docling-parse 7.7.0__tar.gz → 7.8.0__tar.gz

{docling_parse-7.7.0 → docling_parse-7.8.0}/CHANGELOG.md

@@ -1,3 +1,16 @@
+## [v7.8.0](https://github.com/docling-project/docling-parse/releases/tag/v7.8.0) - 2026-07-10
+
+### Feature
+
+* Adding new top-level methods for page content geometry queries ([#297](https://github.com/docling-project/docling-parse/issues/297)) ([`ef41562`](https://github.com/docling-project/docling-parse/commit/ef415624593d546806423e72fb2b812d811d8161))
+
+## [v7.7.1](https://github.com/docling-project/docling-parse/releases/tag/v7.7.1) - 2026-07-10
+
+### Fix
+
+* Fixing the word extraction with spaces as blockers ([#296](https://github.com/docling-project/docling-parse/issues/296)) ([`66e4c33`](https://github.com/docling-project/docling-parse/commit/66e4c339578679fb13410d91261ad5160af5220a))
+* Arabic pdf-rendering ([#295](https://github.com/docling-project/docling-parse/issues/295)) ([`ff00fcf`](https://github.com/docling-project/docling-parse/commit/ff00fcfe7e59a5a81c24fb01332a023c75cc8f0f))
+
 ## [v7.7.0](https://github.com/docling-project/docling-parse/releases/tag/v7.7.0) - 2026-07-08
 
 ### Feature

{docling_parse-7.7.0 → docling_parse-7.8.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling-parse
-Version: 7.7.0
+Version: 7.8.0
 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

{docling_parse-7.7.0 → docling_parse-7.8.0}/app/pybind_parse.cpp

@@ -569,6 +569,21 @@ PYBIND11_MODULE(pdf_parsers, m) {
 	 "Check if word cells have been created")
     .def("has_line_cells", &pdflib::pdf_decoder<pdflib::PAGE>::has_line_cells,
 	 "Check if line cells have been created")
+    .def("intersects_with", &pdflib::pdf_decoder<pdflib::PAGE>::intersects_with,
+         pybind11::arg("bbox"),
+         pybind11::arg("chars") = false,
+         pybind11::arg("shapes") = true,
+         pybind11::arg("bitmaps") = true,
+         "Check whether visible chars, shapes, or bitmaps intersect [left, bottom, right, top]")
+    .def("get_shape_lines", &pdflib::pdf_decoder<pdflib::PAGE>::get_shape_lines,
+         pybind11::arg("horizontal") = true,
+         pybind11::arg("vertical") = true,
+         pybind11::arg("tolerance") = 1e-3,
+         "Return visible horizontal and/or vertical stroked shape segments as [left, bottom, right, top]")
+    .def("get_connected_shape_bounding_boxes",
+         &pdflib::pdf_decoder<pdflib::PAGE>::get_connected_shape_bounding_boxes,
+         pybind11::arg("tolerance") = 0.0,
+         "Return bboxes of visible shapes connected through overlapping bboxes")
     .def("get_timings", [](pdflib::pdf_decoder<pdflib::PAGE>& self) {
 	   // Return as Dict[str, float] (sums) for backward compatibility
 	   return self.get_timings().to_sum_map();

{docling_parse-7.7.0 → docling_parse-7.8.0}/docling_parse/pdf_parser.py

@@ -1221,6 +1221,62 @@ class PageParseResult:
         """Return structured timing data for this page parse."""
         return self._timings
 
