export_pdf_for_test on the proofs build (ExportInput JSON + base64 rasters): the 14 O2 runs give the native digests through wasm; the frozen O11 scanner reads every R4 PDF and the N-1 Zod readers accept its envelope; a stripped R4 PDF answers attachment_stripped; the TypeScript engine reads the R4 PDFs back to the R3 outcome; E2 size ≤ R3 + 5 %. Page resources list the images the page draws (listing every embed cost 9 MB on CAP). Node, stub rasters, TS → Rust: demo 198 → 36 ms, L 772 → 57 ms, CAP 5.7 s → 0.22 s; files 236 KB, 386 KB, 1.6 MB against 243 KB, 463 KB, 2.2 MB. Module with the writer linked (proofs build): 531.6 KB gzip; shipped unchanged at 426 KB until R4.6 links it.
Test fixtures
Golden inputs frozen for the characterisation tests (plan Step 1). They describe the Report v2 document as the current application produces it.
| File | Content |
|---|---|
report-v2.full.json |
One report exercising every zone kind the editor can produce: plain hinge point, bilateral joint with both sides, spine zone with vertebrae, detail-popup zone, catalog findings (with and without vet referral and side), free text, severity, comment, conclusion, recommendations, dated next appointment. boneId values are the real data-anno-id of the legacy Horse.svg / Horse_head.svg now kept under supabase/seed/anatomical-sheets/equine/legacy-sources/; zoneId/path values exist in pathologies.json. |
report-v2.minimal.json |
Smallest valid v2 report (no observations). |
report-v2.full.pdf, report-v2.minimal.pdf |
The two reports exported by the v2 pipeline (commit 199e328, Step 5.0, rasteriser stubbed to a 1×1 PNG, no logos): the archive-like PDFs the v3 importer must read and migrate. Synthetic data only. |
Do not edit these files to make a test pass: they are the behavioural oracle for the Report v2 → v3 adapter and for PDF regression. Add a new fixture instead.
Step 6 freeze (Phase 0, 2026-10-11)
What the PDF pipeline produced right before the Step 6 refactor, recorded by
tests/unit/pdf.goldens.test.ts on the bundled draft sheet
(docs/development/STEP6_PDF_ARCHIVE_ANALYSIS.md §17).
| File | Content |
|---|---|
report-v3.demo.json |
Report v3 on the draft (ids from draft/document.json), 6 observations: main-view points, head-view points, spine zone with vertebrae, bilateral zone with both sides, a vet-referral leaf, a leaf whose label holds an en dash (U+2013), free-text findings, severities, comments, conclusion, recommendations, period next appointment; sheet = the draft ref at freeze time. |
report-v3.bare.pdf |
The demo report exported by the current code (rasteriser stubbed to a 1×1 PNG, no logos, FR): import fixture for the engine. |
goldens/step6/text.<report>.<locale>.<logos>.json |
Every drawText of the export, in drawing order: { page, text, x, y, size, font }, numbers rounded to 2 decimals. |
goldens/step6/images.<report>.<locale>.<logos>.json |
Every drawImage: { page, x, y, width, height }. Recorded per locale too: the footer centres the "powered by" block on the label width, so the ZenQuality logo shifts between FR and EN. |
goldens/step6/links.<report>.<locale>.<logos>.json |
Every /Annot /Link read back from the saved PDF: { page, rect, uri }. Per locale for the same footer reason. |
goldens/step6/views.<report>.json |
Per (observation, view) of the report, the highlighted SVG parsed: viewBox, halo filter present, highlighted elements, visible non-highlighted markers, remaining data-* attributes, <title> count. |
Reports: v2full (report-v2.full.json migrated), demo (report-v3.demo.json),
v2unresolved (legacy sphenoid and ssb leaves plus one frontal leaf, migrated;
zone-less, so its images and views goldens are empty). Logo modes: nologos
(every branding slot empty, 1×1 rasteriser) and logos (the four logo files bundled
under apps/app/src/assets/brand, rasteriser stubbed to a 2×1 PNG).
Never regenerate these files without a user decision; the Step 6 engine must
reproduce them. Generation, once: MAKE_GOLDEN=1 corepack pnpm vitest run tests/unit/pdf.goldens.test.ts from apps/app.
R4 freeze (2026-10-18)
What the TypeScript PDF engine and reader produced right before the Rust port,
frozen on branch feature/rust-r4-pdf-core (docs/development/RUST.md §10.1
lists every oracle with its commit). Generators run behind MAKE_R4_ORACLES=1;
without the flag the same tests assert equality with the committed files.
| Path | Content |
|---|---|
r4/reports/{L,CAP}.json |
Synthetic Report v3 on the draft: 60 observations / 120 views (L) and 500 / 1,000 (CAP), the export memory inputs of plan §B.4 (O13). |
goldens/r4/draw.*.json |
Every drawing call of the export (text, image, line, rectangle, link) at full precision, per page and in call order, for the 12 Step 6 report × locale × logo combinations (O1). |
pdf/r3/ |
The 12 PDFs the R3-era engine produced for those combinations, the store-path export with its report.json, and real/v2full-obs1.fr.logos.pdf with its two browser PNGs and manifest: the byte references of the writer port (O2). |
pdf/step6/envelope.*.pdf |
Three envelope PDFs from before R2 (O3). |
pdf/r4/tolerance/ |
51 PDFs, valid and hostile, with today's import outcome and message in crates/core/tests/oracles/r4/tolerance.json (O8). |
pdf/r4/read/, pdf/r4/inflate/ |
155 reader inputs and 24 zlib framing inputs, the bytes behind read.json and inflate.json; a <slug>.gen.json describes a large input by concatenation (O7, O9). |
Synthetic data only, no real bilan. Never regenerate these files to make a test pass: a difference between the Rust port and an oracle is reported and decided (plan §A.3), or it is a bug in the port.