layout: indent paragraph breaks, heading reclassification, table script filter

Layout heuristics originally developed alongside the base-14 metrics fix and split out of it. These are NOT bug fixes — they are new heuristics, and they apply to every document regardless of vintage. - First-line-indent paragraph breaks: justified documents that separate paragraphs by indentation only (no vertical gap) previously came out as walls of text - Heading classification by dominant alphanumeric font size, so a leading drop cap or oversized math delimiter cannot reclassify a body line; tier computation filters glyph-run lines and requires uniform sizing - Table detection: exclude sub/superscripts attached at a script baseline offset from candidates; reject tiny all-numeric grids in the small-font pass - Embedded drop-cap merge for two-line caps that land as the second line's first item Corpus impact: ~98 of 184 eval documents on top of the font-metrics branch (full stack: 110 of 184; font metrics alone: 12). KNOWN REGRESSIONS — this is why it is split out and left as a draft: 1. Heading hierarchy collapse. In revista63 every heading moves h3 -> h1 because the new tier filters drop that document's larger tiers, so everything at body+2pt becomes top-level. 2 of 60 sampled documents lose heading depth this way. 2. The indent rule fires on non-prose. Its guard only asks whether the previous line ended ragged, which is true of URL lists, file listings, poems, and chat transcripts. In the mythos system card it splits a protein-sequence line and a URL list; the resulting short fragment is then promoted to a heading. Same mechanism yields the junk '#### Rr Fy' heading in 2103_07786. 3. 10 documents lose table rows with no replacement. The semantic composite is flat (+0.0011) but char-accuracy barely registers heading damage ('###' -> '#' is two characters), so that is not evidence these are harmless.
fix(extractor): supply built-in metrics for non-embedded base-14 fonts
2026-08-03 16:16:13 -07:00 · 2026-08-03 16:15:04 -07:00
43 changed files with 991 additions and 4440 deletions
@@ -14,15 +14,13 @@ jobs:
    name: Test
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@4cda84d5c5c54efe2404f9d843567869ab1699d4 # stable
-        with:
-          toolchain: stable
+        uses: dtolnay/rust-toolchain@stable

      - name: Cache cargo
-        uses: Swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 # v2.9.1
+        uses: Swatinem/rust-cache@v2

      - name: Run tests
        run: cargo test --verbose
@@ -34,12 +32,11 @@ jobs:
    name: Format
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@4cda84d5c5c54efe2404f9d843567869ab1699d4 # stable
+        uses: dtolnay/rust-toolchain@stable
        with:
-          toolchain: stable
          components: rustfmt

      - name: Check formatting
@@ -52,16 +49,15 @@ jobs:
    name: Clippy
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@4cda84d5c5c54efe2404f9d843567869ab1699d4 # stable
+        uses: dtolnay/rust-toolchain@stable
        with:
-          toolchain: stable
          components: clippy

      - name: Cache cargo
-        uses: Swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 # v2.9.1
+        uses: Swatinem/rust-cache@v2
        with:
          key: clippy

@@ -75,15 +71,13 @@ jobs:
      matrix:
        os: [ubuntu-latest, macos-latest]
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@4cda84d5c5c54efe2404f9d843567869ab1699d4 # stable
-        with:
-          toolchain: stable
+        uses: dtolnay/rust-toolchain@stable

      - name: Cache cargo
-        uses: Swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 # v2.9.1
+        uses: Swatinem/rust-cache@v2
        with:
          key: build

@@ -94,17 +88,16 @@ jobs:
    name: WebAssembly
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@4cda84d5c5c54efe2404f9d843567869ab1699d4 # stable
+        uses: dtolnay/rust-toolchain@stable
        with:
-          toolchain: stable
          targets: wasm32-unknown-unknown
          components: clippy

      - name: Cache cargo
-        uses: Swatinem/rust-cache@c19371144df3bb44fab255c43d04cbc2ab54d1c4 # v2.9.1
+        uses: Swatinem/rust-cache@v2
        with:
          workspaces: |
            wasm -> target
@@ -24,13 +24,13 @@ jobs:
      name: github-pages
      url: ${{ steps.deploy.outputs.page_url }}
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Upload site artifact
-        uses: actions/upload-pages-artifact@fc324d3547104276b827a68afc52ff2a11cc49c9 # v5.0.0
+        uses: actions/upload-pages-artifact@v3
        with:
          path: site

      - name: Deploy to GitHub Pages
        id: deploy
-        uses: actions/deploy-pages@cd2ce8fcbc39b97be8ca5fce6e763baed58fa128 # v5.0.0
+        uses: actions/deploy-pages@v4
@@ -27,7 +27,7 @@ jobs:
      published: ${{ steps.check.outputs.published }}
      version: ${{ steps.check.outputs.version }}
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4
        with:
          fetch-depth: 2

@@ -85,19 +85,17 @@ jobs:
      contents: read
      id-token: write
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@4cda84d5c5c54efe2404f9d843567869ab1699d4 # stable
-        with:
-          toolchain: stable
+        uses: dtolnay/rust-toolchain@stable

      - name: Verify package
        run: cargo publish --dry-run

      - name: Authenticate with crates.io
        id: auth
-        uses: rust-lang/crates-io-auth-action@c6f97d42243bad5fab37ca0427f495c86d5b1a18 # v1.0.5
+        uses: rust-lang/crates-io-auth-action@v1

      - name: Publish crate
        run: cargo publish
@@ -24,7 +24,7 @@ jobs:
      published: ${{ steps.check.outputs.published }}
      version: ${{ steps.check.outputs.version }}
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4
        with:
          fetch-depth: 2

@@ -98,21 +98,21 @@ jobs:
          - os: windows-latest
            target: x86_64-pc-windows-msvc
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

-      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
+      - uses: actions/setup-python@v5
        with:
          python-version: '3.12'

      - name: Build wheel
-        uses: PyO3/maturin-action@e83996d129638aa358a18fbd1dfb82f0b0fb5d3b # v1.51.0
+        uses: PyO3/maturin-action@v1
        with:
          target: ${{ matrix.target }}
          args: --release --out dist
          manylinux: auto

      - name: Upload wheel
-        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        uses: actions/upload-artifact@v4
        with:
          name: wheels-${{ matrix.target }}
          path: dist/*.whl
@@ -124,16 +124,16 @@ jobs:
    name: Build sdist
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Build sdist
-        uses: PyO3/maturin-action@e83996d129638aa358a18fbd1dfb82f0b0fb5d3b # v1.51.0
+        uses: PyO3/maturin-action@v1
        with:
          command: sdist
          args: --out dist

      - name: Upload sdist
-        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        uses: actions/upload-artifact@v4
        with:
          name: sdist
          path: dist/*.tar.gz
@@ -149,7 +149,7 @@ jobs:
      id-token: write
    steps:
      - name: Download all artifacts
-        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
+        uses: actions/download-artifact@v4
        with:
          path: dist
          merge-multiple: true
@@ -158,7 +158,7 @@ jobs:
        run: ls -la dist/

      - name: Publish to PyPI
-        uses: pypa/gh-action-pypi-publish@dc37677b2e1c63e2034f94d8a5b11f265b73ba33 # release/v1
+        uses: pypa/gh-action-pypi-publish@release/v1
        with:
          packages-dir: dist
          # Tolerate already-uploaded files so a manual re-run can complete
@@ -24,7 +24,7 @@ jobs:
      published: ${{ steps.check.outputs.published }}
      version: ${{ steps.check.outputs.version }}
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4
        with:
          fetch-depth: 2

@@ -71,15 +71,14 @@ jobs:
      contents: read
      id-token: write
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@4cda84d5c5c54efe2404f9d843567869ab1699d4 # stable
+        uses: dtolnay/rust-toolchain@stable
        with:
-          toolchain: stable
          targets: wasm32-unknown-unknown

-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7.0.0
+      - uses: actions/setup-node@v6
        with:
          node-version: '24'
          registry-url: 'https://registry.npmjs.org'
@@ -24,7 +24,7 @@ jobs:
      changed: ${{ steps.check.outputs.changed }}
      version: ${{ steps.check.outputs.version }}
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4
        with:
          fetch-depth: 2

@@ -61,61 +61,31 @@ jobs:
        include:
          - os: ubuntu-latest
            target: x86_64-unknown-linux-gnu
-          # napi-cross builds gnu targets against an old glibc sysroot for
-          # broad distro compatibility; musl targets cross-compile with
-          # zig via cargo-zigbuild (napi's -x flag).
-          - os: ubuntu-latest
-            target: aarch64-unknown-linux-gnu
-            build-flags: --use-napi-cross
-          - os: ubuntu-latest
-            target: x86_64-unknown-linux-musl
-            build-flags: -x
-          - os: ubuntu-latest
-            target: aarch64-unknown-linux-musl
-            build-flags: -x
          - os: macos-14
            target: aarch64-apple-darwin
          - os: windows-latest
            target: x86_64-pc-windows-msvc
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v4

      - name: Install Rust
-        uses: dtolnay/rust-toolchain@4cda84d5c5c54efe2404f9d843567869ab1699d4 # stable
-        with:
-          toolchain: stable
-          targets: ${{ matrix.target }}
+        uses: dtolnay/rust-toolchain@stable

-      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
+      - uses: oven-sh/setup-bun@v2
        with:
          bun-version: latest

-      - name: Install zig
-        if: contains(matrix.target, 'musl')
-        uses: mlugg/setup-zig@d1434d08867e3ee9daa34448df10607b98908d29 # v2.2.1
-        with:
-          version: 0.14.1
-
-      - name: Install cargo-zigbuild
-        if: contains(matrix.target, 'musl')
-        uses: taiki-e/install-action@67729d5c413db75907f0ad1e39bb04b9c868ff60 # v2.85.7
-        env:
-          GITHUB_TOKEN: ${{ github.token }}
-        with:
-          tool: cargo-zigbuild
-
      - name: Cache cargo
-        uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
+        uses: actions/cache@v4
        with:
          path: |
            ~/.cargo/registry/index/
            ~/.cargo/registry/cache/
            ~/.cargo/git/db/
-            ~/.napi-rs
            napi/target/
-          key: ${{ matrix.target }}-cargo-napi-${{ hashFiles('**/Cargo.lock') }}
+          key: ${{ runner.os }}-cargo-napi-${{ hashFiles('**/Cargo.lock') }}
          restore-keys: |
-            ${{ matrix.target }}-cargo-napi-
+            ${{ runner.os }}-cargo-napi-

      - name: Install dependencies
        working-directory: napi
@@ -123,10 +93,10 @@ jobs:

      - name: Build native addon
        working-directory: napi
-        run: bunx napi build --platform --release --target ${{ matrix.target }} ${{ matrix.build-flags }}
+        run: bunx napi build --platform --release

      - name: Upload native binary
-        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        uses: actions/upload-artifact@v4
        with:
          name: bindings-${{ matrix.target }}
          path: napi/*.node
@@ -134,7 +104,7 @@ jobs:

      - name: Upload generated JS bindings
        if: matrix.target == 'x86_64-unknown-linux-gnu'
-        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        uses: actions/upload-artifact@v4
        with:
          name: js-bindings
          path: |
@@ -142,68 +112,23 @@ jobs:
            napi/index.d.ts
          if-no-files-found: error

-  smoke-test:
-    name: Smoke test ${{ matrix.target }}
-    needs: [check-version, build]
-    runs-on: ${{ matrix.os }}
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - os: ubuntu-latest
-            target: x86_64-unknown-linux-gnu
-          - os: ubuntu-latest
-            target: x86_64-unknown-linux-musl
-          - os: ubuntu-24.04-arm
-            target: aarch64-unknown-linux-gnu
-          - os: ubuntu-24.04-arm
-            target: aarch64-unknown-linux-musl
-          - os: macos-14
-            target: aarch64-apple-darwin
-          - os: windows-latest
-            target: x86_64-pc-windows-msvc
-    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
-
-      - name: Download native binary
-        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
-        with:
-          name: bindings-${{ matrix.target }}
-          path: napi
-
-      - name: Download generated JS bindings
-        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
-        with:
-          name: js-bindings
-          path: napi
-
-      # musl binaries must load under a real musl libc, so run inside Alpine.
-      - name: Run smoke test (Alpine)
-        if: contains(matrix.target, 'musl')
-        run: docker run --rm -v "$PWD:/repo" -w /repo/napi node:24-alpine node test.mjs
-
-      - name: Run smoke test
-        if: ${{ !contains(matrix.target, 'musl') }}
-        working-directory: napi
-        run: node test.mjs
-
  publish:
    name: Publish to npm
-    needs: [check-version, build, smoke-test]
+    needs: [check-version, build]
    runs-on: ubuntu-latest
    permissions:
      contents: read
      id-token: write
    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/checkout@v6

-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7.0.0
+      - uses: actions/setup-node@v6
        with:
          node-version: '24'
          registry-url: 'https://registry.npmjs.org'

      - name: Download all artifacts
-        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
+        uses: actions/download-artifact@v4
        with:
          path: napi/artifacts

@@ -229,12 +154,9 @@ jobs:
            node -e '
              const [suffix, version] = process.argv.slice(1)
              const meta = {
-                "linux-x64-gnu":    { os: ["linux"],  cpu: ["x64"],   libc: ["glibc"] },
-                "linux-x64-musl":   { os: ["linux"],  cpu: ["x64"],   libc: ["musl"] },
-                "linux-arm64-gnu":  { os: ["linux"],  cpu: ["arm64"], libc: ["glibc"] },
-                "linux-arm64-musl": { os: ["linux"],  cpu: ["arm64"], libc: ["musl"] },
-                "darwin-arm64":     { os: ["darwin"], cpu: ["arm64"] },
-                "win32-x64-msvc":   { os: ["win32"],  cpu: ["x64"] },
+                "linux-x64-gnu":  { os: ["linux"],  cpu: ["x64"], libc: ["glibc"] },
+                "darwin-arm64":   { os: ["darwin"], cpu: ["arm64"] },
+                "win32-x64-msvc": { os: ["win32"],  cpu: ["x64"] },
              }[suffix]
              if (!meta) {
                console.error(`unknown platform suffix: ${suffix} — add it to the meta map`)
@@ -61,8 +61,7 @@ src/

 - **Unit tests**: inline `#[cfg(test)] mod tests` in each module with synthetic data.
 - **Integration tests**: `tests/integration_tests.rs` with fixture PDFs in `tests/fixtures/`.
- **Regression suite**: sibling repo `pdf-evals` with ~200 snapshot PDFs. Run `cargo build --release` then `bench.py test` in that repo before committing. While iterating, prefer a subset run (`bench.py test -q` for the quick set, or `-s <name>` for a named test set) and save the full `bench.py test` for the final pre-commit check.
- **Semantic quality**: run `bench.py score` in `pdf-evals` for the semantic verdict (TEDS + MHS + reading order + char/word + list preservation, composited). Character-level diff alone misclassifies structural improvements (e.g., column-detection rewrites) as regressions — `score` is the tie-breaker. See `pdf-evals/CLAUDE.md` "Semantic scoring".
+- **Regression suite**: sibling repo `pdf-evals` with 179+ snapshot PDFs. Run `cargo build --release` then `bench.py test` in that repo before committing.

 ## Debugging

@@ -61,7 +61,7 @@ src/

 - **Unit tests**: inline `#[cfg(test)] mod tests` in each module with synthetic data.
 - **Integration tests**: `tests/integration_tests.rs` with fixture PDFs in `tests/fixtures/`.
- **Regression suite**: sibling repo `pdf-evals` with ~200 snapshot PDFs. Run `cargo build --release` then `bench.py test` in that repo before committing. While iterating, prefer a subset run (`bench.py test -q` for the quick set, or `-s <name>` for a named test set) and save the full `bench.py test` for the final pre-commit check.
+- **Regression suite**: sibling repo `pdf-evals` with 187+ snapshot PDFs. Run `cargo build --release` then `bench.py test` in that repo before committing.
 - **Semantic quality**: run `bench.py score` in `pdf-evals` for the semantic verdict (TEDS + MHS + reading order + char/word + list preservation, composited). Character-level diff alone misclassifies structural improvements (e.g., column-detection rewrites) as regressions — `score` is the tie-breaker. See `pdf-evals/CLAUDE.md` "Semantic scoring".

 ## Debugging
@@ -46,14 +46,14 @@ ttf-parser = "0.25"
 # Native builds keep lopdf's parallel parser and CLI logging. Browser WASM is
 # deliberately single-threaded so it works without cross-origin isolation.
 [target.'cfg(not(target_arch = "wasm32"))'.dependencies]
-lopdf = { version = "0.42.0", features = ["rayon"] }
+lopdf = { version = "0.41.0", features = ["rayon"] }
 rayon = "1.10"
 env_logger = "0.11"

 # Browser builds use JavaScript randomness for encrypted PDFs and embed the
 # bundled CMaps because there is no filesystem at runtime.
 [target.'cfg(target_arch = "wasm32")'.dependencies]
-lopdf = { version = "0.42.0", default-features = false, features = ["wasm_js"] }
+lopdf = { version = "0.41.0", default-features = false, features = ["wasm_js"] }
 include_dir = "0.7"

 [dev-dependencies]
@@ -238,7 +238,7 @@ The converter handles:
 |---|---|
 | Headings (H1-H4) | Font size tiers relative to body text, with 0.5pt clustering |
 | Bold/italic | Font name patterns (Bold, Italic, Oblique) |
-| Bullet lists | `•`, `-`, `*`, `○`, `●`, `◦` prefixes |
+| Bullet lists | `*`, `-`, `*`, `○`, `●`, `◦` prefixes |
 | Numbered lists | `1.`, `1)`, `(1)` patterns |
 | Letter lists | `a.`, `a)`, `(a)` patterns |
 | Code blocks | Monospace fonts (Courier, Consolas, Monaco, Menlo, Fira Code, JetBrains Mono) and keyword detection |
@@ -111,8 +111,7 @@ class PdfResult:                     # process_pdf / detect_pdf
    markdown: str | None             # extracted Markdown (None for detect_pdf)
    page_count: int
    processing_time_ms: int
-    pages_needing_ocr: list[int]     # 1-indexed
-    ocr_reasons_by_page: list[PageOcrReasons]
+    pages_needing_ocr: list[int]
    title: str | None
    confidence: float                # 0.0 - 1.0
    is_complex_layout: bool
@@ -120,10 +119,6 @@ class PdfResult:                     # process_pdf / detect_pdf
    pages_with_columns: list[int]
    has_encoding_issues: bool        # broken font encodings — consider OCR fallback

-class PageOcrReasons:                # per-page OCR diagnostics
-    page: int                        # 1-indexed
-    reasons: list[str]               # machine-readable reason identifiers
-
 class PdfClassification:             # classify_pdf
    pdf_type: str
    page_count: int
@@ -145,20 +140,14 @@ class TextItem:                      # extract_text_with_positions
    is_strikeout: bool
    item_type: str

-class RegionText:                    # extract_text_in_regions
-    text: str
-    needs_ocr: bool
-    ocr_reason: str | None           # machine-readable OCR reason
-
 class PageRegionTexts:               # extract_text_in_regions
    page: int                        # 0-indexed
-    regions: list[RegionText]
+    regions: list[RegionText]        # RegionText: text: str, needs_ocr: bool

 class PagesExtractionResult:         # extract_pages_markdown
-    pages: list[PageMarkdown]        # PageMarkdown: page (0-indexed), markdown, needs_ocr, ocr_reason
+    pages: list[PageMarkdown]        # PageMarkdown: page (0-indexed), markdown, needs_ocr
    pages_with_tables: list[int]     # 1-indexed
    pages_with_columns: list[int]    # 1-indexed
    pages_needing_ocr: list[int]     # 1-indexed
-    ocr_reasons_by_page: list[PageOcrReasons]
    is_complex: bool                 # any page has tables or multi-column layout
 ```
@@ -499,13 +499,11 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0de51e6874e94e7bf76d726fc5d13ba782deca734ff60d5bb2fb2607c7406555"
 dependencies = [
 "cfg-if",
- "js-sys",
 "libc",
 "r-efi",
 "rand_core",
 "wasip2",
 "wasip3",
- "wasm-bindgen",
 ]

 [[package]]
@@ -559,25 +557,6 @@ version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3d3067d79b975e8844ca9eb072e16b31c3c1c36928edf9c6789548c524d0d954"

-[[package]]
-name = "include_dir"
-version = "0.7.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "923d117408f1e49d914f1a379a309cffe4f18c05cf4e3d12e613a15fc81bd0dd"
-dependencies = [
- "include_dir_macros",
-]
-
-[[package]]
-name = "include_dir_macros"
-version = "0.7.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7cab85a7ed0bd5f0e76d93846e0147172bed2e2d3f859bcc33a8d9699cad1a75"
-dependencies = [
- "proc-macro2",
- "quote",
-]
-
 [[package]]
 name = "indexmap"
 version = "2.13.0"
@@ -693,9 +672,9 @@ checksum = "5e5032e24019045c762d3c0f28f5b6b8bbf38563a65908389bf7978758920897"

 [[package]]
 name = "lopdf"
-version = "0.42.0"
+version = "0.41.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "25aab26d99567469098e64a02f42679f8965c6401263eefa31d8f2dcc37a221c"
+checksum = "67513274c50a2b51e5f75d9e682fcf4ab064a8a9c9ae2c3c59309084882bb24d"
 dependencies = [
 "aes",
 "bitflags",
@@ -854,7 +833,6 @@ name = "pdf-inspector"
 version = "0.1.7"
 dependencies = [
 "env_logger",
- "include_dir",
 "log",
 "lopdf",
 "once_cell",
@@ -34,7 +34,7 @@ npm install @firecrawl/pdf-inspector
 bun add @firecrawl/pdf-inspector
 ```

-Prebuilt binaries for **Linux x64/ARM64** (glibc and musl/Alpine), **macOS ARM64**, and **Windows x64** — npm installs only the one matching your platform. No Rust toolchain needed.
+Prebuilt binaries for **Linux x64**, **macOS ARM64**, and **Windows x64** — npm installs only the one matching your platform. No Rust toolchain needed.

 ## API

@@ -116,12 +116,9 @@ Prebuilt binaries ship as platform-specific packages installed automatically via

 | Platform | Architecture | Package |
 |----------|-------------|---------|
-| Linux    | x64 (glibc)         | `@firecrawl/pdf-inspector-linux-x64-gnu` |
-| Linux    | x64 (musl/Alpine)   | `@firecrawl/pdf-inspector-linux-x64-musl` |
-| Linux    | ARM64 (glibc)       | `@firecrawl/pdf-inspector-linux-arm64-gnu` |
-| Linux    | ARM64 (musl/Alpine) | `@firecrawl/pdf-inspector-linux-arm64-musl` |
-| macOS    | ARM64               | `@firecrawl/pdf-inspector-darwin-arm64` |
-| Windows  | x64                 | `@firecrawl/pdf-inspector-win32-x64-msvc` |
+| Linux    | x64 (glibc) | `@firecrawl/pdf-inspector-linux-x64-gnu` |
+| macOS    | ARM64       | `@firecrawl/pdf-inspector-darwin-arm64` |
+| Windows  | x64         | `@firecrawl/pdf-inspector-win32-x64-msvc` |

 ## License

@@ -8,12 +8,9 @@
        "@napi-rs/cli": "^3.4.1",
      },
      "optionalDependencies": {
-        "@firecrawl/pdf-inspector-darwin-arm64": "1.12.0",
-        "@firecrawl/pdf-inspector-linux-arm64-gnu": "1.12.0",
-        "@firecrawl/pdf-inspector-linux-arm64-musl": "1.12.0",
-        "@firecrawl/pdf-inspector-linux-x64-gnu": "1.12.0",
-        "@firecrawl/pdf-inspector-linux-x64-musl": "1.12.0",
-        "@firecrawl/pdf-inspector-win32-x64-msvc": "1.12.0",
+        "@firecrawl/pdf-inspector-darwin-arm64": "1.11.0",
+        "@firecrawl/pdf-inspector-linux-x64-gnu": "1.11.0",
+        "@firecrawl/pdf-inspector-win32-x64-msvc": "1.11.0",
      },
    },
  },
@@ -1,6 +1,6 @@
 {
  "name": "@firecrawl/pdf-inspector",
-  "version": "1.12.0",
+  "version": "1.11.2",
  "description": "Fast PDF classification and text extraction. Detect text-based vs scanned PDFs, extract text by region with quality checks. Native Rust performance via napi-rs.",
  "main": "index.js",
  "types": "index.d.ts",
@@ -37,9 +37,6 @@
    "binaryName": "pdf-inspector",
    "targets": [
      "x86_64-unknown-linux-gnu",
-      "x86_64-unknown-linux-musl",
-      "aarch64-unknown-linux-gnu",
-      "aarch64-unknown-linux-musl",
      "aarch64-apple-darwin",
      "x86_64-pc-windows-msvc"
    ]
@@ -52,11 +49,8 @@
    "@napi-rs/cli": "^3.4.1"
  },
  "optionalDependencies": {
-    "@firecrawl/pdf-inspector-linux-x64-gnu": "1.12.0",
-    "@firecrawl/pdf-inspector-linux-x64-musl": "1.12.0",
-    "@firecrawl/pdf-inspector-linux-arm64-gnu": "1.12.0",
-    "@firecrawl/pdf-inspector-linux-arm64-musl": "1.12.0",
-    "@firecrawl/pdf-inspector-darwin-arm64": "1.12.0",
-    "@firecrawl/pdf-inspector-win32-x64-msvc": "1.12.0"
+    "@firecrawl/pdf-inspector-linux-x64-gnu": "1.11.2",
+    "@firecrawl/pdf-inspector-darwin-arm64": "1.11.2",
+    "@firecrawl/pdf-inspector-win32-x64-msvc": "1.11.2"
  }
 }
@@ -10,9 +10,6 @@ class PdfResult:
    page_count: int
    processing_time_ms: int
    pages_needing_ocr: list[int]
-    """1-indexed page numbers that need OCR."""
-    ocr_reasons_by_page: list["PageOcrReasons"]
-    """Machine-readable OCR reasons by 1-indexed page."""
    title: Optional[str]
    confidence: float
    is_complex_layout: bool
@@ -20,13 +17,6 @@ class PdfResult:
    pages_with_columns: list[int]
    has_encoding_issues: bool

-class PageOcrReasons:
-    """OCR reasons for a single 1-indexed page."""
-    page: int
-    """1-indexed page number."""
-    reasons: list[str]
-    """Machine-readable OCR reason identifiers."""
-
 class PdfClassification:
    """Lightweight PDF classification result."""
    pdf_type: str
@@ -57,8 +47,6 @@ class RegionText:
    text: str
    needs_ocr: bool
    """True when the text should not be trusted."""
-    ocr_reason: Optional[str]
-    """Machine-readable OCR reason when the cause is known."""

 class PageRegionTexts:
    """Extracted text for one page's regions."""
@@ -74,8 +62,6 @@ class PageMarkdown:
    """Formatted markdown for this page (empty string when needs_ocr is True)."""
    needs_ocr: bool
    """True when text on this page is unreliable and OCR should be used instead."""
-    ocr_reason: Optional[str]
-    """Machine-readable OCR reason when the cause is known."""

 class PagesExtractionResult:
    """Per-page markdown output with document-wide layout classification."""
@@ -87,8 +73,6 @@ class PagesExtractionResult:
    """1-indexed pages where multi-column layout was detected."""
    pages_needing_ocr: list[int]
    """1-indexed pages that need OCR."""
-    ocr_reasons_by_page: list[PageOcrReasons]
-    """Machine-readable OCR reasons by 1-indexed page."""
    is_complex: bool
    """True if any page has tables or multi-column layout."""

@@ -2,8 +2,8 @@

 use pdf_inspector::extractor::ItemType;
 use pdf_inspector::{
-    extract_text_with_positions_pages_with_password, process_pdf_with_options, LayoutComplexity,
-    PdfOptions, PdfType, ProcessMode, TextItem,
+    extract_text_with_positions_pages, process_pdf_with_options, LayoutComplexity, PdfOptions,
+    PdfType, ProcessMode, TextItem,
 };
 use std::collections::HashSet;
 use std::env;
@@ -103,18 +103,9 @@ fn format_items_json(items: &[TextItem]) -> String {
    )
 }

-fn extract_items_json(
-    pdf_path: &str,
-    page_filter: Option<&HashSet<u32>>,
-    password: Option<&str>,
-) -> Result<String, pdf_inspector::PdfError> {
-    extract_text_with_positions_pages_with_password(pdf_path, page_filter, password)
-        .map(|items| format_items_json(&items))
-}
-
 #[cfg(test)]
 mod tests {
-    use super::{extract_items_json, format_items_json};
+    use super::format_items_json;
    use pdf_inspector::extractor::ItemType;
    use pdf_inspector::TextItem;

@@ -146,24 +137,6 @@ mod tests {
        assert!(json.contains(r#""item_type":"text""#));
        assert!(json.contains(r#""mcid":7"#));
    }
-
-    #[test]
-    fn items_json_uses_supplied_pdf_password() {
-        let path = "tests/fixtures/encrypted-secret123.pdf";
-
-        let without_password = extract_items_json(path, None, None);
-        assert!(
-            without_password.is_err(),
-            "encrypted fixture unexpectedly extracted without a password"
-        );
-
-        let json = extract_items_json(path, None, Some("secret123"))
-            .expect("correct password should decrypt positioned text");
-        assert!(
-            json.contains("Procurement"),
-            "decrypted item JSON should contain fixture text, got {json}"
-        );
-    }
 }

 /// Parse a page specification like "1,3,5-10,20" into a HashSet of page numbers.
@@ -284,8 +257,8 @@ fn main() {
        });

    if items_json_output {
-        match extract_items_json(pdf_path, page_filter.as_ref(), password.as_deref()) {
-            Ok(json) => println!("{}", json),
+        match extract_text_with_positions_pages(pdf_path, page_filter.as_ref()) {
+            Ok(items) => println!("{}", format_items_json(&items)),
            Err(e) => {
                println!(r#"{{"error":"{}"}}"#, json_escape(&e.to_string()));
                process::exit(1);
@@ -150,11 +150,134 @@ static COURIER: &[(char, u16)] = &[
    ('\u{FB02}', 600),
 ];