+    def intersects_with(
+        self,
+        *,
+        bbox: BoundingBox,
+        chars: bool = False,
+        shapes: bool = True,
+        bitmaps: bool = True,
+    ) -> bool:
+        """Return whether visible page content intersects bbox.
+
+        bbox may use top-left or bottom-left coordinates. Native code expects
+        page coordinates with bottom-left origin as [left, bottom, right, top].
+        """
+        bbox_bl = bbox.to_bottom_left_origin(page_height=self.page_height)
+        left = min(bbox_bl.l, bbox_bl.r)
+        right = max(bbox_bl.l, bbox_bl.r)
+        bottom = min(bbox_bl.b, bbox_bl.t)
+        top = max(bbox_bl.b, bbox_bl.t)
+
+        return self._require_page_decoder().intersects_with(
+            [left, bottom, right, top],
+            chars=chars,
+            shapes=shapes,
+            bitmaps=bitmaps,
+        )
+
+    def get_shape_lines(
+        self,
+        *,
+        horizontal: bool = True,
+        vertical: bool = True,
+        tolerance: float = 1e-3,
+    ) -> List[BoundingBox]:
+        """Return visible horizontal and/or vertical stroked shape segments."""
+        return [
+            _to_bounding_box(tuple(bbox))
+            for bbox in self._require_page_decoder().get_shape_lines(
+                horizontal=horizontal,
+                vertical=vertical,
+                tolerance=tolerance,
+            )
+        ]
+
+    def get_connected_shape_bounding_boxes(
+        self,
+        *,
+        tolerance: float = 0.0,
+    ) -> List[BoundingBox]:
+        """Return bboxes of visible shapes connected by overlapping bboxes."""
+        return [
+            _to_bounding_box(tuple(bbox))
+            for bbox in self._require_page_decoder().get_connected_shape_bounding_boxes(
+                tolerance=tolerance,
+            )
+        ]
+
     def _rendering_config(self) -> RenderConfig:
         if self._render_config is None:
             raise RuntimeError(

{docling_parse-7.7.0 → docling_parse-7.8.0}/docling_parse.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling-parse
-Version: 7.7.0
+Version: 7.8.0
 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

{docling_parse-7.7.0 → docling_parse-7.8.0}/docling_parse.egg-info/SOURCES.txt

@@ -662,10 +662,12 @@ src/third_party/pdfium_jbig2/core/fxcrt/numerics/safe_math_shared_impl.h
 tests/README.md
 tests/__init__.py
 tests/conftest.py
+tests/constants.py
 tests/data_utils.py
 tests/rendering_regression.py
 tests/test_embedded_fonts.py
 tests/test_locale_safety.py
 tests/test_parse.py
 tests/test_threaded_parse.py
-tests/test_threaded_render.py
+tests/test_threaded_render.py
+tests/update_groundtruth.py

{docling_parse-7.7.0 → docling_parse-7.8.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "docling-parse"
-version = "7.7.0"
+version = "7.8.0"
 description = "Simple package to extract text with coordinates from programmatic PDFs"
 license = "MIT"
 keywords = ["docling", "pdf", "parser"]

{docling_parse-7.7.0 → docling_parse-7.8.0}/src/parse/page_item_sanitators/cells.h

@@ -31,7 +31,8 @@ namespace pdflib
 		       double horizontal_cell_tolerance, //=1.0,
 		       bool enforce_same_font, //=true,
 		       double space_width_factor_for_merge, //=1.5,
-		       double space_width_factor_for_merge_with_space); //=0.33);
+		       double space_width_factor_for_merge_with_space, //=0.33,
+		       bool block_spaces); // when true, space cells act as hard merge barriers
 
     void sanitize_text(page_item<PAGE_CELLS>& cells);
 
@@ -41,27 +42,31 @@ namespace pdflib
 
     bool applicable_for_merge(page_item<PAGE_CELL>& cell_i,
 			      page_item<PAGE_CELL>& cell_j,
-			      bool enforce_same_font);
+			      bool enforce_same_font,
+			      bool block_spaces);
 
     void contract_cells_into_lines_right_to_left(page_item<PAGE_CELLS>& cells,
 						 double horizontal_cell_tolerance,
 						 bool enforce_same_font,
 						 double space_width_factor_for_merge,
-						 double space_width_factor_for_merge_with_space);
+						 double space_width_factor_for_merge_with_space,
+						 bool block_spaces);
 
