Check pull request comments and evaluate pulls

[toxicphreAK]

Pulls:

• add w:firstLineChars in text for #685 fixed python-openxml/python-docx#941
• https://github.com/pfernique/python-docx/tree/feature/add_svg_picture

[toxicphreAK]

Reviewed in the 2.0.0 feature pass (August 2026) and left open deliberately: this is a standing research task rather than a unit of implementable work, so there is nothing here to close.

Several of the features these threads point at have since landed on main independently — tracked changes (#74), fields and TOC (#80), list numbering (#73), cross-run search and replace (#76), floating images (#85), watermarks (#86) and style copying (#87). Worth re-reading this list against HISTORY.rst before spending time on it, since the overlap is now substantial.

[toxicphreAK]

Worked through this: all 133 open pull requests against python-openxml/python-docx, not just the two linked above, each checked against main rather than against its own description.

The overlap with #11 is large — many of these PRs are the upstream-facing half of the forks surveyed there, by the same authors. Everything genuinely new is now filed.

New issues from this survey

		Upstream PR
#103	w:shd modelled wrongly — we emit invalid XML and raise on valid input	python-openxml#250, python-openxml#1129, python-openxml#908
#104	Character-unit indents and line-unit spacing (w:firstLineChars and friends)	python-openxml#941, upstream issue python-openxml#685
#105	Paragraph mark run properties (w:pPr/w:rPr)	python-openxml#537, python-openxml#417
#106	_Row.repeat_as_header and the rest of w:trPr	python-openxml#226
#107	Table.width (w:tblW), percentage widths, w:tblInd, w:tblCellMar	python-openxml#1529
#108	RTL and vertical text — w:bidi, w:textDirection	python-openxml#307
#109	Table alt text — w:tblCaption, w:tblDescription	adjacent to python-openxml#1530
#110	Style lookup fails on UI-cased built-in style names	python-openxml#958, python-openxml#1043, python-openxml#239
#111	Accept os.PathLike	python-openxml#1571, python-openxml#1168
#112	Copy API for paragraphs, runs, rows and tables	python-openxml#277, python-openxml#1309, python-openxml#429

Two of those are bugs rather than gaps, both reproduced against main:

• w:shd is modelled wrongly: required w:val missing, optional w:fill required #103 — CT_Shd declares w:fill required and does not model w:val at all. The schema (ref/xsd/wml.xsd:467) says the opposite. So font.shading_fill = "FF0000" writes <w:shd w:fill="FF0000"/> with the required w:val missing, and reading a legal <w:shd w:val="pct25" w:color="FF0000"/> raises InvalidXmlError. Every non-solid Word shading preset is in the second category.
• Style lookup fails on documents that store the UI-cased built-in style name #110 — a document whose styles.xml stores <w:name w:val="Heading 1"/> instead of the internal "heading 1" cannot have a heading added to it: add_heading() raises KeyError and "Heading 1" in document.styles is False, with the style sitting right there. Word is case-insensitive here, which is why such documents exist.

The two PRs linked above

• add w:firstLineChars in text for #685 fixed python-openxml/python-docx#941 (w:firstLineChars) — not present, filed as Support character-unit indents and line-unit spacing (w:firstLineChars and friends) #104. Note the PR types the attribute as ST_TwipsMeasure; the schema says ST_DecimalNumber in hundredths of a character unit, so the patch is not directly usable. Widened to the whole *Chars family plus w:spacing/@w:beforeLines|afterLines and the w:start/w:end synonyms, which have the same "reads as unset when Word shows a value" failure.
• pfernique feature/add_svg_picture (upstream Some hack to include SVG python-openxml/python-docx#798, "Some hack to include SVG") — already covered. SVG landed in 2.0.0 with a PNG fallback and MIME_TYPE.SVG handling in StoryPart.get_or_add_image.