-static COURIER_BOLD: &[(char, u16)] = COURIER;
+#[rustfmt::skip]
+static COURIER_BOLD: &[(char, u16)] = &[
+    (' ', 600), ('!', 600), ('"', 600), ('#', 600), ('$', 600), ('%', 600),
+    ('&', 600), ('\'', 600), ('(', 600), (')', 600), ('*', 600), ('+', 600),
+    (',', 600), ('-', 600), ('.', 600), ('/', 600), ('0', 600), ('1', 600),
+    ('2', 600), ('3', 600), ('4', 600), ('5', 600), ('6', 600), ('7', 600),
+    ('8', 600), ('9', 600), (':', 600), (';', 600), ('<', 600), ('=', 600),
+    ('>', 600), ('?', 600), ('@', 600), ('A', 600), ('B', 600), ('C', 600),
+    ('D', 600), ('E', 600), ('F', 600), ('G', 600), ('H', 600), ('I', 600),
+    ('J', 600), ('K', 600), ('L', 600), ('M', 600), ('N', 600), ('O', 600),
+    ('P', 600), ('Q', 600), ('R', 600), ('S', 600), ('T', 600), ('U', 600),
+    ('V', 600), ('W', 600), ('X', 600), ('Y', 600), ('Z', 600), ('[', 600),
+    ('\\', 600), (']', 600), ('^', 600), ('_', 600), ('`', 600), ('a', 600),
+    ('b', 600), ('c', 600), ('d', 600), ('e', 600), ('f', 600), ('g', 600),
+    ('h', 600), ('i', 600), ('j', 600), ('k', 600), ('l', 600), ('m', 600),
+    ('n', 600), ('o', 600), ('p', 600), ('q', 600), ('r', 600), ('s', 600),
+    ('t', 600), ('u', 600), ('v', 600), ('w', 600), ('x', 600), ('y', 600),
+    ('z', 600), ('{', 600), ('|', 600), ('}', 600), ('~', 600), ('\u{00A1}', 600),
+    ('\u{00A2}', 600), ('\u{00A3}', 600), ('\u{00A4}', 600), ('\u{00A5}', 600), ('\u{00A6}', 600), ('\u{00A7}', 600),
+    ('\u{00A8}', 600), ('\u{00A9}', 600), ('\u{00AA}', 600), ('\u{00AB}', 600), ('\u{00AC}', 600), ('\u{00AE}', 600),
+    ('\u{00AF}', 600), ('\u{00B0}', 600), ('\u{00B1}', 600), ('\u{00B2}', 600), ('\u{00B3}', 600), ('\u{00B4}', 600),
+    ('\u{00B5}', 600), ('\u{00B6}', 600), ('\u{00B7}', 600), ('\u{00B8}', 600), ('\u{00B9}', 600), ('\u{00BA}', 600),
+    ('\u{00BB}', 600), ('\u{00BC}', 600), ('\u{00BD}', 600), ('\u{00BE}', 600), ('\u{00BF}', 600), ('\u{00C0}', 600),
+    ('\u{00C1}', 600), ('\u{00C2}', 600), ('\u{00C3}', 600), ('\u{00C4}', 600), ('\u{00C5}', 600), ('\u{00C6}', 600),
+    ('\u{00C7}', 600), ('\u{00C8}', 600), ('\u{00C9}', 600), ('\u{00CA}', 600), ('\u{00CB}', 600), ('\u{00CC}', 600),
+    ('\u{00CD}', 600), ('\u{00CE}', 600), ('\u{00CF}', 600), ('\u{00D0}', 600), ('\u{00D1}', 600), ('\u{00D2}', 600),
+    ('\u{00D3}', 600), ('\u{00D4}', 600), ('\u{00D5}', 600), ('\u{00D6}', 600), ('\u{00D7}', 600), ('\u{00D8}', 600),
+    ('\u{00D9}', 600), ('\u{00DA}', 600), ('\u{00DB}', 600), ('\u{00DC}', 600), ('\u{00DD}', 600), ('\u{00DE}', 600),
+    ('\u{00DF}', 600), ('\u{00E0}', 600), ('\u{00E1}', 600), ('\u{00E2}', 600), ('\u{00E3}', 600), ('\u{00E4}', 600),
+    ('\u{00E5}', 600), ('\u{00E6}', 600), ('\u{00E7}', 600), ('\u{00E8}', 600), ('\u{00E9}', 600), ('\u{00EA}', 600),
+    ('\u{00EB}', 600), ('\u{00EC}', 600), ('\u{00ED}', 600), ('\u{00EE}', 600), ('\u{00EF}', 600), ('\u{00F0}', 600),
+    ('\u{00F1}', 600), ('\u{00F2}', 600), ('\u{00F3}', 600), ('\u{00F4}', 600), ('\u{00F5}', 600), ('\u{00F6}', 600),
+    ('\u{00F7}', 600), ('\u{00F8}', 600), ('\u{00F9}', 600), ('\u{00FA}', 600), ('\u{00FB}', 600), ('\u{00FC}', 600),
+    ('\u{00FD}', 600), ('\u{00FE}', 600), ('\u{00FF}', 600), ('\u{0131}', 600), ('\u{0141}', 600), ('\u{0142}', 600),
+    ('\u{0152}', 600), ('\u{0153}', 600), ('\u{0160}', 600), ('\u{0161}', 600), ('\u{0178}', 600), ('\u{017D}', 600),
+    ('\u{017E}', 600), ('\u{0192}', 600), ('\u{02C6}', 600), ('\u{02C7}', 600), ('\u{02D8}', 600), ('\u{02D9}', 600),
+    ('\u{02DA}', 600), ('\u{02DB}', 600), ('\u{02DC}', 600), ('\u{02DD}', 600), ('\u{2013}', 600), ('\u{2014}', 600),
+    ('\u{2018}', 600), ('\u{2019}', 600), ('\u{201A}', 600), ('\u{201C}', 600), ('\u{201D}', 600), ('\u{201E}', 600),
+    ('\u{2020}', 600), ('\u{2021}', 600), ('\u{2022}', 600), ('\u{2026}', 600), ('\u{2030}', 600), ('\u{2039}', 600),
+    ('\u{203A}', 600), ('\u{2044}', 600), ('\u{20AC}', 600), ('\u{2122}', 600), ('\u{2212}', 600), ('\u{FB01}', 600),
+    ('\u{FB02}', 600),
+];

-static COURIER_OBLIQUE: &[(char, u16)] = COURIER;
+#[rustfmt::skip]
+static COURIER_OBLIQUE: &[(char, u16)] = &[
+    (' ', 600), ('!', 600), ('"', 600), ('#', 600), ('$', 600), ('%', 600),
+    ('&', 600), ('\'', 600), ('(', 600), (')', 600), ('*', 600), ('+', 600),
+    (',', 600), ('-', 600), ('.', 600), ('/', 600), ('0', 600), ('1', 600),
+    ('2', 600), ('3', 600), ('4', 600), ('5', 600), ('6', 600), ('7', 600),
+    ('8', 600), ('9', 600), (':', 600), (';', 600), ('<', 600), ('=', 600),
+    ('>', 600), ('?', 600), ('@', 600), ('A', 600), ('B', 600), ('C', 600),
+    ('D', 600), ('E', 600), ('F', 600), ('G', 600), ('H', 600), ('I', 600),
+    ('J', 600), ('K', 600), ('L', 600), ('M', 600), ('N', 600), ('O', 600),
+    ('P', 600), ('Q', 600), ('R', 600), ('S', 600), ('T', 600), ('U', 600),
+    ('V', 600), ('W', 600), ('X', 600), ('Y', 600), ('Z', 600), ('[', 600),
+    ('\\', 600), (']', 600), ('^', 600), ('_', 600), ('`', 600), ('a', 600),
+    ('b', 600), ('c', 600), ('d', 600), ('e', 600), ('f', 600), ('g', 600),
+    ('h', 600), ('i', 600), ('j', 600), ('k', 600), ('l', 600), ('m', 600),
+    ('n', 600), ('o', 600), ('p', 600), ('q', 600), ('r', 600), ('s', 600),
+    ('t', 600), ('u', 600), ('v', 600), ('w', 600), ('x', 600), ('y', 600),
+    ('z', 600), ('{', 600), ('|', 600), ('}', 600), ('~', 600), ('\u{00A1}', 600),
+    ('\u{00A2}', 600), ('\u{00A3}', 600), ('\u{00A4}', 600), ('\u{00A5}', 600), ('\u{00A6}', 600), ('\u{00A7}', 600),
+    ('\u{00A8}', 600), ('\u{00A9}', 600), ('\u{00AA}', 600), ('\u{00AB}', 600), ('\u{00AC}', 600), ('\u{00AE}', 600),
+    ('\u{00AF}', 600), ('\u{00B0}', 600), ('\u{00B1}', 600), ('\u{00B2}', 600), ('\u{00B3}', 600), ('\u{00B4}', 600),
+    ('\u{00B5}', 600), ('\u{00B6}', 600), ('\u{00B7}', 600), ('\u{00B8}', 600), ('\u{00B9}', 600), ('\u{00BA}', 600),
+    ('\u{00BB}', 600), ('\u{00BC}', 600), ('\u{00BD}', 600), ('\u{00BE}', 600), ('\u{00BF}', 600), ('\u{00C0}', 600),
+    ('\u{00C1}', 600), ('\u{00C2}', 600), ('\u{00C3}', 600), ('\u{00C4}', 600), ('\u{00C5}', 600), ('\u{00C6}', 600),
+    ('\u{00C7}', 600), ('\u{00C8}', 600), ('\u{00C9}', 600), ('\u{00CA}', 600), ('\u{00CB}', 600), ('\u{00CC}', 600),
+    ('\u{00CD}', 600), ('\u{00CE}', 600), ('\u{00CF}', 600), ('\u{00D0}', 600), ('\u{00D1}', 600), ('\u{00D2}', 600),
+    ('\u{00D3}', 600), ('\u{00D4}', 600), ('\u{00D5}', 600), ('\u{00D6}', 600), ('\u{00D7}', 600), ('\u{00D8}', 600),
+    ('\u{00D9}', 600), ('\u{00DA}', 600), ('\u{00DB}', 600), ('\u{00DC}', 600), ('\u{00DD}', 600), ('\u{00DE}', 600),
+    ('\u{00DF}', 600), ('\u{00E0}', 600), ('\u{00E1}', 600), ('\u{00E2}', 600), ('\u{00E3}', 600), ('\u{00E4}', 600),
+    ('\u{00E5}', 600), ('\u{00E6}', 600), ('\u{00E7}', 600), ('\u{00E8}', 600), ('\u{00E9}', 600), ('\u{00EA}', 600),
+    ('\u{00EB}', 600), ('\u{00EC}', 600), ('\u{00ED}', 600), ('\u{00EE}', 600), ('\u{00EF}', 600), ('\u{00F0}', 600),
+    ('\u{00F1}', 600), ('\u{00F2}', 600), ('\u{00F3}', 600), ('\u{00F4}', 600), ('\u{00F5}', 600), ('\u{00F6}', 600),
+    ('\u{00F7}', 600), ('\u{00F8}', 600), ('\u{00F9}', 600), ('\u{00FA}', 600), ('\u{00FB}', 600), ('\u{00FC}', 600),
+    ('\u{00FD}', 600), ('\u{00FE}', 600), ('\u{00FF}', 600), ('\u{0131}', 600), ('\u{0141}', 600), ('\u{0142}', 600),
+    ('\u{0152}', 600), ('\u{0153}', 600), ('\u{0160}', 600), ('\u{0161}', 600), ('\u{0178}', 600), ('\u{017D}', 600),
+    ('\u{017E}', 600), ('\u{0192}', 600), ('\u{02C6}', 600), ('\u{02C7}', 600), ('\u{02D8}', 600), ('\u{02D9}', 600),
+    ('\u{02DA}', 600), ('\u{02DB}', 600), ('\u{02DC}', 600), ('\u{02DD}', 600), ('\u{2013}', 600), ('\u{2014}', 600),
+    ('\u{2018}', 600), ('\u{2019}', 600), ('\u{201A}', 600), ('\u{201C}', 600), ('\u{201D}', 600), ('\u{201E}', 600),
+    ('\u{2020}', 600), ('\u{2021}', 600), ('\u{2022}', 600), ('\u{2026}', 600), ('\u{2030}', 600), ('\u{2039}', 600),
+    ('\u{203A}', 600), ('\u{2044}', 600), ('\u{20AC}', 600), ('\u{2122}', 600), ('\u{2212}', 600), ('\u{FB01}', 600),
+    ('\u{FB02}', 600),
+];

-static COURIER_BOLDOBLIQUE: &[(char, u16)] = COURIER;
+#[rustfmt::skip]
+static COURIER_BOLDOBLIQUE: &[(char, u16)] = &[
+    (' ', 600), ('!', 600), ('"', 600), ('#', 600), ('$', 600), ('%', 600),
+    ('&', 600), ('\'', 600), ('(', 600), (')', 600), ('*', 600), ('+', 600),
+    (',', 600), ('-', 600), ('.', 600), ('/', 600), ('0', 600), ('1', 600),
+    ('2', 600), ('3', 600), ('4', 600), ('5', 600), ('6', 600), ('7', 600),
+    ('8', 600), ('9', 600), (':', 600), (';', 600), ('<', 600), ('=', 600),
+    ('>', 600), ('?', 600), ('@', 600), ('A', 600), ('B', 600), ('C', 600),
+    ('D', 600), ('E', 600), ('F', 600), ('G', 600), ('H', 600), ('I', 600),
+    ('J', 600), ('K', 600), ('L', 600), ('M', 600), ('N', 600), ('O', 600),
+    ('P', 600), ('Q', 600), ('R', 600), ('S', 600), ('T', 600), ('U', 600),
+    ('V', 600), ('W', 600), ('X', 600), ('Y', 600), ('Z', 600), ('[', 600),
+    ('\\', 600), (']', 600), ('^', 600), ('_', 600), ('`', 600), ('a', 600),
+    ('b', 600), ('c', 600), ('d', 600), ('e', 600), ('f', 600), ('g', 600),
+    ('h', 600), ('i', 600), ('j', 600), ('k', 600), ('l', 600), ('m', 600),
+    ('n', 600), ('o', 600), ('p', 600), ('q', 600), ('r', 600), ('s', 600),
+    ('t', 600), ('u', 600), ('v', 600), ('w', 600), ('x', 600), ('y', 600),
+    ('z', 600), ('{', 600), ('|', 600), ('}', 600), ('~', 600), ('\u{00A1}', 600),
+    ('\u{00A2}', 600), ('\u{00A3}', 600), ('\u{00A4}', 600), ('\u{00A5}', 600), ('\u{00A6}', 600), ('\u{00A7}', 600),
+    ('\u{00A8}', 600), ('\u{00A9}', 600), ('\u{00AA}', 600), ('\u{00AB}', 600), ('\u{00AC}', 600), ('\u{00AE}', 600),
+    ('\u{00AF}', 600), ('\u{00B0}', 600), ('\u{00B1}', 600), ('\u{00B2}', 600), ('\u{00B3}', 600), ('\u{00B4}', 600),
+    ('\u{00B5}', 600), ('\u{00B6}', 600), ('\u{00B7}', 600), ('\u{00B8}', 600), ('\u{00B9}', 600), ('\u{00BA}', 600),
+    ('\u{00BB}', 600), ('\u{00BC}', 600), ('\u{00BD}', 600), ('\u{00BE}', 600), ('\u{00BF}', 600), ('\u{00C0}', 600),
+    ('\u{00C1}', 600), ('\u{00C2}', 600), ('\u{00C3}', 600), ('\u{00C4}', 600), ('\u{00C5}', 600), ('\u{00C6}', 600),
+    ('\u{00C7}', 600), ('\u{00C8}', 600), ('\u{00C9}', 600), ('\u{00CA}', 600), ('\u{00CB}', 600), ('\u{00CC}', 600),
+    ('\u{00CD}', 600), ('\u{00CE}', 600), ('\u{00CF}', 600), ('\u{00D0}', 600), ('\u{00D1}', 600), ('\u{00D2}', 600),
+    ('\u{00D3}', 600), ('\u{00D4}', 600), ('\u{00D5}', 600), ('\u{00D6}', 600), ('\u{00D7}', 600), ('\u{00D8}', 600),
+    ('\u{00D9}', 600), ('\u{00DA}', 600), ('\u{00DB}', 600), ('\u{00DC}', 600), ('\u{00DD}', 600), ('\u{00DE}', 600),
+    ('\u{00DF}', 600), ('\u{00E0}', 600), ('\u{00E1}', 600), ('\u{00E2}', 600), ('\u{00E3}', 600), ('\u{00E4}', 600),
+    ('\u{00E5}', 600), ('\u{00E6}', 600), ('\u{00E7}', 600), ('\u{00E8}', 600), ('\u{00E9}', 600), ('\u{00EA}', 600),
+    ('\u{00EB}', 600), ('\u{00EC}', 600), ('\u{00ED}', 600), ('\u{00EE}', 600), ('\u{00EF}', 600), ('\u{00F0}', 600),
+    ('\u{00F1}', 600), ('\u{00F2}', 600), ('\u{00F3}', 600), ('\u{00F4}', 600), ('\u{00F5}', 600), ('\u{00F6}', 600),
+    ('\u{00F7}', 600), ('\u{00F8}', 600), ('\u{00F9}', 600), ('\u{00FA}', 600), ('\u{00FB}', 600), ('\u{00FC}', 600),
+    ('\u{00FD}', 600), ('\u{00FE}', 600), ('\u{00FF}', 600), ('\u{0131}', 600), ('\u{0141}', 600), ('\u{0142}', 600),
+    ('\u{0152}', 600), ('\u{0153}', 600), ('\u{0160}', 600), ('\u{0161}', 600), ('\u{0178}', 600), ('\u{017D}', 600),
+    ('\u{017E}', 600), ('\u{0192}', 600), ('\u{02C6}', 600), ('\u{02C7}', 600), ('\u{02D8}', 600), ('\u{02D9}', 600),
+    ('\u{02DA}', 600), ('\u{02DB}', 600), ('\u{02DC}', 600), ('\u{02DD}', 600), ('\u{2013}', 600), ('\u{2014}', 600),
+    ('\u{2018}', 600), ('\u{2019}', 600), ('\u{201A}', 600), ('\u{201C}', 600), ('\u{201D}', 600), ('\u{201E}', 600),
+    ('\u{2020}', 600), ('\u{2021}', 600), ('\u{2022}', 600), ('\u{2026}', 600), ('\u{2030}', 600), ('\u{2039}', 600),
+    ('\u{203A}', 600), ('\u{2044}', 600), ('\u{20AC}', 600), ('\u{2122}', 600), ('\u{2212}', 600), ('\u{FB01}', 600),
+    ('\u{FB02}', 600),
+];

 #[rustfmt::skip]
 static HELVETICA: &[(char, u16)] = &[
@@ -242,9 +365,91 @@ static HELVETICA_BOLD: &[(char, u16)] = &[
    ('\u{FB02}', 611),
 ];

-static HELVETICA_OBLIQUE: &[(char, u16)] = HELVETICA;
+#[rustfmt::skip]
+static HELVETICA_OBLIQUE: &[(char, u16)] = &[
+    (' ', 278), ('!', 278), ('"', 355), ('#', 556), ('$', 556), ('%', 889),
+    ('&', 667), ('\'', 191), ('(', 333), (')', 333), ('*', 389), ('+', 584),
+    (',', 278), ('-', 333), ('.', 278), ('/', 278), ('0', 556), ('1', 556),
+    ('2', 556), ('3', 556), ('4', 556), ('5', 556), ('6', 556), ('7', 556),
+    ('8', 556), ('9', 556), (':', 278), (';', 278), ('<', 584), ('=', 584),
+    ('>', 584), ('?', 556), ('@', 1015), ('A', 667), ('B', 667), ('C', 722),
+    ('D', 722), ('E', 667), ('F', 611), ('G', 778), ('H', 722), ('I', 278),
+    ('J', 500), ('K', 667), ('L', 556), ('M', 833), ('N', 722), ('O', 778),
+    ('P', 667), ('Q', 778), ('R', 722), ('S', 667), ('T', 611), ('U', 722),
+    ('V', 667), ('W', 944), ('X', 667), ('Y', 667), ('Z', 611), ('[', 278),
+    ('\\', 278), (']', 278), ('^', 469), ('_', 556), ('`', 333), ('a', 556),
+    ('b', 556), ('c', 500), ('d', 556), ('e', 556), ('f', 278), ('g', 556),
+    ('h', 556), ('i', 222), ('j', 222), ('k', 500), ('l', 222), ('m', 833),
+    ('n', 556), ('o', 556), ('p', 556), ('q', 556), ('r', 333), ('s', 500),
+    ('t', 278), ('u', 556), ('v', 500), ('w', 722), ('x', 500), ('y', 500),
+    ('z', 500), ('{', 334), ('|', 260), ('}', 334), ('~', 584), ('\u{00A1}', 333),
+    ('\u{00A2}', 556), ('\u{00A3}', 556), ('\u{00A4}', 556), ('\u{00A5}', 556), ('\u{00A6}', 260), ('\u{00A7}', 556),
+    ('\u{00A8}', 333), ('\u{00A9}', 737), ('\u{00AA}', 370), ('\u{00AB}', 556), ('\u{00AC}', 584), ('\u{00AE}', 737),
+    ('\u{00AF}', 333), ('\u{00B0}', 400), ('\u{00B1}', 584), ('\u{00B2}', 333), ('\u{00B3}', 333), ('\u{00B4}', 333),
+    ('\u{00B5}', 556), ('\u{00B6}', 537), ('\u{00B7}', 278), ('\u{00B8}', 333), ('\u{00B9}', 333), ('\u{00BA}', 365),
+    ('\u{00BB}', 556), ('\u{00BC}', 834), ('\u{00BD}', 834), ('\u{00BE}', 834), ('\u{00BF}', 611), ('\u{00C0}', 667),
+    ('\u{00C1}', 667), ('\u{00C2}', 667), ('\u{00C3}', 667), ('\u{00C4}', 667), ('\u{00C5}', 667), ('\u{00C6}', 1000),
+    ('\u{00C7}', 722), ('\u{00C8}', 667), ('\u{00C9}', 667), ('\u{00CA}', 667), ('\u{00CB}', 667), ('\u{00CC}', 278),
+    ('\u{00CD}', 278), ('\u{00CE}', 278), ('\u{00CF}', 278), ('\u{00D0}', 722), ('\u{00D1}', 722), ('\u{00D2}', 778),
+    ('\u{00D3}', 778), ('\u{00D4}', 778), ('\u{00D5}', 778), ('\u{00D6}', 778), ('\u{00D7}', 584), ('\u{00D8}', 778),
+    ('\u{00D9}', 722), ('\u{00DA}', 722), ('\u{00DB}', 722), ('\u{00DC}', 722), ('\u{00DD}', 667), ('\u{00DE}', 667),
+    ('\u{00DF}', 611), ('\u{00E0}', 556), ('\u{00E1}', 556), ('\u{00E2}', 556), ('\u{00E3}', 556), ('\u{00E4}', 556),
+    ('\u{00E5}', 556), ('\u{00E6}', 889), ('\u{00E7}', 500), ('\u{00E8}', 556), ('\u{00E9}', 556), ('\u{00EA}', 556),
+    ('\u{00EB}', 556), ('\u{00EC}', 278), ('\u{00ED}', 278), ('\u{00EE}', 278), ('\u{00EF}', 278), ('\u{00F0}', 556),
+    ('\u{00F1}', 556), ('\u{00F2}', 556), ('\u{00F3}', 556), ('\u{00F4}', 556), ('\u{00F5}', 556), ('\u{00F6}', 556),
+    ('\u{00F7}', 584), ('\u{00F8}', 611), ('\u{00F9}', 556), ('\u{00FA}', 556), ('\u{00FB}', 556), ('\u{00FC}', 556),
+    ('\u{00FD}', 500), ('\u{00FE}', 556), ('\u{00FF}', 500), ('\u{0131}', 278), ('\u{0141}', 556), ('\u{0142}', 222),
+    ('\u{0152}', 1000), ('\u{0153}', 944), ('\u{0160}', 667), ('\u{0161}', 500), ('\u{0178}', 667), ('\u{017D}', 611),
+    ('\u{017E}', 500), ('\u{0192}', 556), ('\u{02C6}', 333), ('\u{02C7}', 333), ('\u{02D8}', 333), ('\u{02D9}', 333),
+    ('\u{02DA}', 333), ('\u{02DB}', 333), ('\u{02DC}', 333), ('\u{02DD}', 333), ('\u{2013}', 556), ('\u{2014}', 1000),
+    ('\u{2018}', 222), ('\u{2019}', 222), ('\u{201A}', 222), ('\u{201C}', 333), ('\u{201D}', 333), ('\u{201E}', 333),
+    ('\u{2020}', 556), ('\u{2021}', 556), ('\u{2022}', 350), ('\u{2026}', 1000), ('\u{2030}', 1000), ('\u{2039}', 333),
+    ('\u{203A}', 333), ('\u{2044}', 167), ('\u{20AC}', 556), ('\u{2122}', 1000), ('\u{2212}', 584), ('\u{FB01}', 500),
+    ('\u{FB02}', 500),
+];

-static HELVETICA_BOLDOBLIQUE: &[(char, u16)] = HELVETICA_BOLD;
+#[rustfmt::skip]
+static HELVETICA_BOLDOBLIQUE: &[(char, u16)] = &[
+    (' ', 278), ('!', 333), ('"', 474), ('#', 556), ('$', 556), ('%', 889),
+    ('&', 722), ('\'', 238), ('(', 333), (')', 333), ('*', 389), ('+', 584),
+    (',', 278), ('-', 333), ('.', 278), ('/', 278), ('0', 556), ('1', 556),
+    ('2', 556), ('3', 556), ('4', 556), ('5', 556), ('6', 556), ('7', 556),
+    ('8', 556), ('9', 556), (':', 333), (';', 333), ('<', 584), ('=', 584),
+    ('>', 584), ('?', 611), ('@', 975), ('A', 722), ('B', 722), ('C', 722),
+    ('D', 722), ('E', 667), ('F', 611), ('G', 778), ('H', 722), ('I', 278),
+    ('J', 556), ('K', 722), ('L', 611), ('M', 833), ('N', 722), ('O', 778),
+    ('P', 667), ('Q', 778), ('R', 722), ('S', 667), ('T', 611), ('U', 722),
+    ('V', 667), ('W', 944), ('X', 667), ('Y', 667), ('Z', 611), ('[', 333),
+    ('\\', 278), (']', 333), ('^', 584), ('_', 556), ('`', 333), ('a', 556),
+    ('b', 611), ('c', 556), ('d', 611), ('e', 556), ('f', 333), ('g', 611),
+    ('h', 611), ('i', 278), ('j', 278), ('k', 556), ('l', 278), ('m', 889),
+    ('n', 611), ('o', 611), ('p', 611), ('q', 611), ('r', 389), ('s', 556),
+    ('t', 333), ('u', 611), ('v', 556), ('w', 778), ('x', 556), ('y', 556),
+    ('z', 500), ('{', 389), ('|', 280), ('}', 389), ('~', 584), ('\u{00A1}', 333),
+    ('\u{00A2}', 556), ('\u{00A3}', 556), ('\u{00A4}', 556), ('\u{00A5}', 556), ('\u{00A6}', 280), ('\u{00A7}', 556),
+    ('\u{00A8}', 333), ('\u{00A9}', 737), ('\u{00AA}', 370), ('\u{00AB}', 556), ('\u{00AC}', 584), ('\u{00AE}', 737),
+    ('\u{00AF}', 333), ('\u{00B0}', 400), ('\u{00B1}', 584), ('\u{00B2}', 333), ('\u{00B3}', 333), ('\u{00B4}', 333),
+    ('\u{00B5}', 611), ('\u{00B6}', 556), ('\u{00B7}', 278), ('\u{00B8}', 333), ('\u{00B9}', 333), ('\u{00BA}', 365),
+    ('\u{00BB}', 556), ('\u{00BC}', 834), ('\u{00BD}', 834), ('\u{00BE}', 834), ('\u{00BF}', 611), ('\u{00C0}', 722),
+    ('\u{00C1}', 722), ('\u{00C2}', 722), ('\u{00C3}', 722), ('\u{00C4}', 722), ('\u{00C5}', 722), ('\u{00C6}', 1000),
+    ('\u{00C7}', 722), ('\u{00C8}', 667), ('\u{00C9}', 667), ('\u{00CA}', 667), ('\u{00CB}', 667), ('\u{00CC}', 278),
+    ('\u{00CD}', 278), ('\u{00CE}', 278), ('\u{00CF}', 278), ('\u{00D0}', 722), ('\u{00D1}', 722), ('\u{00D2}', 778),
+    ('\u{00D3}', 778), ('\u{00D4}', 778), ('\u{00D5}', 778), ('\u{00D6}', 778), ('\u{00D7}', 584), ('\u{00D8}', 778),
+    ('\u{00D9}', 722), ('\u{00DA}', 722), ('\u{00DB}', 722), ('\u{00DC}', 722), ('\u{00DD}', 667), ('\u{00DE}', 667),
+    ('\u{00DF}', 611), ('\u{00E0}', 556), ('\u{00E1}', 556), ('\u{00E2}', 556), ('\u{00E3}', 556), ('\u{00E4}', 556),
+    ('\u{00E5}', 556), ('\u{00E6}', 889), ('\u{00E7}', 556), ('\u{00E8}', 556), ('\u{00E9}', 556), ('\u{00EA}', 556),
+    ('\u{00EB}', 556), ('\u{00EC}', 278), ('\u{00ED}', 278), ('\u{00EE}', 278), ('\u{00EF}', 278), ('\u{00F0}', 611),
+    ('\u{00F1}', 611), ('\u{00F2}', 611), ('\u{00F3}', 611), ('\u{00F4}', 611), ('\u{00F5}', 611), ('\u{00F6}', 611),
+    ('\u{00F7}', 584), ('\u{00F8}', 611), ('\u{00F9}', 611), ('\u{00FA}', 611), ('\u{00FB}', 611), ('\u{00FC}', 611),
+    ('\u{00FD}', 556), ('\u{00FE}', 611), ('\u{00FF}', 556), ('\u{0131}', 278), ('\u{0141}', 611), ('\u{0142}', 278),
+    ('\u{0152}', 1000), ('\u{0153}', 944), ('\u{0160}', 667), ('\u{0161}', 556), ('\u{0178}', 667), ('\u{017D}', 611),
+    ('\u{017E}', 500), ('\u{0192}', 556), ('\u{02C6}', 333), ('\u{02C7}', 333), ('\u{02D8}', 333), ('\u{02D9}', 333),
+    ('\u{02DA}', 333), ('\u{02DB}', 333), ('\u{02DC}', 333), ('\u{02DD}', 333), ('\u{2013}', 556), ('\u{2014}', 1000),
+    ('\u{2018}', 278), ('\u{2019}', 278), ('\u{201A}', 278), ('\u{201C}', 500), ('\u{201D}', 500), ('\u{201E}', 500),
+    ('\u{2020}', 556), ('\u{2021}', 556), ('\u{2022}', 350), ('\u{2026}', 1000), ('\u{2030}', 1000), ('\u{2039}', 333),
+    ('\u{203A}', 333), ('\u{2044}', 167), ('\u{20AC}', 556), ('\u{2122}', 1000), ('\u{2212}', 584), ('\u{FB01}', 611),
+    ('\u{FB02}', 611),
+];

 #[rustfmt::skip]
 static TIMES_ROMAN: &[(char, u16)] = &[
@@ -566,25 +771,6 @@ static ZAPFDINGBATS_ENCODING: &[(u8, char)] = &[
    (0xFB, '\u{27BB}'), (0xFC, '\u{27BC}'), (0xFD, '\u{27BD}'), (0xFE, '\u{27BE}'),
 ];

-/// Every width table, for exhaustive testing.
-#[cfg(test)]
-static ALL_TABLES: &[(&str, &[(char, u16)])] = &[
-    ("COURIER", COURIER),
-    ("COURIER_BOLD", COURIER_BOLD),
-    ("COURIER_OBLIQUE", COURIER_OBLIQUE),
-    ("COURIER_BOLDOBLIQUE", COURIER_BOLDOBLIQUE),
-    ("HELVETICA", HELVETICA),
-    ("HELVETICA_BOLD", HELVETICA_BOLD),
-    ("HELVETICA_OBLIQUE", HELVETICA_OBLIQUE),
-    ("HELVETICA_BOLDOBLIQUE", HELVETICA_BOLDOBLIQUE),
-    ("TIMES_ROMAN", TIMES_ROMAN),
-    ("TIMES_BOLD", TIMES_BOLD),
-    ("TIMES_ITALIC", TIMES_ITALIC),
-    ("TIMES_BOLDITALIC", TIMES_BOLDITALIC),
-    ("SYMBOL", SYMBOL),
-    ("ZAPFDINGBATS", ZAPFDINGBATS),
-];
-
 #[cfg(test)]
 mod tests {
    use super::*;
@@ -637,13 +823,15 @@ mod tests {

    #[test]
    fn tables_are_sorted_for_binary_search() {
-        // Every table is queried by binary search, so all of them must be
-        // sorted — not just a sample.
-        for (name, table) in ALL_TABLES {
-            assert!(
-                table.windows(2).all(|w| w[0].0 < w[1].0),
-                "{name} is not sorted"
-            );
+        for table in [
+            TIMES_ROMAN,
+            TIMES_ITALIC,
+            HELVETICA,
+            COURIER,
+            SYMBOL,
+            ZAPFDINGBATS,
+        ] {
+            assert!(table.windows(2).all(|w| w[0].0 < w[1].0));
        }
    }
 }
@@ -41,17 +41,6 @@ fn strip_pdf_comments(data: &[u8]) -> Vec<u8> {
    while i < data.len() {
        let b = data[i];
        match b {
-            // Inside a string literal, a backslash escapes the next byte —
-            // `\(`, `\)`, and `\\` must not touch the nesting depth, or a
-            // later `%` glyph inside a string gets stripped as a comment,
-            // corrupting the stream.
-            b'\\' if in_string > 0 => {
-                result.push(b);
-                if let Some(&next) = data.get(i + 1) {
-                    result.push(next);
-                    i += 1;
-                }
-            }
            b'(' if !in_hex_string => {
                in_string += 1;
                result.push(b);
@@ -1865,26 +1854,4 @@ BT 30 700 Tm <41> Tj ET";
            "ET should be preserved after comment stripping"
        );
    }
-
-    #[test]
-    fn test_strip_pdf_comments_escaped_parens() {
-        // An escaped `\)` must not close the string: the `%` after it is
-        // still string content, not a comment (subset fonts routinely map
-        // glyphs to `%` and to escaped parens in the same TJ array).
-        let input = b"[ (a\\)b) 1 (%) 1 (c) ] TJ\n";
-        let output = strip_pdf_comments(input);
-        assert_eq!(output, input.to_vec());
-
-        // Same for an escaped `\(` — must not open a phantom string that
-        // shields a real comment.
-        let input = b"(x\\(y) Tj % real comment\nET\n";
-        let output = strip_pdf_comments(input);
-        assert_eq!(output, b"(x\\(y) Tj  \nET\n");
-
-        // Escaped backslash before a real close-paren: `\\` ends the escape,
-        // the `)` does close the string, and the comment is stripped.
-        let input = b"(x\\\\) Tj % comment\nET\n";
-        let output = strip_pdf_comments(input);
-        assert_eq!(output, b"(x\\\\) Tj  \nET\n");
-    }
 }
@@ -198,27 +198,9 @@ pub(crate) fn build_type3_scales(
        let scale_y = (num(&matrix[2]).powi(2) + num(&matrix[3]).powi(2)).sqrt();
        let bbox_h = (num(&bbox[3]) - num(&bbox[1])).abs();
        let scale = bbox_h * scale_y;
-
-        // `scale` is the glyph box measured in text-space units. For a
-        // self-consistent font it lands near 1.0 — the FontMatrix is the
-        // reciprocal of the glyph-space em by construction — so the Tf
-        // operand is already the rendered size and must be left alone.
-        // A modest deviation is normal and must NOT trigger rescaling:
-        // FontBBox is the glyph bounding box, not the em box, so it is
-        // routinely somewhat smaller (descender..ascender ≈ 0.7) or larger
-        // (tall accents > 1.0).
-        //
-        // Only a wildly inconsistent font gets renormalized. dvips/PK
-        // bitmap fonts declare [1 0 0 -1 0 0] with glyphs spanning
-        // hundreds of units, giving scale ≈ 159 against a nominal size of
-        // 0.12pt — there the declared size carries no information. The
-        // band is deliberately wide so that only that class qualifies,
-        // while any matrix scale (including non-standard ones like 0.005
-        // with a full-em bbox, scale = 5.0) is judged on the product
-        // rather than on the matrix alone.
-        const CONSISTENT_LO: f32 = 0.25;
-        const CONSISTENT_HI: f32 = 4.0;
-        if scale.is_finite() && scale > 0.0 && !(CONSISTENT_LO..=CONSISTENT_HI).contains(&scale) {
+        // Only store meaningful scales; degenerate bboxes ([0 0 0 0] is legal)
+        // and near-1.0 factors keep the nominal size.
+        if scale > 0.01 && (scale - 1.0).abs() > 0.05 {
            scales.insert(String::from_utf8_lossy(font_name).to_string(), scale);
        }
    }
@@ -1654,55 +1636,6 @@ mod tests {
        );
    }

