Hard line breaks are lost: <br> and Word line breaks become spaces #2642

markdownify writes a <br> as two trailing spaces, which is a Markdown hard line break. MarkItDown.convert() then strips trailing whitespace from every output line (the "Normalize the content" step in _markitdown.py). Every hard line break therefore becomes a soft break, which renders as a space. This affects every converter that renders HTML through markdownify, including HTML, DOCX (Word line breaks, Shift+Enter) and RSS. It also affects the OCR text that markitdown-ocr adds for images in DOCX, PPTX and XLSX files.

import io
from markitdown import MarkItDown

html = b"<p>Line one<br>Line two</p>"
print(repr(MarkItDown().convert_stream(io.BytesIO(html), file_extension=".html").markdown))

Actual: 'Line one\nLine two', which renders as "Line one Line two".

Expected: a hard line break, such as 'Line one\\\nLine two'.

The attached DOCX has one paragraph containing a Word line break (<w:br/>) and gives the same result.

markitdown-line-breaks.docx

Related: #2528 reports that the same normalization removes trailing spaces in fenced code blocks. Its PRs, #2529 and #2633, change only code blocks. Open PR #2559 would fix this by keeping lines that end in two or more spaces during the normalization.

Environment: markitdown 0.1.8 (main 4cc9fa1), markdownify 1.2.3, Python 3.14.8, Ubuntu 24.04.5 LTS.