PptxConverter: TypeError "can only concatenate str (not NoneType) to str" when shape/notes text is None

[NikiforovAll]

Summary

PptxConverter.convert performs unguarded + concatenation on shape.text and notes_frame.text. When python-pptx yields None for a text frame (e.g. an <a:r> with no <a:t>, or certain third-party-generated decks), the whole conversion fails with:

PptxConverter threw TypeError with message: can only concatenate str (not "NoneType") to str

One malformed shape fails the entire file.

Affected lines (v0.1.5, also present on main)

packages/markitdown/src/markitdown/converters/_pptx_converter.py:

• L137 — md_content += "# " + shape.text.lstrip() + "\n"
• L139 — md_content += shape.text + "\n"
• L154 — md_content += notes_frame.text

Reproduction

import pptx.text.text
_orig = pptx.text.text.TextFrame.text.fget
pptx.text.text.TextFrame.text = property(
    lambda self: None if _orig(self) == "World" else _orig(self)
)

from markitdown import MarkItDown
MarkItDown().convert("sample.pptx")  # any pptx containing the text "World"
# markitdown._exceptions.FileConversionException:
#  - PptxConverter threw TypeError with message: unsupported operand type(s) for +: 'NoneType' and 'str'

Real-world trigger: a .pptx from a third-party generator with a text-frame run missing its <a:t> child, or a chart/SmartArt element whose title text is None.

Expected

Missing / None text should be treated as empty string. One bad shape shouldn't fail the entire file.

Proposed fix

md_content += "# " + (shape.text or "").lstrip() + "\n"
md_content += (shape.text or "") + "\n"
md_content += (notes_frame.text or "")

Happy to open a PR.

Environment

• markitdown 0.1.5 (markitdown[all])
• python-pptx 1.0.2
• Python 3.12, Ubuntu 24.04

[kelvinelove]

Hey, @NikiforovAll!

I've prepared a fix for this issue and linked a PR of the solution.

Changes Made:

• Line 165: md_content += "# " + (shape.text or "").lstrip() + "\n"
• Line 167: md_content += (shape.text or "") + "\n"
• Line 197: md_content += (notes_frame.text or "")

This implements the exact fix you proposed, using the or "" pattern to safely handle None text values that occur in malformed shapes or third-party-generated PPTX files.

Testing:
I've verified the changes work with the reproduction case you provided. The converter now gracefully handles None values instead of crashing with a TypeError.