Files
cs249r_book/binder/tests/test_pdf_compression.py

135 lines
6.2 KiB
Python

"""Exercise actual PDF preservation and reject damaged compression candidates."""
import importlib.util
import sys
from pathlib import Path
from types import SimpleNamespace
import pymupdf
import pytest
from binder.cli.checks.margin_geometry import scan_page
from binder.cli.commands import _pdf_checks
SCRIPT = Path(__file__).resolve().parents[2] / "books/shared/publish/compress_pdf.py"
spec = importlib.util.spec_from_file_location("book_pdf_compression", SCRIPT)
compression = importlib.util.module_from_spec(spec)
spec.loader.exec_module(compression)
@pytest.fixture
def manuscript(tmp_path):
source = tmp_path / "source.pdf"
with pymupdf.open() as doc:
for title in ("Chapter one", "Chapter two"):
page = doc.new_page()
page.insert_text((72, 72), title)
page.draw_rect(pymupdf.Rect(72, 100, 180, 170), color=(0, 0, 1), fill=(0, 1, 0))
bitmap = pymupdf.Pixmap(pymupdf.csRGB, (0, 0, 20, 20), False)
bitmap.clear_with(180)
page.insert_image(pymupdf.Rect(200, 100, 260, 160), stream=bitmap.tobytes("png"))
doc[0].insert_link({"kind": pymupdf.LINK_GOTO, "from": pymupdf.Rect(72, 72, 180, 90),
"page": 1, "to": pymupdf.Point(72, 72)})
doc[1].insert_link({"kind": pymupdf.LINK_URI, "from": pymupdf.Rect(72, 72, 180, 90),
"uri": "https://mlsysbook.ai/"})
doc.set_toc([[1, "One", 1], [1, "Two", 2]])
doc.set_metadata({"title": "Compression regression fixture", "author": "MLSysBook"})
doc.set_page_labels([{"startpage": 0, "prefix": "", "style": "D", "firstpagenum": 1}])
doc.embfile_add("receipt.txt", b"Constructed regression fixture attachment")
doc.save(source)
return source
def test_lossless_rewrite_preserves_actual_pdf(manuscript, tmp_path):
output = tmp_path / "compressed.pdf"
original = manuscript.read_bytes()
stats = compression.compress_lossless(manuscript, output)
with pymupdf.open(manuscript) as source, pymupdf.open(output) as candidate:
assert stats["verified_pages"] == len(source)
assert compression.verify_lossless_pdf(source, candidate) == len(source)
assert stats["original_size"] == len(original)
assert stats["final_size"] == output.stat().st_size
assert manuscript.read_bytes() == original
@pytest.mark.parametrize("damage", ["page", "text", "drawing", "link", "outline", "metadata", "geometry", "labels", "attachment"])
def test_verifier_rejects_changed_pdf(manuscript, damage):
with pymupdf.open(manuscript) as source, pymupdf.open(manuscript) as candidate:
if damage == "page":
candidate.delete_page(1)
elif damage == "text":
candidate[0].insert_text((72, 200), "Unexpected extra text")
elif damage == "drawing":
candidate[0].draw_rect(pymupdf.Rect(80, 110, 100, 130), fill=(1, 0, 0))
elif damage == "link":
candidate[0].delete_link(candidate[0].get_links()[0])
elif damage == "outline":
candidate.set_toc([])
elif damage == "metadata":
candidate.set_metadata({"title": "Wrong title"})
elif damage == "geometry":
candidate[0].set_cropbox(pymupdf.Rect(0, 0, 500, 700))
elif damage == "labels":
candidate.set_page_labels([])
elif damage == "attachment":
candidate.embfile_del("receipt.txt")
with pytest.raises(RuntimeError, match="Lossless compression changed"):
compression.verify_lossless_pdf(source, candidate)
def test_failed_verification_keeps_existing_destination(manuscript, tmp_path, monkeypatch):
output = tmp_path / "published.pdf"
output.write_bytes(b"Existing destination must survive")
def reject(*args):
raise RuntimeError("Verification rejected candidate")
monkeypatch.setattr(compression, "verify_lossless_pdf", reject)
with pytest.raises(RuntimeError, match="Verification rejected"):
compression.compress_lossless(manuscript, output)
assert output.read_bytes() == b"Existing destination must survive"
assert not list(tmp_path.glob(".pdf-lossless-*"))
def test_refuses_to_overwrite_input(manuscript):
original = manuscript.read_bytes()
with pytest.raises(ValueError, match="different"):
compression.compress_lossless(manuscript, manuscript)
assert manuscript.read_bytes() == original
def test_trim_check_finds_text_omitted_by_normal_extraction():
with pymupdf.open() as document:
page = document.new_page(width=576, height=720)
page.insert_text((100, 745), "Paragraph hidden below the page")
assert "Paragraph hidden" not in page.get_text()
findings = scan_page(page, 10)
assert any(f.issue == "trim-overflow-bottom" and f.bbox[3] > page.rect.height
for f in findings)
@pytest.mark.parametrize("printer_marks", [False, True])
def test_trim_check_uses_actual_page_geometry(printer_marks):
with pymupdf.open() as document:
page = document.new_page(width=612 if printer_marks else 576,
height=756 if printer_marks else 720)
if printer_marks:
page.set_trimbox(pymupdf.Rect(18, 18, 594, 738))
page.insert_text((100, 730 if printer_marks else 700), "Text inside the trim")
assert not any(f.issue.startswith("trim-overflow-") for f in scan_page(page, 10))
def test_existing_cover_printer_label_exemption_survives_unclipped_scan():
with pymupdf.open() as document:
page = document.new_page(width=612, height=756)
page.set_trimbox(pymupdf.Rect(18, 18, 594, 738))
page.insert_text((590, 752), "3")
assert not any(f.issue.startswith("trim-overflow-") for f in scan_page(page, 1))
def test_geometry_scan_error_is_not_silently_skipped(monkeypatch, tmp_path):
def fail_scan(*args, **kwargs):
raise ValueError("Unusable geometry result")
failed_scanner = SimpleNamespace(scan_pdf=fail_scan)
monkeypatch.setitem(sys.modules, "cli.checks.margin_geometry", failed_scanner)
monkeypatch.setitem(sys.modules, "binder.cli.checks.margin_geometry", failed_scanner)
with pytest.raises(RuntimeError, match="geometry scan failed"):
_pdf_checks.scan_margin_geometry_summary(tmp_path / "book.pdf")