docling-parse 7.14.0__tar.gz → 7.16.0__tar.gz

{docling_parse-7.14.0 → docling_parse-7.16.0}/CHANGELOG.md

@@ -1,3 +1,19 @@
+## [v7.16.0](https://github.com/docling-project/docling-parse/releases/tag/v7.16.0) - 2026-08-25
+
+### Feature
+
+* Regression-test overhaul, and the parser/renderer fixes it uncovered ([#325](https://github.com/docling-project/docling-parse/issues/325)) ([`0ff4bea`](https://github.com/docling-project/docling-parse/commit/0ff4beafb45e800f9bd7d1399c281416d2a16ab6))
+
+### Fix
+
+* Open input PDFs with UTF-8 path semantics on Windows ([#326](https://github.com/docling-project/docling-parse/issues/326)) ([`a006162`](https://github.com/docling-project/docling-parse/commit/a0061628a57b8a8089f596744cbe74e6b709fd3f))
+
+## [v7.15.0](https://github.com/docling-project/docling-parse/releases/tag/v7.15.0) - 2026-08-20
+
+### Feature
+
+* **render:** Clip masks, shading patterns and Coons meshes; CCITT, CMYK-JPEG, tiling-pattern and CJK text fixes ([#323](https://github.com/docling-project/docling-parse/issues/323)) ([`850c3c3`](https://github.com/docling-project/docling-parse/commit/850c3c3c6e01591e9c6b19b32f2923258122d6ba))
+
 ## [v7.14.0](https://github.com/docling-project/docling-parse/releases/tag/v7.14.0) - 2026-08-18
 
 ### Feature

{docling_parse-7.14.0 → docling_parse-7.16.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling-parse
-Version: 7.14.0
+Version: 7.16.0
 Summary: Simple package to extract text with coordinates from programmatic PDFs
 Author-email: Peter Staar <taa@zurich.ibm.com>, Christoph Auer <cau@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Panos Vagenas <pva@zurich.ibm.com>, Maxim Lysak <mly@zurich.ibm.com>
 License-Expression: MIT
@@ -246,15 +246,51 @@ options:
   -p PDF, --pdf PDF  Path to the PDF file
 ```
 
+## System fonts
+
+A PDF may name a font without embedding it. There is then nothing in the file
+to draw with, and the renderer substitutes a face installed on the host --
+which is why the same document renders differently on two machines, and why a
+container with no fonts in it renders text as `.notdef` boxes ("tofu").
+
+Install the usual desktop font packages for text, plus a CJK face for Chinese,
+Japanese and Korean documents:
+
+| | |
+|---|---|
+| Debian / Ubuntu | `fonts-dejavu-core fonts-liberation2 fonts-freefont-ttf fonts-urw-base35 fonts-noto-core fonts-noto-cjk` |
+| Fedora / RHEL | `dejavu-sans-fonts liberation-sans-fonts urw-base35-fonts google-noto-sans-fonts google-noto-sans-cjk-fonts` |
+
+On Fedora and RHEL take the packages **without** `-vf-` in the name. The
+variable-font builds carry CFF2 outlines, which the pinned Blend2D cannot read;
+such a file is still drawn, through FreeType and without shaping, but the
+static build is the better path.
+
+Three environment variables override what gets substituted. Each names a font
+**file** (`.ttf`, `.otf`, `.ttc`), and each is tried before the built-in
+candidate list -- a path that does not exist, or that neither Blend2D nor
+FreeType can open, is skipped rather than fatal:
+
+| Variable | Used for |
+|---|---|
+| `DOCLING_PARSE_FALLBACK_FONT` | the general fallback, when a font name resolves to nothing |
+| `DOCLING_PARSE_CJK_FALLBACK_FONT` | requests for a CJK family (`/Batang`, `/SimSun`, `/MS Gothic`, ...) |
+| `DOCLING_PARSE_ARABIC_FALLBACK_FONT` | runs of Arabic script |
+
+Setting them is also how a render is made reproducible across machines: pin the
+same file everywhere and the substituted typeface stops being a property of the
+host. Run with `loglevel="info"` to see which file was chosen for each font.
+
 ## Performance Benchmarks
 
 [`docs/performance_benchmarks.md`](./docs/performance_benchmarks.md) compares `docling-parse` against the other widely used PDF packages on parsing and rendering, with the methodology and the commands to reproduce it.
 
-The tooling behind it lives under [`perf/`](./perf/README.md):
+The tooling behind it lives under [`scripts/benchmarking/`](./scripts/benchmarking/README.md):
 
-- [`perf/run_scaling.py`](./perf/run_scaling.py): measure — thread-scaling sweeps and the cross-package comparison
-- [`perf/run_eval.py`](./perf/run_eval.py): plot — per-page distributions, thread scaling, and per-package correlations
-- [`perf/run_analysis.py`](./perf/run_analysis.py): drill down — replay the slowest pages with C++ stage timings
+- [`scripts/benchmarking/run_performance_benchmarking.py`](./scripts/benchmarking/run_performance_benchmarking.py): measure — thread-scaling sweeps and the cross-package comparison
+- [`scripts/benchmarking/run_performance_eval.py`](./scripts/benchmarking/run_performance_eval.py): plot — per-page distributions, thread scaling, and per-package correlations
+- [`scripts/benchmarking/run_performance_analysis.py`](./scripts/benchmarking/run_performance_analysis.py): drill down — replay the slowest pages with C++ stage timings
+- [`scripts/benchmarking/run_quality_benchmarking.py`](./scripts/benchmarking/run_quality_benchmarking.py): compare quality — textual output and rendered pixels across PDF engines
 
 For historical V1 vs V2 benchmarks, see [legacy_performance_benchmarks.md](./docs/legacy/legacy_performance_benchmarks.md).
 

{docling_parse-7.14.0 → docling_parse-7.16.0}/README.md

@@ -215,15 +215,51 @@ options:
   -p PDF, --pdf PDF  Path to the PDF file
 ```
 