-    /// Build a one-font Type3 document and return its computed scale, if any.
-    #[cfg(test)]
-    fn type3_scale_for(matrix_y: f32, bbox_lo: i64, bbox_hi: i64) -> Option<f32> {
-        use lopdf::{dictionary, Document, Object};
-        let doc = Document::with_version("1.4");
-        let font_dict = dictionary! {
-            "Type" => "Font",
-            "Subtype" => "Type3",
-            "FontMatrix" => vec![
-                Object::Real(matrix_y), Object::Integer(0), Object::Integer(0),
-                Object::Real(matrix_y), Object::Integer(0), Object::Integer(0),
-            ],
-            "FontBBox" => vec![
-                Object::Integer(0), Object::Integer(bbox_lo),
-                Object::Integer(600), Object::Integer(bbox_hi),
-            ],
-        };
-        let mut fonts = std::collections::BTreeMap::new();
-        fonts.insert(b"T9".to_vec(), &font_dict);
-        super::build_type3_scales(&doc, &fonts).get("T9").copied()
-    }
-
-    #[test]
-    fn type3_scale_skips_self_consistent_fonts() {
-        // Conventional 1/1000 matrix with a descender..ascender bbox of 700
-        // units: scale 0.7. The Tf operand is already the rendered size, so
-        // renormalizing would report every size at 0.7x.
-        assert_eq!(type3_scale_for(0.001, -200, 500), None);
-        // Tall-accent bbox slightly over the em (1100 units, scale 1.1).
-        assert_eq!(type3_scale_for(0.001, -100, 1000), None);
-    }
-
-    #[test]
-    fn type3_scale_applies_to_inconsistent_fonts_at_any_matrix_scale() {
-        // Non-standard but valid matrix (0.005) with a full-em bbox:
-        // scale 5.0, so the declared size is off by 5x and must be fixed.
-        let s = type3_scale_for(0.005, 0, 1000).expect("0.005 matrix should rescale");
-        assert!((s - 5.0).abs() < 0.01, "got {s}");
-        // dvips/PK bitmap pattern: unit matrix, glyphs spanning ~159 units.
-        let s = type3_scale_for(1.0, -156, 3).expect("PK pattern should rescale");
-        assert!((s - 159.0).abs() < 0.5, "got {s}");
-    }
-
-    #[test]
-    fn type3_scale_ignores_degenerate_bbox() {
-        // [0 0 0 0] is legal and carries no size information.
-        assert_eq!(type3_scale_for(0.001, 0, 0), None);
-    }
-
    #[test]
    fn texcm_math_symbols_remap() {
        assert_eq!(
@@ -960,713 +960,22 @@ fn spans_multiple_columns(item: &TextItem, columns: &[ColumnRegion]) -> bool {
    overlap_count >= 2
 }

-const PAGE_NUMBER_Y_TOLERANCE: f32 = 3.0;
-const PAGE_NUMBER_CONTEXT_GAP_EM: f32 = 1.5;
-const PAGE_NUMBER_BOTTOM_Y: f32 = 100.0;
-const PAGE_NUMBER_TOP_Y: f32 = 720.0;
-const SPREAD_MIN_CONTENT_WIDTH_EM: f32 = 40.0;
-const SPREAD_EDGE_FRACTION: f32 = 0.25;
-const ADJACENT_PAGE_MIN_CONTENT_WIDTH_EM: f32 = 26.0;
-
-type ContextualCandidateOccurrence = (u32, f32, Vec<(usize, u32)>);
-
-fn page_number_value(item: &TextItem) -> Option<u32> {
-    if !matches!(
-        item.item_type,
-        crate::types::ItemType::Text | crate::types::ItemType::FormField
-    ) {
-        return None;
-    }
-
+/// Check if a text item is likely a page number
+fn is_page_number(item: &TextItem) -> bool {
    let text = item.text.trim();

-    if text.is_empty() || text.len() > 4 || !text.chars().all(|c| c.is_ascii_digit()) {
-        return None;
+    // Must be 1-4 digits only
+    if text.is_empty() || text.len() > 4 {
+        return false;
+    }
+    if !text.chars().all(|c| c.is_ascii_digit()) {
+        return false;
    }

    // Must be at top or bottom of page.
    // US Letter = 792pt, A4 = 841pt. Page numbers are typically in the
    // top ~5% or bottom ~12% of the page.
-    if item.y <= PAGE_NUMBER_TOP_Y && item.y >= PAGE_NUMBER_BOTTOM_Y {
-        return None;
-    }
-
-    text.parse().ok()
-}
-
-/// Mark numeric slots that advance inside a repeated deep-margin line.
-///
-/// A folio can be emitted as part of a footer text run (for example,
-/// `42 Company report`) and therefore look contextual on a single page. Across
-/// the document, however, the surrounding text and Y position repeat while the
-/// numeric slot advances. Require that full signal before treating the slot as
-/// a folio so constant metadata and substantive rows near the page edge remain
-/// untouched.
-fn mark_repeated_folio_candidates(
-    occurrences_by_signature: HashMap<String, Vec<ContextualCandidateOccurrence>>,
-    document_page_count: usize,
-    explicit_folio: &mut [bool],
-) {
-    // Both thresholds are evidence floors: short documents still need four
-    // occurrences, while long documents also need meaningful coverage. Using
-    // `min` here would make one occurrence sufficient in a one-page document.
-    let min_pages = 4usize.max((document_page_count * 30).div_ceil(100));
-
-    for occurrences in occurrences_by_signature.into_values() {
-        if occurrences.len() < min_pages {
-            continue;
-        }
-
-        let distinct_pages: HashSet<u32> = occurrences.iter().map(|(page, _, _)| *page).collect();
-        // Repeated table rows or duplicated drawing labels can share a
-        // signature multiple times on one page. They are not running folios.
-        if distinct_pages.len() != occurrences.len() || distinct_pages.len() < min_pages {
-            continue;
-        }
-
-        let min_y = occurrences
-            .iter()
-            .map(|(_, y, _)| *y)
-            .fold(f32::INFINITY, f32::min);
-        let max_y = occurrences
-            .iter()
-            .map(|(_, y, _)| *y)
-            .fold(f32::NEG_INFINITY, f32::max);
-        if max_y - min_y >= PAGE_NUMBER_Y_TOLERANCE {
-            continue;
-        }
-
-        let slot_count = occurrences[0].2.len();
-        if slot_count == 0
-            || occurrences
-                .iter()
-                .any(|(_, _, candidates)| candidates.len() != slot_count)
-        {
-            continue;
-        }
-
-        for slot in 0..slot_count {
-            let mut values: Vec<(u32, u32, usize)> = occurrences
-                .iter()
-                .map(|(page, _, candidates)| {
-                    let (index, value) = candidates[slot];
-                    (*page, value, index)
-                })
-                .collect();
-            values.sort_by_key(|(page, _, _)| *page);
-
-            let unique_values: HashSet<u32> = values.iter().map(|(_, value, _)| *value).collect();
-            let mostly_unique = unique_values.len() * 5 >= values.len() * 4;
-            let page_tracking_pairs = values
-                .windows(2)
-                .filter(|pair| {
-                    let page_delta = pair[1].0 - pair[0].0;
-                    let value_delta = pair[1].1.saturating_sub(pair[0].1);
-                    value_delta == page_delta || value_delta == page_delta.saturating_mul(2)
-                })
-                .count();
-            let mostly_tracks_page_order = page_tracking_pairs * 5 >= (values.len() - 1) * 4;
-            // A running folio can be offset by front matter or advance twice per
-            // PDF page in a two-page spread, but its magnitude should still be
-            // plausible for the document. This keeps changing metadata such as
-            // a sequence of years from becoming a deletion signal.
-            let max_plausible_folio = (document_page_count as u32).saturating_mul(4).max(100);
-            let plausible_magnitude = values
-                .iter()
-                .all(|(_, value, _)| *value <= max_plausible_folio);
-
-            if mostly_unique && mostly_tracks_page_order && plausible_magnitude {
-                for (_, _, index) in values {
-                    explicit_folio[index] = true;
-                }
-            }
-        }
-    }
-}
-
-/// Mark the contextual half of a facing-page folio pair.
-///
-/// A landscape PDF can contain two printed pages per PDF page. One folio may be
-/// isolated while the other touches footer text; they remain a pair because
-/// they are consecutive, share a deep-margin baseline, and sit on opposite
-/// sides of the spread. The isolated half is strong evidence that the touching
-/// half is also a folio.
-fn mark_spread_folio_pairs(
-    items: &[TextItem],
-    candidate_values: &[Option<u32>],
-    contextual: &[bool],
-    explicit_folio: &mut [bool],
-) {
-    let mut candidates_by_page: HashMap<u32, Vec<usize>> = HashMap::new();
-    let mut page_bounds: HashMap<u32, (f32, f32)> = HashMap::new();
-    for (index, value) in candidate_values.iter().enumerate() {
-        if value.is_some() {
-            candidates_by_page
-                .entry(items[index].page)
-                .or_default()
-                .push(index);
-        }
-    }
-    for item in items.iter().filter(|item| !item.text.trim().is_empty()) {
-        let bounds = page_bounds
-            .entry(item.page)
-            .or_insert((f32::INFINITY, f32::NEG_INFINITY));
-        bounds.0 = bounds.0.min(item.x);
-        bounds.1 = bounds.1.max(item.x + effective_width(item));
-    }
-
-    for (page, page_candidates) in candidates_by_page {
-        let Some(&(page_left, page_right)) = page_bounds.get(&page) else {
-            continue;
-        };
-        let page_width = page_right - page_left;
-        if page_width <= 0.0 {
-            continue;
-        }
-        let left_edge = page_left + page_width * SPREAD_EDGE_FRACTION;
-        let right_edge = page_right - page_width * SPREAD_EDGE_FRACTION;
-        let max_pair_font_size = page_width / SPREAD_MIN_CONTENT_WIDTH_EM;
-        let edge_side = |index: usize| {
-            let center = items[index].x + effective_width(&items[index]) / 2.0;
-            if center <= left_edge {
-                Some(false)
-            } else if center >= right_edge {
-                Some(true)
-            } else {
-                None
-            }
-        };
-
-        // Index strong folio evidence by value and spread edge. Sorted
-        // baselines let each contextual candidate query only the two adjacent
-        // values on the opposite edge in O(log n), rather than comparing every
-        // candidate pair on numeric-heavy pages.
-        let mut known_baselines: HashMap<(u32, bool), Vec<f32>> = HashMap::new();
-        for &index in &page_candidates {
-            if (contextual[index] && !explicit_folio[index])
-                || items[index].font_size >= max_pair_font_size
-            {
-                continue;
-            }
-            let Some(side) = edge_side(index) else {
-                continue;
-            };
-            let value = candidate_values[index].unwrap();
-            known_baselines
-                .entry((value, side))
-                .or_default()
-                .push(items[index].y);
-        }
-        for baselines in known_baselines.values_mut() {
-            baselines.sort_by(f32::total_cmp);
-        }
-
-        for index in page_candidates {
-            if !contextual[index]
-                || explicit_folio[index]
-                || items[index].font_size >= max_pair_font_size
-            {
-                continue;
-            }
-            let Some(value) = candidate_values[index] else {
-                continue;
-            };
-            let Some(side) = edge_side(index) else {
-                continue;
-            };
-            let y = items[index].y;
-            let paired = [value.checked_sub(1), value.checked_add(1)]
-                .into_iter()
-                .flatten()
-                .filter_map(|other_value| known_baselines.get(&(other_value, !side)))
-                .any(|baselines| {
-                    let first = baselines
-                        .partition_point(|baseline| *baseline <= y - PAGE_NUMBER_Y_TOLERANCE);
-                    baselines
-                        .get(first)
-                        .is_some_and(|baseline| *baseline < y + PAGE_NUMBER_Y_TOLERANCE)
-                });
-            if paired {
-                explicit_folio[index] = true;
-            }
-        }
-    }
-}
-
-/// Mark a contextual folio that alternates with an isolated folio on the
-/// neighboring PDF page.
-///
-/// Facing pages commonly put folios on opposite outer edges. A running header
-/// can touch the right-hand folio while the next left-hand folio is isolated.
-/// A single isolated candidate is not enough to remove nearby contextual text.
-/// Require a second pre-existing anchor in the same advancing sequence, along
-/// with a genuinely wide content span, stable baselines/font sizes, and
-/// alternating outer edges.
-fn mark_adjacent_page_folio_pairs(
-    items: &[TextItem],
-    candidate_values: &[Option<u32>],
-    contextual: &[bool],
-    document_page_count: usize,
-    explicit_folio: &mut [bool],
-) {
-    let max_plausible_folio = (document_page_count as u32).saturating_mul(4).max(100);
-    // Do not let newly inferred candidates recursively become evidence for
-    // later candidates; every match must be anchored by evidence established
-    // before this cross-page pass.
-    let strong_folio_evidence = explicit_folio.to_vec();
-    let mut page_bounds: HashMap<u32, (f32, f32)> = HashMap::new();
-    for item in items.iter().filter(|item| !item.text.trim().is_empty()) {
-        let bounds = page_bounds
-            .entry(item.page)
-            .or_insert((f32::INFINITY, f32::NEG_INFINITY));
-        bounds.0 = bounds.0.min(item.x);
-        bounds.1 = bounds.1.max(item.x + effective_width(item));
-    }
-
-    let edge_side = |index: usize| {
-        let &(page_left, page_right) = page_bounds.get(&items[index].page)?;
-        let page_width = page_right - page_left;
-        if page_width < items[index].font_size * ADJACENT_PAGE_MIN_CONTENT_WIDTH_EM {
-            return None;
-        }
-        let center = items[index].x + effective_width(&items[index]) / 2.0;
-        let left_edge = page_left + page_width * SPREAD_EDGE_FRACTION;
-        let right_edge = page_right - page_width * SPREAD_EDGE_FRACTION;
-        if center <= left_edge {
-            Some(false)
-        } else if center >= right_edge {
-            Some(true)
-        } else {
-            None
-        }
-    };
-
-    // Anchor sequences by outer edge and the value/page offset. This enforces
-    // forward page tracking and lets candidates query adjacent pages directly,
-    // while a baseline-sorted index finds a second independent anchor without
-    // a document-wide quadratic scan.
-    let mut anchors_by_page: HashMap<(u32, bool, i64), Vec<usize>> = HashMap::new();
-    let mut anchors_by_sequence: HashMap<(bool, i64), Vec<usize>> = HashMap::new();
-    for (index, value) in candidate_values.iter().enumerate() {
-        let Some(value) = value else {
-            continue;
-        };
-        if *value > max_plausible_folio || (contextual[index] && !strong_folio_evidence[index]) {
-            continue;
-        }
-        let Some(side) = edge_side(index) else {
-            continue;
-        };
-        let page = items[index].page;
-        let offset = i64::from(*value) - i64::from(page);
-        anchors_by_page
-            .entry((page, side, offset))
-            .or_default()
-            .push(index);
-        anchors_by_sequence
-            .entry((side, offset))
-            .or_default()
-            .push(index);
-    }
-    for anchors in anchors_by_sequence.values_mut() {
-        anchors.sort_by(|&left, &right| items[left].y.total_cmp(&items[right].y));
-    }
-
-    for (index, value) in candidate_values.iter().enumerate() {
-        let Some(value) = value else {
-            continue;
-        };
-        if !contextual[index] || explicit_folio[index] || *value > max_plausible_folio {
-            continue;
-        }
-        let Some(side) = edge_side(index) else {
-            continue;
-        };
-
-        let page = items[index].page;
-        let offset = i64::from(*value) - i64::from(page);
-        let Some(sequence_anchors) = anchors_by_sequence.get(&(!side, offset)) else {
-            continue;
-        };
-        let neighbor = [page.checked_sub(1), page.checked_add(1)]
-            .into_iter()
-            .flatten()
-            .filter_map(|neighbor_page| anchors_by_page.get(&(neighbor_page, !side, offset)))
-            .flatten()
-            .copied()
-            .find(|&neighbor_index| {
-                (items[index].y - items[neighbor_index].y).abs() < PAGE_NUMBER_Y_TOLERANCE
-                    && (items[index].font_size - items[neighbor_index].font_size).abs() < 1.0
-            });
-        let Some(neighbor_index) = neighbor else {
-            continue;
-        };
-
-        let first = sequence_anchors.partition_point(|&anchor_index| {
-            items[anchor_index].y <= items[index].y - PAGE_NUMBER_Y_TOLERANCE
-        });
-        let has_second_anchor = sequence_anchors[first..]
-            .iter()
-            .take_while(|&&anchor_index| {
-                items[anchor_index].y < items[index].y + PAGE_NUMBER_Y_TOLERANCE
-            })
-            .any(|&anchor_index| {
-                items[anchor_index].page != page
-                    && items[anchor_index].page != items[neighbor_index].page
-                    && (items[index].font_size - items[anchor_index].font_size).abs() < 1.0
-            });
-        if has_second_anchor {
-            explicit_folio[index] = true;
-        }
-    }
-}
-
-/// Identify page-edge numeric items that belong to a nearby content run.
-///
-/// Numeric candidates on their own do not establish context for one another.
-/// A connected same-baseline run is contextual only when it also contains a
-/// non-candidate item, preserving lines such as `Chapter 1 2026` while still
-/// removing isolated numeric footer clusters.
-fn page_number_context_masks(
-    items: &[TextItem],
-    candidate_values: &[Option<u32>],
-    document_page_count: usize,
-) -> (Vec<bool>, Vec<bool>) {
-    let mut contextual = vec![false; items.len()];
-    let mut explicit_folio = vec![false; items.len()];
-    let mut occurrences_by_signature: HashMap<String, Vec<ContextualCandidateOccurrence>> =
-        HashMap::new();
-    let mut indices_by_page: HashMap<u32, Vec<usize>> = HashMap::new();
-    for (index, item) in items.iter().enumerate() {
-        if matches!(
-            item.item_type,
-            crate::types::ItemType::Text | crate::types::ItemType::FormField
-        ) && !item.text.trim().is_empty()
-        {
-            indices_by_page.entry(item.page).or_default().push(index);
-        }
-    }
-    for mut page_indices in indices_by_page.into_values() {
-        page_indices.sort_by(|&left, &right| {
-            items[right]
-                .y
-                .total_cmp(&items[left].y)
-                .then(items[left].x.total_cmp(&items[right].x))
-        });
-
-        let mut rows: Vec<Vec<usize>> = Vec::new();
-        for index in page_indices {
-            if rows.last().is_some_and(|row| {
-                (items[row[0]].y - items[index].y).abs() < PAGE_NUMBER_Y_TOLERANCE
-            }) {
-                rows.last_mut().unwrap().push(index);
-            } else {
-                rows.push(vec![index]);
-            }
-        }
-
-        for mut row in rows {
-            row.sort_by(|&left, &right| items[left].x.total_cmp(&items[right].x));
-            let mut start = 0;
-            while start < row.len() {
-                let mut end = start + 1;
-                let first = &items[row[start]];
-                let mut group_right = first.x + effective_width(first);
-                let mut group_font_size = first.font_size;
-
-                while end < row.len() {
-                    let item = &items[row[end]];
-                    let gap = item.x - group_right;
-                    if gap > group_font_size.max(item.font_size) * PAGE_NUMBER_CONTEXT_GAP_EM {
-                        break;
-                    }
-                    group_right = group_right.max(item.x + effective_width(item));
-                    group_font_size = group_font_size.max(item.font_size);
-                    end += 1;
-                }
-
-                let group = &row[start..end];
-                let has_lexical_context = group.iter().any(|&index| {
-                    candidate_values[index].is_none()
-                        && items[index]
-                            .text
-                            .chars()
-                            .any(|character| character.is_alphabetic())
-                });
-                // Numeric data near a page edge also needs protection, but a
-                // lone long integer beside a short candidate is not enough to
-                // establish context. Preserve explicit numeric structures
-                // (list markers, ranges, comma-formatted values, dotted index
-                // entries) and dense runs with at least one long integer.
-                let numeric_like = |text: &str| {
-                    text.chars().any(|character| character.is_numeric())
-                        && !text.chars().any(|character| character.is_alphabetic())
-                };
-                let is_structured_numeric_context = |index: usize| {
-                    if candidate_values[index].is_some() {
-                        return false;
-                    }
-                    let text = items[index].text.trim();
-                    numeric_like(text)
-                        && text
-                            .chars()
-                            .any(|character| !character.is_numeric() && !character.is_whitespace())
-                };
-                let has_structured_numeric_context = group
-                    .iter()
-                    .any(|&index| is_structured_numeric_context(index));
-                let numeric_item_count = group
-                    .iter()
-                    .filter(|&&index| numeric_like(items[index].text.trim()))
-                    .count();
-                let has_long_integer = group.iter().any(|&index| {
-                    candidate_values[index].is_none()
-                        && items[index]
-                            .text
-                            .trim()
-                            .chars()
-                            .all(|character| character.is_numeric())
-                });
-                let has_dense_numeric_context = numeric_item_count >= 3 && has_long_integer;
-                let has_context = has_lexical_context
-                    || has_structured_numeric_context
-                    || has_dense_numeric_context;
-                let has_candidate = row[start..end]
-                    .iter()
-                    .any(|&index| candidate_values[index].is_some());
-                // Decorative centered folios have no lexical context, so
-                // recognize the complete delimiter-number-delimiter triplet
-                // before the contextual-content gate. This prevents `- 42 -`
-                // from leaving a malformed `- -` line.
-                if group.len() == 3
-                    && items[group[0]].text.trim() == "-"
-                    && candidate_values[group[1]].is_some()
-                    && items[group[2]].text.trim() == "-"
-                {
-                    for &index in group {
-                        explicit_folio[index] = true;
-                    }
-                }
-                if has_context && has_candidate {
-                    let group = &row[start..end];
-                    let group_text = group
-                        .iter()
-                        .map(|&index| items[index].text.trim())
-                        .filter(|text| !text.is_empty())
-                        .collect::<Vec<_>>()
-                        .join(" ");
-                    let group_is_folio =
-                        crate::text_utils::is_explicit_page_number_expression(&group_text);
-                    let context_text = group
-                        .iter()
-                        .filter(|&&index| candidate_values[index].is_none())
-                        .map(|&index| items[index].text.trim())
-                        .filter(|text| !text.is_empty())
-                        .collect::<Vec<_>>()
-                        .join(" ");
-                    let candidates: Vec<(usize, u32)> = group
-                        .iter()
-                        .filter_map(|&index| candidate_values[index].map(|value| (index, value)))
-                        .collect();
-                    // Recurrence is only evidence for numeric slots at the
-                    // outer boundary of a contextual run. An embedded number
-                    // in repeated prose such as `Page 42 explains the result`
-                    // is substantive content, not a running folio.
-                    let recurrence_candidates: Vec<(usize, u32)> = candidates
-                        .iter()
-                        .copied()
-                        .filter(|(index, _)| {
-                            group.first() == Some(index) || group.last() == Some(index)
-                        })
-                        .collect();
-                    let in_deep_margin = candidates.iter().all(|(index, _)| {
-                        items[*index].y < PAGE_NUMBER_BOTTOM_Y
-                            || items[*index].y > PAGE_NUMBER_TOP_Y
-                    });
-                    if in_deep_margin
-                        && !recurrence_candidates.is_empty()
-                        && context_text
-                            .chars()
-                            .filter(|character| character.is_alphanumeric())
-                            .count()
-                            >= 8
-                    {
-                        let signature = group
-                            .iter()
-                            .map(|&index| {
-                                if candidate_values[index].is_some() {
-                                    "{number}".to_string()
-                                } else {
-                                    items[index]
-                                        .text
-                                        .split_whitespace()
-                                        .collect::<Vec<_>>()
-                                        .join(" ")
-                                        .to_lowercase()
-                                }
-                            })
-                            .collect::<Vec<_>>()
-                            .join(" ");
-                        occurrences_by_signature
-                            .entry(signature)
-                            .or_default()
-                            .push((
-                                items[group[0]].page,
-                                items[group[0]].y,
-                                recurrence_candidates,
-                            ));
-                    }
-                    for (position, &index) in group.iter().enumerate() {
-                        if let Some(value) = candidate_values[index] {
-                            let adjacent_context = [position.checked_sub(1), Some(position + 1)]
-                                .into_iter()
-                                .flatten()
-                                .filter_map(|position| group.get(position).copied())
-                                .any(|adjacent| {
-                                    candidate_values[adjacent].is_none()
-                                        && items[adjacent].text.chars().any(|character| {
-                                            !character.is_numeric() && !character.is_whitespace()
-                                        })
-                                });
-                            let max_plausible_folio =
-                                (document_page_count as u32).saturating_mul(4).max(100);
-                            // Large year/identifier-like values stay attached
-                            // to their lexical run even when a smaller numeric
-                            // candidate sits between them and the text.
-                            let implausible_folio_with_lexical_context =
-                                value > max_plausible_folio && has_lexical_context;
-                            contextual[index] = adjacent_context
-                                || has_dense_numeric_context
-                                || implausible_folio_with_lexical_context;
-                            let previous = position
-                                .checked_sub(1)
-                                .map(|position| items[group[position]].text.trim());
-                            let next = group
-                                .get(position + 1)
-                                .map(|&index| items[index].text.trim());
-                            let follows_page_label =
-                                previous.is_some_and(|text| text.eq_ignore_ascii_case("page"));
-                            let starts_of_expression =
-                                next.is_some_and(|text| text.eq_ignore_ascii_case("of"));
-                            let is_centered_folio = previous == Some("-") && next == Some("-");
-                            explicit_folio[index] |= group_is_folio
-                                && (follows_page_label
-                                    || starts_of_expression
-                                    || is_centered_folio);
-                        }
-                    }
-                    // Remove the complete labeled expression rather than
-                    // leaving fragments such as `Page of 15`. A trailing
-                    // running-header suffix remains untouched.
-                    if group_is_folio
-                        && group.len() >= 4
-                        && items[group[0]].text.trim().eq_ignore_ascii_case("page")
-                        && candidate_values[group[1]].is_some()
-                        && items[group[2]].text.trim().eq_ignore_ascii_case("of")
-                        && items[group[3]]
-                            .text
-                            .trim()
-                            .chars()
-                            .all(|character| character.is_ascii_digit())
-                    {
-                        for &index in &group[..4] {
-                            explicit_folio[index] = true;
-                        }
-                    }
-                }
-                start = end;
-            }
-        }
-    }
-
-    mark_repeated_folio_candidates(
-        occurrences_by_signature,
-        document_page_count,
-        &mut explicit_folio,
-    );
-    mark_spread_folio_pairs(items, candidate_values, &contextual, &mut explicit_folio);
-    mark_adjacent_page_folio_pairs(
-        items,
-        candidate_values,
-        &contextual,
-        document_page_count,
-        &mut explicit_folio,
-    );
-
-    (contextual, explicit_folio)
-}
-
-/// Decide which digit-only page-edge items can be removed before layout.
-///
-/// PDF producers commonly emit one text-showing operation per word. A numeric
-/// item attached to neighboring content on the same baseline is therefore kept.
-/// Complete page-number expressions such as `Page 42` remain removable even
-/// though their numeric item has lexical context.
-fn page_number_removal_mask(items: &[TextItem], document_page_count: usize) -> Vec<bool> {
-    let candidate_values: Vec<Option<u32>> = items.iter().map(page_number_value).collect();
-    let (contextual, explicit_folio) =
-        page_number_context_masks(items, &candidate_values, document_page_count);
-
-    candidate_values
-        .iter()
-        .enumerate()
-        .map(|(index, value)| explicit_folio[index] || (value.is_some() && !contextual[index]))
-        .collect()
-}
-
-/// Return whether selected-page extraction contains a page-edge number whose
-/// folio status depends on evidence from other pages. Isolated and explicitly
-/// labeled folios can be decided locally; only contextual candidates require
-/// a document-wide extraction pass.
-pub(super) fn needs_document_page_number_context(
-    items: &[TextItem],
-    document_page_count: usize,
-) -> bool {
-    let candidate_values: Vec<Option<u32>> = items.iter().map(page_number_value).collect();
-    let (contextual, explicit_folio) =
-        page_number_context_masks(items, &candidate_values, document_page_count);
-
-    candidate_values
-        .iter()
-        .enumerate()
-        .any(|(index, value)| value.is_some() && contextual[index] && !explicit_folio[index])
-}
-
-/// Remove numeric folios using complete document context before downstream
-/// non-table layout partitions could separate the evidence needed to recognize
-/// them.
-#[cfg(test)]
-pub(crate) fn filter_markdown_page_numbers(
-    items: Vec<TextItem>,
-    document_page_count: u32,
-) -> Vec<TextItem> {
-    filter_markdown_page_numbers_with_removed_pages(items, document_page_count).0
-}
-
-/// Filter Markdown folios while retaining the pages where items were removed.
-///
-/// The page set lets downstream table-continuation classification preserve its
-/// pre-filter semantics even though structural layout consumes the cleaned
-/// item collection.
-pub(crate) fn filter_markdown_page_numbers_with_removed_pages(
-    items: Vec<TextItem>,
-    document_page_count: u32,
-) -> (Vec<TextItem>, HashSet<u32>, Vec<bool>) {
-    let remove = page_number_removal_mask(&items, document_page_count as usize);
-    let mut removed_pages = HashSet::new();
-    let items = items
-        .into_iter()
-        .zip(remove.iter().copied())
-        .filter_map(|(item, remove)| {
-            if remove {
-                removed_pages.insert(item.page);
-                None
-            } else {
-                Some(item)
-            }
-        })
-        .collect();
-    (items, removed_pages, remove)
+    item.y > 720.0 || item.y < 100.0
 }

 /// Group text items into lines, with multi-column support
@@ -1896,27 +1205,6 @@ pub(crate) fn group_into_lines_with_thresholds_and_charts(
    )
 }

-/// Group items after document-level page-number filtering has already run.
-///
-/// Partitioned Markdown layout uses this path so a contextual candidate that
-/// was preserved with its complete baseline context is not reconsidered after
-/// its neighboring text lands in another band or chart/prose zone.
-pub(crate) fn group_prefiltered_items_into_lines_with_thresholds_and_charts(
-    items: Vec<TextItem>,
-    page_thresholds: &HashMap<u32, f32>,
-    table_pages: &HashSet<u32>,
-    chart_regions: &HashMap<u32, Vec<(f32, f32, f32, f32)>>,
-) -> Vec<TextLine> {
-    group_into_lines_with_thresholds_and_regions_impl(
-        items,
-        page_thresholds,
-        table_pages,
-        chart_regions,
-        &HashMap::new(),
-        false,
-    )
-}
-
 pub(crate) fn group_into_lines_with_thresholds_and_regions(
    items: Vec<TextItem>,
    page_thresholds: &HashMap<u32, f32>,
@@ -1934,23 +1222,6 @@ pub(crate) fn group_into_lines_with_thresholds_and_regions(
    )
 }

-pub(crate) fn group_prefiltered_items_into_lines_with_thresholds_and_regions(
-    items: Vec<TextItem>,
-    page_thresholds: &HashMap<u32, f32>,
-    table_pages: &HashSet<u32>,
-    chart_regions: &HashMap<u32, Vec<(f32, f32, f32, f32)>>,
-    image_regions: &HashMap<u32, Vec<super::reading_order::ImageRegion>>,
-) -> Vec<TextLine> {
-    group_into_lines_with_thresholds_and_regions_impl(
-        items,
-        page_thresholds,
-        table_pages,
-        chart_regions,
-        image_regions,
-        false,
-    )
-}
-
 fn group_into_lines_with_thresholds_and_regions_impl(
    items: Vec<TextItem>,
    page_thresholds: &HashMap<u32, f32>,
@@ -1963,25 +1234,12 @@ fn group_into_lines_with_thresholds_and_regions_impl(
        return Vec::new();
    }