Already on main

Roughly four fifths of the substantive PRs. Tracked changes (python-openxml#1538, python-openxml#1534, python-openxml#1270) — #74. Footnotes (python-openxml#1392) and comments — upstream v1.2.0 and #34. Fields (python-openxml#1017) — #80. Bookmarks (python-openxml#539, python-openxml#445) — #75. Hyperlinks (python-openxml#784). Custom properties (python-openxml#1273, python-openxml#468) — #72; extended properties (python-openxml#1206). WebP (python-openxml#1423), EMF (python-openxml#196), WMF (python-openxml#731), SVG (python-openxml#798) — #35. .docm / .dotx (python-openxml#1407, python-openxml#1378, python-openxml#716, python-openxml#673). Outline level (python-openxml#1393, python-openxml#1295, python-openxml#1064). East Asian fonts (python-openxml#1437, python-openxml#576, python-openxml#781, python-openxml#329). Borders (python-openxml#326, python-openxml#1129). Section columns (python-openxml#355) — #78. Cell vertical alignment (python-openxml#312). Section.paragraphs (python-openxml#755) — replaced by iter_inner_content(). Reproducible output (python-openxml#810). xpath() namespaces (python-openxml#622). Restart numbering (python-openxml#210). Merged cells (python-openxml#1036), incomplete rows (python-openxml#564), cell-access caching (python-openxml#1051, python-openxml#565). Alt text (python-openxml#1530, python-openxml#227, python-openxml#317). Fractional font sizes (python-openxml#1478), w:highlight w:val="none" (python-openxml#1560), huge attribute values (python-openxml#1174, python-openxml#1272), dangling relationships (python-openxml#1219), corrupt packages (python-openxml#1562), entity expansion (python-openxml#303). Form fields and AltChunk (python-openxml#649). Numbering (python-openxml#582, python-openxml#643) — #73, #98. tblLook (python-openxml#579, python-openxml#1524, python-openxml#462) — #97. Endnotes (python-openxml#860) — #94. Text after tracked changes (python-openxml#734) — Paragraph.original_text.

Also verified fixed independently, so no issue filed: python-openxml#1566 (non-image parts in the sha1 match — package.py:47 skips them), python-openxml#1526 (CT_Tc.bottom on a merge whose origin omits w:vMerge), python-openxml#863 (w:sectPr inside w:sectPrChange is excluded by the anchored xpath at oxml/document.py:45), python-openxml#1024 (next_id takes the max over all @id), python-openxml#1505 (image dpi defaults to 72).

Looked at and deliberately not filed

• [feature]add add_chart for docx using chart from python-pptx. python-openxml/python-docx#392, add_chart — charts mean the DrawingML chart part plus a chart-data workbook. That is python-pptx's chart machinery, not a small addition; out of scope until someone asks with a use case.
• Add CLI-Anything harness for python-docx python-openxml/python-docx#1545 CLI harness, Create Questions bank python-openxml/python-docx#1400 "questions bank" — not library features.
• fix: #609 support for python 2.6 python-openxml/python-docx#841 (Python 2.6), Add StringIO/BytesIO distinction for Python 2/3 python-openxml/python-docx#608 (StringIO/BytesIO for Python 2), Fix "descriptor '__dict__' for '_OxmlElementBase' objects doesn't apply to a 'CT_*' object" in Py3 python-openxml/python-docx#1220 — supported Python is 3.9+.
• fix zipfile.py 'UserWarning: Duplicate name: docProps/core.xml' python-openxml/python-docx#1436 — the duplicate-docProps/core.xml warning does not reproduce on a save / reopen / save round trip; the writer sorts parts by partname and emits each once.
• fixed a bug after using pyinstaller to pack programe in a Linux environment python-openxml/python-docx#1310, Issue176 / Using resource_stream to cover frozen apps python-openxml/python-docx#177 — frozen-app template paths. fixed a bug after using pyinstaller to pack programe in a Linux environment python-openxml/python-docx#1310 is a one-line normalization of parts/hdrftr.py:30 and :50 from dirname/../templates to dirname(dirname)/templates, dropping the .. component that some frozen-app loaders do not resolve. Same result on a normal filesystem, so not worth an issue of its own — take it the next time that file is touched.
• add element as a public property not _element which is private python-openxml/python-docx#1446 (element as a public property), Credit the original author! python-openxml/python-docx#1471, Update Python classifiers. python-openxml/python-docx#1501, and the ~20 docs and typo PRs — either taste, or already accurate here.
• Improving Paragraph.text performance by 500%~ python-openxml/python-docx#1118 ("Paragraph.text 500% faster") — predates the current implementation, which builds text from iter_inner_content() and has different characteristics. Worth revisiting as a measured benchmark rather than as a patch.

Nothing left to track here.