Serialise a QTI item's language as xml:lang

[rtibblesbot]

Summary

• QTI 3.0 defines no language attribute; an item's language now serialises as xml:lang.

References

Fixes #6098. Read-side tolerance: #6095.

Reviewer guidance

• The seven files under tests/utils/qti/fixtures/ are golden expected output compared verbatim by test_convert.py, not assertion-free tests. Mutating one attribute in each fails a distinct test_convert.py case.
• The two exercise_file.checksum updates in test_exercise_creation.py follow from the attribute rename changing the zipped item bytes.
• No data migration: consolidate() converts on read without persisting, so no stored raw_data carries language=.

AI usage

Used Claude Code to implement the change test-first from the plan on the issue. Verified with the full Python test suite and pre-commit.

---

@rtibblesbot's comments are generated by an LLM, and should be evaluated accordingly

How was this generated?

• Ran pre-flight CI checks (lint, format, tests) and verified all pass
• Rebased onto the target branch and resolved any conflicts
• Reorganized commit history into clean, logical commits
• Audited the diff to ensure only issue-relevant files are changed
• Built PR body from the repository's PR template with evidence blocks

Deviations from the issue spec

Criterion	Why not
Run the XSD's embedded Schematron rules in validate_qti_item	Dropped at @AlexVelezLl's request. validation.py is untouched, so nothing here makes validation stricter, and the srcset-on-<img> sample fix it forced went with it. The cost: no regression guard on the root element's attribute names — the XSD's extensionAnyTypeLax.Attr wildcard admits language= as happily as xml:lang=.

🟡 Waiting for feedback Last updated: 2026-08-19 14:31 UTC

[AlexVelezLl]

After reviewing the implications of the schematron, I think it'd be better just not introduce it right now, so please drop that addition.

[rtibblesbot]

After reviewing the implications of the schematron, I think it'd be better just not introduce it right now, so please drop that addition.

Dropped — the whole commit is off the branch (force-pushed), so validation.py and schema/README.md match unstable and the two Schematron tests are gone.

Searched for the rest of that commit's class rather than just the validator: grep -ri schematron|isoschematron over the branch is now 0 hits outside .plans/. The srcset → <picture><source> fix in test_assessmentitem.py's sample item existed only because the Schematron rejected it, so it reverted too — Img permits srcset exactly as it did before this PR.

One line from that commit survives: test_validation.py's sample item builder now writes xml:lang=, not language=, matching what the serialiser emits. grep -rn 'language="' over contentcuration/ leaves two hits, both unrelated — db_tools.py (Django model) and test_convert.py (LegacyAssessmentItem, which keeps the field name).

Worth flagging: nothing now catches the root element's attribute names regressing — the XSD's extensionAnyTypeLax.Attr wildcard admits language= as readily as xml:lang=. PR title and body updated to stop claiming otherwise.

[AlexVelezLl]

Code changes look good, tests provide good confidence.