     void contract_cells_into_lines_left_to_right(page_item<PAGE_CELLS>& cells,
 						 double horizontal_cell_tolerance,
 						 bool enforce_same_font,
 						 double space_width_factor_for_merge,
 						 double space_width_factor_for_merge_with_space,
+						 bool block_spaces,
 						 bool allow_reverse);
-    
+
     // linear
     void contract_cells_into_lines_v1(page_item<PAGE_CELLS>& cells,
 				      double horizontal_cell_tolerance=1.0,
 				      bool enforce_same_font=true,
 				      double space_width_factor_for_merge=1.5,
-				      double space_width_factor_for_merge_with_space=0.33);
+				      double space_width_factor_for_merge_with_space=0.33,
+				      bool block_spaces=false);
 
     // quadratic
     void contract_cells_into_lines_v2(page_item<PAGE_CELLS>& cells,
@@ -140,7 +145,23 @@ namespace pdflib
 
     LOG_S(INFO) << "#-char cells: " << word_cells.size();
 
-    // remove all spaces
+    // Keep the space cells in place and let sanitize_bbox treat them as hard
+    // word-boundary barriers (block_spaces=true). An explicit space glyph is a
+    // far more reliable word separator than the geometric-gap heuristic, which
+    // is ambiguous for tightly-set fonts with narrow spaces. The spaces are
+    // erased afterwards, once the words have been contracted.
+
+    // > space_width_factor_for_merge, so nothing gets merged with a space
+    double space_width_factor_for_merge_with_space = 2.0*config.word_space_width_factor_for_merge;
+
+    sanitize_bbox(word_cells,
+		  config.horizontal_cell_tolerance,
+		  config.enforce_same_font,
+		  config.word_space_width_factor_for_merge,
+		  space_width_factor_for_merge_with_space,
+		  true);
+
+    // remove the space cells that acted as word-boundary barriers
     auto itr = word_cells.begin();
     while(itr!=word_cells.end())
       {
@@ -154,17 +175,6 @@ namespace pdflib
 	  }
       }
 
-    LOG_S(INFO) << "#-char cells (without spaces): " << word_cells.size();
-
-    // > space_width_factor_for_merge, so nothing gets merged with a space
-    double space_width_factor_for_merge_with_space = 2.0*config.word_space_width_factor_for_merge;
-
-    sanitize_bbox(word_cells,
-		  config.horizontal_cell_tolerance,
-		  config.enforce_same_font,
-		  config.word_space_width_factor_for_merge,
-		  space_width_factor_for_merge_with_space);
-
     LOG_S(INFO) << "#-word cells: " << word_cells.size();
 
     //return to_records(word_cells);
@@ -184,12 +194,14 @@ namespace pdflib
 
     LOG_S(INFO) << "# char-cells: " << line_cells.size();
 
+    // lines keep their internal spaces, so spaces are merged normally (block_spaces=false)
     sanitize_bbox(line_cells,
 		  config.horizontal_cell_tolerance,
 		  config.enforce_same_font,
 		  config.line_space_width_factor_for_merge,
-		  config.line_space_width_factor_for_merge_with_space);
-    
+		  config.line_space_width_factor_for_merge_with_space,
+		  false);
+
     LOG_S(INFO) << "# line-cells: " << line_cells.size();
     
     //return to_records(line_cells);
@@ -373,37 +385,40 @@ namespace pdflib
 						double horizontal_cell_tolerance,
 						bool enforce_same_font,
 						double space_width_factor_for_merge,
-						double space_width_factor_for_merge_with_space)
+						double space_width_factor_for_merge_with_space,
+						bool block_spaces)
   {
     contract_cells_into_lines_v1(cells,
 				 horizontal_cell_tolerance,
 				 enforce_same_font,
 				 space_width_factor_for_merge,
-				 space_width_factor_for_merge_with_space);
-
-    /*
-    contract_cells_into_lines_v2(cells,
-				 horizontal_cell_tolerance,
-				 enforce_same_font,
-				 space_width_factor_for_merge,
-				 space_width_factor_for_merge_with_space);
-    */
+				 space_width_factor_for_merge_with_space,
+				 block_spaces);
   }
 
