Book Publishing

End-to-end book production pipeline: Markdown → Pandoc → LuaLaTeX → dual PDF (print + digital).

Scope: Inheritable skill. Covers the complete pipeline for producing professional-quality PDF books from Markdown source, including emoji handling, dual format output, and print-ready configuration.

Pipeline Architecture

Markdown Source
    ↓
Pandoc (--pdf-engine=lualatex)
    ↓
LuaLaTeX Engine
    ↓
├── Print PDF (twoside, crop marks, ISBN)
└── Digital PDF (oneside, hyperlinks, bookmarks)

Key Dependencies

Tool	Version	Purpose
Pandoc	3.x+	Markdown → LaTeX conversion
LuaLaTeX	TeX Live 2024+	PDF rendering (Unicode-native)
Twemoji	Latest	Cross-platform emoji rendering
`needspace` package	LaTeX	Orphan/widow prevention

Why LuaLaTeX: Native Unicode support (XeLaTeX works but LuaLaTeX handles emoji processing more reliably with Lua filters).

Emoji Handling (Critical)

The ZWJ Problem

Zero Width Joiner (ZWJ) sequences combine multiple emoji into one glyph. Sort order is critical:

Emoji	Codepoints	Length
👨‍👩‍👧‍👦	U+1F468 U+200D U+1F469 U+200D U+1F467 U+200D U+1F466	7
👨‍👩‍👧	U+1F468 U+200D U+1F469 U+200D U+1F467	5
👨‍👩	U+1F468 U+200D U+1F469	3
👨	U+1F468	1

CRITICAL RULE: The emoji replacement map MUST be sorted by length descending (longest sequences first). If you process 👨 before 👨‍👩‍👧‍👦, the family emoji gets partially replaced and corrupts the output.

Emoji Map File

Create an explicit emoji-map.json that controls all replacements:

{
  "metadata": {
    "version": "1.0",
    "source": "Twemoji",
    "sortOrder": "length-descending"
  },
  "emojis": [
    {
      "sequence": "👨‍👩‍👧‍👦",
      "codepoints": "1f468-200d-1f469-200d-1f467-200d-1f466",
      "image": "1f468-200d-1f469-200d-1f467-200d-1f466.png",
      "length": 7
    }
  ]
}

Rule: Never rely on automatic emoji detection. Use an explicit map file that you control and sort.

Twemoji Base64 Embedding

Embed Twemoji images directly as base64 in the LaTeX output to avoid external file dependencies:

-- Pandoc Lua filter for emoji replacement
function Str(elem)
  -- Process emoji map (length-descending order)
  for _, entry in ipairs(emoji_map) do
    if elem.text:find(entry.sequence) then
      local img = pandoc.Image("", entry.base64_data_uri)
      -- Set size to match surrounding text
      img.attributes.height = "1em"
      return img
    end
  end
end

Windows Emoji Limitation

Windows cannot natively render flag emoji (🇺🇸, 🇬🇧, etc.) in many contexts. Solutions:

Approach	Result
Twemoji replacement in PDF	Full flag rendering
HTML output with Twemoji CSS	Full flag rendering
Windows terminal/editor	Broken or missing flags

Rule: Always preview emoji-heavy content in the PDF output, not in the editor.

Dual PDF Configuration

Print Edition

# pandoc-print.yaml
pdf-engine: lualatex
variables:
  documentclass: book
  classoption:
    - twoside            # Different left/right margins
    - openright           # Chapters start on right pages
  geometry:
    - inner=2.5cm        # Binding side (wider)
    - outer=2cm
    - top=2.5cm
    - bottom=2.5cm
  fontsize: 11pt
  mainfont: "Linux Libertine O"
  monofont: "Fira Code"
  linestretch: 1.15

Digital Edition

# pandoc-digital.yaml
pdf-engine: lualatex
variables:
  documentclass: book
  classoption:
    - oneside            # Symmetric margins
  geometry:
    - margin=2cm
  fontsize: 11pt
  colorlinks: true
  linkcolor: blue
  urlcolor: blue
  toccolor: blue

