[rtibblesbot]

Summary

Moves Perseus input-answer parsing onto QTI's format_number, replacing #6255's isfinite filter and #6259's _perseus_number; Perseus and QTI now read an input answer the same way.

• Mixed nodes (native QTI + legacy input questions) publish QTI only when a legacy input answer is not a finite number (e.g. Sphere, +50%, 1e400); otherwise they keep their Perseus archive.
• An input question answered only by an image publish drops counts as answerless, so it is left out of the Perseus archive and assessment metadata.
• Every Perseus archive, Perseus-only nodes included, keeps JSON-number answers (4, 0), reads +3 / 1E3 as numbers, and no longer publishes "correct": false answers as correct.

References

Closes #6208. Closes #6266. Refs #6268. Refs #6232.

Reviewer guidance

• The third bullet changes Perseus-only exports too; if that is out of scope, the gate must instead make the int-4 fixture QTI only.
• QTI's input _process_answers override moved into the base generator (utils/assessment/base.py), so both formats keep a JSON 0.
• Perseus logs no warning of its own for a skipped input; the QTI generator already warns.

AI usage

Used Claude Code to reproduce the bug with a scratch publish test, write the fix and its tests, and draft this description. Verified with the export channel, exercise creation and QTI test suites, and pre-commit.

🤖 Generated with Claude Code

---

Written by rtibblesbot, an LLM-based coding agent.

🟡 Waiting for feedback Last updated: 2026-10-16 22:35 UTC

[rtibblesbot]

Outside this PR's scope — tracked under #5998:

• [QTI] Perseus archive of an exercise with raw Perseus questions keeps legacy input questions whose answers are not numbers #6276 [QTI] Perseus archive of an exercise with raw Perseus questions keeps legacy input questions whose answers are not numbers

---

@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

[rtibblesbot]

Outside this PR's scope — tracked in:

• Restoring a published QTI item stores lxml's reserialization instead of the published XML #6305 Restoring a published QTI item stores lxml's reserialization instead of the published XML

---

Written by rtibblesbot, an LLM-based coding agent.

[rtibbles]

Please rebase this on latest unstable, it seems awfully similar to a recently merged PR.

[rtibblesbot]

@rtibbles Rebased on unstable. This was stacked on #6262, which has merged, so the diff is now three commits:

• fix(perseus): read numeric input answers as Perseus renders them — Perseus export of a legacy input question drops JSON-number answers and accepts answers marked incorrect #6266, Perseus export of a legacy input question accepts answers marked incorrect #6268
• fix(perseus): count input questions answered only by a dropped image as answerless — an answer of only an unpackageable image still left a numeric-input with answers: [] after Leave answerless legacy input questions out of Perseus archives #6262
• fix(qti): publish mixed exercises QTI only when Perseus can't read an input — [QTI] Perseus archive of a mixed exercise keeps legacy input questions with no numeric answer #6208

The description's "Stacked on #6262" line and its second bullet describe #6262's work, not this diff.

[rtibbles]

As written, the description reads like a revert of merged work. Please rebase on latest unstable (#6282, #6302 and #6306 have merged since), then rewrite the description to match the diff:

• Say up front that this moves Perseus input-answer parsing onto format_number, replacing #6255's isfinite filter and #6259's _perseus_number.
• Drop "Supersedes #6255". It's already merged.
• Drop "Not covered: +3 / 1E3 (#6224, #6259)". This diff deletes #6259's code and covers that case.

[rtibblesbot]

Rebased on unstable (98b560e, includes #6282, #6302, #6306); clean, export/exercise-creation/QTI suites pass. Description rewritten:

• Leads with the move onto format_number, replacing Drop non-finite numeric answers from Perseus input questions #6255's isfinite filter and Normalise xsd:double text entry answers for Perseus derivation #6259's _perseus_number.
• Dropped "Supersedes Drop non-finite numeric answers from Perseus input questions #6255" and the "Not covered: +3 / 1E3" line; the Perseus bullet now says those parse.

[rtibblesbot]

Quick check on rtibblesbot's work on this PR: react 👍 helpful · 😕 mixed · 👎 not helpful. Reply with anything specific.