-    // Markdown output omits standalone numeric headers/footers. Determine
-    // standalone status from rough baseline context before layout analysis so
-    // removed page numbers cannot affect column detection. Plain-text callers
-    // opt out because dropping extracted text violates that API.
+    // Markdown output omits standalone numeric headers/footers. Plain-text
+    // callers opt out because dropping extracted text violates that API.
    let items = if filter_page_numbers {
-        // Item-only grouping has no document metadata, so use the highest
-        // observed 1-based page as its best available coverage denominator.
-        // The Markdown document path passes the authoritative PDF page count
-        // through `filter_markdown_page_numbers` before reaching this helper.
-        let observed_page_count = items
-            .iter()
-            .map(|item| item.page as usize)
-            .max()
-            .unwrap_or(0);
-        let remove = page_number_removal_mask(&items, observed_page_count);
        items
            .into_iter()
-            .zip(remove)
-            .filter_map(|(item, remove)| (!remove).then_some(item))
+            .filter(|item| !is_page_number(item))
            .collect()
    } else {
        items
@@ -1996,15 +1254,6 @@ fn group_into_lines_with_thresholds_and_regions_impl(

    for page in pages {
        let page_items: Vec<TextItem> = items.iter().filter(|i| i.page == page).cloned().collect();
-        // Page-edge numeric runs are weak evidence for column geometry. Keep
-        // contextual values for line assembly, but prevent their preservation
-        // from changing the page's inferred layout.
-        let column_detection_items: Vec<TextItem> = page_items
-            .iter()
-            .filter(|item| page_number_value(item).is_none())
-            .cloned()
-            .collect();
-        let column_detection_items = column_detection_items.as_slice();

        // Use pre-computed threshold from fix_letterspaced_items if available
        // (computed before embedded-space removal, with full signal).
@@ -2016,9 +1265,9 @@ fn group_into_lines_with_thresholds_and_regions_impl(
        // their own positioned-region ordering and therefore stay on that path.
        if !chart_regions.contains_key(&page) {
            let preliminary_columns =
-                detect_columns(column_detection_items, page, table_pages.contains(&page));
+                detect_columns(&page_items, page, table_pages.contains(&page));
            let detected_split =
-                (preliminary_columns.len() == 2).then(|| preliminary_columns[0].x_max);
+                (preliminary_columns.len() == 2).then_some(preliminary_columns[0].x_max);
            if let Some(band) = image_regions.get(&page).and_then(|regions| {
                super::reading_order::infer_image_anchored_flow(
                    &page_items,
@@ -2058,9 +1307,6 @@ fn group_into_lines_with_thresholds_and_regions_impl(
                let col_input: Vec<TextItem> = page_items
                    .iter()
                    .filter(|it| {
-                        if page_number_value(it).is_some() {
-                            return false;
-                        }
                        let cx = it.x + it.width / 2.0;
                        // Tight bounds: this only blinds the histogram to
                        // chart-internal text; rows adjacent to the chart
@@ -2073,7 +1319,7 @@ fn group_into_lines_with_thresholds_and_regions_impl(
                    .collect();
                detect_columns(&col_input, page, table_pages.contains(&page))
            }
-            None => detect_columns(column_detection_items, page, table_pages.contains(&page)),
+            None => detect_columns(&page_items, page, table_pages.contains(&page)),
        };

        if columns.len() <= 1 {
@@ -153,54 +153,16 @@ pub(crate) fn extract_form_fields(
        },
        Err(_) => return items,
    };
-    if fields.is_empty() {
-        return items;
-    }
-    let annotation_pages = annotation_page_map(doc, page_map);

    for field_obj in &fields {
        if let Ok(field_ref) = field_obj.as_reference() {
-            walk_form_fields(
-                doc,
-                field_ref,
-                None,
-                "",
-                page_map,
-                &annotation_pages,
-                &mut items,
-            );
+            walk_form_fields(doc, field_ref, None, "", page_map, &mut items);
        }
    }

    items
 }

-/// Map widget annotation objects back to the page whose `/Annots` array owns
-/// them. Some valid widgets omit `/P`, so the page tree is the only reliable
-/// ownership signal available for page-filtered extraction.
-fn annotation_page_map(
-    doc: &Document,
-    page_map: &HashMap<ObjectId, u32>,
-) -> HashMap<ObjectId, u32> {
-    let mut annotation_pages = HashMap::new();
-    for (&page_id, &page_num) in page_map {
-        let Some(annotations) = doc
-            .get_dictionary(page_id)
-            .ok()
-            .and_then(|page| page.get(b"Annots").ok())
-            .and_then(|annotations| resolve_array(doc, annotations))
-        else {
-            continue;
-        };
-        for annotation in annotations {
-            if let Ok(annotation_id) = annotation.as_reference() {
-                annotation_pages.insert(annotation_id, page_num);
-            }
-        }
-    }
-    annotation_pages
-}
-
 /// Recursively walk the form field tree, extracting leaf field values.
 pub(crate) fn walk_form_fields(
    doc: &Document,
@@ -208,7 +170,6 @@ pub(crate) fn walk_form_fields(
    parent_ft: Option<&[u8]>,
    parent_name: &str,
    page_map: &HashMap<ObjectId, u32>,
-    annotation_pages: &HashMap<ObjectId, u32>,
    items: &mut Vec<TextItem>,
 ) {
    let field_dict = match doc.get_dictionary(field_id) {
@@ -245,15 +206,7 @@ pub(crate) fn walk_form_fields(
            let kids = kids.clone();
            for kid in &kids {
                if let Ok(kid_ref) = kid.as_reference() {
-                    walk_form_fields(
-                        doc,
-                        kid_ref,
-                        ft,
-                        &full_name,
-                        page_map,
-                        annotation_pages,
-                        items,
-                    );
+                    walk_form_fields(doc, kid_ref, ft, &full_name, page_map, items);
                }
            }
            return;
@@ -346,7 +299,6 @@ pub(crate) fn walk_form_fields(
        .ok()
        .and_then(|o| o.as_reference().ok())
        .and_then(|p| page_map.get(&p).copied())
-        .or_else(|| annotation_pages.get(&field_id).copied())
        .unwrap_or(1);

    let text = if full_name.is_empty() {
@@ -372,43 +324,3 @@ pub(crate) fn walk_form_fields(
        mcid: None,
    });
 }
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use lopdf::{dictionary, Object};
-
-    #[test]
-    fn widget_without_page_reference_uses_owning_page_annotation() {
-        let mut doc = Document::new();
-        let widget_id = doc.add_object(dictionary! {
-            "Type" => "Annot",
-            "Subtype" => "Widget",
-            "FT" => "Tx",
-            "T" => Object::string_literal("customer"),
-            "V" => Object::string_literal("Alice"),
-            "Rect" => vec![10.into(), 20.into(), 110.into(), 40.into()],
-        });
-        let page_one_id = doc.add_object(dictionary! {
-            "Type" => "Page",
-        });
-        let page_two_id = doc.add_object(dictionary! {
-            "Type" => "Page",
-            "Annots" => vec![Object::Reference(widget_id)],
-        });
-        let catalog_id = doc.add_object(dictionary! {
-            "Type" => "Catalog",
-            "AcroForm" => dictionary! {
-                "Fields" => vec![Object::Reference(widget_id)],
-            },
-        });
-        doc.trailer.set("Root", Object::Reference(catalog_id));
-
-        let page_map = HashMap::from([(page_one_id, 1), (page_two_id, 2)]);
-        let items = extract_form_fields(&doc, &page_map);
-
-        assert_eq!(items.len(), 1);
-        assert_eq!(items[0].page, 2);
-        assert_eq!(items[0].text, "customer: Alice");
-    }
-}
@@ -28,12 +28,9 @@ pub use crate::text_utils::{is_bold_font, is_italic_font};
 pub use crate::types::{ItemType, TextLine};
 pub(crate) use fonts::FontStyleCache;
 pub(crate) use layout::detect_columns;
-#[cfg(test)]
-use layout::filter_markdown_page_numbers;
-pub(crate) use layout::filter_markdown_page_numbers_with_removed_pages;
 pub(crate) use layout::group_into_lines_with_thresholds;
-pub(crate) use layout::group_prefiltered_items_into_lines_with_thresholds_and_charts;
-pub(crate) use layout::group_prefiltered_items_into_lines_with_thresholds_and_regions;
+pub(crate) use layout::group_into_lines_with_thresholds_and_charts;
+pub(crate) use layout::group_into_lines_with_thresholds_and_regions;
 pub(crate) use layout::is_newspaper_layout;
 pub(crate) use layout::ColumnRegion;
 pub use layout::{group_into_lines, group_into_lines_preserving_all_text};
@@ -85,33 +82,17 @@ pub fn extract_text_with_positions_pages<P: AsRef<Path>>(
    path: P,
    page_filter: Option<&HashSet<u32>>,
 ) -> Result<Vec<TextItem>, PdfError> {
-    let (items, _rects, _lines) =
-        extract_text_with_positions_and_rects_with_password(path, page_filter, None)?;
+    let (items, _rects, _lines) = extract_text_with_positions_and_rects(path, page_filter)?;
    Ok(items)
 }

-/// Extract text with positions from a file, limited to specific pages and
-/// decrypting with `password` when the PDF is encrypted.
-///
-/// `page_filter` is an optional set of 1-indexed page numbers to process.
-/// When `None`, all pages are processed.
-pub fn extract_text_with_positions_pages_with_password<P: AsRef<Path>>(
+/// Extract text with positions and rectangles from a file.
+pub(crate) fn extract_text_with_positions_and_rects<P: AsRef<Path>>(
    path: P,
    page_filter: Option<&HashSet<u32>>,
-    password: Option<&str>,
-) -> Result<Vec<TextItem>, PdfError> {
-    let (items, _rects, _lines) =
-        extract_text_with_positions_and_rects_with_password(path, page_filter, password)?;
-    Ok(items)
-}
-
-pub(crate) fn extract_text_with_positions_and_rects_with_password<P: AsRef<Path>>(
-    path: P,
-    page_filter: Option<&HashSet<u32>>,
-    password: Option<&str>,
 ) -> Result<PageExtraction, PdfError> {
    crate::validate_pdf_file(&path)?;
-    let (doc, _) = crate::load_document_from_path_with_password(&path, password)?;
+    let (doc, _) = crate::load_document_from_path(&path)?;
    let font_cmaps = FontCMaps::from_doc(&doc);
    let (extraction, _thresholds, _gid_pages) =
        extract_positioned_text_from_doc(&doc, &font_cmaps, page_filter)?;
@@ -160,92 +141,17 @@ pub(crate) fn extract_positioned_text_from_doc(
    font_cmaps: &FontCMaps,
    page_filter: Option<&HashSet<u32>>,
 ) -> Result<(PageExtraction, PageThresholds, HashSet<u32>), PdfError> {
-    extract_positioned_text_impl(doc, font_cmaps, page_filter, false, None)
+    extract_positioned_text_impl(doc, font_cmaps, page_filter, false)
 }

-/// Extract selected pages and gather document-wide folio evidence only when a
-/// selected page contains an ambiguous contextual page-edge number. Errors on
-/// selected pages remain fatal; errors on context-only pages are skipped.
-pub(crate) fn extract_positioned_text_with_folio_context(
+/// Extract with option to include invisible (Tr=3) text.
+/// Used for Mixed/template PDFs where the OCR text layer is invisible.
+pub(crate) fn extract_positioned_text_include_invisible(
    doc: &Document,
    font_cmaps: &FontCMaps,
    page_filter: Option<&HashSet<u32>>,
 ) -> Result<(PageExtraction, PageThresholds, HashSet<u32>), PdfError> {
-    extract_positioned_text_with_folio_context_impl(doc, font_cmaps, page_filter, false)
-}
-
-/// Invisible-text variant of [`extract_positioned_text_with_folio_context`].
-pub(crate) fn extract_positioned_text_include_invisible_with_folio_context(
-    doc: &Document,
-    font_cmaps: &FontCMaps,
-    page_filter: Option<&HashSet<u32>>,
-) -> Result<(PageExtraction, PageThresholds, HashSet<u32>), PdfError> {
-    extract_positioned_text_with_folio_context_impl(doc, font_cmaps, page_filter, true)
-}
-
-fn extract_positioned_text_with_folio_context_impl(
-    doc: &Document,
-    font_cmaps: &FontCMaps,
-    page_filter: Option<&HashSet<u32>>,
-    include_invisible: bool,
-) -> Result<(PageExtraction, PageThresholds, HashSet<u32>), PdfError> {
-    let Some(required_pages) = page_filter else {
-        return extract_positioned_text_impl(doc, font_cmaps, None, include_invisible, None);
-    };
-
-    let (
-        (mut selected_items, mut selected_rects, mut selected_lines),
-        mut page_thresholds,
-        mut gid_encoded_pages,
-    ) = extract_positioned_text_impl(
-        doc,
-        font_cmaps,
-        Some(required_pages),
-        include_invisible,
-        None,
-    )?;
-    if !layout::needs_document_page_number_context(&selected_items, doc.get_pages().len()) {
-        return Ok((
-            (selected_items, selected_rects, selected_lines),
-            page_thresholds,
-            gid_encoded_pages,
-        ));
-    }
-
-    let context_pages: HashSet<u32> = doc
-        .get_pages()
-        .keys()
-        .copied()
-        .filter(|page| !required_pages.contains(page))
-        .collect();
-    let ((context_items, context_rects, context_lines), context_thresholds, context_gid_pages) =
-        extract_positioned_text_impl(
-            doc,
-            font_cmaps,
-            Some(&context_pages),
-            include_invisible,
-            Some(required_pages),
-        )?;
-    selected_items.extend(context_items);
-    selected_rects.extend(context_rects);
-    selected_lines.extend(context_lines);
-    page_thresholds.extend(context_thresholds);
-    gid_encoded_pages.extend(context_gid_pages);
-    Ok((
-        (selected_items, selected_rects, selected_lines),
-        page_thresholds,
-        gid_encoded_pages,
-    ))
-}
-
-/// Extract all pages for document-wide analysis while allowing malformed
-/// unselected pages to be skipped. Any requested page still fails normally.
-pub(crate) fn extract_positioned_text_for_document_analysis(
-    doc: &Document,
-    font_cmaps: &FontCMaps,
-    required_pages: &HashSet<u32>,
-) -> Result<(PageExtraction, PageThresholds, HashSet<u32>), PdfError> {
-    extract_positioned_text_impl(doc, font_cmaps, None, false, Some(required_pages))
+    extract_positioned_text_impl(doc, font_cmaps, page_filter, true)
 }

 fn extract_positioned_text_impl(
@@ -253,7 +159,6 @@ fn extract_positioned_text_impl(
    font_cmaps: &FontCMaps,
    page_filter: Option<&HashSet<u32>>,
    include_invisible: bool,
-    required_pages: Option<&HashSet<u32>>,
 ) -> Result<(PageExtraction, PageThresholds, HashSet<u32>), PdfError> {
    let pages = doc.get_pages();
    let mut all_items = Vec::new();
@@ -275,25 +180,15 @@ fn extract_positioned_text_impl(
                continue;
            }
        }
-        let page_result = extract_page_text_items(
-            doc,
-            page_id,
-            *page_num,
-            font_cmaps,
-            include_invisible,
-            &mut style_cache,
-        );
-        let ((mut items, mut rects, mut lines), has_gid_fonts, coords_rotated) = match page_result {
-            Ok(extraction) => extraction,
-            Err(error) if required_pages.is_some_and(|required| !required.contains(page_num)) => {
-                debug!(
-                    "page {}: skipping context-only extraction error: {}",
-                    page_num, error
-                );
-                continue;
-            }
-            Err(error) => return Err(error),
-        };
+        let ((mut items, mut rects, mut lines), has_gid_fonts, coords_rotated) =
+            extract_page_text_items(
+                doc,
+                page_id,
+                *page_num,
+                font_cmaps,
+                include_invisible,
+                &mut style_cache,
+            )?;
        // Clip to the visible page box: single-page extracts and imposed
        // spreads keep neighboring pages' content in the stream, positioned
        // outside the CropBox. Extracting it interleaves invisible text into
@@ -423,9 +318,7 @@ fn extract_positioned_text_impl(
    }

    // Extract AcroForm field values
-    let form_items = extract_form_fields(doc, &page_id_to_num)
-        .into_iter()
-        .filter(|item| page_filter.is_none_or(|filter| filter.contains(&item.page)));
+    let form_items = extract_form_fields(doc, &page_id_to_num);
    all_items.extend(form_items);

    Ok((
@@ -1639,724 +1532,6 @@ mod tests {
        assert_eq!(lines[0].text(), "42");
    }

-    #[test]
-    fn inline_numeric_run_near_page_edge_is_not_removed() {
-        let mut items = vec![
-            make_merge_item("Total", 100.0, 30.0),
-            make_merge_item("730", 136.0, 18.0),
-            make_merge_item("seats", 160.0, 30.0),
-        ];
-        for item in &mut items {
-            item.y = 780.0;
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "Total 730 seats");
-    }
-
-    #[test]
-    fn numeric_page_footer_separated_from_label_is_removed() {
-        let mut page_number = make_merge_item("42", 25.0, 12.0);
-        page_number.y = 50.0;
-        let mut footer_label = make_merge_item("DOCUMENT FOOTER", 60.0, 100.0);
-        footer_label.y = 50.0;
-
-        let lines = group_into_lines(vec![page_number, footer_label]);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "DOCUMENT FOOTER");
-    }
-
-    #[test]
-    fn decorative_marker_does_not_contextualize_numeric_page_footer() {
-        let mut marker = make_merge_item("•", 19.0, 6.0);
-        marker.y = 30.0;
-        let mut page_number = make_merge_item("42", 37.0, 10.0);
-        page_number.y = 30.0;
-        let mut footer_label = make_merge_item("Company report footer", 68.0, 120.0);
-        footer_label.y = 30.0;
-
-        let lines = group_into_lines(vec![marker, page_number, footer_label]);
-
-        assert!(lines.iter().all(|line| !line.text().contains("42")));
-        assert!(lines
-            .iter()
-            .any(|line| line.text().contains("Company report footer")));
-    }
-
-    #[test]
-    fn labeled_page_number_is_removed_in_a_short_document() {
-        let mut label = make_merge_item("Page", 25.0, 28.0);
-        label.y = 50.0;
-        let mut page_number = make_merge_item("42", 57.0, 12.0);
-        page_number.y = 50.0;
-
-        let lines = group_into_lines(vec![label, page_number]);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "Page");
-    }
-
-    #[test]
-    fn labeled_page_number_with_running_header_suffix_is_removed() {
-        let mut items = vec![
-            make_merge_item("Page", 25.0, 28.0),
-            make_merge_item("42", 57.0, 12.0),
-            make_merge_item("of", 73.0, 12.0),
-            make_merge_item("100", 89.0, 18.0),
-            make_merge_item("Report header", 111.0, 78.0),
-        ];
-        for item in &mut items {
-            item.y = 50.0;
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "Report header");
-    }
-
-    #[test]
-    fn page_of_total_expression_is_removed_without_leaving_fragments() {
-        let mut items = vec![
-            make_merge_item("Page", 482.0, 27.0),
-            make_merge_item("1", 513.0, 6.0),
-            make_merge_item("of", 523.0, 10.0),
-            make_merge_item("15", 537.0, 12.0),
-        ];
-        for item in &mut items {
-            item.y = 46.0;
-        }
-
-        let lines = group_into_lines(items);
-
-        assert!(lines.is_empty());
-    }
-
-    #[test]
-    fn document_folio_filter_survives_per_page_layout_splitting() {
-        let mut items = Vec::new();
-        for (page, value) in [(1, "42"), (2, "43"), (3, "44")] {
-            let mut label = make_merge_item("Page", 25.0, 28.0);
-            label.page = page;
-            label.y = 50.0;
-            let mut page_number = make_merge_item(value, 57.0, 12.0);
-            page_number.page = page;
-            page_number.y = 50.0;
-            items.extend([label, page_number]);
-        }
-
-        let filtered = filter_markdown_page_numbers(items, 3);
-        assert!(filtered
-            .iter()
-            .all(|item| !matches!(item.text.as_str(), "42" | "43" | "44")));
-        let mut lines = Vec::new();
-        for page in 1..=3 {
-            let page_items = filtered
-                .iter()
-                .filter(|item| item.page == page)
-                .cloned()
-                .collect();
-            lines.extend(
-                group_prefiltered_items_into_lines_with_thresholds_and_charts(
-                    page_items,
-                    &HashMap::new(),
-                    &HashSet::new(),
-                    &HashMap::new(),
-                ),
-            );
-        }
-
-        assert_eq!(lines.len(), 3);
-        assert!(lines.iter().all(|line| line.text() == "Page"));
-    }
-
-    #[test]
-    fn numeric_only_page_edge_runs_do_not_contextualize_folios() {
-        let mut page_number = make_merge_item("42", 25.0, 12.0);
-        page_number.y = 50.0;
-        let mut long_number = make_merge_item("12345", 43.0, 30.0);
-        long_number.y = 50.0;
-
-        let lines = group_into_lines(vec![page_number, long_number]);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "12345");
-    }
-
-    #[test]
-    fn structured_and_dense_numeric_page_edge_runs_are_preserved() {
-        let mut list_marker = make_merge_item("11)", 25.0, 18.0);
-        list_marker.y = 50.0;
-        let mut chapter = make_merge_item("13", 47.0, 12.0);
-        chapter.y = 50.0;
-
-        let mut isbn_prefix = make_merge_item("9", 25.0, 6.0);
-        isbn_prefix.page = 2;
-        isbn_prefix.y = 50.0;
-        let mut isbn_mid = make_merge_item("780113", 35.0, 36.0);
-        isbn_mid.page = 2;
-        isbn_mid.y = 50.0;
-        let mut isbn_end = make_merge_item("227426", 75.0, 36.0);
-        isbn_end.page = 2;
-        isbn_end.y = 50.0;
-
-        let lines = group_into_lines(vec![list_marker, chapter, isbn_prefix, isbn_mid, isbn_end]);
-
-        assert_eq!(lines.len(), 2);
-        assert_eq!(lines[0].text(), "11) 13");
-        assert_eq!(lines[1].text(), "9 780113 227426");
-    }
-
-    #[test]
-    fn incrementing_numeric_body_column_is_not_treated_as_a_folio() {
-        let mut items = Vec::new();
-        for (page, value) in [(1, "13"), (2, "14"), (3, "15")] {
-            let mut row_number = make_merge_item(value, 72.0, 12.0);
-            row_number.page = page;
-            row_number.y = 730.0;
-            let mut name = make_merge_item("Person", 90.0, 42.0);
-            name.page = page;
-            name.y = 730.0;
-            items.extend([row_number, name]);
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 3);
-        assert_eq!(lines[0].text(), "13 Person");
-        assert_eq!(lines[1].text(), "14 Person");
-        assert_eq!(lines[2].text(), "15 Person");
-    }
-
-    #[test]
-    fn advancing_number_in_repeated_deep_margin_footer_is_removed() {
-        let mut items = Vec::new();
-        for (page, value) in [(1, "2"), (2, "4"), (3, "6"), (4, "8")] {
-            let mut page_number = make_merge_item(value, 25.0, 12.0);
-            page_number.page = page;
-            page_number.y = 30.0;
-            let mut footer = make_merge_item("Company report footer", 41.0, 120.0);
-            footer.page = page;
-            footer.y = 30.0;
-            items.extend([page_number, footer]);
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 4);
-        assert!(lines
-            .iter()
-            .all(|line| line.text() == "Company report footer"));
-    }
-
-    #[test]
-    fn repeated_substantive_page_number_prose_is_preserved() {
-        let mut items = Vec::new();
-        for (page, value) in [(1, "42"), (2, "43"), (3, "44"), (4, "45")] {
-            let mut page_label = make_merge_item("Page", 25.0, 28.0);
-            page_label.page = page;
-            page_label.y = 30.0;
-            let mut number = make_merge_item(value, 57.0, 12.0);
-            number.page = page;
-            number.y = 30.0;
-            let mut explanation = make_merge_item("explains the result", 73.0, 108.0);
-            explanation.page = page;
-            explanation.y = 30.0;
-            items.extend([page_label, number, explanation]);
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 4);
-        for (line, value) in lines.iter().zip(["42", "43", "44", "45"]) {
-            assert_eq!(line.text(), format!("Page {value} explains the result"));
-        }
-    }
-
-    #[test]
-    fn numeric_candidates_do_not_bridge_lexical_context() {
-        let mut report = make_merge_item("Report", 25.0, 40.0);
-        report.y = 30.0;
-        let mut year = make_merge_item("2026", 69.0, 24.0);
-        year.y = 30.0;
-        let mut folio = make_merge_item("42", 97.0, 12.0);
-        folio.y = 30.0;
-
-        let lines = group_into_lines(vec![report, year, folio]);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "Report 2026");
-    }
-
-    #[test]
-    fn centered_folio_delimiters_are_removed_with_the_number() {
-        let mut left = make_merge_item("-", 270.0, 6.0);
-        left.y = 30.0;
-        let mut number = make_merge_item("42", 280.0, 12.0);
-        number.y = 30.0;
-        let mut right = make_merge_item("-", 296.0, 6.0);
-        right.y = 30.0;
-
-        let lines = group_into_lines(vec![left, number, right]);
-
-        assert!(lines.is_empty());
-    }
-
-    #[test]
-    fn centered_delimiters_inside_substantive_text_are_preserved() {
-        let mut items = vec![
-            make_merge_item("Result", 240.0, 36.0),
-            make_merge_item("-", 280.0, 6.0),
-            make_merge_item("42", 290.0, 12.0),
-            make_merge_item("-", 306.0, 6.0),
-            make_merge_item("approved", 316.0, 48.0),
-        ];
-        for item in &mut items {
-            item.y = 30.0;
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "Result-42-approved");
-    }
-
-    #[test]
-    fn changing_year_in_repeated_deep_margin_header_is_preserved() {
-        let mut items = Vec::new();
-        for (page, year) in [(1, "2020"), (2, "2021"), (3, "2022"), (4, "2023")] {
-            let mut year = make_merge_item(year, 25.0, 24.0);
-            year.page = page;
-            year.y = 780.0;
-            let mut header = make_merge_item("Annual report", 53.0, 78.0);
-            header.page = page;
-            header.y = 780.0;
-            items.extend([year, header]);
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 4);
-        assert_eq!(lines[0].text(), "2020 Annual report");
-        assert_eq!(lines[3].text(), "2023 Annual report");
-    }
-
-    #[test]
-    fn sparse_repeated_margin_numbers_do_not_meet_the_folio_evidence_floor() {
-        let mut items = Vec::new();
-        for page in 1..=4 {
-            if page <= 2 {
-                let value = if page == 1 { "2" } else { "4" };
-                let mut number = make_merge_item(value, 25.0, 12.0);
-                number.page = page;
-                number.y = 30.0;
-                let mut footer = make_merge_item("Company report footer", 41.0, 120.0);
-                footer.page = page;
-                footer.y = 30.0;
-                items.extend([number, footer]);
-            } else {
-                let mut body = make_merge_item("Body text", 72.0, 54.0);
-                body.page = page;
-                body.y = 400.0;
-                items.push(body);
-            }
-        }
-
-        let lines = group_into_lines(items);
-
-        assert!(lines
-            .iter()
-            .any(|line| line.text() == "2 Company report footer"));
-        assert!(lines
-            .iter()
-            .any(|line| line.text() == "4 Company report footer"));
-    }
-
-    #[test]
-    fn sparse_document_pages_count_toward_repeated_folio_coverage() {
-        let mut items = Vec::new();
-        for (page, value) in [(1, "1"), (10, "10"), (19, "19"), (28, "28")] {
-            let mut number = make_merge_item(value, 25.0, 12.0);
-            number.page = page;
-            number.y = 30.0;
-            let mut footer = make_merge_item("Company report footer", 41.0, 120.0);
-            footer.page = page;
-            footer.y = 30.0;
-            items.extend([number, footer]);
-        }
-
-        let lines = group_into_lines(items);
-
-        assert!(lines
-            .iter()
-            .any(|line| line.text() == "1 Company report footer"));
-        assert!(lines
-            .iter()
-            .any(|line| line.text() == "28 Company report footer"));
-    }
-
-    #[test]
-    fn trailing_blank_pages_count_toward_repeated_folio_coverage() {
-        let mut items = Vec::new();
-        for (page, value) in [(1, "1"), (2, "2"), (3, "3"), (4, "4")] {
-            let mut number = make_merge_item(value, 25.0, 12.0);
-            number.page = page;
-            number.y = 30.0;
-            let mut footer = make_merge_item("Company report footer", 41.0, 120.0);
-            footer.page = page;
-            footer.y = 30.0;
-            items.extend([number, footer]);
-        }
-
-        let filtered = filter_markdown_page_numbers(items, 20);
-
-        assert!(filtered.iter().any(|item| item.text == "1"));
-        assert!(filtered.iter().any(|item| item.text == "4"));
-    }
-
-    #[test]
-    fn prefiltered_contextual_number_survives_layout_partitioning() {
-        let mut items = vec![
-            make_merge_item("Total", 100.0, 30.0),
-            make_merge_item("730", 136.0, 18.0),
-            make_merge_item("seats", 160.0, 30.0),
-        ];
-        for item in &mut items {
-            item.y = 780.0;
-        }
-
-        let filtered = filter_markdown_page_numbers(items, 1);
-        let partitioned_number: Vec<TextItem> = filtered
-            .into_iter()
-            .filter(|item| item.text == "730")
-            .collect();
-        let lines = group_prefiltered_items_into_lines_with_thresholds_and_charts(
-            partitioned_number,
-            &HashMap::new(),
-            &HashSet::new(),
-            &HashMap::new(),
-        );
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "730");
-    }
-
-    #[test]
-    fn numeric_only_partition_does_not_define_columns() {
-        let mut items = Vec::new();
-        for row in 0..20 {
-            let y = 90.0 - row as f32 * 4.0;
-            let mut left = make_merge_item(&(row + 1).to_string(), 50.0, 20.0);
-            left.y = y;
-            let mut right = make_merge_item(&(row + 101).to_string(), 350.0, 20.0);
-            right.y = y;
-            items.extend([left, right]);
-        }
-        assert_eq!(detect_columns(&items, 1, false).len(), 2);
-
-        let lines = group_prefiltered_items_into_lines_with_thresholds_and_charts(
-            items,
-            &HashMap::new(),
-            &HashSet::new(),
-            &HashMap::new(),
-        );
-
-        assert_eq!(lines.len(), 20);
-        assert!(lines.iter().all(|line| line.items.len() == 2));
-    }
-
-    #[test]
-    fn separated_content_is_not_treated_as_a_spread_folio_pair() {
-        let mut value = make_merge_item("12", 100.0, 12.0);
-        value.y = 30.0;
-        let mut label = make_merge_item("Total", 116.0, 30.0);
-        label.y = 30.0;
-        let mut unrelated_number = make_merge_item("13", 300.0, 12.0);
-        unrelated_number.y = 30.0;
-
-        let lines = group_into_lines(vec![value, label, unrelated_number]);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "12 Total");
-    }
-
-    #[test]
-    fn repeated_folio_uses_the_full_page_edge_band() {
-        let mut items = Vec::new();
-        for (page, value) in [(1, "2"), (2, "4"), (3, "6"), (4, "8")] {
-            let mut page_number = make_merge_item(value, 25.0, 12.0);
-            page_number.page = page;
-            page_number.y = 80.0;
-            let mut footer = make_merge_item("Company report footer", 41.0, 120.0);
-            footer.page = page;
-            footer.y = 80.0;
-            items.extend([page_number, footer]);
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 4);
-        assert!(lines
-            .iter()
-            .all(|line| line.text() == "Company report footer"));
-    }
-
-    #[test]
-    fn contextual_folio_on_facing_page_spread_is_removed() {
-        let mut marker = make_merge_item("•", 19.0, 6.0);
-        marker.y = 30.0;
-        let mut left_folio = make_merge_item("326", 35.0, 17.0);
-        left_folio.y = 30.0;
-        let mut footer = make_merge_item("Company report footer", 61.0, 120.0);
-        footer.y = 30.0;
-        let mut right_folio = make_merge_item("327", 1148.0, 17.0);
-        right_folio.y = 30.0;
-
-        let lines = group_into_lines(vec![marker, left_folio, footer, right_folio]);
-
-        assert!(lines
-            .iter()
-            .all(|line| !line.text().contains("326") && !line.text().contains("327")));
-        assert!(lines
-            .iter()
-            .any(|line| line.text().contains("Company report footer")));
-    }
-
-    #[test]
-    fn contextual_folios_alternating_across_pages_are_removed() {
-        let headers = [
-            "Letter to shareholders",
-            "Corporate governance report",
-            "Business environment overview",
-            "Consolidated financial statements",
-        ];
-        let mut items = Vec::new();
-        for page in 1..=8 {
-            let mut body = make_merge_item("Body text", 50.0, 500.0);
-            body.page = page;
-            body.y = 400.0;
-            items.push(body);
-
-            let mut folio = make_merge_item(&(page + 22).to_string(), 0.0, 14.0);
-            folio.page = page;
-            folio.y = 780.0;
-            if page % 2 == 0 {
-                folio.x = 50.0;
-                items.push(folio);
-            } else {
-                let mut header = make_merge_item(headers[(page / 2) as usize], 350.0, 180.0);
-                header.page = page;
-                header.y = 780.0;
-                folio.x = 536.0;
-                items.extend([header, folio]);
-            }
-        }
-
-        let filtered = filter_markdown_page_numbers(items, 8);
-
-        assert!(filtered.iter().all(|item| {
-            !matches!(
-                item.text.as_str(),
-                "23" | "24" | "25" | "26" | "27" | "28" | "29" | "30"
-            )
-        }));
-        assert!(headers
-            .iter()
-            .all(|header| filtered.iter().any(|item| item.text == *header)));
-    }
-
-    #[test]
-    fn one_isolated_neighbor_does_not_remove_contextual_number() {
-        let mut body_one = make_merge_item("Body text", 50.0, 500.0);
-        body_one.y = 400.0;
-        let mut label = make_merge_item("Report", 450.0, 70.0);
-        label.y = 780.0;
-        let mut contextual = make_merge_item("1", 526.0, 7.0);
-        contextual.y = 780.0;
-
-        let mut body_two = body_one.clone();
-        body_two.page = 2;
-        let mut isolated = make_merge_item("2", 50.0, 7.0);
-        isolated.page = 2;
-        isolated.y = 780.0;
-
-        let filtered =
-            filter_markdown_page_numbers(vec![body_one, label, contextual, body_two, isolated], 2);
-
-        assert!(filtered.iter().any(|item| item.text == "Report"));
-        assert!(filtered.iter().any(|item| item.text == "1"));
-        assert!(filtered.iter().all(|item| item.text != "2"));
-    }
-
-    #[test]
-    fn narrow_content_span_does_not_establish_adjacent_page_edges() {
-        let mut body_one = make_merge_item("Body text", 100.0, 120.0);
-        body_one.y = 400.0;
-        let mut label = make_merge_item("Report", 170.0, 60.0);
-        label.y = 780.0;
-        let mut contextual = make_merge_item("1", 235.0, 7.0);
-        contextual.y = 780.0;
-
-        let mut body_two = body_one.clone();
-        body_two.page = 2;
-        let mut isolated_two = make_merge_item("2", 100.0, 7.0);
-        isolated_two.page = 2;
-        isolated_two.y = 780.0;
-
-        let mut body_four = body_one.clone();
-        body_four.page = 4;
-        let mut isolated_four = make_merge_item("4", 100.0, 7.0);
-        isolated_four.page = 4;
-        isolated_four.y = 780.0;
-
-        let filtered = filter_markdown_page_numbers(
-            vec![
-                body_one,
-                label,
-                contextual,
-                body_two,
-                isolated_two,
-                body_four,
-                isolated_four,
-            ],
-            4,
-        );
-
-        assert!(filtered.iter().any(|item| item.text == "Report"));
-        assert!(filtered.iter().any(|item| item.text == "1"));
-    }
-
-    #[test]
-    fn same_edge_number_on_an_adjacent_page_is_not_folio_evidence() {
-        let mut body_one = make_merge_item("Body text", 50.0, 500.0);
-        body_one.y = 400.0;
-        let mut isolated = make_merge_item("42", 50.0, 14.0);
-        isolated.y = 780.0;
-
-        let mut body_two = body_one.clone();
-        body_two.page = 2;
-        let mut contextual = make_merge_item("43", 50.0, 14.0);
-        contextual.page = 2;
-        contextual.y = 780.0;
-        let mut label = make_merge_item("cases reviewed", 70.0, 90.0);
-        label.page = 2;
-        label.y = 780.0;
-
-        let filtered =
-            filter_markdown_page_numbers(vec![body_one, isolated, body_two, contextual, label], 2);
-
-        assert!(filtered.iter().any(|item| item.text == "43"));
-        assert!(filtered.iter().any(|item| item.text == "cases reviewed"));
-    }
-
-    #[test]
-    fn constant_number_in_repeated_deep_margin_header_is_preserved() {
-        let mut items = Vec::new();
-        for page in 1..=4 {
-            let mut year = make_merge_item("2026", 25.0, 24.0);
-            year.page = page;
-            year.y = 780.0;
-            let mut header = make_merge_item("Annual report", 53.0, 78.0);
-            header.page = page;
-            header.y = 780.0;
-            items.extend([year, header]);
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 4);
-        assert!(lines.iter().all(|line| line.text() == "2026 Annual report"));
-    }
-
-    #[test]
-    fn page_number_prefix_does_not_remove_substantive_text_during_layout() {
-        let mut items = vec![
-            make_merge_item("Page", 25.0, 28.0),
-            make_merge_item("42", 57.0, 12.0),
-            make_merge_item("explains", 73.0, 44.0),
-            make_merge_item("the result", 121.0, 55.0),
-        ];
-        for item in &mut items {
-            item.y = 50.0;
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "Page 42 explains the result");
-    }
-
-    #[test]
-    fn repeated_page_number_prefix_with_substantive_text_is_preserved_during_layout() {
-        let mut items = Vec::new();
-        for (page, value, chapter) in [(1, "42", "Chapter 1"), (2, "43", "Chapter 2")] {
-            let mut label = make_merge_item("Page", 25.0, 28.0);
-            label.page = page;
-            label.y = 50.0;
-            let mut page_number = make_merge_item(value, 57.0, 12.0);
-            page_number.page = page;
-            page_number.y = 50.0;
-            let mut suffix = make_merge_item(chapter, 73.0, 58.0);
-            suffix.page = page;
-            suffix.y = 50.0;
-            items.extend([label, page_number, suffix]);
-        }
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 2);
-        assert_eq!(lines[0].text(), "Page 42 Chapter 1");
-        assert_eq!(lines[1].text(), "Page 43 Chapter 2");
-    }
-
-    #[test]
-    fn page_number_phrase_in_the_page_body_is_preserved() {
-        let items = vec![
-            make_merge_item("Page", 25.0, 28.0),
-            make_merge_item("42", 57.0, 12.0),
-        ];
-
-        let lines = group_into_lines(items);
-
-        assert_eq!(lines.len(), 1);
-        assert_eq!(lines[0].text(), "Page 42");
-    }
-
-    #[test]
-    fn short_numeric_context_near_page_edge_is_preserved() {
-        let mut chapter = make_merge_item("Chapter", 100.0, 45.0);
-        chapter.y = 760.0;
-        let mut chapter_number = make_merge_item("1", 151.0, 6.0);
-        chapter_number.y = 760.0;
-
-        let chapter_lines = group_into_lines(vec![chapter, chapter_number]);
-        assert_eq!(chapter_lines.len(), 1);
-        assert_eq!(chapter_lines[0].text(), "Chapter 1");
-
-        let mut year = make_merge_item("2026", 100.0, 24.0);
-        year.y = 760.0;
-        let mut report = make_merge_item("Report", 130.0, 36.0);
-        report.y = 760.0;
-
-        let report_lines = group_into_lines(vec![year, report]);
-        assert_eq!(report_lines.len(), 1);
-        assert_eq!(report_lines[0].text(), "2026 Report");
-
-        let mut chapter = make_merge_item("Chapter", 100.0, 45.0);
-        chapter.y = 760.0;
-        let mut chapter_number = make_merge_item("1", 151.0, 6.0);
-        chapter_number.y = 760.0;
-        let mut edition_year = make_merge_item("2026", 163.0, 24.0);
-        edition_year.y = 760.0;
-
-        let chained_lines = group_into_lines(vec![chapter, chapter_number, edition_year]);
-        assert_eq!(chained_lines.len(), 1);
-        assert_eq!(chained_lines[0].text(), "Chapter 1 2026");
-    }
-
    #[test]
    fn test_bold_italic_detection() {
        // Test bold detection
@@ -44,16 +44,6 @@ const MIN_TABULAR_RULE_ITEMS: usize = 3;
 const MIN_TABULAR_RULE_GAPS: usize = 2;
 const TABULAR_RULE_GAP_EM: f32 = 2.0;