+## System fonts
+
+A PDF may name a font without embedding it. There is then nothing in the file
+to draw with, and the renderer substitutes a face installed on the host --
+which is why the same document renders differently on two machines, and why a
+container with no fonts in it renders text as `.notdef` boxes ("tofu").
+
+Install the usual desktop font packages for text, plus a CJK face for Chinese,
+Japanese and Korean documents:
+
+| | |
+|---|---|
+| Debian / Ubuntu | `fonts-dejavu-core fonts-liberation2 fonts-freefont-ttf fonts-urw-base35 fonts-noto-core fonts-noto-cjk` |
+| Fedora / RHEL | `dejavu-sans-fonts liberation-sans-fonts urw-base35-fonts google-noto-sans-fonts google-noto-sans-cjk-fonts` |
+
+On Fedora and RHEL take the packages **without** `-vf-` in the name. The
+variable-font builds carry CFF2 outlines, which the pinned Blend2D cannot read;
+such a file is still drawn, through FreeType and without shaping, but the
+static build is the better path.
+
+Three environment variables override what gets substituted. Each names a font
+**file** (`.ttf`, `.otf`, `.ttc`), and each is tried before the built-in
+candidate list -- a path that does not exist, or that neither Blend2D nor
+FreeType can open, is skipped rather than fatal:
+
+| Variable | Used for |
+|---|---|
+| `DOCLING_PARSE_FALLBACK_FONT` | the general fallback, when a font name resolves to nothing |
+| `DOCLING_PARSE_CJK_FALLBACK_FONT` | requests for a CJK family (`/Batang`, `/SimSun`, `/MS Gothic`, ...) |
+| `DOCLING_PARSE_ARABIC_FALLBACK_FONT` | runs of Arabic script |
+
+Setting them is also how a render is made reproducible across machines: pin the
+same file everywhere and the substituted typeface stops being a property of the
+host. Run with `loglevel="info"` to see which file was chosen for each font.
+
 ## Performance Benchmarks
 
 [`docs/performance_benchmarks.md`](./docs/performance_benchmarks.md) compares `docling-parse` against the other widely used PDF packages on parsing and rendering, with the methodology and the commands to reproduce it.
 
