from __future__ import annotations
import csv
import hashlib
import json
import os
import random
import shutil
import subprocess
import sys
from pathlib import Path
from zipfile import ZIP_DEFLATED, ZIP_STORED, ZipFile, ZipInfo
import PIL
from PIL import Image, ImageDraw, ImageFilter, ImageFont
ROOT = Path(__file__).parent
PAGES = ROOT / "page-images"
OCR = ROOT / "raw-ocr"
OUTPUT = ROOT / "lantern-ledger-scan-to-epub.epub"
FONT_PATH = Path(
os.environ.get(
"BOOK_FIXTURE_FONT",
"/System/Library/Fonts/Supplemental/Georgia.ttf",
)
)
FIXED_TIME = (2026, 9, 19, 0, 0, 0)
TITLE = "The Lantern Ledger"
CHAPTERS = [
(
"chapter-01.xhtml",
"Chapter One: The Quay",
[
"At first light, Mara carried the brass lantern to the empty quay. The tide had erased every footprint except one narrow trail beside the warehouse door.",
"She opened the ledger and found a name written twice: once in black ink, once in blue. The difference looked small, but the harbour master treated it as a warning.",
],
),
(
"chapter-02.xhtml",
"Chapter Two: The Inventory",
[
"The lower shelf held three maps, a cracked compass, and a sealed packet. Mara recorded each object before moving it, because order mattered more than speed.",
"A page torn from the back of the ledger listed a destination beyond the northern shoal. The final line stopped after the word river, as if the writer had been interrupted.",
],
),
(
"chapter-03.xhtml",
"Chapter Three: The Crossing",
[
"By noon the fog had thinned. Mara followed the marked channel and kept the lantern covered until the boat passed the last buoy.",
"She did not know whether the repeated name described one traveller or two. She knew only that the same uncertainty should not multiply in every later copy.",
],
),
(
"chapter-04.xhtml",
"Chapter Four: The Return",
[
"The packet contained no treasure, only a corrected manifest and a short note. The blue name was the approved form; the black name had survived from an earlier draft.",
"Mara returned the ledger before dusk. She left the lantern on the quay, where the next reader could begin with the right record instead of repeating her search.",
],
),
]
MIMETYPE = "application/epub+zip"
CONTAINER = """
"""
STYLE = "body { font-family: serif; line-height: 1.55; margin: 5%; } h1 { break-before: page; }"
def sha256(path: Path) -> str:
return hashlib.sha256(path.read_bytes()).hexdigest()
def preflight() -> None:
if not FONT_PATH.is_file():
raise SystemExit(
"Missing fixture font. Set BOOK_FIXTURE_FONT to a readable TrueType "
"font path; the committed artifact used macOS Georgia.ttf."
)
if not shutil.which("tesseract"):
raise SystemExit(
"Missing tesseract executable. Install Tesseract 5.5.3 with the eng "
"traineddata package, or place a compatible tesseract on PATH."
)
languages = subprocess.run(
["tesseract", "--list-langs"],
check=True,
capture_output=True,
text=True,
).stdout.splitlines()
if "eng" not in {language.strip() for language in languages}:
raise SystemExit(
"Missing Tesseract eng traineddata. Install it before running the fixture."
)
def wrap(draw: ImageDraw.ImageDraw, text: str, font: ImageFont.FreeTypeFont, width: int) -> list[str]:
words = text.split()
lines: list[str] = []
current: list[str] = []
for word in words:
candidate = " ".join([*current, word])
if draw.textlength(candidate, font=font) <= width:
current.append(word)
else:
lines.append(" ".join(current))
current = [word]
if current:
lines.append(" ".join(current))
return lines
def render_pages() -> list[Path]:
PAGES.mkdir(parents=True, exist_ok=True)
body = ImageFont.truetype(str(FONT_PATH), 38)
heading = ImageFont.truetype(str(FONT_PATH), 50)
small = ImageFont.truetype(str(FONT_PATH), 25)
created: list[Path] = []
random.seed(20260919)
for index, (_, chapter, paragraphs) in enumerate(CHAPTERS, 1):
image = Image.new("L", (1200, 1600), 243)
pixels = image.load()
for y in range(image.height):
for x in range(image.width):
pixels[x, y] = max(224, min(252, pixels[x, y] + random.choice((-2, -1, 0, 0, 0, 1, 2))))
draw = ImageDraw.Draw(image)
draw.text((90, 55), TITLE.upper(), font=small, fill=92)
draw.line((90, 95, 1110, 95), fill=170, width=2)
y = 175
for line in wrap(draw, chapter, heading, 980):
draw.text((110, y), line, font=heading, fill=38)
y += 68
y += 45
for paragraph in paragraphs:
lines = wrap(draw, paragraph, body, 960)
# Force one source line break with a printed hyphen to demonstrate that
# OCR output requires editorial reconstruction before EPUB packaging.
if index == 2 and paragraph == paragraphs[1] and len(lines) >= 3:
lines[1] = lines[1].rstrip() + "-"
for line in lines:
draw.text((120, y), line, font=body, fill=45)
y += 55
y += 35
draw.text((585, 1515), str(index), font=small, fill=95)
image = image.rotate((-0.35, 0.25, -0.2, 0.3)[index - 1], resample=Image.Resampling.BICUBIC, fillcolor=243)
image = image.filter(ImageFilter.GaussianBlur(0.25))
path = PAGES / f"page-{index:02d}.png"
image.save(path, optimize=True)
created.append(path)
return created
def run_ocr(pages: list[Path]) -> list[Path]:
OCR.mkdir(parents=True, exist_ok=True)
outputs: list[Path] = []
for page in pages:
base = OCR / page.stem
subprocess.run(
["tesseract", str(page), str(base), "-l", "eng", "--oem", "1", "--psm", "6", "--dpi", "300"],
check=True,
capture_output=True,
text=True,
)
outputs.append(base.with_suffix(".txt"))
return outputs
def xhtml(title: str, paragraphs: list[str]) -> str:
body = "".join(f"
{paragraph}
" for paragraph in paragraphs)
return f"""
{title}