-/// Strikeout decorations are text-sized. Diagram connectors, signature
-/// lines, and chart rules often cross glyphs too, but extend well beyond the
-/// text they happen to intersect.
-const STRIKE_OWNER_PAD_EM: f32 = 0.75;
-const STRIKE_OWNER_MIN_PAD: f32 = 4.0;
-const STRIKE_ROW_Y_TOLERANCE_EM: f32 = 0.15;
-const STRIKE_ROW_Y_TOLERANCE_MIN: f32 = 5.0;
-const GRAPHIC_CONNECTION_EPS: f32 = 2.0;
-const GRAPHIC_CONNECTOR_MAX_THICKNESS: f32 = 4.0;
-
 #[derive(Clone)]
 pub(crate) struct UnderlineLine {
    pub(crate) x1: f32,
@@ -397,206 +387,6 @@ fn rule_strikes_item(rule: &Rule, item: &TextItem) -> bool {
    overlap >= min_overlap
 }

-fn is_bare_list_marker(text: &str) -> bool {
-    matches!(
-        text.trim(),
-        "•" | "◦" | "▪" | "▫" | "‣" | "⁃" | "●" | "○" | "■" | "□" | "-" | "*"
-    )
-}
-
-fn same_strike_row(left: &TextItem, right: &TextItem) -> bool {
-    let font_size = left.font_size.max(right.font_size);
-    let tolerance = (font_size * STRIKE_ROW_Y_TOLERANCE_EM).max(STRIKE_ROW_Y_TOLERANCE_MIN);
-    (left.y - right.y).abs() <= tolerance
-}
-
-fn is_inline_script(rule: &Rule, candidate: &TextItem, parent: &TextItem) -> bool {
-    if !is_underline_candidate(candidate)
-        || is_bare_list_marker(&candidate.text)
-        || candidate.font_size <= 0.0
-        || candidate.font_size >= parent.font_size * 0.75
-        || candidate.text.len() > 4
-        || !candidate.text.chars().all(|c| c.is_ascii_digit())
-        || (candidate.y - parent.y).abs() > 5.0
-    {
-        return false;
-    }
-
-    let parent_ends_with_letter = parent.text.chars().last().is_some_and(char::is_alphabetic);
-    if !parent_ends_with_letter {
-        return false;
-    }
-
-    let parent_right = parent.x + parent.width;
-    let gap = candidate.x - parent_right;
-    if gap >= parent.font_size * 0.2 || gap <= -parent.font_size * 0.3 {
-        return false;
-    }
-
-    let overlap = rule.x2.min(candidate.x + candidate.width) - rule.x1.max(candidate.x);
-    overlap >= candidate.width * MIN_X_OVERLAP
-}
-
-/// Return the items owned by a snug mid-glyph rule.
-///
-/// Real strikeout decorations track the width of the deleted text, including
-/// runs split by font/style changes and adjacent numeric super/subscripts.
-/// Non-text graphics can cross the same vertical window, but arrow shafts,
-/// signature lines, fraction bars, and chart rules extend materially beyond
-/// the intersected glyphs. Requiring the rule to stay within a small em-sized
-/// pad of a contiguous matched row separates those cases without relying on
-/// document-specific fonts or coordinates.
-///
-/// Ownership is computed once per rule. This keeps the strikeout pass at the
-/// same rule-by-item scale as underline detection instead of rescanning the
-/// whole page for every matching item.
-fn snug_strike_owner_indices(rule: &Rule, items: &[TextItem]) -> Vec<usize> {
-    let mut struck_indices: Vec<usize> = items
-        .iter()
-        .enumerate()
-        .filter_map(|(index, candidate)| {
-            (is_underline_candidate(candidate)
-                && !is_bare_list_marker(&candidate.text)
-                && rule_strikes_item(rule, candidate))
-            .then_some(index)
-        })
-        .collect();
-    if struck_indices.is_empty() {
-        return Vec::new();
-    }
-    struck_indices.sort_by(|&left, &right| items[left].y.total_cmp(&items[right].y));
-
-    let mut rows: Vec<Vec<usize>> = Vec::new();
-    for index in struck_indices {
-        if let Some(row) = rows
-            .last_mut()
-            .filter(|row| same_strike_row(&items[row[0]], &items[index]))
-        {
-            row.push(index);
-        } else {
-            rows.push(vec![index]);
-        }
-    }
-
-    let mut owned_indices = Vec::new();
-    for mut row in rows {
-        row.sort_by(|&left, &right| items[left].x.total_cmp(&items[right].x));
-
-        // Underline detection runs before the extractor's script-merging
-        // pass. Include the same tightly adjacent numeric script shape here
-        // when the rule spans it, so the owner width and semantic mark both
-        // survive that later merge.
-        let scripts: Vec<usize> = items
-            .iter()
-            .enumerate()
-            .filter_map(|(index, candidate)| {
-                let parent_pos =
-                    row.partition_point(|&row_index| items[row_index].x <= candidate.x);
-                let parent_index = parent_pos.checked_sub(1).map(|pos| row[pos])?;
-                is_inline_script(rule, candidate, &items[parent_index]).then_some(index)
-            })
-            .collect();
-        row.extend(scripts);
-        row.sort_by(|&left, &right| items[left].x.total_cmp(&items[right].x));
-        row.dedup();
-
-        let x1 = row
-            .iter()
-            .map(|&index| items[index].x)
-            .fold(f32::INFINITY, f32::min);
-        let x2 = row
-            .iter()
-            .map(|&index| items[index].x + items[index].width)
-            .fold(f32::NEG_INFINITY, f32::max);
-        let max_font_size = row
-            .iter()
-            .map(|&index| items[index].font_size)
-            .fold(0.0, f32::max);
-        let pad = (max_font_size * STRIKE_OWNER_PAD_EM).max(STRIKE_OWNER_MIN_PAD);
-
-        if rule.x1 < x1 - pad || rule.x2 > x2 + pad {
-            continue;
-        }
-
-        let contiguous = row.windows(2).all(|pair| {
-            let gap = items[pair[1]].x - (items[pair[0]].x + items[pair[0]].width);
-            gap <= (max_font_size * 2.0).max(12.0)
-        });
-        if contiguous {
-            owned_indices.extend(row);
-        }
-    }
-
-    owned_indices.sort_unstable();
-    owned_indices.dedup();
-    owned_indices
-}
-
-/// Diagram and table rules participate in larger path geometry. A vertical
-/// or diagonal segment meeting the candidate rule is strong evidence that
-/// the horizontal segment is a connector, border, arrow, or symbol rather
-/// than an isolated text decoration.
-fn has_connected_nonhorizontal_segment(rule: &Rule, lines: &[UnderlineLine], page: u32) -> bool {
-    lines.iter().any(|line| {
-        if line.page != page {
-            return false;
-        }
-
-        let dx = line.x2 - line.x1;
-        let dy = line.y2 - line.y1;
-        if dy.abs() <= MAX_RULE_THICKNESS {
-            return false;
-        }
-
-        let y_min = line.y1.min(line.y2) - GRAPHIC_CONNECTION_EPS;
-        let y_max = line.y1.max(line.y2) + GRAPHIC_CONNECTION_EPS;
-        if rule.y < y_min || rule.y > y_max {
-            return false;
-        }
-
-        let t = (rule.y - line.y1) / dy;
-        if !(-0.05..=1.05).contains(&t) {
-            return false;
-        }
-        let intersection_x = line.x1 + t * dx;
-        intersection_x >= rule.x1 - GRAPHIC_CONNECTION_EPS
-            && intersection_x <= rule.x2 + GRAPHIC_CONNECTION_EPS
-    })
-}
-
-/// Filled diagrams often build connectors from intersecting thin rectangles
-/// instead of stroked path segments. Treat only narrow, vertically elongated
-/// rectangles as connector geometry; broad fills can legitimately sit behind
-/// struck text and must not veto its decoration.
-fn has_connected_nonhorizontal_rect(rule: &Rule, rects: &[PdfRect], page: u32) -> bool {
-    rects.iter().any(|rect| {
-        if rect.page != page {
-            return false;
-        }
-
-        let (x1, x2) = if rect.width >= 0.0 {
-            (rect.x, rect.x + rect.width)
-        } else {
-            (rect.x + rect.width, rect.x)
-        };
-        let (y1, y2) = if rect.height >= 0.0 {
-            (rect.y, rect.y + rect.height)
-        } else {
-            (rect.y + rect.height, rect.y)
-        };
-        let width = x2 - x1;
-        let height = y2 - y1;
-
-        width > 0.0
-            && width <= GRAPHIC_CONNECTOR_MAX_THICKNESS
-            && height > width * 2.0
-            && rule.y >= y1 - GRAPHIC_CONNECTION_EPS
-            && rule.y <= y2 + GRAPHIC_CONNECTION_EPS
-            && x2 >= rule.x1 - GRAPHIC_CONNECTION_EPS
-            && x1 <= rule.x2 + GRAPHIC_CONNECTION_EPS
-    })
-}
-
 /// Mark `is_underline` on text items that have a horizontal rule just
 /// below their baseline, and `is_strikeout` on items whose glyphs a rule
 /// crosses at mid x-height. `items`, `rects`, and `lines` are a single
@@ -650,43 +440,27 @@ pub(crate) fn mark_underlined_items(
        .map(|(i, _)| i)
        .collect();

-    let mut strikeout_items = vec![false; items.len()];
-    for (rule_idx, rule) in rules.iter().enumerate() {
-        if tabular_rules.contains(&rule_idx)
-            || has_connected_nonhorizontal_segment(rule, lines, page)
-            || has_connected_nonhorizontal_rect(rule, rects, page)
-        {
-            continue;
-        }
-        for item_idx in snug_strike_owner_indices(rule, items) {
-            strikeout_items[item_idx] = true;
-        }
-    }
-
-    let underlined_items: HashSet<usize> = items
-        .iter()
-        .enumerate()
-        .filter(|(_, item)| {
-            is_underline_candidate(item)
-                && rules.iter().enumerate().any(|(rule_idx, rule)| {
-                    !tabular_rules.contains(&rule_idx)
-                        && !fraction_rules.contains(&rule_idx)
-                        && rule_matches_item(rule, item)
-                })
-        })
-        .map(|(item_idx, _)| item_idx)
-        .collect();
-
-    for (item_idx, item) in items.iter_mut().enumerate() {
+    for item in items.iter_mut() {
        if !is_underline_candidate(item) {
            continue;
        }

-        if strikeout_items[item_idx] {
-            item.is_strikeout = true;
-        }
-        if underlined_items.contains(&item_idx) {
-            item.is_underline = true;
+        for (rule_idx, rule) in rules.iter().enumerate() {
+            if tabular_rules.contains(&rule_idx) {
+                continue;
+            }
+            // The fraction guard only gates UNDERLINE marking — a rule that
+            // reads as a fraction bar from below can still legitimately
+            // strike through a line above it.
+            if !fraction_rules.contains(&rule_idx) && rule_matches_item(rule, item) {
+                item.is_underline = true;
+            }
+            if rule_strikes_item(rule, item) {
+                item.is_strikeout = true;
+            }
+            if item.is_underline && item.is_strikeout {
+                break;
+            }
        }
    }
 }
@@ -806,148 +580,6 @@ mod tests {
        assert!(!items[0].is_underline);
    }

-    #[test]
-    fn long_connector_crossing_text_is_not_a_strikeout() {
-        let mut items = vec![item("diagram label", 160.0, 500.0, 60.0, 10.0)];
-        let lines = vec![hline(100.0, 280.0, 503.0)];
-
-        mark_underlined_items(&mut items, &[], &lines, 1);
-
-        assert!(!items[0].is_strikeout);
-    }
-
-    #[test]
-    fn chart_rule_ending_inside_short_label_is_not_a_strikeout() {
-        let mut items = vec![item("T 18", 300.0, 500.0, 20.0, 10.0)];
-        let lines = vec![hline(100.0, 315.0, 503.0)];
-
-        mark_underlined_items(&mut items, &[], &lines, 1);
-
-        assert!(!items[0].is_strikeout);
-    }
-
-    #[test]
-    fn connected_diagram_segment_is_not_a_strikeout() {
-        let mut items = vec![item("V8", 100.0, 500.0, 12.0, 10.0)];
-        let lines = vec![
-            hline(99.0, 113.0, 503.0),
-            UnderlineLine {
-                x1: 106.0,
-                y1: 496.0,
-                x2: 109.0,
-                y2: 510.0,
-                stroke_width: 1.0,
-                page: 1,
-            },
-        ];
-
-        mark_underlined_items(&mut items, &[], &lines, 1);
-
-        assert!(!items[0].is_strikeout);
-    }
-
-    #[test]
-    fn connected_filled_rect_is_not_a_strikeout() {
-        let mut items = vec![item("V8", 100.0, 500.0, 12.0, 10.0)];
-        let rects = vec![
-            thin_rect(99.0, 502.6, 14.0),
-            PdfRect {
-                x: 106.0,
-                y: 496.0,
-                width: 2.0,
-                height: 14.0,
-                page: 1,
-            },
-        ];
-
-        mark_underlined_items(&mut items, &rects, &[], 1);
-
-        assert!(!items[0].is_strikeout);
-    }
-
-    #[test]
-    fn broad_fill_behind_text_does_not_block_strikeout() {
-        let mut items = vec![item("deleted", 100.0, 500.0, 40.0, 10.0)];
-        let rects = vec![
-            thin_rect(99.0, 502.6, 42.0),
-            PdfRect {
-                x: 90.0,
-                y: 490.0,
-                width: 100.0,
-                height: 20.0,
-                page: 1,
-            },
-        ];
-
-        mark_underlined_items(&mut items, &rects, &[], 1);
-
-        assert!(items[0].is_strikeout);
-    }
-
-    #[test]
-    fn list_bullet_is_not_a_strikeout() {
-        for marker in ["•", "-", "*"] {
-            let mut items = vec![item(marker, 100.0, 500.0, 6.0, 10.0)];
-            let lines = vec![hline(99.0, 107.0, 503.0)];
-
-            mark_underlined_items(&mut items, &[], &lines, 1);
-
-            assert!(!items[0].is_strikeout, "marker {marker:?}");
-        }
-    }
-
-    #[test]
-    fn snug_rule_marks_adjacent_split_runs_as_strikeout() {
-        let mut items = vec![
-            item("deleted", 100.0, 500.0, 40.0, 10.0),
-            item("text", 142.0, 500.0, 25.0, 10.0),
-        ];
-        let lines = vec![hline(99.0, 168.0, 503.0)];
-
-        mark_underlined_items(&mut items, &[], &lines, 1);
-
-        assert!(items.iter().all(|item| item.is_strikeout));
-    }
-
-    #[test]
-    fn snug_rule_groups_split_runs_with_baseline_drift() {
-        let mut items = vec![
-            item("deleted", 100.0, 500.0, 40.0, 10.0),
-            item("text", 142.0, 498.0, 25.0, 10.0),
-        ];
-        let lines = vec![hline(99.0, 168.0, 503.0)];
-
-        mark_underlined_items(&mut items, &[], &lines, 1);
-
-        assert!(items.iter().all(|item| item.is_strikeout));
-    }
-
-    #[test]
-    fn snug_rule_keeps_inline_superscript_in_strike_owner() {
-        let mut items = vec![
-            item("deleted", 100.0, 500.0, 40.0, 10.0),
-            item("2", 140.5, 503.0, 4.0, 6.0),
-        ];
-        let lines = vec![hline(99.0, 145.0, 503.0)];
-
-        mark_underlined_items(&mut items, &[], &lines, 1);
-
-        assert!(items.iter().all(|item| item.is_strikeout));
-    }
-
-    #[test]
-    fn snug_rule_keeps_inline_subscript_in_strike_owner() {
-        let mut items = vec![
-            item("deleted", 100.0, 500.0, 40.0, 10.0),
-            item("2", 140.5, 497.0, 4.0, 6.0),
-        ];
-        let lines = vec![hline(99.0, 145.0, 503.0)];
-
-        mark_underlined_items(&mut items, &[], &lines, 1);
-
-        assert!(items.iter().all(|item| item.is_strikeout));
-    }
-
    #[test]
    fn baseline_rule_marks_underline_not_strikeout() {
        let mut items = vec![item("underlined", 100.0, 500.0, 60.0, 10.0)];
@@ -50,11 +50,11 @@ pub use detector::{
 };
 pub use extractor::{
    extract_text, extract_text_with_positions, extract_text_with_positions_mem,
-    extract_text_with_positions_pages, extract_text_with_positions_pages_with_password,
+    extract_text_with_positions_pages,
 };
 pub use markdown::{
-    to_markdown, to_markdown_from_items, to_markdown_from_items_with_rects,
-    to_markdown_from_items_with_rects_and_page_count, MarkdownOptions, MarkdownProfile,
+    to_markdown, to_markdown_from_items, to_markdown_from_items_with_rects, MarkdownOptions,
+    MarkdownProfile,
 };
 pub use process_mode::ProcessMode;
 pub use types::{LayoutComplexity, PdfLine, PdfRect, TextItem};
@@ -462,39 +462,16 @@ pub fn extract_pages_markdown_mem(
    let (doc, page_count) = load_document_from_mem(buffer)?;
    let font_cmaps = FontCMaps::from_doc(&doc);

-    // Extract ALL pages to get accurate, document-wide font stats. A malformed
-    // unselected page cannot make a valid requested page fail, but errors on a
-    // requested page retain the normal extraction semantics.
-    let required_pages: Option<HashSet<u32>> = pages.map(|pages| {
-        pages
-            .iter()
-            .filter_map(|page| page.checked_add(1))
-            .collect()
-    });
+    // Extract ALL pages to get accurate, document-wide font stats.
    let ((all_items, all_rects, all_lines), page_thresholds, gid_pages) =
-        if let Some(required_pages) = required_pages.as_ref() {
-            extractor::extract_positioned_text_for_document_analysis(
-                &doc,
-                &font_cmaps,
-                required_pages,
-            )?
-        } else {
-            extractor::extract_positioned_text_from_doc(&doc, &font_cmaps, None)?
-        };
+        extractor::extract_positioned_text_from_doc(&doc, &font_cmaps, None)?;
    let text_quality = analyze_text_quality(&all_items);

-    // Resolve page numbers with full-document context before partitioning.
-    // Per-page Markdown receives the original items plus these decisions so
-    // table detection can retain legitimate numeric cells.
-    let (filtered_items, removed_page_number_pages, page_number_removal_mask) =
-        extractor::filter_markdown_page_numbers_with_removed_pages(all_items.clone(), page_count);
-
-    // Tables need the original numeric cells; columns use folio-cleaned
-    // evidence so removed page numbers cannot create false layout metadata.
-    let complexity = compute_layout_complexity(&all_items, &filtered_items, &all_rects, &all_lines);
+    // Compute layout complexity from full document (near-zero cost).
+    let complexity = compute_layout_complexity(&all_items, &all_rects, &all_lines);

    // Compute font stats from full document (cross-page consistency).
-    let font_stats = markdown::analysis::calculate_font_stats_from_items(&filtered_items);
+    let font_stats = markdown::analysis::calculate_font_stats_from_items(&all_items);

    // When caller doesn't specify pages, return every page in document order.
    let all_pages: Vec<u32>;
@@ -525,13 +502,12 @@ pub fn extract_pages_markdown_mem(

        let page_1idx = page_0idx + 1;

-        // Partition items, removal decisions, and rects for this page only.
-        let (page_items, page_number_removal_mask): (Vec<TextItem>, Vec<bool>) = all_items
+        // Filter items/rects for this page only
+        let page_items: Vec<TextItem> = all_items
            .iter()
-            .zip(&page_number_removal_mask)
-            .filter(|(item, _)| item.page == page_1idx)
-            .map(|(item, remove)| (item.clone(), *remove))
-            .unzip();
+            .filter(|i| i.page == page_1idx)
+            .cloned()
+            .collect();

        let page_rects: Vec<PdfRect> = all_rects
            .iter()
@@ -558,14 +534,9 @@ pub fn extract_pages_markdown_mem(
                options,
                &page_rects,
                &[],
-                markdown::MarkdownDocumentContext {
-                    page_thresholds: &page_thresholds,
-                    struct_roles: None,
-                    struct_tables: &[],
-                    page_count,
-                    prefiltered_page_number_pages: Some(&removed_page_number_pages),
-                    prefiltered_page_number_mask: Some(&page_number_removal_mask),
-                },
+                &page_thresholds,
+                None,
+                &[],
            )
        };

@@ -1090,12 +1061,7 @@ pub fn extract_tables_in_regions_mem(
                    candidates.push(candidate);
                }
            }
-            let detected = tables::detect_tables_with_page_width(
-                &matched,
-                base_font_size,
-                false,
-                items.map_or(1.0, |items| tables::content_width(items)),
-            );
+            let detected = tables::detect_tables(&matched, base_font_size, false);
            if let Some(candidate) = detected
                .iter()
                .find_map(|t| evaluate(TableCandidateSource::Heuristic, t))
@@ -3639,10 +3605,7 @@ fn process_document(
    // Step 2 — Extraction (reuses the already-loaded document)
    let extracted = {
        let font_cmaps = FontCMaps::from_doc(&doc);
-        // Most page-filtered requests extract only the selected pages. Gather
-        // other pages only when a selected contextual folio needs cross-page
-        // evidence; failures on those context-only pages are non-fatal.
-        let result = extractor::extract_positioned_text_with_folio_context(
+        let result = extractor::extract_positioned_text_from_doc(
            &doc,
            &font_cmaps,
            options.page_filter.as_ref(),
@@ -3653,19 +3616,9 @@ fn process_document(
        // This unlocks OCR text layers behind scanned images.
        if pdf_type == PdfType::Mixed {
            if let Ok((ref items, _, _)) = result.as_ref().map(|(e, _, _)| e) {
-                let sample: String = items
-                    .iter()
-                    .filter(|item| {
-                        options
-                            .page_filter
-                            .as_ref()
-                            .is_none_or(|filter| filter.contains(&item.page))
-                    })
-                    .take(200)
-                    .map(|item| item.text.as_str())
-                    .collect();
+                let sample: String = items.iter().take(200).map(|i| i.text.as_str()).collect();
                if is_garbage_text(&sample) || sample.trim().is_empty() {
-                    extractor::extract_positioned_text_include_invisible_with_folio_context(
+                    extractor::extract_positioned_text_include_invisible(
                        &doc,
                        &font_cmaps,
                        options.page_filter.as_ref(),
@@ -3675,7 +3628,7 @@ fn process_document(
                }
            } else {
                // Normal extraction failed — try invisible as fallback
-                extractor::extract_positioned_text_include_invisible_with_folio_context(
+                extractor::extract_positioned_text_include_invisible(
                    &doc,
                    &font_cmaps,
                    options.page_filter.as_ref(),
@@ -3739,13 +3692,6 @@ fn process_document(
                    let mut garbage_pages: std::collections::HashSet<u32> =
                        std::collections::HashSet::new();
                    for &pg in &ocr_set {
-                        if options
-                            .page_filter
-                            .as_ref()
-                            .is_some_and(|filter| !filter.contains(&pg))
-                        {
-                            continue;
-                        }
                        let page_text: String = items
                            .iter()
                            .filter(|i| i.page == pg)
@@ -3785,38 +3731,9 @@ fn process_document(
                    }
                };

-            let selected_page = |page: u32| {
-                options
-                    .page_filter
-                    .as_ref()
-                    .is_none_or(|filter| filter.contains(&page))
-            };
-            let rects: Vec<_> = rects
-                .into_iter()
-                .filter(|rect| selected_page(rect.page))
-                .collect();
-            let lines: Vec<_> = lines
-                .into_iter()
-                .filter(|line| selected_page(line.page))
-                .collect();
-            let gid_encoded_pages: HashSet<_> = gid_encoded_pages
-                .into_iter()
-                .filter(|page| selected_page(*page))
-                .collect();
-            let FolioFilteredItems {
-                items,
-                layout_items,
-                removal_mask,
-                removed_pages,
-            } = select_items_with_document_folio_context(
-                items,
-                page_count,
-                options.page_filter.as_ref(),
-            );
-
            let text_quality = analyze_text_quality(&items);
            merge_ocr_reasons(&mut ocr_reasons_by_page, text_quality.reasons_by_page);
-            let layout = compute_layout_complexity(&items, &layout_items, &rects, &lines);
+            let layout = compute_layout_complexity(&items, &rects, &lines);

            let md = if options.mode == ProcessMode::Analyze {
                None
@@ -3826,14 +3743,9 @@ fn process_document(
                    options.markdown,
                    &rects,
                    &lines,
-                    markdown::MarkdownDocumentContext {
-                        page_thresholds: &page_thresholds,
-                        struct_roles: struct_roles.as_ref(),
-                        struct_tables: &struct_tables,
-                        page_count,
-                        prefiltered_page_number_pages: Some(&removed_pages),
-                        prefiltered_page_number_mask: Some(removal_mask.as_slice()),
-                    },
+                    &page_thresholds,
+                    struct_roles.as_ref(),
+                    &struct_tables,
                ))
            };

