docling 2.41.0__tar.gz → 2.42.0__tar.gz

{docling-2.41.0 → docling-2.42.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling
-Version: 2.41.0
+Version: 2.42.0
 Summary: SDK and CLI for parsing PDF, DOCX, HTML, and more, to a unified document representation for powering downstream workflows such as gen AI applications.
 Author-email: Christoph Auer <cau@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Maxim Lysak <mly@zurich.ibm.com>, Nikos Livathinos <nli@zurich.ibm.com>, Ahmed Nassar <ahn@zurich.ibm.com>, Panos Vagenas <pva@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>
 License-Expression: MIT
@@ -50,6 +50,7 @@ Requires-Dist: tqdm<5.0.0,>=4.65.0
 Requires-Dist: pluggy<2.0.0,>=1.0.0
 Requires-Dist: pylatexenc<3.0,>=2.10
 Requires-Dist: scipy<2.0.0,>=1.6.0
+Requires-Dist: accelerate<2,>=1.0.0
 Provides-Extra: tesserocr
 Requires-Dist: tesserocr<3.0.0,>=2.7.1; extra == "tesserocr"
 Provides-Extra: ocrmac

{docling-2.41.0 → docling-2.42.0}/docling/backend/docx/latex/omml.py

@@ -260,7 +260,15 @@ class oMath2Latex(Tag2Method):
         the fraction object
         """
         c_dict = self.process_children_dict(elm)
-        pr = c_dict["fPr"]
+        pr = c_dict.get("fPr")
+        if pr is None:
+            # Handle missing fPr element gracefully
+            _log.debug("Missing fPr element in fraction, using default formatting")
+            latex_s = F_DEFAULT
+            return latex_s.format(
+                num=c_dict.get("num"),
+                den=c_dict.get("den"),
+            )
         latex_s = get_val(pr.type, default=F_DEFAULT, store=F)
         return pr.text + latex_s.format(num=c_dict.get("num"), den=c_dict.get("den"))
 

{docling-2.41.0 → docling-2.42.0}/docling/backend/html_backend.py

@@ -379,6 +379,25 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
         else:
             _log.debug(f"list-item has no text: {element}")
 
+    @staticmethod
+    def _get_cell_spans(cell: Tag) -> tuple[int, int]:
+        """Extract colspan and rowspan values from a table cell tag.
+
+        This function retrieves the 'colspan' and 'rowspan' attributes from a given
+        table cell tag.
+        If the attribute does not exist or it is not numeric, it defaults to 1.
+        """
+        raw_spans: tuple[str, str] = (
+            str(cell.get("colspan", "1")),
+            str(cell.get("rowspan", "1")),
+        )
+        int_spans: tuple[int, int] = (
+            int(raw_spans[0]) if raw_spans[0].isnumeric() else 1,
+            int(raw_spans[1]) if raw_spans[0].isnumeric() else 1,
+        )
+
+        return int_spans
+
     @staticmethod
     def parse_table_data(element: Tag) -> Optional[TableData]:  # noqa: C901
         nested_tables = element.find("table")
@@ -398,10 +417,9 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
                 if not isinstance(row, Tag):
                     continue
                 cell_tag = cast(Tag, cell)
-                val = cell_tag.get("colspan", "1")
-                colspan = int(val) if (isinstance(val, str) and val.isnumeric()) else 1
-                col_count += colspan
-                if cell_tag.name == "td" or cell_tag.get("rowspan") is None:
+                col_span, row_span = HTMLDocumentBackend._get_cell_spans(cell_tag)
+                col_count += col_span
+                if cell_tag.name == "td" or row_span == 1:
                     is_row_header = False
             num_cols = max(num_cols, col_count)
             if not is_row_header:
@@ -428,10 +446,11 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
             row_header = True
             for html_cell in cells:
                 if isinstance(html_cell, Tag):
+                    _, row_span = HTMLDocumentBackend._get_cell_spans(html_cell)
                     if html_cell.name == "td":
                         col_header = False
                         row_header = False
-                    elif html_cell.get("rowspan") is None:
+                    elif row_span == 1:
                         row_header = False
             if not row_header:
                 row_idx += 1
@@ -456,18 +475,7 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
                 text = html_cell.text
 
