fix(pptx): move hidden-shape content to invisible layer

docling/backend/mspowerpoint_backend.py

@@ -12,6 +12,21 @@
 from tempfile import mkdtemp
 from typing import Any, Callable, Final, Iterable, Iterator, Literal, Optional, Union
 
+from docling.backend.abstract_backend import (
+    DeclarativeDocumentBackend,
+    PaginatedDocumentBackend,
+)
+from docling.backend.docx.drawingml.utils import (
+    convert_to_modern_format,
+    crop_whitespace,
+    get_docx_to_pdf_converter,
+)
+from docling.backend.docx.latex.omml import oMath2Latex
+from docling.backend.utils.image import normalize_image_for_png
+from docling.datamodel.backend_options import MsPowerpointBackendOptions
+from docling.datamodel.base_models import FormatToMimeType, InputFormat
+from docling.datamodel.document import InputDocument
+from docling.exceptions import DocumentLoadError
 from docling_core.types.doc import (
     BoundingBox,
     CoordOrigin,
@@ -38,22 +53,6 @@
 from PIL import Image, UnidentifiedImageError
 from typing_extensions import override
 
-from docling.backend.abstract_backend import (
-    DeclarativeDocumentBackend,
-    PaginatedDocumentBackend,
-)
-from docling.backend.docx.drawingml.utils import (
-    convert_to_modern_format,
-    crop_whitespace,
-    get_docx_to_pdf_converter,
-)
-from docling.backend.docx.latex.omml import oMath2Latex
-from docling.backend.utils.image import normalize_image_for_png
-from docling.datamodel.backend_options import MsPowerpointBackendOptions
-from docling.datamodel.base_models import FormatToMimeType, InputFormat
-from docling.datamodel.document import InputDocument
-from docling.exceptions import DocumentLoadError
-
 _log = logging.getLogger(__name__)
 
 _PPTX_AVAILABLE: bool = False
@@ -1687,7 +1686,25 @@ def handle_groups(shape, parent_slide, slide_ind, doc, slide_size):
 
             # Loop through each shape in the slide
             for shape in self._iter_shapes_by_position(slide.shapes):
-                handle_shapes(shape, parent_slide, slide_ind, doc, slide_size)
+                if self._is_hidden_shape(shape):
+                    # Like a hidden slide, keep the content but move it to
+                    # the invisible layer instead of the body layer.
+                    seen = {
+                        id(item)
+                        for item, _ in doc.iterate_items(
+                            root=parent_slide, with_groups=True
+                        )
+                    }
+                    handle_shapes(shape, parent_slide, slide_ind, doc, slide_size)
+                    for item, _ in doc.iterate_items(
+                        root=parent_slide,
+                        with_groups=True,
+                        included_content_layers={ContentLayer.BODY},
+                    ):
+                        if id(item) not in seen:
+                            item.content_layer = ContentLayer.INVISIBLE
+                else:
+                    handle_shapes(shape, parent_slide, slide_ind, doc, slide_size)
 
             # Handle notes slide
             if slide.has_notes_slide:
@@ -1719,6 +1736,15 @@ def handle_groups(shape, parent_slide, slide_ind, doc, slide_size):
 
         return doc
 
+    def _is_hidden_shape(self, shape) -> bool:
+        """Whether a shape is hidden in PowerPoint (Selection Pane > Hide).
+
+        A shape whose ``p:cNvPr`` carries ``hidden="1"`` is never presented,
+        so its content belongs on the invisible layer, like a hidden slide.
+        """
+        cNvPr = shape._element.find(".//p:cNvPr", namespaces=self.NAMESPACES)
+        return cNvPr is not None and cNvPr.get("hidden") == "1"
+
     @staticmethod
     def _hide_slide(doc: DoclingDocument, parent_slide: GroupItem) -> None:
         """Move the content of a hidden slide to the invisible layer.

tests/test_backend_pptx.py

@@ -12,6 +12,14 @@
 from types import SimpleNamespace
 
 import pytest
+from docling.backend.docx.drawingml.utils import get_libreoffice_cmd
+from docling.datamodel.backend_options import MsPowerpointBackendOptions
+from docling.datamodel.base_models import InputFormat, ItemAndImageEnrichmentElement
+from docling.datamodel.document import ConversionResult, DoclingDocument, InputDocument
+from docling.datamodel.pipeline_options import ConvertPipelineOptions
+from docling.document_converter import DocumentConverter, PowerpointFormatOption
+from docling.models.base_model import BaseItemAndImageEnrichmentModel
+from docling.pipeline.simple_pipeline import SimplePipeline
 from docling_core.types.doc import (
     ContentLayer,
     GroupItem,
@@ -21,20 +29,12 @@
     TextItem,
 )
 
-from docling.backend.docx.drawingml.utils import get_libreoffice_cmd
 from docling.backend.mspowerpoint_backend import (
     _MAX_CHART_TABLE_CELLS,
     MsPowerpointDocumentBackend,
     _is_metafile,
     _last_populated_row,
 )
-from docling.datamodel.backend_options import MsPowerpointBackendOptions
-from docling.datamodel.base_models import InputFormat, ItemAndImageEnrichmentElement
-from docling.datamodel.document import ConversionResult, DoclingDocument, InputDocument
-from docling.datamodel.pipeline_options import ConvertPipelineOptions
-from docling.document_converter import DocumentConverter, PowerpointFormatOption
-from docling.models.base_model import BaseItemAndImageEnrichmentModel
-from docling.pipeline.simple_pipeline import SimplePipeline
 
 from .test_data_gen_flag import GEN_TEST_DATA
 from .verify_utils import verify_document, verify_export
@@ -1345,6 +1345,42 @@ def texts(layer: ContentLayer) -> list[str]:
     ]
 
 
+def test_pptx_hidden_shape_goes_to_invisible_layer(tmp_path: Path):
+    """A shape hidden in PowerPoint (``p:cNvPr/@hidden="1"``) on a shown slide
+    is left out of the default export.
+
+    Its content stays available in ``ContentLayer.INVISIBLE``, like a hidden
+    slide.
+    """
+    from pptx import Presentation
+    from pptx.util import Inches
+
+    prs = Presentation()
+    slide = prs.slides.add_slide(prs.slide_layouts[6])
+    slide.shapes.add_textbox(
+        Inches(1), Inches(1), Inches(6), Inches(1)
+    ).text_frame.text = "Shown shape text"
+    hidden = slide.shapes.add_textbox(Inches(1), Inches(3), Inches(6), Inches(1))
+    hidden.text_frame.text = "Hidden shape text"
+    hidden._element.nvSpPr.cNvPr.set("hidden", "1")
+
+    pptx_path = tmp_path / "hidden_shape.pptx"
+    prs.save(pptx_path)
+
+    doc = convert_with_pptx_backend(pptx_path)
+
+    def texts(layer: ContentLayer) -> list[str]:
+        return [
+            item.text
+            for item, _ in doc.iterate_items(included_content_layers={layer})
+            if isinstance(item, TextItem)
+        ]
+
+    assert "Hidden shape text" not in doc.export_to_markdown()
+    assert "Shown shape text" in doc.export_to_markdown()
+    assert texts(ContentLayer.INVISIBLE) == ["Hidden shape text"]
+
+
 @pytest.mark.parametrize(
     ("layout_marker", "override_layer", "expected"),
     [