fix: PptxConverter tolerates None shape.text and notes text

[alvinttang]

Refs #1808.

PptxConverter.convert does unguarded + / .lstrip() on shape.text and notes_text_frame.text. python-pptx returns None for a text frame whose <a:r> run has no <a:t> child, and for certain third-party-generated decks. One malformed shape then fails the whole file with AttributeError: 'NoneType' object has no attribute 'lstrip' or TypeError: can only concatenate str (not "NoneType") to str.

Fix: treat None text as empty string at all three sites (title shape, body shape, notes frame).

RED

FAILED tests/test_pptx_none_text.py::test_pptx_converter_handles_none_shape_text
- PptxConverter threw AttributeError: 'NoneType' object has no attribute 'lstrip'
FAILED tests/test_pptx_none_text.py::test_pptx_converter_handles_none_notes_text
- PptxConverter threw TypeError: can only concatenate str (not "NoneType") to str

GREEN

tests/test_pptx_none_text.py::test_pptx_converter_handles_none_shape_text PASSED
tests/test_pptx_none_text.py::test_pptx_converter_handles_none_notes_text PASSED
2 passed in 0.65s

Full test_module_vectors.py (109 tests, covers real pptx conversion) still passes.

[noezhiya-dot]

Solid fix for a real-world crash. python-pptx can return None for shape.text when a run has no text child, and third-party PPTX generators produce these files in the wild.

The fix is clean: treat None as empty string at all three access points (title, body, notes). The RED→GREEN test approach is good — you demonstrated the exact failures before fixing.

The regression tests are well-crafted with monkeypatch to simulate the None cases without needing fixture files that are hard to create. LGTM.

[noezhiya-dot]

Solid defensive fix. The root cause is clear — shape.text can return None when runs have no <a:t> child element, and the existing code assumed it was always a string.

The fix is minimal and correct:

• text = shape.text or "" normalizes None to empty string in one place
• Both title and body paths use the normalized text variable
• Notes text frame gets the same treatment

Tests are well-structured:

• Uses monkeypatch to simulate None returns, which is the right approach for unit testing
• Covers both shape.text and notes_text_frame.text paths
• The in-memory PPTX builder for the notes test is a nice touch

One small note: the test file name test_pptx_none_text.py follows a different convention than the existing test_pptx.py — consider whether the maintainers prefer separate test files or appending to the existing one. Either way, the tests themselves are good.

Approve.