@@ -5592,50 +5504,9 @@ mod looks_like_partial_table_tests {
    }
 }

-struct FolioFilteredItems {
-    items: Vec<types::TextItem>,
-    layout_items: Vec<types::TextItem>,
-    removal_mask: Vec<bool>,
-    removed_pages: HashSet<u32>,
-}
-
-/// Resolve folios with complete document context, then select the caller's
-/// requested pages without losing those decisions.
-fn select_items_with_document_folio_context(
-    all_items: Vec<types::TextItem>,
-    page_count: u32,
-    page_filter: Option<&HashSet<u32>>,
-) -> FolioFilteredItems {
-    let (all_layout_items, all_removed_pages, all_removal_mask) =
-        extractor::filter_markdown_page_numbers_with_removed_pages(all_items.clone(), page_count);
-    let selected_page = |page: u32| page_filter.is_none_or(|filter| filter.contains(&page));
-
-    let (items, removal_mask) = all_items
-        .into_iter()
-        .zip(all_removal_mask)
-        .filter(|(item, _)| selected_page(item.page))
-        .unzip();
-    let layout_items = all_layout_items
-        .into_iter()
-        .filter(|item| selected_page(item.page))
-        .collect();
-    let removed_pages = all_removed_pages
-        .into_iter()
-        .filter(|page| selected_page(*page))
-        .collect();
-
-    FolioFilteredItems {
-        items,
-        layout_items,
-        removal_mask,
-        removed_pages,
-    }
-}
-
 /// Analyse extracted items and rects for layout complexity.
 fn compute_layout_complexity(
    items: &[types::TextItem],
-    column_items: &[types::TextItem],
    rects: &[types::PdfRect],
    lines: &[types::PdfLine],
 ) -> LayoutComplexity {
@@ -5657,7 +5528,6 @@ fn compute_layout_complexity(

        // Check for side-by-side layout
        let owned_items: Vec<types::TextItem> = page_items.iter().map(|i| (*i).clone()).collect();
-        let page_content_width = tables::content_width(&owned_items);
        let bands = markdown::split_side_by_side(&owned_items);

        let band_ranges: Vec<(f32, f32)> = if bands.is_empty() {
@@ -5708,12 +5578,7 @@ fn compute_layout_complexity(
                break;
            }
            // Heuristic fallback for borderless tables
-            let heuristic_tables = tables::detect_tables_with_page_width(
-                &band_items,
-                base_size,
-                false,
-                page_content_width,
-            );
+            let heuristic_tables = tables::detect_tables(&band_items, base_size, false);
            if has_data_table(&heuristic_tables) {
                found_table = true;
                break;
@@ -5726,7 +5591,7 @@ fn compute_layout_complexity(

    let mut pages_with_columns: Vec<u32> = Vec::new();
    for page in seen_pages {
-        let cols = extractor::detect_columns(column_items, page, pages_with_tables.contains(&page));
+        let cols = extractor::detect_columns(items, page, pages_with_tables.contains(&page));
        if cols.len() >= 2 {
            pages_with_columns.push(page);
        }
@@ -5938,68 +5803,6 @@ mod tests {
        }
    }

-    #[test]
-    fn removed_sparse_folios_leave_no_layout_evidence() {
-        let items = vec![
-            test_item("1", 25.0, 20.0, 12.0, 10.0),
-            test_item("2", 520.0, 60.0, 12.0, 10.0),
-        ];
-        let (filtered, _, _) =
-            extractor::filter_markdown_page_numbers_with_removed_pages(items.clone(), 1);
-        assert!(filtered.is_empty());
-
-        let filtered = compute_layout_complexity(&items, &filtered, &[], &[]);
-
-        assert!(!filtered.is_complex);
-        assert!(filtered.pages_with_tables.is_empty());
-        assert!(filtered.pages_with_columns.is_empty());
-    }
-
-    #[test]
-    fn page_selection_keeps_document_wide_folio_layout_decisions() {
-        let mut items = Vec::new();
-        for page in 1..=4 {
-            for row in 0..8 {
-                let y = 20.0 + row as f32 * 8.0;
-                let mut folio = test_item(&(row * 10 + page).to_string(), 25.0, y, 12.0, 10.0);
-                folio.page = page;
-                let mut footer =
-                    test_item(&format!("Footer row {row} summary"), 43.0, y, 470.0, 10.0);
-                footer.page = page;
-                let mut body = test_item(&format!("Body{row}"), 530.0, y, 55.0, 10.0);
-                body.page = page;
-                items.extend([folio, footer, body]);
-            }
-        }
-
-        let page_one_items: Vec<_> = items
-            .iter()
-            .filter(|item| item.page == 1)
-            .cloned()
-            .collect();
-        let (page_local_layout, _, _) =
-            extractor::filter_markdown_page_numbers_with_removed_pages(page_one_items.clone(), 4);
-        let page_local = compute_layout_complexity(&page_one_items, &page_local_layout, &[], &[]);
-        assert!(
-            page_local.pages_with_columns.contains(&1),
-            "fixture must reproduce page-local folio column evidence"
-        );
-
-        let selected =
-            select_items_with_document_folio_context(items, 4, Some(&HashSet::from([1])));
-        assert_eq!(
-            selected
-                .removal_mask
-                .iter()
-                .filter(|remove| **remove)
-                .count(),
-            8
-        );
-        let document_wide =
-            compute_layout_complexity(&selected.items, &selected.layout_items, &[], &[]);
-        assert!(!document_wide.pages_with_columns.contains(&1));
-    }
-
    #[test]
    fn test_detect_encoding_issues_fffd() {
        assert!(detect_encoding_issues(
@@ -171,74 +171,6 @@ pub(crate) fn is_toc_marker_heading(text: &str) -> bool {
 /// equation and absent from name-plus-number headings. A bare trailing colon
 /// is NOT a fragment signal either: real headings frequently end with colons
 /// ("Procedure:", "Steps for Using the Microscope:").
-/// True when the line opens with a section number ("3.", "2.1.4", "IV)").
-///
-/// Mirrors the acceptance of `heading::parse_numbering` rather than the
-/// stricter `convert::starts_with_section_number`, which deliberately
-/// requires two components because it bypasses isolation checks. Here a
-/// single "1." counts: numbering is independent evidence of a heading, and
-/// `heading.rs` applies its numbered-prefix allowance *after* consulting
-/// `is_heading_fragment`, so without this exemption a numbered
-/// sentence-case heading would be vetoed before that allowance can run.
-fn starts_with_numbering_prefix(t: &str) -> bool {
-    let Some(first) = t.split_whitespace().next() else {
-        return false;
-    };
-    let has_delimiter = first.ends_with(['.', ')', ':']);
-    let token = first.trim_end_matches(['.', ')', ':']);
-    if token.is_empty() {
-        return false;
-    }
-    let parts: Vec<&str> = token.split('.').collect();
-    let decimal = parts
-        .iter()
-        .all(|p| !p.is_empty() && p.len() <= 3 && p.chars().all(|c| c.is_ascii_digit()));
-    if decimal {
-        // "1." / "2.1." carry a delimiter; "2.3 Title" is written without
-        // one, so a multi-component number is accepted bare. A bare single
-        // number ("3 apples") is not — that is ordinary prose.
-        return has_delimiter || parts.len() >= 2;
-    }
-    // Roman numerals go through the heading parser's own grammar so the two
-    // agree: uppercase I/V/X/L/C only, at most 8 characters. A looser rule
-    // here would exempt markers the parser rejects — "iv)" or "d)" from an
-    // alphabetical list — letting an ordinary list item bypass the veto and
-    // reach heading promotion.
-    //
-    // A delimiter is also required: a bare leading "I" is the pronoun far
-    // more often than a section number.
-    has_delimiter && crate::markdown::heading::roman_value(token).is_some()
-}
-
-/// True when the line reads as a title rather than a sentence: every
-/// content word (ignoring minor words) starts uppercase. Used to spare real
-/// headings from the dangling-verb veto — "Bond Yields" is a section title,
-/// "the method yields" is a stranded clause, and only the casing tells them
-/// apart.
-fn looks_title_case(t: &str) -> bool {
-    const MINOR: &[&str] = &[
-        "a", "an", "the", "of", "and", "or", "for", "to", "in", "on", "at", "by", "with", "from",
-        "as", "is", "are", "that", "than", "into",
-    ];
-    let mut content = 0usize;
-    let mut capitalized = 0usize;
-    for w in t.split_whitespace() {
-        let cleaned: String = w.chars().filter(|c| c.is_alphabetic()).collect();
-        if cleaned.is_empty() {
-            continue;
-        }
-        if MINOR.contains(&cleaned.to_lowercase().as_str()) {
-            continue;
-        }
-        content += 1;
-        if cleaned.chars().next().is_some_and(char::is_uppercase) {
-            capitalized += 1;
-        }
-    }
-    // A single content word ("Yields") is a title by default.
-    content == 0 || capitalized == content
-}
-
 pub(crate) fn is_heading_fragment(text: &str) -> bool {
    let t = text.trim_end();

@@ -312,133 +244,9 @@ pub(crate) fn is_heading_fragment(text: &str) -> bool {
    if t.ends_with(':') && t.split_whitespace().any(is_equation_number) {
        return true;
    }
-
-    // Dangling clause: a stranded sentence lead-in ends on a relational
-    // verb with no terminal punctuation — "Note that the exact error equals"
-    // left ahead of its formula when a phantom table dissolved.
-    //
-    // Gated on the line reading as prose rather than a title. Case is the
-    // discriminator the trailing word alone cannot provide: a heading is
-    // title case ("Bond Yields", "The Method Yields") while a stranded
-    // lead-in is sentence case ("the method yields"). Without this gate the
-    // veto eats real headings — "Bond Yields", "Crop Yields" and any wrapped
-    // title-case heading the preprocessor failed to merge.
-    if !t.ends_with(['.', '!', '?', ':', ';', ')', ']'])
-        && !looks_title_case(t)
-        && !starts_with_numbering_prefix(t)
-    {
-        if let Some(last) = t.split_whitespace().next_back() {
-            let word: String = last
-                .trim_matches(|c: char| !c.is_alphanumeric())
-                .to_lowercase();
-            // Relational verbs only, and only those with no common noun
-            // sense. "yields" was dropped for exactly that reason: "Bond
-            // Yields" is a real section title. Function words, copulas and
-            // auxiliaries were measured and rejected outright — a heading
-            // that wraps across lines ends on those, and suppressing them
-            // destroyed real IRS Publication 17 headings.
-            const DANGLING_TAIL: &[&str] =
-                &["equals", "denotes", "implies", "satisfies", "signifies"];
-            if DANGLING_TAIL.contains(&word.as_str()) {
-                return true;
-            }
-        }
-    }
    false
 }

-#[cfg(test)]
-mod fragment_heading_tests {
-    use super::is_heading_fragment;
-
-    #[test]
-    fn dangling_tail_marks_stranded_clause() {
-        // opendataloader 01030000000144: left behind when a phantom table
-        // dissolved, ahead of its formula on the next line.
-        assert!(is_heading_fragment("Note that the exact error equals"));
-        assert!(is_heading_fragment("The remainder term satisfies"));
-        assert!(is_heading_fragment("we conclude that the sum equals"));
-    }
-
-    #[test]
-    fn real_headings_survive() {
-        assert!(!is_heading_fragment("Introduction"));
-        assert!(!is_heading_fragment("Error Analysis"));
-        assert!(!is_heading_fragment("Materials and Methods"));
-        assert!(!is_heading_fragment("Results"));
-        assert!(!is_heading_fragment("3.2 Richardson Extrapolation"));
-        assert!(!is_heading_fragment("Discussion and Conclusions"));
-        // Terminal punctuation means the clause is complete.
-        assert!(!is_heading_fragment("What is a Derivative?"));
-        assert!(!is_heading_fragment("Procedure:"));
-        assert!(!is_heading_fragment("Note that this is important."));
-    }
-
-    #[test]
-    fn title_case_headings_ending_in_a_verb_survive() {
-        // "yields" is also a plural noun; these are real section titles.
-        assert!(!is_heading_fragment("Bond Yields"));
-        assert!(!is_heading_fragment("Crop Yields"));
-        assert!(!is_heading_fragment("Dividend Yields"));
-        assert!(!is_heading_fragment("Yields"));
-        // A wrapped title-case heading whose first line ends on a listed
-        // verb must survive even if the preprocessor failed to merge it.
-        assert!(!is_heading_fragment("The Theorem Implies"));
-        assert!(!is_heading_fragment("What This Denotes"));
-    }
-
-    #[test]
-    fn numbered_sentence_case_headings_survive() {
-        // heading.rs consults is_heading_fragment BEFORE applying its
-        // numbered-prefix allowance, so the veto must not pre-empt it.
-        assert!(!is_heading_fragment("1. What the model implies"));
-        assert!(!is_heading_fragment("2.3 How the estimator satisfies"));
-        assert!(!is_heading_fragment("IV) What this denotes"));
-        // Without numbering the same wording is still a stranded clause.
-        assert!(is_heading_fragment("What the model implies"));
-        // A bare leading number or pronoun is prose, not numbering.
-        assert!(is_heading_fragment("3 apples and what that implies"));
-        assert!(is_heading_fragment("I think the model implies"));
-        // Markers heading::parse_numbering rejects must not be exempted
-        // either, or an ordinary list item bypasses the veto: lowercase
-        // roman, alphabetical markers, and over-long tokens.
-        assert!(is_heading_fragment("iv) the estimator satisfies"));
-        assert!(is_heading_fragment("d) the value implies"));
-        // Unsupported character (M is outside the parser's I/V/X/L/C set).
-        assert!(is_heading_fragment("MMMM. the value implies"));
-        // Over-long token: nine valid characters, so this exercises the
-        // 8-character bound rather than the character set.
-        assert!(is_heading_fragment("IIIIIIIII. the value implies"));
-        // Eight is still within the bound and stays exempt.
-        assert!(!is_heading_fragment("IIIIIIII. What this implies"));
-        // Uppercase roman within the parser's grammar is still exempt.
-        assert!(!is_heading_fragment("IV. What this denotes"));
-        assert!(!is_heading_fragment("XII) What this implies"));
-    }
-
-    #[test]
-    fn wrapped_headings_are_not_fragments() {
-        // A heading that wraps across lines ends on a function word. These
-        // are real headings from IRS Publication 17 and must survive.
-        assert!(!is_heading_fragment("Casualty and"));
-        assert!(!is_heading_fragment("Rule 10. You Must Be at"));
-        assert!(!is_heading_fragment("Higher Standard Deduction for"));
-        assert!(!is_heading_fragment("Qualifying Child of"));
-        assert!(!is_heading_fragment("When Can I Withdraw or"));
-        // Copulas and auxiliaries also end real wrapped headings.
-        assert!(!is_heading_fragment("Rule 15. Your AGI Must Be"));
-        assert!(!is_heading_fragment("What Medical Expenses Are"));
-        assert!(!is_heading_fragment("Rule 13. You Must Have"));
-        assert!(!is_heading_fragment("When Can a Roth IRA Be"));
-    }
-
-    #[test]
-    fn dangling_check_is_case_insensitive() {
-        // All-caps is not sentence case, so the veto must not fire there.
-        assert!(!is_heading_fragment("THE REMAINDER EQUALS"));
-    }
-}
-
 /// Compute the Y-gap threshold for paragraph break detection.
 ///
 /// Instead of using a fixed multiple of base_size (which fails for double-spaced
@@ -560,23 +368,95 @@ pub(crate) fn compute_paragraph_threshold(lines: &[TextLine], base_size: f32) ->
 /// Discover distinct heading font-size tiers in the document.
 /// Returns tiers sorted largest-first (tier 0 = H1, tier 1 = H2, …).
 /// Sizes within 0.5pt are clustered into the same tier. Capped at 4 tiers.
+/// Dominant font size of a line: the size cluster (0.5pt tolerance) covering
+/// the most alphanumeric characters. A line led by an oversized ornament
+/// (drop cap, Type3 display-math delimiter) must be classified by its body
+/// text, not its first glyph — and operators/delimiters don't get a vote,
+/// so a display equation whose parens and plus signs render larger than its
+/// variables still classifies at the variables' size. Falls back to counting
+/// all non-whitespace chars for lines with no alphanumerics. Returns `None`
+/// for lines with no text.
+pub(crate) fn line_dominant_font_size(line: &TextLine) -> Option<f32> {
+    fn dominant(line: &TextLine, count_chars: fn(&str) -> usize) -> Option<f32> {
+        let mut clusters: Vec<(f32, usize)> = Vec::new();
+        for item in &line.items {
+            let chars = count_chars(&item.text);
+            if chars == 0 {
+                continue;
+            }
+            if let Some((_, count)) = clusters
+                .iter_mut()
+                .find(|(s, _)| (*s - item.font_size).abs() < 0.5)
+            {
+                *count += chars;
+            } else {
+                clusters.push((item.font_size, chars));
+            }
+        }
+        clusters
+            .into_iter()
+            .max_by_key(|&(_, count)| count)
+            .map(|(size, _)| size)
+    }
+    dominant(line, |t| t.chars().filter(|c| c.is_alphanumeric()).count())
+        .or_else(|| dominant(line, |t| t.chars().filter(|c| !c.is_whitespace()).count()))
+}
+
 pub(crate) fn compute_heading_tiers(lines: &[TextLine], base_size: f32) -> Vec<f32> {
    let mut heading_sizes: Vec<f32> = Vec::new();

    for line in lines {
-        if let Some(first) = line.items.first() {
-            if first.font_size / base_size >= 1.2 {
-                // Digit-only lines (page numbers, issue numbers) must not
-                // define heading tiers: a large bold folio claims tier 0 and
-                // blocks the bold-size fallback for the document's real
-                // same-size headings.
-                let text = line.text();
-                let t = text.trim();
-                if !t.is_empty() && t.chars().all(|c| !c.is_alphabetic()) {
-                    continue;
-                }
-                heading_sizes.push(first.font_size);
+        // Use the dominant (alphanumeric-weighted) size — the same notion
+        // heading DETECTION matches against tiers — so a large leading
+        // ornament or section marker can't register a tier at a size no
+        // line will ever be classified with.
+        let Some(dominant) = line_dominant_font_size(line) else {
+            continue;
+        };
+        if dominant / base_size >= 1.2 {
+            // Digit-only lines (page numbers, issue numbers) must not
+            // define heading tiers: a large bold folio claims tier 0 and
+            // blocks the bold-size fallback for the document's real
+            // same-size headings.
+            let text = line.text();
+            let t = text.trim();
+            if !t.is_empty() && t.chars().all(|c| !c.is_alphabetic()) {
+                continue;
            }
+            // Near-empty lines (oversized glyph runs — display-math
+            // operators like ∫∫, ⟨⟩ or √ that decode as "ZZ"/"h i"/"p p"
+            // in TeX bitmap fonts, drop caps) must not define tiers
+            // either. Real headings have at least three letters, two of
+            // them distinct.
+            let mut letters: Vec<char> = t
+                .chars()
+                .filter(|c| c.is_alphabetic())
+                .flat_map(|c| c.to_lowercase())
+                .collect();
+            let total_letters = letters.len();
+            letters.sort_unstable();
+            letters.dedup();
+            if total_letters < 3 || letters.len() < 2 {
+                continue;
+            }
+            // A heading line is uniformly sized: every item within ±20% of
+            // the dominant size. Mixed-size lines (a drop cap or oversized
+            // math glyph anywhere in the line — leading OR trailing) are
+            // body content, not headings.
+            let uniform = line
+                .items
+                .iter()
+                .all(|i| (i.font_size - dominant).abs() <= dominant * 0.2);
+            if !uniform {
+                continue;
+            }
+            log::debug!(
+                "heading tier candidate: fs={:.1} page={} {:?}",
+                dominant,
+                line.page,
+                t.chars().take(60).collect::<String>()
+            );
+            heading_sizes.push(dominant);
        }
    }

@@ -458,6 +458,83 @@ fn is_body_size_all_bold_line(line: &TextLine, base_size: f32) -> bool {
            .all(|item| item.is_bold && (item.font_size - first.font_size).abs() < 0.5)
 }

+/// First-line-indent paragraph break. Justified documents (TeX/dvips output)
+/// often separate paragraphs only by indenting the first line — the vertical
+/// gap stays at normal line spacing, so the Y-gap rule never fires and whole
+/// sections merge into one wall of text.
+///
+/// Fires only when all of:
+/// - the line advance is a normal in-paragraph step (`0 < y_gap <= para_threshold`)
+/// - the left margin is established (2+ consecutive lines at the same X), so
+///   list continuations and ragged layouts don't trigger
+/// - this line starts 0.8–4 em deeper than that margin (a first-line indent,
+///   not a column jump)
+/// - the previous line ended short of the established right edge (a paragraph's
+///   last line is ragged; hanging-indent wraps like bibliography continuation
+///   lines follow a full-width line and must not split) — OR ended at a
+///   sentence boundary, which covers paragraphs whose justified last line
+///   happens to run flush (a citation's first line almost never wraps exactly
+///   at a sentence end)
+#[allow(clippy::too_many_arguments)]
+fn is_indent_paragraph_break(
+    line_x: f32,
+    y_gap: f32,
+    para_threshold: f32,
+    base_size: f32,
+    margin_x: f32,
+    margin_run: u32,
+    prev_right: f32,
+    max_right: f32,
+    prev_ends_sentence: bool,
+) -> bool {
+    let indent = line_x - margin_x;
+    margin_run >= 2
+        && y_gap > 0.0
+        && y_gap <= para_threshold
+        && indent >= base_size * 0.8
+        && indent <= base_size * 4.0
+        && (prev_right <= max_right - base_size || prev_ends_sentence)
+}
+
+/// True when a line's text ends at a sentence boundary: terminal punctuation,
+/// optionally followed by a closing quote/bracket.
+fn line_ends_sentence(text: &str) -> bool {
+    let trimmed = text.trim_end();
+    let trimmed = trimmed.trim_end_matches(['"', '\u{201D}', '\u{2019}', ')', ']']);
+    trimmed.ends_with(['.', '!', '?'])
+}
+
+/// True when math/operator symbols rival the alphanumeric content of a line —
+/// the signature of a display equation rather than prose or a heading.
+fn is_symbol_dominated(text: &str) -> bool {
+    let alnum = text.chars().filter(|c| c.is_alphanumeric()).count();
+    // Grouping characters are deliberately excluded: parenthesized short
+    // headings like "(R-12)" or "[DRAFT]" are common and are not math.
+    let mathy = text
+        .chars()
+        .filter(|c| {
+            matches!(
+                c,
+                '+' | '\u{2212}'
+                    | '='
+                    | '/'
+                    | '^'
+                    | '|'
+                    | '<'
+                    | '>'
+                    | '\u{00B1}'
+                    | '\u{00D7}'
+                    | '\u{00B7}'
+                    | '\u{2211}'
+                    | '\u{222B}'
+                    | '\u{221A}'
+                    | '\u{221E}'
+            )
+        })
+        .count();
+    mathy >= 2 && mathy * 2 >= alnum.max(1)
+}
+
 fn is_wrapped_same_style_line(prev: &TextLine, next: &TextLine, para_threshold: f32) -> bool {
    if prev.page != next.page {
        return false;
@@ -777,6 +854,12 @@ pub(super) fn to_markdown_from_lines_with_tables_and_images(
    let mut toc_suppress_page: Option<u32> = None;
    let mut inserted_tables: HashSet<(u32, usize)> = HashSet::new();
    let mut inserted_images: HashSet<(u32, usize)> = HashSet::new();
+    // First-line-indent paragraph detection state
+    let mut left_margin_x = f32::MAX;
+    let mut left_margin_run = 0u32;
+    let mut prev_right = 0.0f32;
+    let mut page_max_right = 0.0f32;
+    let mut prev_ends_sentence = false;

    // Collect all pages that have tables or images (including image-only pages)
    let mut all_content_pages: Vec<u32> = page_tables
@@ -851,6 +934,11 @@ pub(super) fn to_markdown_from_lines_with_tables_and_images(
            prev_y = f32::MAX;
            prev_x = 0.0;
            paragraph_in_wrapped_bold_run = false;
+            left_margin_x = f32::MAX;
+            left_margin_run = 0;
+            prev_right = 0.0;
+            page_max_right = 0.0;
+            prev_ends_sentence = false;

            if options.include_page_numbers {
                output.push_str(&format!("<!-- Page {} -->\n\n", current_page));
@@ -907,7 +995,22 @@ pub(super) fn to_markdown_from_lines_with_tables_and_images(
            && !line_all_bold
            && y_gap > base_size * 1.2
            && y_gap <= para_threshold;
-        if (is_para_break || is_band_switch || is_bold_to_regular_break) && in_paragraph {
+        let is_indent_break = in_paragraph
+            && !in_list
+            && is_indent_paragraph_break(
+                line_x,
+                y_gap,
+                para_threshold,
+                base_size,
+                left_margin_x,
+                left_margin_run,
+                prev_right,
+                page_max_right,
+                prev_ends_sentence,
+            );
+        if (is_para_break || is_band_switch || is_bold_to_regular_break || is_indent_break)
+            && in_paragraph
+        {
            output.push_str("\n\n");
            in_paragraph = false;
            paragraph_in_wrapped_bold_run = false;
@@ -916,6 +1019,21 @@ pub(super) fn to_markdown_from_lines_with_tables_and_images(
        // Let the continuation check below decide if we're still in a list
        prev_y = line.y;
        prev_x = line_x;
+        // Track the stable left margin and line right edges for
+        // first-line-indent paragraph detection
+        if (line_x - left_margin_x).abs() <= 2.0 {
+            left_margin_run += 1;
+        } else {
+            left_margin_x = line_x;
+            left_margin_run = 1;
+        }
+        prev_right = line
+            .items
+            .iter()
+            .map(|i| i.x + i.width)
+            .fold(0.0f32, f32::max);
+        page_max_right = page_max_right.max(prev_right);
+        prev_ends_sentence = line_ends_sentence(&line.text());

        // Get text with optional bold/italic formatting
        let text = line.text_with_formatting(
@@ -1007,8 +1125,15 @@ pub(super) fn to_markdown_from_lines_with_tables_and_images(
            && !is_toc_entry_line(plain_trimmed)
            && !is_heading_fragment(plain_trimmed)
            && toc_suppress_page != Some(line.page)
+            // Display equations must never become headings — through the
+            // font-tier path or the rarity fallback. They are standalone and
+            // isolated by construction, so without this gate they score
+            // straight past both.
+            && !plain_trimmed.contains('=')
+            && !is_symbol_dominated(plain_trimmed)
        {
-            let line_font_size = line.items.first().map(|i| i.font_size).unwrap_or(base_size);
+            let line_font_size =
+                crate::markdown::analysis::line_dominant_font_size(line).unwrap_or(base_size);
            detect_header_level(
                line_font_size,
                base_size,
@@ -1289,6 +1414,12 @@ pub fn to_markdown_from_lines(lines: Vec<TextLine>, options: MarkdownOptions) ->
    let mut prev_had_dot_leaders = false;
    let mut paragraph_in_wrapped_bold_run = false;
    let mut toc_suppress_page: Option<u32> = None;
+    // First-line-indent paragraph detection state
+    let mut left_margin_x = f32::MAX;
+    let mut left_margin_run = 0u32;
+    let mut prev_right = 0.0f32;
+    let mut page_max_right = 0.0f32;
+    let mut prev_ends_sentence = false;

    for (line_idx, line) in lines.iter().enumerate() {
        // Page break
@@ -1306,6 +1437,11 @@ pub fn to_markdown_from_lines(lines: Vec<TextLine>, options: MarkdownOptions) ->
            last_list_x = None;
            prev_had_dot_leaders = false;
            paragraph_in_wrapped_bold_run = false;
+            left_margin_x = f32::MAX;
+            left_margin_run = 0;
+            prev_right = 0.0;
+            page_max_right = 0.0;
+            prev_ends_sentence = false;

            if options.include_page_numbers {
                output.push_str(&format!("<!-- Page {} -->\n\n", current_page));
@@ -1315,6 +1451,7 @@ pub fn to_markdown_from_lines(lines: Vec<TextLine>, options: MarkdownOptions) ->
        // Paragraph break: large forward Y gap (normal) or large backward jump
        // (newspaper columns emitted sequentially on the same page).
        let y_gap = prev_y - line.y;
+        let line_x = line.items.first().map(|i| i.x).unwrap_or(0.0);
        let is_para_break = y_gap.abs() > para_threshold;
        let line_all_bold = !line.items.is_empty() && line.items.iter().all(|item| item.is_bold);
        let line_in_wrapped_bold_run = wrapped_bold_paragraph_lines.contains(&line_idx);
@@ -1324,7 +1461,20 @@ pub fn to_markdown_from_lines(lines: Vec<TextLine>, options: MarkdownOptions) ->
            && !line_all_bold
            && y_gap > base_size * 1.2
            && y_gap <= para_threshold;
-        if (is_para_break || is_bold_to_regular_break) && in_paragraph {
+        let is_indent_break = in_paragraph
+            && !in_list
+            && is_indent_paragraph_break(
+                line_x,
+                y_gap,
+                para_threshold,
+                base_size,
+                left_margin_x,
+                left_margin_run,
+                prev_right,
+                page_max_right,
+                prev_ends_sentence,
+            );
+        if (is_para_break || is_bold_to_regular_break || is_indent_break) && in_paragraph {
            output.push_str("\n\n");
            in_paragraph = false;
            paragraph_in_wrapped_bold_run = false;
@@ -1332,6 +1482,21 @@ pub fn to_markdown_from_lines(lines: Vec<TextLine>, options: MarkdownOptions) ->
        // Don't immediately end list on paragraph break
        // Let the continuation check below decide if we're still in a list
        prev_y = line.y;
+        // Track the stable left margin and line right edges for
+        // first-line-indent paragraph detection
+        if (line_x - left_margin_x).abs() <= 2.0 {
+            left_margin_run += 1;
+        } else {
+            left_margin_x = line_x;
+            left_margin_run = 1;
+        }
+        prev_right = line
+            .items
+            .iter()
+            .map(|i| i.x + i.width)
+            .fold(0.0f32, f32::max);
+        page_max_right = page_max_right.max(prev_right);
+        prev_ends_sentence = line_ends_sentence(&line.text());

        // Get text with optional bold/italic formatting
        let text = line.text_with_formatting(
@@ -1371,8 +1536,12 @@ pub fn to_markdown_from_lines(lines: Vec<TextLine>, options: MarkdownOptions) ->
            && !is_heading_fragment(plain_trimmed)
            && toc_suppress_page != Some(line.page)
            && !(options.detect_code && line.items.iter().any(|i| is_monospace_font(&i.font)))
+            // Display equations must never become headings (see main path).
+            && !plain_trimmed.contains('=')
+            && !is_symbol_dominated(plain_trimmed)
        {
-            let line_font_size = line.items.first().map(|i| i.font_size).unwrap_or(base_size);
+            let line_font_size =
+                crate::markdown::analysis::line_dominant_font_size(line).unwrap_or(base_size);
            if let Some(header_level) = detect_header_level(
                line_font_size,
                base_size,
@@ -127,9 +127,7 @@ fn visual_style(line: &TextLine) -> Option<VisualStyle> {
    })
 }

-/// Shared with `analysis::starts_with_numbering_prefix` so the veto
-/// exemption and the heading parser agree on what a roman numeral is.
-pub(super) fn roman_value(token: &str) -> Option<u32> {
+fn roman_value(token: &str) -> Option<u32> {
    if token.is_empty() || token.len() > 8 {
        return None;
    }
@@ -975,54 +975,18 @@ pub fn to_markdown_from_items_with_rects(
    items: Vec<TextItem>,
    options: MarkdownOptions,
    rects: &[crate::types::PdfRect],
-) -> String {
-    let document_page_count = items.iter().map(|item| item.page).max().unwrap_or(0);
-    to_markdown_from_items_with_rects_and_page_count(items, options, rects, document_page_count)
-}
-
-/// Convert positioned text items to Markdown with an authoritative PDF page count.
-///
-/// Use this overload when the owning PDF is available so trailing blank or
-/// unextracted pages are included in document-level header and folio coverage.
-/// Item-only callers can continue using [`to_markdown_from_items_with_rects`],
-/// which falls back to the highest observed item page.
-pub fn to_markdown_from_items_with_rects_and_page_count(
-    items: Vec<TextItem>,
-    options: MarkdownOptions,
-    rects: &[crate::types::PdfRect],
-    document_page_count: u32,
 ) -> String {
    to_markdown_from_items_with_rects_and_lines(
        items,
        options,
        rects,
        &[],
-        MarkdownDocumentContext {
-            page_thresholds: &HashMap::new(),
-            struct_roles: None,
-            struct_tables: &[],
-            page_count: document_page_count,
-            prefiltered_page_number_pages: None,
-            prefiltered_page_number_mask: None,
-        },
+        &HashMap::new(),
+        None,
+        &[],
    )
 }

