symbols(appendix) (ROOT / "appendix.tex").write_text(appendix) if __name__ == "__main__": prepare()