-The tooling behind it lives under [`perf/`](./perf/README.md):
+The tooling behind it lives under [`scripts/benchmarking/`](./scripts/benchmarking/README.md):
 
-- [`perf/run_scaling.py`](./perf/run_scaling.py): measure — thread-scaling sweeps and the cross-package comparison
-- [`perf/run_eval.py`](./perf/run_eval.py): plot — per-page distributions, thread scaling, and per-package correlations
-- [`perf/run_analysis.py`](./perf/run_analysis.py): drill down — replay the slowest pages with C++ stage timings
+- [`scripts/benchmarking/run_performance_benchmarking.py`](./scripts/benchmarking/run_performance_benchmarking.py): measure — thread-scaling sweeps and the cross-package comparison
+- [`scripts/benchmarking/run_performance_eval.py`](./scripts/benchmarking/run_performance_eval.py): plot — per-page distributions, thread scaling, and per-package correlations
+- [`scripts/benchmarking/run_performance_analysis.py`](./scripts/benchmarking/run_performance_analysis.py): drill down — replay the slowest pages with C++ stage timings
+- [`scripts/benchmarking/run_quality_benchmarking.py`](./scripts/benchmarking/run_quality_benchmarking.py): compare quality — textual output and rendered pixels across PDF engines
 
 For historical V1 vs V2 benchmarks, see [legacy_performance_benchmarks.md](./docs/legacy/legacy_performance_benchmarks.md).
 

{docling_parse-7.14.0 → docling_parse-7.16.0}/app/pybind_parse.cpp

@@ -241,6 +241,11 @@ namespace
       row["index"] = bitmap_index;
       row["xobject_key"] = instr.get_key();
       row["source"] = pdflib::to_string(instr.get_source());
+      // The decode path the samples came out of, as the PDF spells it. It is
+      // what says whether these bytes are reproducible: only /JPXDecode is
+      // inverse-transformed in floating point, and only it can round a sample
+      // differently on a different machine.
+      row["filters"] = instr.get_filters();
       row["shape"] = instr.get_shape();
       row["pixel_format"] = pixel_format_name(instr.get_pixel_format());
       row["image_mask"] = instr.is_image_mask();
@@ -408,10 +413,14 @@ PYBIND11_MODULE(pdf_parsers, m) {
 
   // PdfShape - graphic shape with coordinates
   pybind11::class_<pdflib::page_item<pdflib::PAGE_SHAPE>>(m, "PdfShape")
-    .def("get_x", &pdflib::page_item<pdflib::PAGE_SHAPE>::get_x,
+    .def("get_x",
+         static_cast<std::vector<double>& (pdflib::page_item<pdflib::PAGE_SHAPE>::*)()>(
+           &pdflib::page_item<pdflib::PAGE_SHAPE>::get_x),
 	 pybind11::return_value_policy::reference_internal,
 	 "Get x coordinates of shape points")
-    .def("get_y", &pdflib::page_item<pdflib::PAGE_SHAPE>::get_y,
+    .def("get_y",
+         static_cast<std::vector<double>& (pdflib::page_item<pdflib::PAGE_SHAPE>::*)()>(
+           &pdflib::page_item<pdflib::PAGE_SHAPE>::get_y),
 	 pybind11::return_value_policy::reference_internal,
 	 "Get y coordinates of shape points")
     .def("get_i", &pdflib::page_item<pdflib::PAGE_SHAPE>::get_i,
@@ -514,7 +523,9 @@ PYBIND11_MODULE(pdf_parsers, m) {
 
   // PdfShapes - iterable container of PdfShape objects
   pybind11::class_<pdflib::page_item<pdflib::PAGE_SHAPES>>(m, "PdfShapes")
-    .def("__len__", &pdflib::page_item<pdflib::PAGE_SHAPES>::size)
+    .def("__len__",
+         static_cast<size_t (pdflib::page_item<pdflib::PAGE_SHAPES>::*)()>(
+           &pdflib::page_item<pdflib::PAGE_SHAPES>::size))
     .def("__getitem__", [](pdflib::page_item<pdflib::PAGE_SHAPES>& self, size_t i)
 	 -> pdflib::page_item<pdflib::PAGE_SHAPE>& {
 	   if (i >= self.size()) {
@@ -1140,6 +1151,9 @@ PYBIND11_MODULE(pdf_parsers, m) {
 