-pub(crate) struct MarkdownDocumentContext<'a> {
-    pub(crate) page_thresholds: &'a HashMap<u32, f32>,
-    pub(crate) struct_roles:
-        Option<&'a HashMap<u32, HashMap<i64, crate::structure_tree::StructRole>>>,
-    pub(crate) struct_tables: &'a [crate::structure_tree::StructTable],
-    pub(crate) page_count: u32,
-    /// Pages where an upstream document-level pass removed folios. This keeps
-    /// table-continuation classification consistent after masked items drop.
-    pub(crate) prefiltered_page_number_pages: Option<&'a HashSet<u32>>,
-    /// Document-level removal decisions aligned with this call's input items.
-    /// Table detection consumes the original items; the mask is applied only
-    /// after table claims have been established.
-    pub(crate) prefiltered_page_number_mask: Option<&'a [bool]>,
-}
-
 /// Convert positioned text items to markdown, using rectangles and line segments for table detection.
 ///
 /// Line-based detection runs first (strongest structural evidence), then rect-based,
@@ -1032,43 +996,27 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
    options: MarkdownOptions,
    rects: &[crate::types::PdfRect],
    pdf_lines: &[crate::types::PdfLine],
-    context: MarkdownDocumentContext<'_>,
+    page_thresholds: &HashMap<u32, f32>,
+    struct_roles: Option<&HashMap<u32, HashMap<i64, crate::structure_tree::StructRole>>>,
+    struct_tables: &[crate::structure_tree::StructTable],
 ) -> String {
    use crate::tables::{
-        content_width, detect_tables_from_lines, detect_tables_from_rects,
-        detect_tables_from_struct_tree, detect_tables_with_page_width, try_build_rect_guided_table,
+        detect_tables, detect_tables_from_lines, detect_tables_from_rects,
+        detect_tables_from_struct_tree, try_build_rect_guided_table,
    };
    use crate::types::ItemType;

-    let MarkdownDocumentContext {
-        page_thresholds,
-        struct_roles,
-        struct_tables,
-        page_count: document_page_count,
-        prefiltered_page_number_pages,
-        prefiltered_page_number_mask,
-    } = context;
-
    if items.is_empty() {
        return String::new();
    }

-    // Table detection must retain the original collection because short
-    // numeric table cells can be indistinguishable from folios until
-    // structural context is available. A precomputed mask carries the
-    // document-wide decision without removing items before table claims.
-    debug_assert!(prefiltered_page_number_mask.is_none_or(|mask| mask.len() == items.len()));
-    let has_precomputed_page_number_mask = prefiltered_page_number_mask.is_some();
-    let removed_page_number_pages = prefiltered_page_number_pages.cloned().unwrap_or_default();
-
    // Separate images and links from text items
    let mut images: Vec<TextItem> = Vec::new();
    let mut page_image_regions: HashMap<u32, Vec<(f32, f32, f32, f32)>> = HashMap::new();
    let mut links: Vec<TextItem> = Vec::new();
    let mut text_items: Vec<TextItem> = Vec::new();
-    let mut text_item_page_number_mask: Vec<bool> = Vec::new();

-    for (input_index, item) in items.into_iter().enumerate() {
+    for item in items {
        match &item.item_type {
            ItemType::Image => {
                page_image_regions.entry(item.page).or_default().push((
@@ -1087,12 +1035,6 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
                }
            }
            ItemType::Text | ItemType::FormField => {
-                text_item_page_number_mask.push(
-                    prefiltered_page_number_mask
-                        .and_then(|mask| mask.get(input_index))
-                        .copied()
-                        .unwrap_or(false),
-                );
                text_items.push(item);
            }
        }
@@ -1133,6 +1075,7 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(

    let mut pages: Vec<u32> = page_groups.keys().copied().collect();
    pages.sort();
+    let page_count = pages.last().copied().unwrap_or(0) + 1;

    // Track band splits per page so we can split non-table items later
    let mut page_band_splits: HashMap<u32, Vec<(f32, f32)>> = HashMap::new();
@@ -1144,7 +1087,6 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
    for page in pages {
        let group = page_groups.get(&page).unwrap();
        let page_items: Vec<TextItem> = group.iter().map(|(_, item)| (*item).clone()).collect();
-        let page_content_width = content_width(&page_items);

        // Chart-bar regions: bar charts drawn as filled rects read as cell
        // rects or aligned text and get gridded into phantom tables. Their
@@ -1186,10 +1128,7 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
        });
        let chart_prose_columns = chart_prose_split.is_some();

-        // Check for side-by-side table layout using the original items. Sparse
-        // numeric cells need table context before they can be distinguished
-        // safely from folios; cleaned evidence is reserved for column and
-        // final non-table layout decisions.
+        // Check for side-by-side layout (e.g. two tables placed left and right)
        let mut bands = split_side_by_side(&page_items);
        // A rect table crossing a proposed split boundary means the "gutter"
        // is really the gap between ruled and borderless table columns —
@@ -1429,12 +1368,7 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
                    // table can share the prose anchors. Reject only candidates
                    // whose cells prove they are parallel prose fragments.
                    let reject_parallel_prose = chart_prose_columns && !was_split;
-                    let tables = detect_tables_with_page_width(
-                        subset_items,
-                        base_size,
-                        false,
-                        page_content_width,
-                    );
+                    let tables = detect_tables(subset_items, base_size, false);
                    for table in tables {
                        if reject_parallel_prose && is_parallel_prose_table(&table) {
                            log::debug!(
@@ -1615,12 +1549,7 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
            // and reject chart-page prose candidates individually below.
            let skip_body_font =
                merged_retry_skips_body_font(detected_columns, !chart_regions.is_empty());
-            let heuristic_tables = detect_tables_with_page_width(
-                &chart_free,
-                base_size,
-                skip_body_font,
-                page_content_width,
-            );
+            let heuristic_tables = detect_tables(&chart_free, base_size, skip_body_font);
            for table in &heuristic_tables {
                if !chart_regions.is_empty() && is_parallel_prose_table(table) {
                    log::debug!(
@@ -1705,20 +1634,16 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
    };

    // Filter out table items and process the rest
-    let non_table_items: Vec<(usize, TextItem)> = text_items
+    let non_table_items: Vec<TextItem> = text_items
        .into_iter()
        .enumerate()
        .filter(|(idx, _)| !table_items.contains(idx))
+        .map(|(_, item)| item)
        .collect();

    // Find pages that are table-only (no remaining non-table text)
    let table_only_pages: HashSet<u32> = {
-        let mut pages_with_text: HashSet<u32> =
-            non_table_items.iter().map(|(_, item)| item.page).collect();
-        // Preserve the pre-filter continuation classification: a page that
-        // originally also contained a folio does not become table-only merely
-        // because an upstream document-level pass removed it.
-        pages_with_text.extend(removed_page_number_pages);
+        let pages_with_text: HashSet<u32> = non_table_items.iter().map(|i| i.page).collect();
        page_tables
            .keys()
            .filter(|p| !pages_with_text.contains(p))
@@ -1733,25 +1658,11 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
    // column detection on pages where table column gaps would be misidentified.
    let table_page_set: HashSet<u32> = page_tables.keys().copied().collect();

-    let non_table_items = if has_precomputed_page_number_mask {
-        non_table_items
-            .into_iter()
-            .filter(|(index, _)| !text_item_page_number_mask[*index])
-            .map(|(_, item)| item)
-            .collect()
-    } else {
-        crate::extractor::filter_markdown_page_numbers_with_removed_pages(
-            non_table_items.into_iter().map(|(_, item)| item).collect(),
-            document_page_count,
-        )
-        .0
-    };
-
    // Split non-table items by band boundaries before line grouping so that
    // items from different side-by-side zones (e.g. left/right month columns
    // in a calendar) don't merge into the same line.
    let lines = if page_band_splits.is_empty() && page_chart_prose_splits.is_empty() {
-        crate::extractor::group_prefiltered_items_into_lines_with_thresholds_and_regions(
+        crate::extractor::group_into_lines_with_thresholds_and_regions(
            non_table_items,
            page_thresholds,
            &table_page_set,
@@ -1779,14 +1690,13 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
            }
        }
        // Process unsplit pages normally
-        let mut all_lines =
-            crate::extractor::group_prefiltered_items_into_lines_with_thresholds_and_regions(
-                unsplit_items,
-                page_thresholds,
-                &table_page_set,
-                &page_chart_map,
-                &page_image_regions,
-            );
+        let mut all_lines = crate::extractor::group_into_lines_with_thresholds_and_regions(
+            unsplit_items,
+            page_thresholds,
+            &table_page_set,
+            &page_chart_map,
+            &page_image_regions,
+        );
        // Process each split page's bands independently, then interleave
        // by Y position so paired zones (e.g. left/right months) appear together.
        let mut split_pages: Vec<u32> = split_page_items.keys().copied().collect();
@@ -1804,7 +1714,7 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
                    .collect();
                if !band_items.is_empty() {
                    page_lines.extend(
-                        crate::extractor::group_prefiltered_items_into_lines_with_thresholds_and_charts(
+                        crate::extractor::group_into_lines_with_thresholds_and_charts(
                            band_items,
                            page_thresholds,
                            &table_page_set,
@@ -1838,7 +1748,7 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
                        .collect();
                    if !column_items.is_empty() {
                        zone_lines.extend(
-                            crate::extractor::group_prefiltered_items_into_lines_with_thresholds_and_charts(
+                            crate::extractor::group_into_lines_with_thresholds_and_charts(
                                column_items,
                                page_thresholds,
                                &table_page_set,
@@ -1879,7 +1789,7 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(
                        item.y >= low || item_is_in_chart_region(item, chart_regions)
                    });
                all_lines.extend(
-                    crate::extractor::group_prefiltered_items_into_lines_with_thresholds_and_charts(
+                    crate::extractor::group_into_lines_with_thresholds_and_charts(
                        chart_zone,
                        page_thresholds,
                        &table_page_set,
@@ -1898,7 +1808,7 @@ pub(crate) fn to_markdown_from_items_with_rects_and_lines(

    // Strip repeated headers/footers before conversion
    let lines = if options.strip_headers_footers {
-        preprocess::strip_repeated_lines(lines, document_page_count)
+        preprocess::strip_repeated_lines(lines, page_count)
    } else {
        lines
    };
@@ -2011,53 +1921,6 @@ mod tests {
        it
    }

-    #[test]
-    fn precomputed_folio_mask_preserves_numeric_table_cells() {
-        let mut items = Vec::new();
-        let mut rects = Vec::new();
-        for row in 0..4 {
-            for column in 0..2 {
-                let mut item = make_item_w(
-                    110.0 + column as f32 * 100.0,
-                    30.0 + row as f32 * 20.0,
-                    20.0,
-                    1,
-                );
-                item.text = (row * 2 + column + 1).to_string();
-                items.push(item);
-                rects.push(PdfRect {
-                    x: 100.0 + column as f32 * 100.0,
-                    y: 20.0 + row as f32 * 20.0,
-                    width: 100.0,
-                    height: 20.0,
-                    page: 1,
-                });
-            }
-        }
-
-        // Simulate document-level folio decisions that would remove every
-        // short numeric item if applied before structural table detection.
-        let removal_mask = vec![true; items.len()];
-        let removed_pages = HashSet::from([1]);
-        let markdown = to_markdown_from_items_with_rects_and_lines(
-            items,
-            MarkdownOptions::default(),
-            &rects,
-            &[],
-            MarkdownDocumentContext {
-                page_thresholds: &HashMap::new(),
-                struct_roles: None,
-                struct_tables: &[],
-                page_count: 1,
-                prefiltered_page_number_pages: Some(&removed_pages),
-                prefiltered_page_number_mask: Some(&removal_mask),
-            },
-        );
-
-        assert!(markdown.contains("|1|2|"), "{markdown}");
-        assert!(markdown.contains("|7|8|"), "{markdown}");
-    }
-
    #[test]
    fn early_layout_excludes_chart_items_before_column_detection() {
        let mut items = Vec::new();
@@ -3,7 +3,6 @@
 use regex::Regex;

 use super::{MarkdownOptions, MarkdownProfile};
-use crate::text_utils::is_page_number_line;

 /// Clean up markdown output with post-processing
 pub(crate) fn clean_markdown(mut text: String, options: &MarkdownOptions) -> String {
@@ -146,7 +145,7 @@ fn fix_hyphenation(text: &str) -> String {
    result
 }

-/// Remove isolated page-number expressions from Markdown.
+/// Remove standalone page numbers (lines that are just 1-4 digit numbers)
 fn remove_page_numbers(text: &str) -> String {
    let mut result = Vec::new();
    let lines: Vec<&str> = text.lines().collect();
@@ -184,6 +183,69 @@ fn remove_page_numbers(text: &str) -> String {
    result.join("\n")
 }

+/// Check if a line looks like a page number
+fn is_page_number_line(trimmed: &str) -> bool {
+    // Empty lines are not page numbers
+    if trimmed.is_empty() {
+        return false;
+    }
+
+    // Pattern 1: Just a number (1-4 digits)
+    if trimmed.len() <= 4 && trimmed.chars().all(|c| c.is_ascii_digit()) {
+        return true;
+    }
+
+    // Pattern 2: "Page X of Y" or "Page X" or "Page   of" (placeholder)
+    let lower = trimmed.to_lowercase();
+    if let Some(rest) = lower.strip_prefix("page") {
+        let rest = rest.trim();
+        // "Page   of" (empty page numbers)
+        if rest == "of" || rest.starts_with("of ") {
+            return true;
+        }
+        // "Page X" or "Page X of Y"
+        if rest
+            .chars()
+            .next()
+            .map(|c| c.is_ascii_digit())
+            .unwrap_or(false)
+        {
+            return true;
+        }
+        // Just "Page" followed by whitespace and maybe "of"
+        if rest.is_empty()
+            || rest
+                .split_whitespace()
+                .all(|w| w == "of" || w.chars().all(|c| c.is_ascii_digit()))
+        {
+            return true;
+        }
+    }
+
+    // Pattern 3: "X of Y" where X and Y are numbers
+    if let Some(of_idx) = trimmed.find(" of ") {
+        let before = trimmed[..of_idx].trim();
+        let after = trimmed[of_idx + 4..].trim();
+        if before.chars().all(|c| c.is_ascii_digit())
+            && after.chars().all(|c| c.is_ascii_digit())
+            && !before.is_empty()
+            && !after.is_empty()
+        {
+            return true;
+        }
+    }
+
+    // Pattern 4: "- X -" centered page number
+    if trimmed.len() >= 3 && trimmed.starts_with('-') && trimmed.ends_with('-') {
+        let inner = trimmed[1..trimmed.len() - 1].trim();
+        if inner.chars().all(|c| c.is_ascii_digit()) && !inner.is_empty() {
+            return true;
+        }
+    }
+
+    false
+}
+
 /// Convert URLs to markdown links
 fn format_urls(text: &str) -> String {
    use once_cell::sync::Lazy;
@@ -427,14 +489,12 @@ mod tests {
    fn test_is_page_number_page_x() {
        assert!(is_page_number_line("Page 5"));
        assert!(is_page_number_line("page 12"));
-        assert!(is_page_number_line("Page123"));
    }

    #[test]
    fn test_is_page_number_page_x_of_y() {
        assert!(is_page_number_line("Page 3 of 10"));
        assert!(is_page_number_line("page 1 of 5"));
-        assert!(is_page_number_line("Page 3 of 10 Report header"));
    }

    #[test]
@@ -466,12 +526,6 @@ mod tests {
        assert!(!is_page_number_line("Total: 500"));
    }

-    #[test]
-    fn test_is_page_number_labeled_running_header() {
-        assert!(is_page_number_line("Page 42 Chapter 5"));
-        assert!(is_page_number_line("Page 42 explains the result"));
-    }
-
    // --- remove_page_numbers ---

    #[test]
@@ -497,16 +551,6 @@ mod tests {
        assert!(result.contains("42"));
    }

-    #[test]
-    fn test_remove_page_numbers_labeled_header_with_content() {
-        let input = "Content\n\nPage 42 explains the result\n---\nEnd";
-        let result = remove_page_numbers(input);
-
-        assert!(!result.contains("Page 42 explains the result"));
-        assert!(result.contains("Content"));
-        assert!(result.contains("End"));
-    }
-
    #[test]
    fn test_remove_page_numbers_multiple_patterns() {
        let input = "\n1\n\nContent\n\n2\n\n---\nMore\n\n3\n";
@@ -40,8 +40,9 @@ fn effective_heading_level(
        }
    }

-    // Fall back to font-size heuristic
-    let font = line.items.first().map(|i| i.font_size).unwrap_or(base_size);
+    // Fall back to font-size heuristic (dominant size, so a leading drop cap
+    // or oversized math delimiter doesn't reclassify a body line)
+    let font = crate::markdown::analysis::line_dominant_font_size(line).unwrap_or(base_size);
    detect_header_level(
        font,
        base_size,
@@ -72,7 +73,10 @@ pub(crate) fn merge_heading_lines(

    for line in lines {
        let line_level = effective_heading_level(&line, base_size, heading_tiers, struct_roles);
-        let line_font = line.items.first().map(|i| i.font_size).unwrap_or(base_size);
+        // Dominant size, consistent with effective_heading_level: a large
+        // leading delimiter must not inflate the merge gap threshold.
+        let line_font =
+            crate::markdown::analysis::line_dominant_font_size(&line).unwrap_or(base_size);

        // Check if the previous line is a heading at the same level on the same page
        let should_merge = if let (Some(prev), Some(curr_level)) = (result.last(), line_level) {
@@ -169,6 +173,87 @@ pub(crate) fn merge_drop_caps(lines: Vec<TextLine>, base_size: f32) -> Vec<TextL
                .map(|c| c.is_uppercase())
                .unwrap_or(false);

+        // Embedded drop cap: a two-line drop cap's baseline aligns with the
+        // paragraph's SECOND line, so Y-grouping puts the glyph at the start
+        // of that line instead of on its own line. Left in place, the cap
+        // ends up mid-sentence after paragraph joining ("...which exchange T
+        // bandwidth..."). Detect it, prepend the char to the previous line
+        // (the paragraph's first line, typeset beside the cap), and drop it.
+        // The size gate is 1.8x (not 2.5x): bitmap (Type3) caps measure their
+        // glyph bbox, not the em box, so a two-line cap can be as small as
+        // ~1.9x the body size.
+        if line.items.len() > 1 {
+            let first = &line.items[0];
+            // The rest of the line must be a substantive body-text run —
+            // a lone label or math fragment beside a big glyph is not a
+            // drop-cap paragraph and must not be rewritten.
+            let rest_letters: usize = line.items[1..]
+                .iter()
+                .map(|i| i.text.chars().filter(|c| c.is_alphabetic()).count())
+                .sum();
+            let is_embedded_cap = first.font_size >= base_size * 1.8
+                && first.text.trim().chars().count() == 1
+                && first
+                    .text
+                    .trim()
+                    .chars()
+                    .next()
+                    .is_some_and(|c| c.is_uppercase())
+                && line.items[1..]
+                    .iter()
+                    .all(|i| i.font_size < base_size * 1.5)
+                && line.items[1..].iter().any(|i| i.x > first.x)
+                && rest_letters >= 8;
+            if is_embedded_cap {
+                let drop_char = first.text.trim().chars().next().unwrap();
+                let cap_x = first.x;
+                let cap_span = first.font_size * 1.5;
+                let line_y = line.y;
+                // Only the IMMEDIATELY preceding line qualifies: the
+                // paragraph's first line sits directly above, in the same
+                // column, indented past the cap glyph by roughly the cap's
+                // width. It must also read as a paragraph first line —
+                // substantive body text starting with a letter — so
+                // headings, labels, or table fragments that merely fall in
+                // the geometric window are never rewritten.
+                let target = result.last_mut().filter(|prev| {
+                    let prev_text = prev.text();
+                    let prev_trimmed = prev_text.trim();
+                    prev.page == line.page
+                        && prev.y > line_y
+                        && prev.y - line_y <= cap_span
+                        && prev
+                            .items
+                            .first()
+                            .is_some_and(|i| i.x > cap_x && i.x - cap_x <= first.font_size * 2.5)
+                        && prev_trimmed
+                            .chars()
+                            .next()
+                            .is_some_and(|c| c.is_alphabetic())
+                        && prev_trimmed.chars().filter(|c| c.is_alphabetic()).count() >= 8
+                });
+                if let Some(prev_line) = target {
+                    if let Some(first_item) = prev_line.items.first_mut() {
+                        // A mid-word cap ("T" + "HE recent…") joins directly.
+                        // Leading whitespace on the remainder means the cap
+                        // is a standalone word ("A" + " long time ago…") —
+                        // preserve exactly one separating space.
+                        let had_leading_ws = first_item.text.starts_with(char::is_whitespace);
+                        let rest = first_item.text.trim_start().to_string();
+                        first_item.text = if had_leading_ws {
+                            format!("{} {}", drop_char, rest)
+                        } else {
+                            format!("{}{}", drop_char, rest)
+                        };
+                    }
+                    let mut line = line.clone();
+                    line.items.remove(0);
+                    result.push(line);
+                    continue;
+                }
+            }
+        }
+
        if is_drop_cap {
            let drop_char = trimmed.chars().next().unwrap();

@@ -598,6 +683,51 @@ mod tests {
        }
    }

+    fn make_item_at(text: &str, font_size: f32, x: f32) -> TextItem {
+        let mut item = make_item(text, font_size, None);
+        item.x = x;
+        item.width = text.len() as f32 * font_size * 0.5;
+        item
+    }
+
+    #[test]
+    fn test_embedded_drop_cap_moves_to_paragraph_start() {
+        // Two-line drop cap: the big "T" baseline-aligns with the paragraph's
+        // second line, so it lands as that line's first item (Shannon
+        // entropy.pdf page 1 pattern).
+        let first_line = TextLine {
+            items: vec![make_item_at(
+                "HE recent development which exchange",
+                10.0,
+                90.0,
+            )],
+            y: 712.0,
+            page: 1,
+            adaptive_threshold: 0.10,
+        };
+        let second_line = TextLine {
+            items: vec![
+                make_item_at("T", 25.0, 72.0),
+                make_item_at("bandwidth for signal-to-noise ratio", 10.0, 90.0),
+            ],
+            y: 700.0,
+            page: 1,
+            adaptive_threshold: 0.10,
+        };
+        let result = merge_drop_caps(vec![first_line, second_line], 10.0);
+        assert_eq!(result.len(), 2);
+        assert!(
+            result[0].text().starts_with("THE recent"),
+            "cap should prepend to paragraph start: {}",
+            result[0].text()
+        );
+        assert!(
+            result[1].text().starts_with("bandwidth"),
+            "cap item should be removed from second line: {}",
+            result[1].text()
+        );
+    }
+
    #[test]
    fn test_merge_struct_tree_headings() {
        // Two consecutive lines tagged as H2 via struct tree, same font size as body
@@ -436,8 +436,7 @@ pub(crate) fn recover_header_row(
        .iter()
        .enumerate()
        .filter(|(_, item)| {
-            !item.is_strikeout
-                && item.font_size > small_font_threshold
+            item.font_size > small_font_threshold
                && item.y > first_row_y
                && item.y <= first_row_y + row_gap_limit
        })
@@ -805,31 +804,6 @@ mod tests {
        assert_eq!(table.rows.len(), rows_before);
    }

-    #[test]
-    fn test_recover_header_row_skips_strikeout_candidates() {
-        let mut old_col1 = make_item("Old Col1", 100.0, 520.0, 12.0);
-        old_col1.is_strikeout = true;
-        let mut old_col2 = make_item("Old Col2", 200.0, 520.0, 12.0);
-        old_col2.is_strikeout = true;
-        let all_items = vec![
-            old_col1,
-            old_col2,
-            make_item("A", 100.0, 500.0, 8.0),
-            make_item("B", 200.0, 500.0, 8.0),
-        ];
-        let mut table = Table {
-            columns: vec![100.0, 200.0],
-            rows: vec![500.0, 480.0],
-            cells: vec![vec!["A".into(), "B".into()], vec!["C".into(), "D".into()]],
-            item_indices: vec![2, 3],
-            kind: TableKind::Data,
-        };
-
-        recover_header_row(&mut table, &all_items, 9.0);
-        assert_eq!(table.rows.len(), 2);
-        assert_eq!(table.cells[0], vec!["A", "B"]);
-    }
-
    #[test]
    fn test_recover_header_row_too_far_above() {
        let all_items = vec![
@@ -12,9 +12,7 @@ mod grid;
 pub mod structured;

 pub use detect_heuristic::detect_tables;
-pub(crate) use detect_heuristic::{
-    content_width, detect_tables_with_page_width, is_table_of_contents,
-};
+pub(crate) use detect_heuristic::is_table_of_contents;
 pub use detect_lines::detect_tables_from_lines;
 pub(crate) use detect_lines::detect_vector_grid_tables_from_lines;
 pub(crate) use detect_rects::cluster_rects;
@@ -7,76 +7,6 @@
 use crate::types::TextItem;
 use unicode_normalization::UnicodeNormalization;

-/// Return whether text is an explicit page-number expression.
-///
-/// This strict form is suitable before layout, where removing one numeric item
-/// from substantive text such as `Page 42 explains the result` would lose data.
-pub(crate) fn is_explicit_page_number_expression(text: &str) -> bool {
-    let trimmed = text.trim();
-    if trimmed.is_empty() {
-        return false;
-    }
-
-    let is_number = |value: &str| {
-        !value.is_empty() && value.chars().all(|character| character.is_ascii_digit())
-    };
-
-    if trimmed.len() <= 4 && is_number(trimmed) {
-        return true;
-    }
-
-    if trimmed.len() >= 3 && trimmed.starts_with('-') && trimmed.ends_with('-') {
-        let inner = trimmed[1..trimmed.len() - 1].trim();
-        if is_number(inner) {
-            return true;
-        }
-    }
-
-    let lowercase = trimmed.to_ascii_lowercase();
-    if let Some(rest) = lowercase.strip_prefix("page") {
-        let words: Vec<&str> = rest.split_whitespace().collect();
-        if words.len() >= 3 && is_number(words[0]) && words[1] == "of" && is_number(words[2]) {
-            return true;
-        }
-        if words.len() >= 2 && words[0] == "of" && is_number(words[1]) {
-            return true;
-        }
-        return match words.as_slice() {
-            [] | ["of"] => true,
-            [number] => is_number(number),
-            ["of", total] => is_number(total),
-            [number, "of", total] => is_number(number) && is_number(total),
-            _ => false,
-        };
-    }
-
-    let words: Vec<&str> = lowercase.split_whitespace().collect();
-    match words.as_slice() {
-        [number, "of", total] => is_number(number) && is_number(total),
-        _ => false,
-    }
-}
-
-/// Return whether a completed Markdown line looks like a page number or a
-/// labeled running header.
-///
-/// At this stage the complete line and surrounding breaks are available, so a
-/// leading `Page N` remains compatible with the existing header cleanup even
-/// when the PDF appends a chapter or document title.
-pub(crate) fn is_page_number_line(text: &str) -> bool {
-    if is_explicit_page_number_expression(text) {
-        return true;
-    }
-
-    let lowercase = text.trim().to_ascii_lowercase();
-    lowercase.strip_prefix("page").is_some_and(|rest| {
-        rest.trim_start()
-            .chars()
-            .next()
-            .is_some_and(|character| character.is_ascii_digit())
-    })
-}
-
 /// Check if a character is CJK (Chinese, Japanese, Korean).
 /// CJK languages don't use spaces between words, so word-boundary
 /// heuristics should not apply when CJK characters are involved.
@@ -148,17 +148,14 @@ impl TextLine {
        self.text_with_formatting(false, false, false)
    }

-    /// Get text with optional bold/italic/decorative markdown formatting.
-    ///
-    /// `format_decorations` enables both geometrically detected source
-    /// decorations: underline (`<u>`) and strikeout (`<s>`).
+    /// Get text with optional bold/italic/underline markdown formatting
    pub fn text_with_formatting(
        &self,
        format_bold: bool,
        format_italic: bool,
-        format_decorations: bool,
+        format_underline: bool,
    ) -> String {
-        if !format_bold && !format_italic && !format_decorations {
+        if !format_bold && !format_italic && !format_underline {
            return self.text_plain();
        }

@@ -168,7 +165,6 @@ impl TextLine {
        let mut current_bold = false;
        let mut current_italic = false;
        let mut current_underline = false;
-        let mut current_strikeout = false;

        for (i, item) in self.items.iter().enumerate() {
            let text = item.text.as_str();
@@ -194,16 +190,13 @@ impl TextLine {
            // we push text_trimmed below (which strips it).
            let has_leading_space = text.starts_with(' ');

-            // Check for style changes. Source decorations are exclusive:
-            // `<u>`/`<s>` content stays free of `**`/`*` markers — consumers
-            // (and the eval harnesses this feeds) match tag content literally,
-            // and mixed nesting breaks that. A struck-and-underlined item is
-            // emitted as struck text because deletion is the stronger semantic
-            // distinction in redline documents.
-            let item_strikeout = format_decorations && item.is_strikeout;
-            let item_underline = format_decorations && item.is_underline && !item_strikeout;
-            let item_bold = format_bold && item.is_bold && !item_underline && !item_strikeout;
-            let item_italic = format_italic && item.is_italic && !item_underline && !item_strikeout;
+            // Check for style changes. Underline is exclusive: `<u>` content
+            // stays free of `**`/`*` markers — consumers (and the eval
+            // harnesses this feeds) match the tag content literally, and
+            // mixed `<u>**x**</u>` nesting breaks that.
+            let item_underline = format_underline && item.is_underline;
+            let item_bold = format_bold && item.is_bold && !item_underline;
+            let item_italic = format_italic && item.is_italic && !item_underline;

            // Close previous styles if they change
            if current_italic && !item_italic {
@@ -218,10 +211,6 @@ impl TextLine {
                result.push_str("</u>");
                current_underline = false;
            }
-            if current_strikeout && !item_strikeout {
-                result.push_str("</s>");
-                current_strikeout = false;
-            }

            // Add space: either from spacing logic or preserved from item text
            if needs_space || (has_leading_space && !result.is_empty() && !result.ends_with(' ')) {
@@ -233,10 +222,6 @@ impl TextLine {
                result.push_str("<u>");
                current_underline = true;
            }
-            if item_strikeout && !current_strikeout {
-                result.push_str("<s>");
-                current_strikeout = true;
-            }
            if item_bold && !current_bold {
                result.push_str("**");
                current_bold = true;
@@ -259,9 +244,6 @@ impl TextLine {
        if current_underline {
            result.push_str("</u>");
        }
-        if current_strikeout {
-            result.push_str("</s>");
-        }

        result
    }
@@ -327,88 +309,3 @@ impl TextLine {
            || space_already_exists)
    }
 }
-
-#[cfg(test)]
-mod formatting_tests {
-    use super::{ItemType, TextItem, TextLine};
-
-    fn item(text: &str, x: f32, width: f32, strikeout: bool) -> TextItem {
-        TextItem {
-            text: text.to_string(),
-            x,
-            y: 100.0,
-            width,
-            height: 12.0,
-            font: "F1".to_string(),
-            font_size: 12.0,
-            page: 1,
-            is_bold: false,
-            is_italic: false,
-            is_underline: false,
-            is_strikeout: strikeout,
-            item_type: ItemType::Text,
-            mcid: None,
-        }
-    }
-
-    fn line(items: Vec<TextItem>) -> TextLine {
-        TextLine {
-            items,
-            y: 100.0,
-            page: 1,
-            adaptive_threshold: 0.1,
-        }
-    }
-
-    #[test]
-    fn formatting_emits_semantic_strikeout() {
-        let line = line(vec![item("deleted", 10.0, 42.0, true)]);
-
-        assert_eq!(
-            line.text_with_formatting(true, true, true),
-            "<s>deleted</s>"
-        );
-    }
-
-    #[test]
-    fn formatting_closes_strikeout_before_live_text() {
-        let line = line(vec![
-            item("keep", 10.0, 24.0, false),
-            item("remove", 40.0, 42.0, true),
-            item("keep", 88.0, 24.0, false),
-        ]);
-
-        assert_eq!(
-            line.text_with_formatting(true, true, true),
-            "keep <s>remove</s> keep"
-        );
-    }
-
-    #[test]
-    fn formatting_coalesces_adjacent_struck_items() {
-        let line = line(vec![
-            item("deleted", 10.0, 42.0, true),
-            item("words", 58.0, 30.0, true),
-        ]);
-
-        assert_eq!(
-            line.text_with_formatting(true, true, true),
-            "<s>deleted words</s>"
-        );
-    }
-
-    #[test]
-    fn strikeout_takes_precedence_over_other_styles() {
-        let mut decorated = item("deleted", 10.0, 42.0, true);
-        decorated.is_bold = true;
-        decorated.is_italic = true;
-        decorated.is_underline = true;
-        let line = line(vec![decorated]);
-
-        assert_eq!(
-            line.text_with_formatting(true, true, true),
-            "<s>deleted</s>"
-        );
-        assert_eq!(line.text(), "deleted");
-    }
-}
@@ -8,13 +8,12 @@ use pdf_inspector::{
    detect_pdf_type, detect_vector_grid_in_region_mem, extract_pages_markdown,
    extract_pages_markdown_mem, extract_tables_in_regions_mem, extract_text,
    extract_text_in_regions_mem, extract_text_with_positions, extract_text_with_positions_mem,
-    process_pdf_mem, process_pdf_mem_with_options, process_pdf_with_options, to_markdown,
-    to_markdown_from_items_with_rects_and_page_count, MarkdownOptions, PdfError, PdfOptions,
+    process_pdf_mem, process_pdf_with_options, to_markdown, MarkdownOptions, PdfError, PdfOptions,
    PdfType, TextItem,
 };
 use std::collections::HashSet;

-fn make_text_pdf(content: &str, media_box: &str) -> Vec<u8> {
+fn make_minimal_text_pdf() -> Vec<u8> {
    let mut pdf = b"%PDF-1.4\n".to_vec();
    let mut offsets = vec![0usize];

