jedarden/pdftract
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
pdftract/notes/bf-4zuss.md
jedarden d5d7249ea6 docs(bf-4zuss): add corpus manifest verification note 
Verify all acceptance criteria for grep-corpus manifest and validation:
- manifest.csv exists with proper schema (1000 entries)
- validate-corpus.sh is executable and functional
- make validate-corpus runs successfully (1000/1000 files valid)
- Corpus contains 10,590 pages across 6.5MB of synthetic PDFs

Closes bf-4zuss.
2026-07-06 08:53:01 -04:00

3.1 KiB
Raw Blame History

bf-4zuss: Corpus manifest and validation

Summary

Verified that the grep-corpus manifest and validation infrastructure is fully operational.

Acceptance Criteria Verification

PASS: tests/fixtures/grep-corpus/manifest.csv exists with proper schema

  • Location: /home/coding/pdftract/tests/fixtures/grep-corpus/manifest.csv
  • Schema: filename,source_url,page_count,file_size,checksum,license
  • Sample entry: synthetic_1000.pdf,synthetic-generation,5,3751,<sha256>,public-domain

PASS: scripts/validate-corpus.sh exists and is executable

  • Location: /home/coding/pdftract/scripts/validate-corpus.sh
  • Permissions: -rwxr-xr-x (executable)
  • Lines: 183 lines of bash validation logic

PASS: make validate-corpus runs validation successfully

$ make validate-corpus
Validating grep-corpus...
[INFO] Validated 200 files...
[INFO] Validated 400 files...
[INFO] Validated 600 files...
[INFO] Validated 800 files...
[INFO] Validated 1000 files...

=== Corpus Validation Summary ===
Total files in manifest: 1000
Valid files:             1000

Corpus metrics (for valid files):
  Total pages:  10590
  Total size:   6870643 bytes

[✓] VALIDATION PASSED

PASS: Manifest contains ≥1000 entries with all required fields

  • Entry count: 1000 PDF files
  • All fields present: filename, source_url, page_count, file_size, checksum, license
  • All entries validated: 0 missing files, 0 size mismatches, 0 checksum mismatches

PASS: Validation script confirms corpus meets size/count targets

  • Files validated: 1000/1000
  • Total pages: 10,590 pages
  • Total size: 6,870,643 bytes (~6.5 MB)
  • Licenses: All synthetic PDFs marked as public-domain

Related Files

Core Implementation

  • tests/fixtures/grep-corpus/manifest.csv - Corpus manifest with 1000 entries
  • tests/fixtures/grep-corpus/corpus/ - Directory containing 1000 synthetic PDFs
  • scripts/validate-corpus.sh - Validation script that verifies:
    • File existence
    • File size matching
    • SHA256 checksum verification
    • License presence
    • Corpus quality metrics

Supporting Scripts

  • scripts/download-grep-corpus.sh - Downloads PDFs and generates manifest entries
  • scripts/grep-corpus-generate-manifest.sh - Bootstraps manifest for existing corpus

Makefile Integration

  • Makefile:22-24 - validate-corpus target that calls the validation script

Corpus Quality Metrics

  • Corpus type: Synthetic PDFs generated via ReportLab (public-domain)
  • Page count distribution: 1-20 pages per PDF
  • Average file size: ~6.9 KB per PDF
  • Total corpus size: ~6.5 MB for 1000 PDFs
  • Validation status: All 1000 files valid with correct checksums

Implementation Notes

The corpus generation uses synthetic PDFs rather than downloaded content to ensure:

  1. Determinism: Same script produces identical corpus
  2. Known license: Public domain (no licensing concerns)
  3. Controlled variety: Mix of page counts (1-20) and content
  4. Fast regeneration: No network dependencies

References

  • Parent bead: bf-38sa3
  • Phase 7.8 (plan lines 2796–2824)