d0f52751ce
The indent trigger was using .abs() which fired on both increased indent (non-indented → indented) AND decreased indent (indented → non-indented). This caused drop-cap style paragraphs (indented first line, flush-left continuation) to incorrectly split into two blocks. Per plan Phase 4.4 heuristic #2, indent change should only trigger when the current line is MORE indented (to the right, larger x0) than the block average - i.e., a new paragraph starting after non-indented text. It should NOT trigger for decreased indent (first line indented, rest flush-left). Fix: Remove .abs() and only check if line_x0 - block_avg_x0 > threshold. Tests: - test_indented_first_line_new_block: PASS (non-indented → indented splits) - test_indented_first_line_of_paragraph_not_split: PASS (drop cap stays together) - All 179 line module tests: PASS |
||
|---|---|---|
| .. | ||
| __pycache__ | ||
| acrobat_resave | ||
| byte_identical | ||
| content_edit_one_glyph | ||
| content_edit_one_paragraph | ||
| linearization_toggle | ||
| metadata_only | ||
| pdftk_resave | ||
| qpdf_resave | ||
| .clean_source.pdf | ||
| check_compression.py | ||
| check_trailer.py | ||
| create_fixtures.py | ||
| debug_content_streams.py | ||
| generate_fingerprint_fixtures.py | ||
| generate_fingerprint_fixtures_pikepdf.py | ||
| inspect_fixtures.py | ||
| README.md | ||
Fingerprint Reproducibility Test Fixtures
This directory contains fixture pairs that verify the fingerprint algorithm's reproducibility and content-sensitivity properties.
Fixture Provenance
All fixtures are generated from a clean source PDF (.clean_source.pdf) created using pikepdf, a Python library for PDF manipulation. The source is a 3-page PDF with Lorem Ipsum text, created with minimal metadata.
Generation
Fixtures are generated using generate_fingerprint_fixtures.py, which requires:
- Python 3.11+
pikepdflibrary (install via nix-shell or pip)
nix-shell --pure --packages python3 python3Packages.pikepdf --run \
'python3 tests/fingerprint/fixtures/generate_fingerprint_fixtures.py'
Fixture Pairs
Each fixture pair contains:
v1.pdf- Original or first variantv2.pdf- Second variant (modified copy or re-saved version)expected.txt- Either "MATCH" (fingerprints should be identical) or "DIFFER" (fingerprints should differ)
1. byte_identical
Expected: MATCH
- Same PDF copied twice (verifies fingerprint determinism)
2. acrobat_resave
Expected: MATCH
- Simulates Acrobat re-save using qpdf
- Changes
/CreationDate,/ID, and xref byte layout - Preserves content (metadata-only changes should not affect fingerprint per ADR-008)
3. pdftk_resave
Expected: MATCH
- Simulates pdftk re-save using qpdf
- Changes object stream layout and compression
- Content should produce identical fingerprint
4. qpdf_resave
Expected: MATCH
- Same source through qpdf with
--object-streams=preserve --normalize-content=y - Verifies qpdf re-save produces same fingerprint
5. linearization_toggle
Expected: MATCH (KU-7)
- Unlinearized PDF vs
qpdf --linearizeoutput - Different byte layouts but same content
- Verifies linearization independence (KU-7 requirement)
6. metadata_only
Expected: MATCH (ADR-008)
- Original vs copy with changed
/Title,/Author,/Producer,/CreationDate - Verifies metadata independence per ADR-008
7. content_edit_one_glyph
Expected: DIFFER
- "Hello World" vs "Hello Worl" (one character removed)
- Verifies content-sensitivity: removing a single glyph changes fingerprint
8. content_edit_one_paragraph
Expected: DIFFER
- Original paragraph vs variant with one word changed
- Verifies content-sensitivity: paragraph edit changes fingerprint
License
The fixture PDFs are generated using MIT-licensed tools (pikepdf, qpdf) and contain public-domain text (Lorem Ipsum). Fixtures are MIT-licensed.
References
- ADR-008: Metadata independence
- KU-7: Linearization independence
- INV-3: Fingerprint reproducibility (100 invocations produce identical results)
- INV-13: Fingerprint format (
^pdftract-v1:[0-9a-f]{64}$)