[chrikrah]

Two places in the JATS backend write None into text a reader sees.

_add_figure_captions sets caption = None when a <fig> carries a <label> and no <caption>. That combination is legal, and common for a numbered but uncaptioned figure. The f-string at :917 then renders the four characters None straight after the label:

<fig id="f1"><label>Figure 1.</label><graphic/></fig>

caption text: 'Figure 1.None'

_add_tables reads table["label"] = node.xpath("label")[0].text with no guard at :1175. An empty <label/> has .text of None in lxml, so a table with neither a real label nor a caption still produces a caption item:

<table-wrap id="t1"><label/><table><tbody><tr><td>A</td><td>B</td></tr></tbody></table></table-wrap>

caption text: 'None'

After the change the first reads Figure 1. and the second emits no caption item at all.

The file already has the pattern

:634 reads etal_node[0].text or DEFAULT_TEXT_ETAL. Both sites now keep the empty-string default the surrounding code already uses.

Tests

$ python -m pytest tests/test_backend_jats.py -q
59 passed

$ python -m ruff check docling/backend/xml/jats_backend.py tests/test_backend_jats.py
All checks passed!
$ python -m ruff format --check docling/backend/xml/jats_backend.py tests/test_backend_jats.py
2 files already formatted

Two new cases. Revert jats_backend.py alone, keep them, and both fail.

No groundtruth file changed. grep -rl None tests/data/jats/groundtruth/ returns nothing, so no committed fixture reaches either path today, which is why this went uncaught.

Scope

The <fig> case needs label present and caption absent. A figure with both, or with neither, already rendered correctly. The table case needs an empty <label/> element rather than a missing one.

Duplicate search

Three open pull requests touch docling/backend/xml/jats_backend.py: #4272, #3891 and #3791. None mentions _add_figure_captions, table["label"] or caption = None.

gh issue list --state all --search "jats caption None" and gh pr list --state all --search "jats figure caption" return nothing covering this.

Checklist:

• Documentation has been updated, if necessary.
  No document describes how a caption is composed from <label> and <caption>.
• Examples have been added, if necessary.
  The change corrects existing output rather than adding a capability.
• Tests have been added, if necessary.

[github-actions]

✅ DCO Check Passed

Thanks @chrikrah, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

🟢 Merge protection satisfied — ready to merge.

Show 1 satisfied protection

🟢 Enforce conventional commit

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[ceberam]

Thanks @chrikrah for your contribution!
Both fixes are correct. The _add_figure_captions change properly defaults caption to "" instead of None, preventing the literal string "None" from appearing after the label in the f-string. The _add_tables change applies the same or "" guard already used at line 638 for <etal>. Scope is narrow — no regressions, no performance impact.

I have added a minor style comment below.
In addition, please review the inline comments. The three-line comment before caption: str = "" in _add_figure_captions explains why None was not used — that is authoring context, and the type annotation already says it all. Remove it. The single-line comment in _add_tables is more defensible (lxml's None-on-empty-element is non-obvious), but trim the second clause to remove the fix-reasoning, for instance: # lxml sets .text to None for an empty element; fall back to "".

[ceberam]

All items from doc.texts are TextItem instances and always have a .label field. The project coding standards prohibit broad getattr(...)/hasattr(...) patterns. The rest of the test file uses direct attribute access. Replace with:

Suggested change

	def _caption_texts(doc: DoclingDocument) -> list[str]:
	return [
	t.text for t in doc.texts if str(getattr(t, "label", "")).endswith("caption")
	]
	def _caption_texts(doc: DoclingDocument) -> list[str]:
	return [t.text for t in doc.texts if t.label == DocItemLabel.CAPTION]

[chrikrah]

All three taken, in c9f9b96.

The three-line comment in _add_figure_captions is gone. It was authoring context, as you say. The annotation caption: str = "" states what the value is, and why it beats None belongs in the pull request rather than the file.

The _add_tables comment is now the single line you suggested, with the fix-reasoning clause dropped:

# lxml sets .text to None for an empty element; fall back to "".

And _caption_texts uses direct attribute access:

return [t.text for t in doc.texts if t.label == DocItemLabel.CAPTION]

That file already imports DocItemLabel, so nothing else moved.

59 passed. Revert jats_backend.py alone, keep the tests, and 2 still fail, so the two cases pin the production change rather than the helper.

ruff check and ruff format --check are clean on both files.

[chrikrah]

Rebased onto 2d5c590 to clear a conflict. #4272 landed a test in the same place in tests/test_backend_jats.py, so both sets of tests now sit side by side; nothing of either was dropped. Head is dab5b14.

64 passed

Revert jats_backend.py alone and keep the tests:

FAILED test_jats_labelled_figure_without_a_caption_omits_the_word_none
FAILED test_jats_table_with_an_empty_label_produces_no_caption
2 failed, 62 passed

ruff check and ruff format --check are clean on both files.

[dolfim-ibm]

lgtm