     Attributes:
         render_text (bool): Render glyph outlines for text cells [default=true].
+        render_shapes (bool): Paint vector shape instructions [default=true].
+        render_shadings (bool): Paint shading instructions [default=true].
+        render_non_rect_clip_masks (bool): Apply non-rectangular clip masks [default=true].
         min_stroke_width (float): Minimum stroke width in device pixels for hairlines and sub-pixel vector strokes [default=1.0].
         draw_text_bbox (bool): Draw bounding quad for each text cell [default=false].
         draw_text_basepoint (bool): Draw the text baseline origin as a small red dot [default=false].
@@ -1155,6 +1169,9 @@ PYBIND11_MODULE(pdf_parsers, m) {
     )")
     .def(pybind11::init<>())
     .def_readwrite("render_text",             &pdflib::render_config::render_text)
+    .def_readwrite("render_shapes",           &pdflib::render_config::render_shapes)
+    .def_readwrite("render_shadings",         &pdflib::render_config::render_shadings)
+    .def_readwrite("render_non_rect_clip_masks", &pdflib::render_config::render_non_rect_clip_masks)
     .def_readwrite("min_stroke_width",        &pdflib::render_config::min_stroke_width)
     .def_readwrite("draw_text_bbox",          &pdflib::render_config::draw_text_bbox)
     .def_readwrite("draw_text_basepoint",     &pdflib::render_config::draw_text_basepoint)

{docling_parse-7.14.0 → docling_parse-7.16.0}/app/render.cpp

@@ -404,7 +404,14 @@ int main(int argc, char* argv[])
                                         export_page_pdf_files,
                                         std::filesystem::path(page_pdf_dir),
                                         ifile)) { return 1; }
-              rnd.show();
+              if (result.count("output"))
+                {
+                  rnd.save(ofile);
+                }
+              else
+                {
+                  rnd.show();
+                }
             }
           else
             {

{docling_parse-7.14.0 → docling_parse-7.16.0}/cmake/extlib_blend2d.cmake

@@ -14,11 +14,18 @@ else()
     # --- AsmJit (embedded by blend2d for JIT pipeline generation) ---
     # Use the inline form of FetchContent_Populate (not deprecated) to download
     # the source only. blend2d will add it via add_subdirectory using ASMJIT_DIR.
+    #
+    # Pinned alongside the blend2d commit below: asmjit generates blend2d's
+    # rasterization pipelines, so tracking master let an unrelated asmjit change
+    # shift rendered pixels between builds -- the renderer groundtruth is
+    # compared against stored PNGs, so that surfaces as a spurious regression.
+    # asmjit publishes no tags at all, so the pin has to be a commit SHA, and
+    # GIT_SHALLOW cannot stay: a depth-1 clone only contains the tip, so it
+    # cannot check out an arbitrary commit when this is next bumped.
     FetchContent_Populate(
         asmjit
         GIT_REPOSITORY https://github.com/asmjit/asmjit.git
-        GIT_TAG        master
-        GIT_SHALLOW    TRUE
+        GIT_TAG        0bd5787b54b575ed94bf32ac452153b34385c514
     )
     # Point blend2d to the downloaded asmjit source.
     set(ASMJIT_DIR "${asmjit_SOURCE_DIR}" CACHE PATH "Path to AsmJit source" FORCE)