Key Differences

Feature	Print	Digital
Page layout	`twoside` (inner/outer margins)	`oneside` (symmetric)
Chapters	`openright` (start on right page)	No forced page side
Links	Black text (no hyperlinks on paper)	Blue, clickable
Crop marks	Yes (for professional printing)	No
Blank pages	Inserted for chapter openings	None

Page Numbering

Front Matter vs. Body

Section	Numbering	Style
Title page	None	—
Copyright, dedication	None	—
Table of Contents	Roman (i, ii, iii)	`\pagenumbering{roman}`
Foreword, preface	Roman (continues)	—
Chapter 1+	Arabic (1, 2, 3)	`\pagenumbering{arabic}`
Appendices	Arabic (continues) or lettered	`\appendix`
Index	Arabic (continues)	—

Rule: The transition from Roman to Arabic numbering resets at page 1 for the first chapter.

LaTeX auto-generates TOC from headings. Configure depth:

\setcounter{tocdepth}{2}    % Include down to subsections
\setcounter{secnumdepth}{2} % Number down to subsections

Heading Lint Before Build

Run a heading validation pass before PDF generation:

Check	Rule	Why
No skipped levels	H1 → H2 → H3 (not H1 → H3)	TOC structure breaks
Unique within chapter	No duplicate H2 headings in same chapter	Anchor collisions
No trailing punctuation	"Getting Started" not "Getting Started."	TOC formatting
Consistent casing	Title Case or Sentence case, not mixed	Professional appearance

Orphan/Widow Prevention

The `needspace` Package

Prevents headings from appearing at the bottom of a page with no following content:

\usepackage{needspace}
\needspace{4\baselineskip}  % Ensure 4 lines available before heading

Configuration

Element	Minimum Space
Chapter title	New page (automatic in `book` class)
Section heading	4 `\baselineskip`
Subsection heading	3 `\baselineskip`
Code block	5 `\baselineskip`
Figure/table	Full figure height + caption

Build Pipeline

Complete Build Script

#!/bin/bash
# build-book.sh — Dual PDF build

set -e

# Step 1: Lint headings
echo "Linting headings..."
python scripts/lint-headings.py chapters/*.md

# Step 2: Generate emoji map (sorted by length descending)
echo "Generating emoji map..."
python scripts/build-emoji-map.py --sort-by-length-desc

# Step 3: Build print edition
echo "Building print PDF..."
pandoc chapters/*.md \
  --defaults=pandoc-print.yaml \
  --lua-filter=filters/emoji.lua \
  --lua-filter=filters/needspace.lua \
  --toc \
  --output=output/book-print.pdf

# Step 4: Build digital edition
echo "Building digital PDF..."
pandoc chapters/*.md \
  --defaults=pandoc-digital.yaml \
  --lua-filter=filters/emoji.lua \
  --toc \
  --output=output/book-digital.pdf

echo "Build complete: output/book-print.pdf, output/book-digital.pdf"

Build Validation

After building, verify:

Check	How
Page count reasonable	Compare to previous build
TOC links work	Click in PDF reader
Emoji render correctly	Visual spot-check family/flag emoji
No orphan headings	Scan for headings at page bottom
Front matter numbering	Roman numerals before Chapter 1
Chapter openings (print)	Always on right-hand page

Project Structure

book/
├── chapters/
│   ├── 00-frontmatter.md
│   ├── 01-introduction.md
│   ├── 02-foundations.md
│   └── ...
├── filters/
│   ├── emoji.lua
│   └── needspace.lua
├── scripts/
│   ├── build-emoji-map.py
│   └── lint-headings.py
├── assets/
│   ├── emoji/              # Twemoji PNGs
│   └── images/             # Book images
├── output/                 # Generated PDFs (gitignored)
├── emoji-map.json          # Explicit emoji map
├── pandoc-print.yaml       # Print edition config
├── pandoc-digital.yaml     # Digital edition config
└── build-book.sh           # Build script

book-publishing