@@ -41,11 +40,10 @@ fn make_text_pdf(content: &str, media_box: &str) -> Vec<u8> {
        &mut pdf,
        &mut offsets,
        3,
-        &format!(
-            "<< /Type /Page /Parent 2 0 R /MediaBox [{media_box}] /Resources << /Font << /F1 5 0 R >> >> /Contents 4 0 R >>"
-        ),
+        "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792] /Resources << /Font << /F1 5 0 R >> >> /Contents 4 0 R >>",
    );

+    let content = "BT /F1 12 Tf 100 700 Td (Hello World) Tj 0 -14 Td (Second Line) Tj 0 -14 Td (Third Line) Tj ET";
    add_object(
        &mut pdf,
        &mut offsets,
@@ -81,186 +79,6 @@ fn make_text_pdf(content: &str, media_box: &str) -> Vec<u8> {
    pdf
 }

-fn make_recurring_contextual_folio_pdf() -> Vec<u8> {
-    let mut pdf = b"%PDF-1.4\n".to_vec();
-    let mut offsets = vec![0usize];
-
-    fn add_object(pdf: &mut Vec<u8>, offsets: &mut Vec<usize>, id: usize, body: &str) {
-        offsets.push(pdf.len());
-        pdf.extend_from_slice(format!("{id} 0 obj\n").as_bytes());
-        pdf.extend_from_slice(body.as_bytes());
-        pdf.extend_from_slice(b"\nendobj\n");
-    }
-
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        1,
-        "<< /Type /Catalog /Pages 2 0 R >>",
-    );
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        2,
-        "<< /Type /Pages /Kids [3 0 R 5 0 R 7 0 R 9 0 R] /Count 4 >>",
-    );
-    for page_index in 0..4 {
-        let page_id = 3 + page_index * 2;
-        let content_id = page_id + 1;
-        add_object(
-            &mut pdf,
-            &mut offsets,
-            page_id,
-            &format!(
-                "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792] /Resources << /Font << /F1 11 0 R >> >> /Contents {content_id} 0 R >>"
-            ),
-        );
-        let page_number = page_index + 1;
-        let content = format!(
-            "BT /F1 12 Tf 1 0 0 1 25 30 Tm ({page_number}) Tj 1 0 0 1 41 30 Tm (Company report footer) Tj 1 0 0 1 72 700 Tm (Body page {page_number}) Tj ET"
-        );
-        add_object(
-            &mut pdf,
-            &mut offsets,
-            content_id,
-            &format!(
-                "<< /Length {} >>\nstream\n{}\nendstream",
-                content.len(),
-                content
-            ),
-        );
-    }
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        11,
-        "<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica >>",
-    );
-
-    let xref_start = pdf.len();
-    pdf.extend_from_slice(format!("xref\n0 {}\n", offsets.len()).as_bytes());
-    pdf.extend_from_slice(b"0000000000 65535 f \n");
-    for offset in offsets.iter().skip(1) {
-        pdf.extend_from_slice(format!("{offset:010} 00000 n \n").as_bytes());
-    }
-    pdf.extend_from_slice(
-        format!(
-            "trailer\n<< /Size {} /Root 1 0 R >>\nstartxref\n{}\n%%EOF",
-            offsets.len(),
-            xref_start
-        )
-        .as_bytes(),
-    );
-
-    pdf
-}
-
-fn make_pdf_with_malformed_unselected_page() -> Vec<u8> {
-    let mut pdf = b"%PDF-1.4\n".to_vec();
-    let mut offsets = vec![0usize];
-
-    fn add_object(pdf: &mut Vec<u8>, offsets: &mut Vec<usize>, id: usize, body: &str) {
-        offsets.push(pdf.len());
-        pdf.extend_from_slice(format!("{id} 0 obj\n").as_bytes());
-        pdf.extend_from_slice(body.as_bytes());
-        pdf.extend_from_slice(b"\nendobj\n");
-    }
-
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        1,
-        "<< /Type /Catalog /Pages 2 0 R >>",
-    );
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        2,
-        "<< /Type /Pages /Kids [3 0 R 5 0 R] /Count 2 >>",
-    );
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        3,
-        "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792] /Resources << /Font << /F1 7 0 R >> >> /Contents 4 0 R >>",
-    );
-    let content = "BT /F1 12 Tf 1 0 0 1 25 30 Tm (1) Tj 1 0 0 1 41 30 Tm (Company report footer) Tj 1 0 0 1 72 700 Tm (Selected page text) Tj 0 -16 Td (More selected text) Tj 0 -16 Td (Still selected text) Tj ET";
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        4,
-        &format!(
-            "<< /Length {} >>\nstream\n{}\nendstream",
-            content.len(),
-            content
-        ),
-    );
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        5,
-        "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792] /Resources << /Font << /F1 7 0 R >> >> /Contents 6 0 R >>",
-    );
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        6,
-        "<< /Length 3 >>\nstream\nBI \nendstream",
-    );
-    add_object(
-        &mut pdf,
-        &mut offsets,
-        7,
-        "<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica >>",
-    );
-
-    let xref_start = pdf.len();
-    pdf.extend_from_slice(format!("xref\n0 {}\n", offsets.len()).as_bytes());
-    pdf.extend_from_slice(b"0000000000 65535 f \n");
-    for offset in offsets.iter().skip(1) {
-        pdf.extend_from_slice(format!("{offset:010} 00000 n \n").as_bytes());
-    }
-    pdf.extend_from_slice(
-        format!(
-            "trailer\n<< /Size {} /Root 1 0 R >>\nstartxref\n{}\n%%EOF",
-            offsets.len(),
-            xref_start
-        )
-        .as_bytes(),
-    );
-
-    pdf
-}
-
-fn make_minimal_text_pdf() -> Vec<u8> {
-    make_text_pdf(
-        "BT /F1 12 Tf 100 700 Td (Hello World) Tj 0 -14 Td (Second Line) Tj 0 -14 Td (Third Line) Tj ET",
-        "0 0 612 792",
-    )
-}
-
-fn make_digit_run_repro_pdf() -> Vec<u8> {
-    let content = r#"BT
-/F1 12 Tf
-1 0 0 1 72 780 Tm (A\)) Tj
-1 0 0 1 96 780 Tm (The) Tj
-1 0 0 1 126 780 Tm (total) Tj
-1 0 0 1 166 780 Tm (of) Tj
-1 0 0 1 186 780 Tm (730) Tj
-1 0 0 1 220 780 Tm (seats) Tj
-1 0 0 1 262 780 Tm (was) Tj
-1 0 0 1 296 780 Tm (approved.) Tj
-1 0 0 1 72 755 Tm (B\)) Tj
-1 0 0 1 96 755 Tm (let) Tj
-1 0 0 1 120 755 Tm (log) Tj
-1 0 0 1 150 755 Tm (2) Tj
-1 0 0 1 164 755 Tm (=) Tj
-1 0 0 1 180 755 Tm (a) Tj
-1 0 0 1 72 720 Tm (C\) Control: The total of 730 seats was approved. let log 2 = a) Tj
-ET"#;
-    make_text_pdf(content, "0 0 595 842")
-}
-
 fn truncate_eof_marker(mut pdf: Vec<u8>) -> Vec<u8> {
    assert!(pdf.ends_with(b"%%EOF"));
    pdf.pop();
@@ -512,21 +330,6 @@ fn test_group_into_lines_sorting_by_x() {
    assert_eq!(lines[0].text(), "First Second Third");
 }

-#[test]
-fn test_digit_only_text_runs_are_preserved_in_markdown() {
-    let pdf = make_digit_run_repro_pdf();
-
-    let items = extract_text_with_positions_mem(&pdf).expect("extract positioned text");
-    assert!(items.iter().any(|item| item.text == "730"));
-    assert!(items.iter().any(|item| item.text == "2"));
-
-    let result = process_pdf_mem(&pdf).expect("convert PDF to markdown");
-    assert_eq!(
-        result.markdown.expect("markdown output").trim(),
-        "A) The total of 730 seats was approved.\nB) let log 2 = a\nC) Control: The total of 730 seats was approved. let log 2 = a"
-    );
-}
-
 // ============================================================================
 // MarkdownOptions Tests
 // ============================================================================
@@ -744,30 +547,6 @@ fn test_markdown_from_items_page_breaks() {
    assert!(md.contains("Content on second page"));
 }

-#[test]
-fn test_markdown_page_count_overload_includes_trailing_blank_pages_in_folio_coverage() {
-    let mut items = Vec::new();
-    for (page, value) in [(1, "1"), (2, "2"), (3, "3"), (4, "4")] {
-        items.push(make_text_item(value, 25.0, 30.0, 12.0, page));
-        items.push(make_text_item(
-            "Company report footer",
-            41.0,
-            30.0,
-            12.0,
-            page,
-        ));
-    }
-    let options = MarkdownOptions {
-        strip_headers_footers: false,
-        ..MarkdownOptions::default()
-    };
-
-    let md = to_markdown_from_items_with_rects_and_page_count(items, options, &[], 20);
-
-    assert!(md.contains("1 Company report footer"));
-    assert!(md.contains("4 Company report footer"));
-}
-
 // ============================================================================
 // Markdown From Lines Tests
 // ============================================================================
@@ -3148,57 +2927,6 @@ fn test_extract_pages_markdown_basic() {
    assert!(!result.pages[0].needs_ocr);
 }

-#[test]
-fn test_extract_pages_markdown_uses_document_wide_folio_context() {
-    let pdf = make_recurring_contextual_folio_pdf();
-    let result = extract_pages_markdown_mem(&pdf, None).unwrap();
-
-    assert_eq!(result.pages.len(), 4);
-    for (index, page) in result.pages.iter().enumerate() {
-        assert!(page.markdown.contains("Company report footer"));
-        assert!(
-            !page
-                .markdown
-                .contains(&format!("{} Company report footer", index + 1)),
-            "recurring contextual folio survived on page {}: {}",
-            index + 1,
-            page.markdown
-        );
-    }
-}
-
-#[test]
-fn test_process_pdf_page_filter_uses_document_wide_folio_context() {
-    let pdf = make_recurring_contextual_folio_pdf();
-    let result = process_pdf_mem_with_options(&pdf, PdfOptions::new().pages([1])).unwrap();
-    let markdown = result.markdown.unwrap();
-
-    assert!(markdown.contains("Company report footer"));
-    assert!(!markdown.contains("1 Company report footer"), "{markdown}");
-    assert!(markdown.contains("Body page 1"));
-    assert!(!markdown.contains("Body page 2"));
-}
-
-#[test]
-fn test_selected_page_ignores_context_only_extraction_failure() {
-    let pdf = make_pdf_with_malformed_unselected_page();
-
-    let pages = extract_pages_markdown_mem(&pdf, Some(&[0])).unwrap();
-    assert_eq!(pages.pages.len(), 1);
-    assert!(pages.pages[0].markdown.contains("Selected page text"));
-
-    let result = process_pdf_mem_with_options(&pdf, PdfOptions::new().pages([1])).unwrap();
-    let markdown = result.markdown.unwrap();
-    assert!(markdown.contains("Selected page text"));
-}
-
-#[test]
-fn test_requested_page_extraction_failure_remains_fatal() {
-    let pdf = make_pdf_with_malformed_unselected_page();
-
-    assert!(extract_pages_markdown_mem(&pdf, Some(&[1])).is_err());
-}
-
 #[test]
 fn test_extract_pages_markdown_page_ordering() {
    let buf = std::fs::read("tests/fixtures/nexo-price-en.pdf").unwrap();
@@ -6,4 +6,7 @@
 |Exclusive Special|83,500,000 75,909,091(7,590,909) with 3.5% individual consumption tax applied 82,232,000|79,275,000 with 3.5% individual consumption tax applied 79,275,000|▶ Standard equipment of Exclusive plus • Smart Safety Technology: Forward Collision-avoidance Assist(intersection crossing/changing lanes in oncoming traffic/approaching from either side/ evasive steering assist), Highway Driving Assist 2, Navigation-based Smart Cruise Control(access road) • Exterior: Roof rack • Interior: Metallic pedal, Driving mode-dependent ambient mood lighting(crash pad, 1st/2nd-row door trim) • Seat: Synthetic leather seats(patch applied), Power-adjustable driver's seat(8-way, lumbar support, and Integrated Memory System(driver's seat and outside mirror connected)), Power-adjustable front passenger’s seat(8-way), Ventilated 1st-row seats, Heated 2nd-row seats • Convenience: Hi-pass(e hi-pass), In-car fingerprint authentication system(personalization, startup, payment, and etc.), Smart power tailgate|▶ [600,000] Built-in Cam 2 Plus, Augmented reality navigation ▶ [850,000] Indoor/outdoor V2L ▶ [950,000] Parking Assist ▶ [1,150,000] Audio by BANG & OLUFSEN sound system ▶ [250,000] 19-inch alloy wheels & tires|
 |Prestige|87,893,000 79,902,727(7,990,273) with 3.5% individual consumption tax applied 86,559,000|83,445,000 with 3.5% individual consumption tax applied 83,445,000|▶ Standard equipment of Exclusive Special plus • Smart Safety Technology: Remote Smart Parking Assist 2, Parking Collison- avoidance Assist(front/side/rear) • Exterior: Intelligent Front-Lighting System(IFS), Dynamic welcome/escort lighting(1 type), Sequential turn signals(front and rear), Ambient lighting auto flush door handles, Two-tone door garnish, Glossy black rear diffuser • Interior: Recycled PET suede interior materials(headlining/sunvisor), Fabric upholstered crash pad • Seat: BIO-processed natural leather seats(metal patch applied, embossed design punching), Passenger's seat walk-in device, 1st-row relaxation comfort seats(leg rest included), Ventilated 2nd-row seats • Convenience: Parking Distance Warning-Side, Head-Up Display, Digital key 2, Wireless phone charger(dual), Surround View Monitor, Blind-spot View Monitor, LED reverse light guide • Infotainment: Audio by BANG & OLUFSEN sound system(14 speakers, including external amp), Active road noise control, Active Sound Design|▶ [600,000] Built-in Cam 2 Plus, Augmented reality navigation ▶ [850,000] Indoor/outdoor V2L ▶ [900,000] Vision roof ▶ [1,380,000] Digital side mirror ▶ [750,000] Camera package ▶ [250,000] 19-inch alloy wheels & tires|

-**Classification Details** **Indoor/outdoor V2L** Indoor V2L, Outdoor V2L(connectorless type) **Parking Assist** Surround View Monitor, Blind-spot View Monitor, Parking Distance Warning-Side, Parking Collison-avoidance Assist-Rear **Audio by BANG & OLUFSEN** Audio by BANG & OLUFSEN sound system(14 speakers, including external amp.), Active road noise control, Active Sound Design **sound system** **Camera package** Digital center mirror(with camera sensor cleaning system), Driver monitoring system THE ALL-NEW NEXO /// ECO-FRIENDLY CAR
+**Classification Details** **Indoor/outdoor V2L** Indoor V2L, Outdoor V2L(connectorless type) **Parking Assist** Surround View Monitor, Blind-spot View Monitor, Parking Distance Warning-Side, Parking Collison-avoidance Assist-Rear **Audio by BANG & OLUFSEN**
+
+Audio by BANG & OLUFSEN sound system(14 speakers, including external amp.), Active road noise control, Active Sound Design **sound system** **Camera package** Digital center mirror(with camera sensor cleaning system), Driver monitoring system THE ALL-NEW NEXO /// ECO-FRIENDLY CAR
+
@@ -8,7 +8,7 @@ Department of the Treasury **Internal Revenue Service**

 # and Report to Employer

-### This publication contains:
+## This publication contains:

 **Form 4070A,** Employee’s Daily Record of Tips **Form 4070,** Employee’s Report of Tips to Employer

@@ -22,7 +22,7 @@ Name and address of employee

 **Publication 1244 (Rev. 7-96)** Cat. No. 44472W

-### Instructions
+## Instructions

 You must keep sufficient proof to show the amount of your tip income for the year. A daily record of your tip income is considered sufficient proof. Keep a daily record for each workday showing the amount of cash and credit card tips received directly from customers or other employees. Also keep a record of the amount of tips, if any, you paid to other employees through tip sharing, tip pooling or other arrangements, and the names of employees to whom you paid tips. Show the date that each entry is made. This date should be on or near the date you received the tip income. You may use **Form 4070A**, Employee’s Daily Record of Tips, or any other daily record to record your tips. **Reporting Tips to Your Employer.—**If you receive tips that total $20 or more for any month while working for one employer, you must report the tips to your employer. Tips include cash left by customers, tips customers add to credit card charges, and tips you receive from other employees. You must report your tips for any one month by the 10th day of the next month. If the 10th day falls on a Saturday, Sunday, or legal holiday, you may give the report to your employer on the next business day that is not a Saturday, Sunday, or legal holiday. You must report tips that total $20 or more every month regardless of your total wages and tips for the year. You may use **Form 4070**, Employee’s Report of Tips to Employer, to report your tips to your employer. See the instructions on the back of Form 4070. You must include all tips, including tips not reported to your employer, as wages on your income tax return. You may use the last page of this publication to total your tips for the year. Your employer must withhold income, social security, and Medicare (or railroad retirement) taxes on tips you report. Your employer usually deducts the withholding due on tips from your regular wages.

@@ -56,11 +56,7 @@ tips of directly from customers received other employees paid tips rec’d. entr
 **2.** Report total credit card tips (col. **b**) on Form 4070, line **2.**
 **3.** Report total tips paid out (col. **c**) on Form 4070, line **3.** **Page 4**

-Form **4070** Employee’s Report (Rev. July 1996)
-
-## of Tips to EmployerOMB No. 1545-0065
-
-Department of the Treasury Internal Revenue Service' **For Paperwork Reduction Act Notice, see back of form.** Employee’s name and address **Social security number**
+Form Employee’s Report (Rev. July 1996) of Tips to EmployerOMB No. 1545-0065 Department of the Treasury Internal Revenue Service' **For Paperwork Reduction Act Notice, see back of form.** Employee’s name and address **Social security number**

 Employer’s name and address (include establishment name, if different) **1** Cash tips received

@@ -78,6 +74,7 @@ forms simpler, we would be happy to hear from you. You can write to the Tax Form

 **Unreported Tips.—**If you received tips of $20 or more for any month while working for one employer but did not report them to your employer, you must figure and pay social security and Medicare taxes on the unreported tips when you file your tax return. If you have unreported tips, you **must** use Form 1040 and **Form 4137,** Social Security and Medicare Tax on Unreported Tip Income, to report them. You may **not** use Form 1040A or 1040EZ. Employees subject to the Railroad Retirement Tax Act **cannot** use Form 4137 to pay railroad retirement tax on unreported tips. To get railroad retirement credit, you must report tips to your employer. If you do not report tips to your employer as required, you may be charged a penalty of 50% of the social security and Medicare taxes (or railroad retirement tax) due on the unreported tips unless there was reasonable cause for not reporting them. **Additional Information.—**Get **Pub. 531,** Reporting Tip Income, and Form 4137 for more information on tips. If you are an employee of certain large food or beverage establishments, see Pub. 531 for tip allocation rules. **Recordkeeping.—**If you do not keep a daily record of tips, you must keep other reliable proof of the tip income you received. This proof includes copies of restaurant bills and credit card charges that show amounts customers added as tips. Keep your tip income records for as long as the information on them may be needed in the administration of any Internal Revenue law.

-### Instructions (continued)
+## Instructions (continued)

 Use this space to total your tips for the year
+
@@ -46,9 +46,19 @@ more about investing in tax losses than burst, cap rates spreads steadily com- r

 8 6 Z E L L / L U R I E R E A L E S T A T E C E N T E R

-ingdebtspreadswerepartoforiginalpro formamodels.Thiscapratespreadcom- pressionoffsetweakcashflowsinapost- recessionary economy from 2002 to 2005, while continued compression, combined with improved cash flows, pushed property values skyward in 2006 throughmid-2007. Cap rate compression reduced the importance of the ability to add value. After all, if all you had to do to make moneywastoleveragetothehiltwhilecap ratesfell,whytakeontheextraworkand riskofattemptingtoaddvalue?Stateddif- ferently: Why print money if it is laying everywhereonthestreets? In Tables III and IV, we demonstrate thepowerofcapratecompressionviavery simple pro forma cash flow analyses that assume Year 1 NOI of $100; a going-in cap rate of 9 percent; an LTV of 70 per- cent; and an interest rate of 7 percent. Withineachfigure,wedisplaytwoscenar- ios, which vary based on NOI growth assumptions.ScenarioIassumesthatNOI growsby3percentperyear,whileScenario IIassumesavalue-addNOIgrowthof20 percentbetweenyearstwoandthree. The only other difference between TablesIIIandIVisinresidualcaprates, which are assumed to be 6 percent and 9 percent, respectively. Based on these assumptions, we calculate the equity IRRs. It is clear that cap rate compres- sion is a significant factor in driving
+ingdebtspreadswerepartoforiginalpro formamodels.Thiscapratespreadcom- pressionoffsetweakcashflowsinapost- recessionary economy from 2002 to 2005, while continued compression, combined with improved cash flows, pushed property values skyward in 2006 throughmid-2007.

-returns. That is, cap rate compression from 9 percent to 6 percent increased IRR on leveraged stabilized properties by 250 percent, to a staggering 57 per- cent. Who needs to take on value add riskatthisreturnforstabilizedassets? Intheearly1980s,moneywasmadein real estate by mastering the creation and syndication of tax gimmicks. In the late 1980s, one made money by mastering bank and S&L connections to over-lever- age.Intheearly1990s,onemademoneyin realestatebyhavingaccesstoequity—the morethebetter.Duringthelate1990s,one made money from real estate by realizing large spreads between cap rates and debt costs.And,overthepastfiveyears,theway to make money in real estate was to own realestateonahighlyleveragedbasisascap ratesplunged. Theclassicassetpricingmodelisthe capital asset pricing model (CAPM). CAPM is a simple, yet elegant, model that relates asset pricing to the risk-free rate(F),theabilityofanassettoreduce portfolio variance (B), and the expected rate of return on the market bundle of investableassets(M).CAPMisfarfrom perfect,butprovidesacrudebenchmark for asset pricing, around which discrep- ancies and novelties arise. Specifically, CAPM states that an asset’s price is set suchthattheexpectedreturnforanasset
+Cap rate compression reduced the importance of the ability to add value. After all, if all you had to do to make moneywastoleveragetothehiltwhilecap ratesfell,whytakeontheextraworkand riskofattemptingtoaddvalue?Stateddif- ferently: Why print money if it is laying everywhereonthestreets?
+
+In Tables III and IV, we demonstrate thepowerofcapratecompressionviavery simple pro forma cash flow analyses that assume Year 1 NOI of $100; a going-in cap rate of 9 percent; an LTV of 70 per- cent; and an interest rate of 7 percent. Withineachfigure,wedisplaytwoscenar- ios, which vary based on NOI growth assumptions.ScenarioIassumesthatNOI growsby3percentperyear,whileScenario IIassumesavalue-addNOIgrowthof20 percentbetweenyearstwoandthree.
+
+The only other difference between TablesIIIandIVisinresidualcaprates, which are assumed to be 6 percent and 9 percent, respectively. Based on these assumptions, we calculate the equity IRRs. It is clear that cap rate compres- sion is a significant factor in driving
+
+returns. That is, cap rate compression from 9 percent to 6 percent increased IRR on leveraged stabilized properties by 250 percent, to a staggering 57 per- cent. Who needs to take on value add riskatthisreturnforstabilizedassets?
+
+Intheearly1980s,moneywasmadein real estate by mastering the creation and syndication of tax gimmicks. In the late 1980s, one made money by mastering bank and S&L connections to over-lever- age.Intheearly1990s,onemademoneyin realestatebyhavingaccesstoequity—the morethebetter.Duringthelate1990s,one made money from real estate by realizing large spreads between cap rates and debt costs.And,overthepastfiveyears,theway to make money in real estate was to own realestateonahighlyleveragedbasisascap ratesplunged.
+
+Theclassicassetpricingmodelisthe capital asset pricing model (CAPM). CAPM is a simple, yet elegant, model that relates asset pricing to the risk-free rate(F),theabilityofanassettoreduce portfolio variance (B), and the expected rate of return on the market bundle of investableassets(M).CAPMisfarfrom perfect,butprovidesacrudebenchmark for asset pricing, around which discrep- ancies and novelties arise. Specifically, CAPM states that an asset’s price is set suchthattheexpectedreturnforanasset

 (R)is R=F+ β(M-F).
 R E V I E W 8 7
@@ -1,16 +1,18 @@
 Reprinted with corrections from *The Bell System Technical Journal,* Vol. 27, pp. 379–423, 623–656, July, October, 1948.

-## A Mathematical Theory of Communication
+# A Mathematical Theory of Communication

-### By C. E. SHANNON
+## By C. E. SHANNON

 INTRODUCTION

-HE recent development of various methods of modulation such as PCM and PPM which exchange
+THE recent development of various methods of modulation such as PCM and PPM which exchange bandwidth for signal-to-noise ratio has intensified the interest in a general theory of communication. A basis for such a theory is contained in the important papers of Nyquist¹ and Hartley² on this subject. In the present paper we will extend the theory to include a number of new factors, in particular the effect of noise in the channel, and the savings possible due to the statistical structure of the original message and due to the nature of the final destination of the information.

-# Tbandwidth for signal-to-noise ratio has intensified the interest in a general theory of communication. A
+The fundamental problem of communication is that of reproducing at one point either exactly or ap- proximately a message selected at another point. Frequently the messages have *meaning*; that is they refer to or are correlated according to some system with certain physical or conceptual entities. These semantic aspects of communication are irrelevant to the engineering problem. The significant aspect is that the actual message is one *selected from a set* of possible messages. The system must be designed to operate for each possible selection, not just the one which will actually be chosen since this is unknown at the time of design.

-basis for such a theory is contained in the important papers of Nyquist¹ and Hartley² on this subject. In the present paper we will extend the theory to include a number of new factors, in particular the effect of noise in the channel, and the savings possible due to the statistical structure of the original message and due to the nature of the final destination of the information. The fundamental problem of communication is that of reproducing at one point either exactly or ap- proximately a message selected at another point. Frequently the messages have *meaning*; that is they refer to or are correlated according to some system with certain physical or conceptual entities. These semantic aspects of communication are irrelevant to the engineering problem. The significant aspect is that the actual message is one *selected from a set* of possible messages. The system must be designed to operate for each possible selection, not just the one which will actually be chosen since this is unknown at the time of design. If the number of messages in the set is finite then this number or any monotonic function of this number can be regarded as a measure of the information produced when one message is chosen from the set, all choices being equally likely. As was pointed out by Hartley the most natural choice is the logarithmic function. Although this definition must be generalized considerably when we consider the influence of the statistics of the message and when we have a continuous range of messages, we will in all cases use an essentially logarithmic measure. The logarithmic measure is more convenient for various reasons:
+If the number of messages in the set is finite then this number or any monotonic function of this number can be regarded as a measure of the information produced when one message is chosen from the set, all choices being equally likely. As was pointed out by Hartley the most natural choice is the logarithmic function. Although this definition must be generalized considerably when we consider the influence of the statistics of the message and when we have a continuous range of messages, we will in all cases use an essentially logarithmic measure.
+
+The logarithmic measure is more convenient for various reasons:

 1. It is practically more useful. Parameters of engineering importance such as time, bandwidth, number of relays, etc., tend to vary linearly with the logarithm of the number of possibilities. For example, adding one relay to a group doubles the number of possible states of the relays. It adds 1 to the base 2 logarithm of this number. Doubling the time roughly squares the number of possible messages, or doubles the logarithm, etc.
 2. It is nearer to our intuitive feeling as to the proper measure. This is closely related to (1) since we in- tuitively measures entities by linear comparison with common standards. One feels, for example, that two punched cards should have twice the capacity of one for information storage, and two identical channels twice the capacity of one for transmitting information.
@@ -22,7 +24,7 @@ J. W. Tukey. A device with two stable positions, such as a relay or a flip-flop

 log₂*M* = log₁₀*M*= log₁₀2 = 3:32 log₁₀*M*;

-1 Nyquist, H., “Certain Factors Affecting Telegraph Speed,” *Bell System Technical Journal,* April 1924, p. 324; “Certain Topics in Telegraph Transmission Theory,” *A.I.E.E. Trans.,* v. 47, April 1928, p. 617. 2 Hartley, R. V. L., “Transmission of Information,” *Bell System Technical Journal,* July 1928, p. 535.
+1 Nyquist, H., “Certain Factors Affecting Telegraph Speed,” *Bell System Technical Journal,* April 1924, p. 324; “Certain Topics in Telegraph Transmission Theory,” *A.I.E.E. Trans.,* v. 47, April 1928, p. 617. Hartley, R. V. L., “Transmission of Information,” *Bell System Technical Journal,* July 1928, p. 535.

 INFORMATION SOURCE TRANSMITTER RECEIVER DESTINATION

@@ -32,7 +34,9 @@ NOISE SOURCE

 Fig. 1 — Schematic diagram of a general communication system.

-a decimal digit is about 3 13 bits. A digit wheel on a desk computing machine has ten stable positions and therefore has a storage capacity of one decimal digit. In analytical work where integration and differentiation are involved the base *e* is sometimes useful. The resulting units of information will be called natural units. Change from the base *a* to base *b* merely requires multiplication by log*ba*. By a communication system we will mean a system of the type indicated schematically in Fig. 1. It consists of essentially five parts:
+a decimal digit is about 3 13 bits. A digit wheel on a desk computing machine has ten stable positions and therefore has a storage capacity of one decimal digit. In analytical work where integration and differentiation are involved the base *e* is sometimes useful. The resulting units of information will be called natural units. Change from the base *a* to base *b* merely requires multiplication by log*ba*.
+
+By a communication system we will mean a system of the type indicated schematically in Fig. 1. It consists of essentially five parts:

 1. An *information source* which produces a message or sequence of messages to be communicated to the receiving terminal. The message may be of various types: (a) A sequence of letters as in a telegraph of teletype system; (b) A single function of time *f* (*t*) as in radio or telephony; (c) A function of time and other variables as in black and white television — here the message may be thought of as a function *f* (*x*; *y*;*t*) of two space coordinates and time, the light intensity at point (*x*; *y*) and time *t* on a pickup tube plate; (d) Two or more functions of time, say *f* (*t*), *g*(*t*), *h*(*t*) — this is the case in “three- dimensional” sound transmission or if the system is intended to service several individual channels in multiplex; (e) Several functions of several variables — in color television the message consists of three functions *f* (*x*; *y*;*t*), *g*(*x*; *y*;*t*), *h*(*x*; *y*;*t*) defined in a three-dimensional continuum — we may also think of these three functions as components of a vector field defined in the region — similarly, several black and white television sources would produce “messages” consisting of a number of functions of three variables; (f) Various combinations also occur, for example in television with an associated audio channel.
 2. A *transmitter* which operates on the message in some way to produce a signal suitable for trans- mission over the channel. In telephony this operation consists merely of changing sound pressure into a proportional electrical current. In telegraphy we have an encoding operation which produces a sequence of dots, dashes and spaces on the channel corresponding to the message. In a multiplex PCM system the different speech functions must be sampled, compressed, quantized and encoded, and finally interleaved properly to construct the signal. Vocoder systems, television and frequency modulation are other examples of complex operations applied to the message to obtain the signal.
@@ -1,18 +1,20 @@
-##### Technical Information
+### Technical Information

-## l T-12 SI
+#### l T-12 SI

-##### DuPont Fluorochemicals
+#### DuPont Fluorochemicals

-#### Thermodynamic Properties
+## Thermodynamic Properties

-**of** ®
+**of**
+
+®

 # Freon 12

-##### (R-12)
+### (R-12)

-##### Technical Information Technical Information
+#### Technical Information Technical Information

 **®** **Thermodynamic Properties of Freon 12 Refrigerant** **(R-12)** **SI Units**

@@ -22,11 +24,11 @@ S.A., Lemmon, E.W., and Peskin, Vf = Fluid (liquid) specific volume
 A.P., NIST Standard Reference in cubic meters per kilogram Database 23, NIST thermodynamic and transport properties of Vg = Vapour (gas) specific volume refrigerants and refrigerant in cubic meters per kilogram mixtures – REFPROP version 6.01, Standard Reference Data Program, df and dg = Fluid and Vapour National Institute of Standards and (respectively) densities in Technology, 1998).
 kilograms per cubic meter

-##### H = Enthalpy (kJ/kg)
+H = Enthalpy (kJ/kg)

-##### S = Entropy (kJ/kg.K)
+S = Entropy (kJ/kg.K)

-##### Physical Properties
+#### Physical Properties

 |Chemical Formula|CCl₂F₂|
 |---|---|
@@ -605,9 +605,9 @@ checksum = "0ceec5bc11778974d1bcb055b18002eba7f4b3518b6a0081b3af5f21666da9ad"

 [[package]]
 name = "lopdf"
-version = "0.42.0"
+version = "0.41.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "25aab26d99567469098e64a02f42679f8965c6401263eefa31d8f2dcc37a221c"
+checksum = "67513274c50a2b51e5f75d9e682fcf4ab064a8a9c9ae2c3c59309084882bb24d"
 dependencies = [
 "aes",
 "bitflags 2.13.1",