   bool page_item_sanitator<PAGE_CELLS>::applicable_for_merge(page_item<PAGE_CELL>& cell_i,
 						       page_item<PAGE_CELL>& cell_j,
-						       bool enforce_same_font)
+						       bool enforce_same_font,
+						       bool block_spaces)
   {
     if(not cell_i.active)
       {
 	return false;
       }
-    
+
     if(not cell_j.active)
       {
 	return false;
       }
-    
+
+    // An explicit space glyph is a hard word boundary: never merge a space cell
+    // with anything, nor across one. The space cells are removed afterwards.
+    if(block_spaces and
+       (utils::string::is_space(cell_i.text) or utils::string::is_space(cell_j.text)))
+      {
+	return false;
+      }
+
     if(enforce_same_font and cell_i.font_name!=cell_j.font_name)
       {
 	// Exception: ligature glyphs are often encoded in a different font than
@@ -427,13 +442,14 @@ namespace pdflib
 							       double horizontal_cell_tolerance,
 							       bool enforce_same_font,
 							       double space_width_factor_for_merge,
-							       double space_width_factor_for_merge_with_space)
+							       double space_width_factor_for_merge_with_space,
+							       bool block_spaces)
   {
-    contract_cells_into_lines_left_to_right(cells, horizontal_cell_tolerance, enforce_same_font, space_width_factor_for_merge, space_width_factor_for_merge_with_space, false);
-    
-    contract_cells_into_lines_right_to_left(cells, horizontal_cell_tolerance, enforce_same_font, space_width_factor_for_merge, space_width_factor_for_merge_with_space);
-    
-    contract_cells_into_lines_left_to_right(cells, horizontal_cell_tolerance, enforce_same_font, space_width_factor_for_merge, space_width_factor_for_merge_with_space, true);
+    contract_cells_into_lines_left_to_right(cells, horizontal_cell_tolerance, enforce_same_font, space_width_factor_for_merge, space_width_factor_for_merge_with_space, block_spaces, false);
+
+    contract_cells_into_lines_right_to_left(cells, horizontal_cell_tolerance, enforce_same_font, space_width_factor_for_merge, space_width_factor_for_merge_with_space, block_spaces);
+
+    contract_cells_into_lines_left_to_right(cells, horizontal_cell_tolerance, enforce_same_font, space_width_factor_for_merge, space_width_factor_for_merge_with_space, block_spaces, true);
   }
   
   void page_item_sanitator<PAGE_CELLS>::contract_cells_into_lines_left_to_right(page_item<PAGE_CELLS>& cells,
@@ -441,6 +457,7 @@ namespace pdflib
 									  bool enforce_same_font,
 									  double space_width_factor_for_merge,
 									  double space_width_factor_for_merge_with_space,
+									  bool block_spaces,
 									  bool allow_reverse)
   {
     // take care for left to right printing
@@ -454,7 +471,7 @@ namespace pdflib
 
 	for(int j=i+1; j<cells.size(); j++)
 	  {
-	    if(not applicable_for_merge(cells[i], cells[j], enforce_same_font))
+	    if(not applicable_for_merge(cells[i], cells[j], enforce_same_font, block_spaces))
 	      {
 		break;
 	      }
@@ -528,8 +545,9 @@ namespace pdflib
 									  double horizontal_cell_tolerance,
 									  bool enforce_same_font,
 									  double space_width_factor_for_merge,
-									  double space_width_factor_for_merge_with_space)
-  {    
+									  double space_width_factor_for_merge_with_space,
+									  bool block_spaces)
+  {
     // take care for right to left printing
     for(int i=cells.size()-1; i>=0; i--)
       {
@@ -541,7 +559,7 @@ namespace pdflib
 
 	for(int j=i-1; j>=0; j--)
 	  {
-	    if(not applicable_for_merge(cells[i], cells[j], enforce_same_font))
+	    if(not applicable_for_merge(cells[i], cells[j], enforce_same_font, block_spaces))
 	      {
 		break;
 	      }