{docling_parse-7.14.0 → docling_parse-7.16.0}/docling_parse/pdf_parser.py

@@ -1416,6 +1416,12 @@ class PageParseResult:
 def _copy_render_config(src: RenderConfig) -> RenderConfig:
     dst = RenderConfig()
     dst.render_text = src.render_text
+    if hasattr(src, "render_shapes"):
+        dst.render_shapes = src.render_shapes
+    if hasattr(src, "render_shadings"):
+        dst.render_shadings = src.render_shadings
+    if hasattr(src, "render_non_rect_clip_masks"):
+        dst.render_non_rect_clip_masks = src.render_non_rect_clip_masks
     dst.min_stroke_width = src.min_stroke_width
     dst.draw_text_bbox = src.draw_text_bbox
     dst.resolve_fonts = src.resolve_fonts

{docling_parse-7.14.0 → docling_parse-7.16.0}/docling_parse.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling-parse
-Version: 7.14.0
+Version: 7.16.0
 Summary: Simple package to extract text with coordinates from programmatic PDFs
 Author-email: Peter Staar <taa@zurich.ibm.com>, Christoph Auer <cau@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Panos Vagenas <pva@zurich.ibm.com>, Maxim Lysak <mly@zurich.ibm.com>
 License-Expression: MIT
@@ -246,15 +246,51 @@ options:
   -p PDF, --pdf PDF  Path to the PDF file
 ```
 
+## System fonts
+
+A PDF may name a font without embedding it. There is then nothing in the file
+to draw with, and the renderer substitutes a face installed on the host --
+which is why the same document renders differently on two machines, and why a
+container with no fonts in it renders text as `.notdef` boxes ("tofu").
+
+Install the usual desktop font packages for text, plus a CJK face for Chinese,
+Japanese and Korean documents:
+
+| | |
+|---|---|
+| Debian / Ubuntu | `fonts-dejavu-core fonts-liberation2 fonts-freefont-ttf fonts-urw-base35 fonts-noto-core fonts-noto-cjk` |
+| Fedora / RHEL | `dejavu-sans-fonts liberation-sans-fonts urw-base35-fonts google-noto-sans-fonts google-noto-sans-cjk-fonts` |
+
+On Fedora and RHEL take the packages **without** `-vf-` in the name. The
+variable-font builds carry CFF2 outlines, which the pinned Blend2D cannot read;
+such a file is still drawn, through FreeType and without shaping, but the
+static build is the better path.
+
+Three environment variables override what gets substituted. Each names a font
+**file** (`.ttf`, `.otf`, `.ttc`), and each is tried before the built-in
+candidate list -- a path that does not exist, or that neither Blend2D nor
+FreeType can open, is skipped rather than fatal:
+
+| Variable | Used for |
+|---|---|
+| `DOCLING_PARSE_FALLBACK_FONT` | the general fallback, when a font name resolves to nothing |
+| `DOCLING_PARSE_CJK_FALLBACK_FONT` | requests for a CJK family (`/Batang`, `/SimSun`, `/MS Gothic`, ...) |
+| `DOCLING_PARSE_ARABIC_FALLBACK_FONT` | runs of Arabic script |
+
+Setting them is also how a render is made reproducible across machines: pin the
+same file everywhere and the substituted typeface stops being a property of the
+host. Run with `loglevel="info"` to see which file was chosen for each font.
+
 ## Performance Benchmarks
 
 [`docs/performance_benchmarks.md`](./docs/performance_benchmarks.md) compares `docling-parse` against the other widely used PDF packages on parsing and rendering, with the methodology and the commands to reproduce it.
 
-The tooling behind it lives under [`perf/`](./perf/README.md):
+The tooling behind it lives under [`scripts/benchmarking/`](./scripts/benchmarking/README.md):
 
