fix malformed pptx issue (#761)

* fix malformed pptx issue

Added a new test to check for the ability to partition a malformed PowerPoint file. Modified the `partition_pptx` function to skip processing shapes that are not on the actual slide, but only if they have top and left positions. Also modified `_order_shapes` function to handle cases where shapes do not have top or left positions.

* update changelog

* fix lint issue SIM102 nested ifs

* fix black linting

CHANGELOG.md

@@ -17,6 +17,8 @@
 * Page number defaults to `None` instead of `1` when page number is not present in the metadata.
   A page number of `None` indicates that page numbers are not being tracked for the document
   or that page numbers do not apply to the element in question..
+* Fixes an issue with some pptx files. Assume pptx shapes are found in top left position of slide
+  in case the shape.top and shape.left attributes are `None`.
 
 ## 0.7.5
 

test_unstructured/partition/test_pptx.py

@@ -190,3 +190,11 @@ def test_partition_pptx_many_pages():
 
     # The page_number of PageBreak is None
     assert set(filter(None, (elt.metadata.page_number for elt in elements))) == {1, 2}
+
+
+def test_partition_pptx_malformed():
+    filename = os.path.join(EXAMPLE_DOCS_DIRECTORY, "fake-power-point-malformed.pptx")
+    elements = partition_pptx(filename=filename)
+
+    assert elements[0].text == "Problem Date Placeholder"
+    assert elements[1].text == "Test Slide"

unstructured/partition/pptx.py

@@ -84,7 +84,8 @@ def partition_pptx(
             if not shape.has_text_frame:
                 continue
             # NOTE(robinson) - avoid processing shapes that are not on the actual slide
-            if shape.top < 0 or shape.left < 0:
+            # NOTE - skip check if no top or left position (shape displayed top left)
+            if (shape.top and shape.left) and (shape.top < 0 or shape.left < 0):
                 continue
             for paragraph in shape.text_frame.paragraphs:
                 text = paragraph.text
@@ -107,7 +108,7 @@ def partition_pptx(
 
 def _order_shapes(shapes):
     """Orders the shapes from top to bottom and left to right."""
-    return sorted(shapes, key=lambda x: (x.top, x.left))
+    return sorted(shapes, key=lambda x: (x.top or 0, x.left or 0))
 
 
 def _is_bulleted_paragraph(paragraph) -> bool: