v0.4.0-dev feat: add coordinate-aware vector OCR routing

.github/workflows/ci.yml

@@ -145,7 +145,7 @@ jobs:
           sudo apt-get install -y tesseract-ocr tesseract-ocr-eng poppler-utils
           tesseract --version
           pdftoppm -v
-          node --test integrations/javascript/*.test.mjs
+          node --test --test-concurrency=1 integrations/javascript/*.test.mjs
 
   python:
     name: Python ${{ matrix.python-version }} bindings and wheel

CHANGELOG.md

@@ -0,0 +1,34 @@
+# Changelog
+
+## Unreleased: 0.4.0-dev
+
+### Added
+
+- Schema 0.18 page geometry and pixel-to-PDF transforms preserve raw PDF
+  coordinates through quarter-turn OCR, CropBox and UserUnit handling.
+- Schema 0.19 page and region route evidence explains bounded vector candidates
+  and their regional OCR requests. Public records retain counters, limits and
+  exact candidate bounds, including candidates sharing a bitmap crop.
+- JavaScript adapters validate the new contracts and expose coordinate mapping
+  helpers while preserving text, chunk origins and source locators.
+- Small reproducible outlined-text fixtures and negative controls cover native
+  clipping, transformed Forms, rules, highlights, logos and resource refusal.
+
+### Fixed
+
+- Rotated OCR preserves recognized word order. Inferred tables on rotated pages
+  remain withheld until their topology is qualified.
+- Partially clipped native spans no longer suppress outlines outside their clip.
+- Exhausted native coverage accounting abstains instead of publishing zero
+  coverage as complete evidence.
+- OCR text elsewhere in a reused bitmap crop cannot mark a disjoint vector
+  candidate recovered. Recovery remains text availability, with completeness
+  unverified.
+
+### Changed
+
+- CI runs adapter test files serially to avoid process-start deadline contention
+  observed during concurrent local verification.
+- Vector collection and classification retain explicit work bounds. These
+  bounds exclude initial PDF decompression and do not establish a whole-process
+  memory limit.

README.md

@@ -548,8 +548,9 @@ The minimal specialist config shape is a JSON object with optional `ocr`,
 `enabled`, `executable`, `args`, `rasterizer_executable`, `crop_dpi`,
 `crop_padding_points`, `crop_grayscale`, and `timeout_ms`. Formula invocation
 is opt-in and model-neutral; no recognizer package is bundled. Memory-opened
-documents and rotated pages produce explicit crop-stage outcomes rather than
-silently invoking a text-only fallback.
+documents, invalid geometry and unsupported rotations produce explicit
+crop-stage outcomes. The development schema `0.18` maps declared quarter turns
+through shared crop transforms; scan orientation detection is separate.
 
 The document manifest is the top-level intelligence summary: input SHA256,
 parser/schema versions, page count, encrypted/corrupt flags, route counts,
@@ -669,7 +670,13 @@ and ruling lines, invokes OCR only for scanned routes, then reconciles native,
 OCR, table, formula, and form spans with typed provenance.
 