-- [`perf/run_scaling.py`](./perf/run_scaling.py): measure — thread-scaling sweeps and the cross-package comparison
-- [`perf/run_eval.py`](./perf/run_eval.py): plot — per-page distributions, thread scaling, and per-package correlations
-- [`perf/run_analysis.py`](./perf/run_analysis.py): drill down — replay the slowest pages with C++ stage timings
+- [`scripts/benchmarking/run_performance_benchmarking.py`](./scripts/benchmarking/run_performance_benchmarking.py): measure — thread-scaling sweeps and the cross-package comparison
+- [`scripts/benchmarking/run_performance_eval.py`](./scripts/benchmarking/run_performance_eval.py): plot — per-page distributions, thread scaling, and per-package correlations
+- [`scripts/benchmarking/run_performance_analysis.py`](./scripts/benchmarking/run_performance_analysis.py): drill down — replay the slowest pages with C++ stage timings
+- [`scripts/benchmarking/run_quality_benchmarking.py`](./scripts/benchmarking/run_quality_benchmarking.py): compare quality — textual output and rendered pixels across PDF engines
 
 For historical V1 vs V2 benchmarks, see [legacy_performance_benchmarks.md](./docs/legacy/legacy_performance_benchmarks.md).
 

{docling_parse-7.14.0 → docling_parse-7.16.0}/docling_parse.egg-info/SOURCES.txt

@@ -555,6 +555,7 @@ src/parse/qpdf/logger.h
 src/parse/qpdf/object.h
 src/parse/qpdf/qpdf_compat.h
 src/parse/qpdf/stream_decoder.h
+src/parse/qpdf/stream_filters.h
 src/parse/qpdf/stream_instruction.h
 src/parse/qpdf/to_json.h
 src/parse/utils/files.h
@@ -582,7 +583,7 @@ src/render/blend2d_embedded_font_cache.h
 src/render/blend2d_font_resolver.h
 src/render/blend2d_renderer.h
 src/render/config.h
-src/render/freetype_embedded_font_cache.h
+src/render/freetype_font_cache.h
 src/render/naive_renderer.h
 src/render/template_renderer.h
 src/third_party/pdfium_jbig2.h
@@ -671,32 +672,43 @@ tests/__init__.py
 tests/conftest.py
 tests/constants.py
 tests/data_utils.py
+tests/groundtruth_io.py
 tests/pdf_builder.py
+tests/regression_page_selection.py
 tests/rendering_regression.py
-tests/test_actual_text.py
-tests/test_clipping.py
-tests/test_cmap_encoding.py
-tests/test_colorspace.py
-tests/test_colorspaces.py
-tests/test_embedded_fonts.py
-tests/test_font_fallback.py
-tests/test_font_name_resolution.py
-tests/test_gid_glyph_names.py
-tests/test_indexed_images.py
-tests/test_load_failure.py
-tests/test_locale_safety.py
-tests/test_parse.py
-tests/test_patterns.py
-tests/test_pypdfium_render.py
-tests/test_shading.py
-tests/test_shadings.py
-tests/test_standard_font_widths.py
-tests/test_text_render_modes.py
-tests/test_threaded_parse.py
-tests/test_threaded_render.py
-tests/test_tounicode_fallback.py
-tests/test_transparency.py
-tests/test_type3_fonts.py
-tests/test_vertical_writing.py
-tests/test_widget_appearances.py
-tests/update_groundtruth.py
+tests/test_regression_embedded_fonts.py
+tests/test_regression_locale_safety.py
+tests/test_regression_parse.py
+tests/test_regression_pypdfium_render.py
+tests/test_regression_threaded_parse.py
+tests/test_regression_threaded_render.py
+tests/test_unit_actual_text.py
+tests/test_unit_bitmap_artifacts.py
+tests/test_unit_ccitt_images.py
+tests/test_unit_cjk_fonts.py
+tests/test_unit_clipping.py
+tests/test_unit_cmap_encoding.py
+tests/test_unit_colorspace.py
+tests/test_unit_colorspaces.py
+tests/test_unit_font_fallback.py
+tests/test_unit_font_name_resolution.py
+tests/test_unit_gid_glyph_names.py
+tests/test_unit_glyph_proportions.py
+tests/test_unit_indexed_images.py