                 # label = html_cell.name
-                col_val = html_cell.get("colspan", "1")
-                col_span = (
-                    int(col_val)
-                    if isinstance(col_val, str) and col_val.isnumeric()
-                    else 1
-                )
-                row_val = html_cell.get("rowspan", "1")
-                row_span = (
-                    int(row_val)
-                    if isinstance(row_val, str) and row_val.isnumeric()
-                    else 1
-                )
+                col_span, row_span = HTMLDocumentBackend._get_cell_spans(html_cell)
                 if row_header:
                     row_span -= 1
                 while (

{docling-2.41.0 → docling-2.42.0}/docling/backend/xml/jats_backend.py

@@ -93,8 +93,8 @@ class JatsDocumentBackend(DeclarativeDocumentBackend):
 
         # Initialize the root of the document hierarchy
         self.root: Optional[NodeItem] = None
-
-        self.valid = False
+        self.hlevel: int = 0
+        self.valid: bool = False
         try:
             if isinstance(self.path_or_stream, BytesIO):
                 self.path_or_stream.seek(0)
@@ -147,6 +147,7 @@ class JatsDocumentBackend(DeclarativeDocumentBackend):
                 binary_hash=self.document_hash,
             )
             doc = DoclingDocument(name=self.file.stem or "file", origin=origin)
+            self.hlevel = 0
 
             # Get metadata XML components
             xml_components: XMLComponents = self._parse_metadata()
@@ -304,7 +305,9 @@ class JatsDocumentBackend(DeclarativeDocumentBackend):
             title: str = abstract["label"] or DEFAULT_HEADER_ABSTRACT
             if not text:
                 continue
-            parent = doc.add_heading(parent=self.root, text=title)
+            parent = doc.add_heading(
+                parent=self.root, text=title, level=self.hlevel + 1
+            )
             doc.add_text(
                 parent=parent,
                 text=text,
@@ -637,7 +640,10 @@ class JatsDocumentBackend(DeclarativeDocumentBackend):
                 elif child.tag == "ack":
                     text = DEFAULT_HEADER_ACKNOWLEDGMENTS
                 if text:
-                    new_parent = doc.add_heading(text=text, parent=parent)
+                    self.hlevel += 1
+                    new_parent = doc.add_heading(
+                        text=text, parent=parent, level=self.hlevel
+                    )
             elif child.tag == "list":
                 new_parent = doc.add_group(
                     label=GroupLabel.LIST, name="list", parent=parent
@@ -694,6 +700,8 @@ class JatsDocumentBackend(DeclarativeDocumentBackend):
                 new_text = self._walk_linear(doc, new_parent, child)
                 if not (node.getparent().tag == "p" and node.tag in flush_tags):
                     node_text += new_text
+                if child.tag in ("sec", "ack") and text:
+                    self.hlevel -= 1
 
             # pick up the tail text
             node_text += child.tail.replace("\n", " ") if child.tail else ""

{docling-2.41.0 → docling-2.42.0}/docling/datamodel/pipeline_options.py

@@ -217,7 +217,7 @@ smolvlm_picture_description = PictureDescriptionVlmOptions(
 
 # GraniteVision
 granite_picture_description = PictureDescriptionVlmOptions(
-    repo_id="ibm-granite/granite-vision-3.2-2b-preview",
+    repo_id="ibm-granite/granite-vision-3.3-2b",
     prompt="What is shown in this image?",
 )
 
@@ -279,6 +279,9 @@ class LayoutOptions(BaseModel):
     """Options for layout processing."""
 
     create_orphan_clusters: bool = True  # Whether to create clusters for orphaned cells
+    keep_empty_clusters: bool = (
+        False  # Whether to keep clusters that contain no text cells
+    )
     model_spec: LayoutModelConfig = DOCLING_LAYOUT_V2
 