 The versioned JSON, artifact JSONL, and streaming JSONL schema is currently
-`0.17.0`, with parser build identity `0.4.0-dev` after the `v0.3.0` release.
+`0.19.0` in local development, with parser build identity `0.4.0-dev` after the
+`v0.3.0` release. Schema `0.17.0` landed in [PR #6](https://github.com/Tom-R-Main/pdf-parser/pull/6)
+at `9cac608`, with all six CI checks green. Schemas `0.18` and `0.19` are locally
+checked and committed on `codex/coordinate-vector-ocr`; remote CI and merge
+remain separate. The `0.19` vector-routing continuation's 23 outlined-text and
+negative-control fixtures, 13 real CLI/SDK tests and both JSONL readers are
+qualified in the [implementation evidence](docs/parser-host-implementation-plan.md#bounded-vector-routing-019-locally-checked).
 Every emitted record carries a `provenance` envelope with document and
 source identity, input hash context, artifact id, page/bbox, source kind,
 confidence, related span/block/chunk ids, route trace ids, and route reasons.
@@ -692,6 +699,36 @@ Schema `0.17.0` adds content-stream and Form invocation scope to spans.
 ActualText and classified text boundaries survive adaptive reconstruction; Form
 text uses scoped resources and transformed geometry. Native and adaptive text
 quality gates run against the same born-digital and hard-font floors.
+Schema `0.18.0` adds nullable page geometry and OCR/formula pixel transforms for
+declared quarter-turn pages. Source bboxes remain raw, unrotated PDF user-space
+coordinates; UserUnit determines their physical size. Display geometry uses the
+effective CropBox in physical points, while rasterization uses MediaBox and
+retains bounded crop budgets. Legacy PageInfo dimensions and the manifest's
+`display_transform:null` remain compatible. These mappings do not detect scan
+orientation or establish visual completeness.
+Rotated OCR retains recognized word order; inferred tables on those pages are
+not promoted to table artifacts until their topology is qualified.
+Schema `0.19.0` adds nullable vector evidence to page and region routes. Bounded
+painted-path collection and geometric grouping can request regional OCR for
+unexplained vector contours. Rules, simple boxes, native-accounted shapes and
+isolated symbols can be withheld. The collector caps retained shapes at 512,
+segments per path at 512, operators at 100,000, cumulative vector-lexer input at
+8 MiB, graphics stack depth at 32, Form invocations at 256 and candidate regions
+at 64. Page summaries retain operator and byte counters. These lexer bounds
+exclude initial PDF decompression and do not establish whole-process memory
+limits. Native accounting uses rectangle unions capped at 64 intersections per
+coverage query and 1,000,000 native-span checks during page classification;
+candidate-merge coverage has its own bounded check budget. Exhaustion produces
+partial evidence and abstention.
+Partial or limited evidence remains explicit. Grouping uses known display axes;
+candidate bboxes stay in raw PDF coordinates. Exact candidate bounds can be
+retained within a larger image OCR route to avoid duplicate crops. These
+candidates require retained OCR text intersecting the exact candidate extent
+before recovery is reported; recognized text elsewhere in the bitmap crop does
+not qualify. Recovery and merged route bounds leave completeness unverified.
+Vector candidates do not establish recognized glyphs, final visibility or
+complete coverage, and use the existing
+OCR engine and crop budgets.
 Host adoption remains explicit; accepted schema families and parser/compiler
 pins must be updated together.
 

benchmark/eval/vector_routing/.gitignore

@@ -0,0 +1,4 @@
+# Small generated correctness fixtures are intentionally versioned here.
+!*.pdf
+!font-license.txt
+!truth/*.txt

benchmark/eval/vector_routing/font-license.txt

@@ -0,0 +1,57 @@
+Bitstream Vera Fonts Copyright
+
+The fonts have a generous copyright, allowing derivative works (as
+long as "Bitstream" or "Vera" are not in the names), and full
+redistribution (so long as they are not *sold* by themselves). They
+can be be bundled, redistributed and sold with any software.
+
+The fonts are distributed under the following copyright:
+
+Copyright
+=========
+
+Copyright (c) 2003 by Bitstream, Inc. All Rights Reserved. Bitstream
+Vera is a trademark of Bitstream, Inc.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of the fonts accompanying this license ("Fonts") and associated
+documentation files (the "Font Software"), to reproduce and distribute
+the Font Software, including without limitation the rights to use,
+copy, merge, publish, distribute, and/or sell copies of the Font
+Software, and to permit persons to whom the Font Software is furnished
+to do so, subject to the following conditions:
+
+The above copyright and trademark notices and this permission notice
+shall be included in all copies of one or more of the Font Software
+typefaces.
+
+The Font Software may be modified, altered, or added to, and in
+particular the designs of glyphs or characters in the Fonts may be
+modified and additional glyphs or characters may be added to the
+Fonts, only if the fonts are renamed to names not containing either
+the words "Bitstream" or the word "Vera".
+
+This License becomes null and void to the extent applicable to Fonts
+or Font Software that has been modified and is distributed under the
+"Bitstream Vera" names.
+
+The Font Software may be sold as part of a larger software package but
+no copy of one or more of the Font Software typefaces may be sold by
+itself.
+
+THE FONT SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT
+OF COPYRIGHT, PATENT, TRADEMARK, OR OTHER RIGHT. IN NO EVENT SHALL
+BITSTREAM OR THE GNOME FOUNDATION BE LIABLE FOR ANY CLAIM, DAMAGES OR
+OTHER LIABILITY, INCLUDING ANY GENERAL, SPECIAL, INDIRECT, INCIDENTAL,
+OR CONSEQUENTIAL DAMAGES, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF THE USE OR INABILITY TO USE THE FONT
+SOFTWARE OR FROM OTHER DEALINGS IN THE FONT SOFTWARE.
+
+Except as contained in this notice, the names of Gnome, the Gnome
+Foundation, and Bitstream Inc., shall not be used in advertising or
+otherwise to promote the sale, use or other dealings in this Font
+Software without prior written authorization from the Gnome Foundation
+or Bitstream Inc., respectively. For further information, contact:
+fonts at gnome dot org.