+tests/test_unit_jpeg_images.py
+tests/test_unit_load_failure.py
+tests/test_unit_patterns.py
+tests/test_unit_shading.py
+tests/test_unit_shadings.py
+tests/test_unit_standard_font_widths.py
+tests/test_unit_text_render_modes.py
+tests/test_unit_tounicode_fallback.py
+tests/test_unit_transparency.py
+tests/test_unit_type3_fonts.py
+tests/test_unit_unicode_paths.py
+tests/test_unit_vertical_writing.py
+tests/test_unit_widget_appearances.py
+tests/update_groundtruth.py
+tests/tools/prune_groundtruth.py
+tests/tools/recompress_groundtruth.py
+tests/tools/select_regression_pages.py

{docling_parse-7.14.0 → docling_parse-7.16.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "docling-parse"
-version = "7.14.0"
+version = "7.16.0"
 description = "Simple package to extract text with coordinates from programmatic PDFs"
 license = "MIT"
 keywords = ["docling", "pdf", "parser"]
@@ -67,8 +67,6 @@ dev = [
     "pre-commit>=3.7.1,<4.0.0",
     "mypy>=1.13.0,<2.0.0",
     "tqdm>=4.67.0,<5.0.0",
-    "boto>=2.49.0,<3.0.0",
-    "boto3>=1.35.67,<2.0.0",
     "huggingface-hub>=1.11.0",
     "tabulate>=0.9.0,<1.0.0",    
 ]
@@ -166,9 +164,10 @@ python_version = "3.10"
 [[tool.mypy.overrides]]
 module = [
     "tabulate.*",
-    "botocore.*",
-    "boto3.*",
-    "pypdfium2.*"
+    "pypdfium2.*",
+    "tqdm.*",
+    "fitz.*",
+    "fontTools.*"
 ]
 ignore_missing_imports = true
 

{docling_parse-7.14.0 → docling_parse-7.16.0}/src/parse/enums.h

@@ -76,7 +76,11 @@ namespace pdflib
     IDENTITY_H,
     IDENTITY_V,
 
-    CMAP_RESOURCES
+    // /Encoding names one of Adobe's predefined CMaps
+    CMAP_RESOURCES,
+
+    // /Encoding is a CMap program embedded in the file
+    CMAP_STREAM
   };
 
   font_encoding_name to_encoding_name(std::string name)
@@ -88,6 +92,7 @@ namespace pdflib
     else if(name=="IDENTITY_H" or name=="/Identity-H"       ) { return IDENTITY_H; }
     else if(name=="IDENTITY_V" or name=="/Identity-V"       ) { return IDENTITY_V; }
     else if(name=="CMAP_RESOURCES"                          ) { return CMAP_RESOURCES; }
+    else if(name=="CMAP_STREAM"                             ) { return CMAP_STREAM; }
     else 
       {
         LOG_S(ERROR) << __FILE__ << ":" << __LINE__ << " --> unknown encoding " << name;
@@ -106,6 +111,7 @@ namespace pdflib
       case IDENTITY_H: { return "IDENTITY_H"; } break;
       case IDENTITY_V: { return "IDENTITY_V"; } break;
       case CMAP_RESOURCES: { return "CMAP_RESOURCES"; } break;
+      case CMAP_STREAM:    { return "CMAP_STREAM"; } break;
 
       default:
         {

{docling_parse-7.14.0 → docling_parse-7.16.0}/src/parse/page_item_sanitators/page_dimension.h

@@ -303,13 +303,17 @@ namespace pdflib
 	double x1 = std::min(page_bbox[2], image_bbox[2]); // Min of top-right x
 	double y1 = std::min(page_bbox[3], image_bbox[3]); // Min of top-right y
 	
-	// Validate the intersection
-	if (x0 > x1 || y0 > y1)
+	// Validate the intersection. An empty one means the image is off-page,
+	// and a degenerate one (zero width or height) means the image matrix
+	// maps the unit square of ISO 32000-1, 8.9.5.2 onto a line or a point,
+	// so the image covers no area and paints nothing. Either way there is
+	// nothing to report as a page image.
+	if (x0 >= x1 || y0 >= y1)
 	  {
 	    LOG_S(INFO) << "deleting image: ("
 			<< itr->x0 << ", " << itr->y0 << ", " << itr->x1 << ", " << itr->y1 << ")";
 	    
-	    // No intersection, remove
+	    // No visible area, remove
 	    itr = images.erase(itr);
 	  }
 	else