feat(wasm+py+forensic): Lanes 11, 15, 17 — WASM parity, PyO3 tests, forensic inspection

[jacob-cotten]

Summary

Three lanes shipped together as they share pdfplumber-core infrastructure:

Lane 11 — WASM Bindings (full API parity)

What was missing:

• WasmCroppedPage — entirely absent from WASM despite existing in PyO3
• lines(), rects(), curves(), images() — no geometry API on any WASM type
• hyperlinks(), annots(), bookmarks() — no annotation/link access
• rotation, bbox, mediaBox — missing page properties
• crop(), within_bbox(), outside_bbox() — no spatial filtering

What was added:

• WasmCroppedPage struct with full extraction API (chars, words, tables, geometry, crop chain)
• All geometry methods on WasmPage
• WasmPdf::bookmarks() for TOC access
• TypeScript .d.ts: PdfLine, PdfRect, PdfCurve, PdfImage, PdfBookmark, PdfHyperlink, WasmCroppedPage class
• package.json for wasm-pack npm packaging
• Complete browser demo rewrite: metadata display, bookmarks, per-page nav, crop demo (header/body split), geometry inspector, hyperlinks section, WASM load indicator
• 26 new Rust unit tests

Lane 17 — PyO3 Python Bindings (test suite)

What was missing: Zero Python integration tests despite 98 Rust unit tests in the crate.

What was added:

• tests/conftest.py: pure-Python hand-crafted PDF byte fixture (no external deps — works before maturin develop)
• tests/test_basic.py: 50+ Python integration tests covering PDF.open_bytes, PDF.open, pages, metadata, bookmarks, Page properties, all extraction methods, crop/within_bbox/outside_bbox, CroppedPage methods

Lane 15 — Forensic Metadata Inspection

New module: pdfplumber-core::forensic (~650 lines, 40+ tests)

• ProducerKind: 18-variant enum fingerprinting known PDF producers — AdobeAcrobat, MicrosoftWord, LibreOffice, GoogleDocs, Latex, Pdf24 (online converter 🚩), Smallpdf (online converter 🚩), and 11 more
• detect_incremental_updates(bytes): pure byte-scan for startxref markers — each = one PDF revision. Detects post-signing modifications.
• WatermarkFinding / WatermarkKind: LowOpacityText, InvisibleText, RepeatedTextBlock, LowOpacityOverlay
• PageGeometryAnomaly: unusual rotation, non-standard dimensions
• MetadataFinding: scrubbed fields, creation=mod date (common in online converter output)
• ForensicReport::build(): assembles all findings, computes risk score (0=clean, accumulates per finding type)
• format_text(): human-readable multi-section forensic report
• Pdf::inspect(&raw_bytes): public API entry point — never fails, all errors produce sensible defaults

New CLI subcommand: pdfplumber inspect <file> [--format text|json]

• Non-zero exit code when risk_score > 0 (CI pipeline friendly)

Note on CI jobs: This PR does not include the CI workflow additions (requires workflow OAuth scope). CI jobs for WASM (wasm-pack build) and PyO3 (maturin develop + pytest) will be added in a follow-up PR.

Test plan

• cargo check -p pdfplumber -p pdfplumber-core -p pdfplumber-cli — verify all imports resolve
• cargo test -p pdfplumber-core — 40+ new forensic tests
• cargo check -p pdfplumber-wasm --target wasm32-unknown-unknown — WASM compile check
• cargo test -p pdfplumber-py --lib --features pyo3/auto-initialize — PyO3 unit tests
• maturin develop && pytest crates/pdfplumber-py/tests/ — Python integration tests

🤖 Generated with Claude Code