 

{docling-2.41.0 → docling-2.42.0}/docling/document_converter.py

@@ -1,6 +1,7 @@
 import hashlib
 import logging
 import sys
+import threading
 import time
 from collections.abc import Iterable, Iterator
 from functools import partial
@@ -49,6 +50,7 @@ from docling.pipeline.standard_pdf_pipeline import StandardPdfPipeline
 from docling.utils.utils import chunkify
 
 _log = logging.getLogger(__name__)
+_PIPELINE_CACHE_LOCK = threading.Lock()
 
 
 class FormatOption(BaseModel):
@@ -315,17 +317,18 @@ class DocumentConverter:
         # Use a composite key to cache pipelines
         cache_key = (pipeline_class, options_hash)
 
-        if cache_key not in self.initialized_pipelines:
-            _log.info(
-                f"Initializing pipeline for {pipeline_class.__name__} with options hash {options_hash}"
-            )
-            self.initialized_pipelines[cache_key] = pipeline_class(
-                pipeline_options=pipeline_options
-            )
-        else:
-            _log.debug(
-                f"Reusing cached pipeline for {pipeline_class.__name__} with options hash {options_hash}"
-            )
+        with _PIPELINE_CACHE_LOCK:
+            if cache_key not in self.initialized_pipelines:
+                _log.info(
+                    f"Initializing pipeline for {pipeline_class.__name__} with options hash {options_hash}"
+                )
+                self.initialized_pipelines[cache_key] = pipeline_class(
+                    pipeline_options=pipeline_options
+                )
+            else:
+                _log.debug(
+                    f"Reusing cached pipeline for {pipeline_class.__name__} with options hash {options_hash}"
+                )
 
         return self.initialized_pipelines[cache_key]
 

{docling-2.41.0 → docling-2.42.0}/docling/models/picture_description_vlm_model.py

@@ -65,6 +65,7 @@ class PictureDescriptionVlmModel(
                 self.processor = AutoProcessor.from_pretrained(artifacts_path)
                 self.model = AutoModelForVision2Seq.from_pretrained(
                     artifacts_path,
+                    device_map=self.device,
                     torch_dtype=torch.bfloat16,
                     _attn_implementation=(
                         "flash_attention_2"
@@ -72,7 +73,7 @@ class PictureDescriptionVlmModel(
                         and accelerator_options.cuda_use_flash_attention2
                         else "eager"
                     ),
-                ).to(self.device)
+                )
 
             self.provenance = f"{self.options.repo_id}"
 

{docling-2.41.0 → docling-2.42.0}/docling/utils/layout_postprocessor.py

@@ -267,8 +267,9 @@ class LayoutPostprocessor:
         # Initial cell assignment
         clusters = self._assign_cells_to_clusters(clusters)
 
-        # Remove clusters with no cells
-        clusters = [cluster for cluster in clusters if cluster.cells]
+        # Remove clusters with no cells (if keep_empty_clusters is False)
+        if not self.options.keep_empty_clusters:
+            clusters = [cluster for cluster in clusters if cluster.cells]
 
         # Handle orphaned cells
         unassigned = self._find_unassigned_cells(clusters)

{docling-2.41.0 → docling-2.42.0}/docling.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling
-Version: 2.41.0
+Version: 2.42.0
 Summary: SDK and CLI for parsing PDF, DOCX, HTML, and more, to a unified document representation for powering downstream workflows such as gen AI applications.
 Author-email: Christoph Auer <cau@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Maxim Lysak <mly@zurich.ibm.com>, Nikos Livathinos <nli@zurich.ibm.com>, Ahmed Nassar <ahn@zurich.ibm.com>, Panos Vagenas <pva@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>
 License-Expression: MIT
@@ -50,6 +50,7 @@ Requires-Dist: tqdm<5.0.0,>=4.65.0
 Requires-Dist: pluggy<2.0.0,>=1.0.0
 Requires-Dist: pylatexenc<3.0,>=2.10
 Requires-Dist: scipy<2.0.0,>=1.6.0
+Requires-Dist: accelerate<2,>=1.0.0
 Provides-Extra: tesserocr
 Requires-Dist: tesserocr<3.0.0,>=2.7.1; extra == "tesserocr"
 Provides-Extra: ocrmac

