fix: preserve the order of shapes in partition_pptx output (#193)

* order the shapes top to bottom and left to right

* added tests for ordering

* update change log and bump version

* more tests

* don't need enumerate

* n -> on

Author: MthwRobinson

CHANGELOG.md

@@ -1,4 +1,4 @@
-## 0.4.5-dev5
+## 0.4.6
 
 * Loosen the default cap threshold to `0.5`.
 * Add a `UNSTRUCTURED_NARRATIVE_TEXT_CAP_THRESHOLD` environment variable for controlling
@@ -13,6 +13,7 @@
 * Checks that titles and narrative text are at least 50% alpha characters.
 * Restricts titles to a maximum word length. Adds a `UNSTRUCTURED_TITLE_MAX_WORD_LENGTH`
   environment variable for controlling the max number of words in a title.
+* Updated `partition_pptx` to order the elements on the page
 
 ## 0.4.4
 

test_unstructured/partition/test_pptx.py

@@ -2,8 +2,10 @@
 import pathlib
 import pytest
 
+import pptx
+
 from unstructured.partition.pptx import partition_pptx
-from unstructured.documents.elements import ListItem, NarrativeText, Title
+from unstructured.documents.elements import ListItem, NarrativeText, Text, Title
 
 DIRECTORY = pathlib.Path(__file__).parent.resolve()
 EXAMPLE_DOCS_DIRECTORY = os.path.join(DIRECTORY, "..", "..", "example-docs")
@@ -41,3 +43,47 @@ def test_partition_pptx_raises_with_both_specified():
 def test_partition_pptx_raises_with_neither():
     with pytest.raises(ValueError):
         partition_pptx()
+
+
+def test_partition_pptx_orders_elements(tmpdir):
+    filename = os.path.join(tmpdir, "test-ordering.pptx")
+
+    presentation = pptx.Presentation()
+    blank_slide_layout = presentation.slide_layouts[6]
+    slide = presentation.slides.add_slide(blank_slide_layout)
+
+    left = top = width = height = pptx.util.Inches(2)
+    txBox = slide.shapes.add_textbox(left, top, width, height)
+    tf = txBox.text_frame
+    tf.text = "This is lower and should come second"
+
+    left = top = width = height = pptx.util.Inches(1)
+    left = top = pptx.util.Inches(-10)
+    txBox = slide.shapes.add_textbox(left, top, width, height)
+    tf = txBox.text_frame
+    tf.text = "This is off the page and shouldn't appear"
+
+    left = top = width = height = pptx.util.Inches(2)
+    txBox = slide.shapes.add_textbox(left, top, width, height)
+    tf = txBox.text_frame
+    tf.text = ""
+
+    left = top = width = height = pptx.util.Inches(1)
+    txBox = slide.shapes.add_textbox(left, top, width, height)
+    tf = txBox.text_frame
+    tf.text = "This is higher and should come first"
+
+    top = width = height = pptx.util.Inches(1)
+    left = pptx.util.Inches(0.5)
+    txBox = slide.shapes.add_textbox(left, top, width, height)
+    tf = txBox.text_frame
+    tf.text = "-------------TOP-------------"
+
+    presentation.save(filename)
+
+    elements = partition_pptx(filename=filename)
+    assert elements == [
+        Text("-------------TOP-------------"),
+        NarrativeText("This is higher and should come first"),
+        NarrativeText("This is lower and should come second"),
+    ]

unstructured/__version__.py

@@ -1 +1 @@
-__version__ = "0.4.5-dev5"  # pragma: no cover
+__version__ = "0.4.6"  # pragma: no cover

unstructured/partition/pptx.py

@@ -2,7 +2,7 @@
 
 import pptx
 
-from unstructured.documents.elements import Element, ListItem, NarrativeText, Title
+from unstructured.documents.elements import Element, ListItem, NarrativeText, Text, Title
 from unstructured.partition.text_type import (
     is_possible_narrative_text,
     is_possible_title,
@@ -35,9 +35,16 @@ def partition_pptx(filename: Optional[str] = None, file: Optional[IO] = None) ->
 
     elements: List[Element] = list()
     for slide in presentation.slides:
-        for shape in slide.shapes:
+        for shape in _order_shapes(slide.shapes):
+            # NOTE(robinson) - we don't deal with tables yet, but so future humans can find
+            # it again, here are docs on how to deal with tables. The check for tables should
+            # be `if shape.has_table`
+            # ref: https://python-pptx.readthedocs.io/en/latest/user/table.html#adding-a-table
             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:
+                continue
             for paragraph in shape.text_frame.paragraphs:
                 text = paragraph.text
                 if text.strip() == "":
@@ -48,10 +55,17 @@ def partition_pptx(filename: Optional[str] = None, file: Optional[IO] = None) ->
                     elements.append(NarrativeText(text=text))
                 elif is_possible_title(text):
                     elements.append(Title(text=text))
+                else:
+                    elements.append(Text(text=text))
 
     return elements
 
 
+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))
+
+
 def _is_bulleted_paragraph(paragraph) -> bool:
     """Determines if the paragraph is bulleted by looking for a bullet character prefix. Bullet
     characters in the openxml schema are represented by buChar"""