{docling-2.41.0 → docling-2.42.0}/docling.egg-info/requires.txt

@@ -23,6 +23,7 @@ tqdm<5.0.0,>=4.65.0
 pluggy<2.0.0,>=1.0.0
 pylatexenc<3.0,>=2.10
 scipy<2.0.0,>=1.6.0
+accelerate<2,>=1.0.0
 
 [asr]
 openai-whisper>=20250625

{docling-2.41.0 → docling-2.42.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "docling"
-version = "2.41.0"  # DO NOT EDIT, updated automatically
+version = "2.42.0"  # DO NOT EDIT, updated automatically
 description = "SDK and CLI for parsing PDF, DOCX, HTML, and more, to a unified document representation for powering downstream workflows such as gen AI applications."
 license = "MIT"
 keywords = [
@@ -70,6 +70,7 @@ dependencies = [
   'scipy (>=1.6.0,<2.0.0)',
   # 'scipy (>=1.6.0,<2.0.0) ; python_version >= "3.10"',
   # 'scipy (>=1.6.0,<1.14.0) ; python_version < "3.10"',
+  "accelerate>=1.0.0,<2",
 ]
 
 [project.urls]

{docling-2.41.0 → docling-2.42.0}/tests/test_backend_jats.py

@@ -14,9 +14,9 @@ from .verify_utils import verify_document, verify_export
 GENERATE = GEN_TEST_DATA
 
 
-def get_pubmed_paths():
-    directory = Path(os.path.dirname(__file__) + "/data/pubmed/")
-    xml_files = sorted(directory.rglob("*.xml"))
+def get_jats_paths():
+    directory = Path(os.path.dirname(__file__) + "/data/jats/")
+    xml_files = sorted(directory.rglob("*.nxml"))
     return xml_files
 
 
@@ -25,20 +25,20 @@ def get_converter():
     return converter
 
 
-def test_e2e_pubmed_conversions(use_stream=False):
-    pubmed_paths = get_pubmed_paths()
+def test_e2e_jats_conversions(use_stream=False):
+    jats_paths = get_jats_paths()
     converter = get_converter()
 
-    for pubmed_path in pubmed_paths:
+    for jats_path in jats_paths:
         gt_path = (
-            pubmed_path.parent.parent / "groundtruth" / "docling_v2" / pubmed_path.name
+            jats_path.parent.parent / "groundtruth" / "docling_v2" / jats_path.name
         )
         if use_stream:
-            buf = BytesIO(pubmed_path.open("rb").read())
-            stream = DocumentStream(name=pubmed_path.name, stream=buf)
+            buf = BytesIO(jats_path.open("rb").read())
+            stream = DocumentStream(name=jats_path.name, stream=buf)
             conv_result: ConversionResult = converter.convert(stream)
         else:
-            conv_result: ConversionResult = converter.convert(pubmed_path)
+            conv_result: ConversionResult = converter.convert(jats_path)
         doc: DoclingDocument = conv_result.document
 
         pred_md: str = doc.export_to_markdown()
@@ -54,9 +54,9 @@ def test_e2e_pubmed_conversions(use_stream=False):
         assert verify_document(doc, str(gt_path) + ".json", GENERATE), "export to json"
 
 
-def test_e2e_pubmed_conversions_stream():
-    test_e2e_pubmed_conversions(use_stream=True)
+def test_e2e_jats_conversions_stream():
+    test_e2e_jats_conversions(use_stream=True)
 
 
-def test_e2e_pubmed_conversions_no_stream():
-    test_e2e_pubmed_conversions(use_stream=False)
+def test_e2e_jats_conversions_no_stream():
+    test_e2e_jats_conversions(use_stream=False)