Validate QTI numeric answers the same way inline and headlessly by rtibblesbot · Pull Request #6258 · learningequality/studio

rtibblesbot · 2026-10-01T04:26:40Z

Summary

Inline and headless validation accept only finite xsd:double Numeric answers, compared by value (21 duplicates 21.0).
Incomplete count matches inline errors.
Numeric and Text entry flag every repeated answer, the first included.

References

Fixes #6150. First seen: #6095 (comment)

Reviewer guidance

Answers reopen as authored (21.0), not normalized. Reformat on load?

QA steps

Setup: pnpm devsetup, admin login. New question: "Published Channel" → "Topic 2" → "Add" → "New exercise" → "Questions" → "New question", plus a prompt. QTI demo: /channels/<channel-id>/#/qti-demo.

New question, Numeric: 5, -5, +5, 5., .5, 1e-5, 1E5, 2.3e+10. Only 5, +5, 5. show "Duplicate answers are not allowed".
New question, Numeric: e, -, +, 1e, 1e2e3, 1.2.3, 1,234, INF, NaN, 1e400. Each shows "Must be a valid number (e.g. 12, 0.5, -3.14)"; "Close" shows "Incomplete".
New question, Numeric: 21, 21.0, 1e2, 100. All show "Duplicate answers are not allowed".
New question, Numeric: e, -: no duplicate error. Change - to e: both show "Duplicate answers are not allowed".
New question, Numeric: 21.0, 1e-5; "Close", "Edit", edit the prompt, "Close", "Edit": still 21.0 and 1e-5.
New question, Text entry: Paris, paris, "Case-sensitive" unticked: both show "Duplicate answers are not allowed"; ticking "Case-sensitive" on both clears them.
QTI demo, "Question 3 of 8 — Numeric" → "Edit": 21, 21.0, then "Add acceptable answer" for 2.1e1, 5, e, e. All but 5 show "Duplicate answers are not allowed". "Delete answer 3": 21, 21.0 stay flagged; "Delete answer 2": 21 clears. "Close": Question 3 shows "Incomplete"; "Edit": both e flagged. Delete one e, set the other to 7, "Close": no "Incomplete".
QTI demo, "Question 4 of 8 — Text entry" → "Edit": Paris, paris, PARIS unticked; "Add acceptable answer" twice for Paris, "Case-sensitive" ticked. All five show "Duplicate answers are not allowed". Tick "Case-sensitive" on paris: only it clears. Tick it on the first Paris: PARIS clears. "Close": only Question 4 shows "Incomplete"; "Edit": the three Paris flagged. "Delete answer 5", "Delete answer 4", "Close": no "Incomplete".

Evidence

Numeric repeated answers

Step	Screenshot
Closed: "Incomplete"
Fixed: no "Incomplete"

numeric-duplicates-flow.webm

Text entry repeated answers

Step	Screenshot
All five flagged
paris clears
PARIS clears
Only Question 4 "Incomplete"
Fixed: no "Incomplete"

More captures (4)

Step	Screenshot
Numeric: all but 5 flagged
Numeric fixed: 21, 5, 7
Text entry on open
Text entry reopened

AI usage

Claude Code wrote the fix and tests; verified with Jest and pre-commit.

Deviations from the issue spec

Criterion	Why not
Accept every valid `xsd:double`	`1e400` overflows to `Infinity` and is rejected like `INF`

@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

	🟡 Waiting for feedback Last updated: 2026-10-01 22:02 UTC

AlexVelezLl · 2026-10-01T14:09:04Z

-
      if (val) {
        if (seen.has(lookupKey)) {
          errors.push({ code: ValidationError.DUPLICATE_ANSWER_CONTENT, id: answer.id });


Could you fix the error here that only flags the latter value as a duplicate instead of flagging all of them?

Fixed in c820430: every answer in a repeated group is now flagged, matching choice/validation.js.

Searched every DUPLICATE_* producer under QTIEditor/interactions/: 5 matched. Choice and ordering already flag all copies by id; associate and match flag by text. Text entry was the only one flagging just the later copy.

Tests now expect both ids (numeric and text entry), plus the headless validateItem cases.

AlexVelezLl · 2026-10-01T14:09:21Z

+  checked.querySelectorAll(':scope > qti-correct-response, :scope > qti-mapping').forEach(el => {
+    el.remove();
+  });
+  QTIDeclaration.fromXML(checked);


It's very difficult to understand what this is doing, and I'm not completely sure yet. Could you add more comments, please? Isn't there a better way to do it? Given that the QTIDeclaration.fromXML response isn't being stored anywhere, it makes me think this is rather asserting something indirectly, which is why the intention here is not clear.

It asserts that the declaration parses. The answers are left out because fromXML coerces floats, which would throw on or truncate an invalid answer. The rest is still checked: identifier, cardinality and default value. A malformed float declaration then yields [], as on the string path.

Moved it into a named helper, assertDeclarationParses, with a docstring saying this (54f1f81).

Constructing QTIDeclaration from the attributes alone would be more direct. It would skip the default-value check, though, and the validateItem test "reports a non-numeric default value as no correct answer" fails without it. The other way is a raw-read mode in QTIDeclaration, which would change shared serialization code for one caller.

Replaced the helper with the direct checks in 8baa10b: extractNumericAnswers constructs QTIDeclaration from the attributes (identifier, cardinality), then rejects a non-numeric default value explicitly. No fromXML call on the float path now.

Now in 2b0f97a: the default value goes through QTIDeclaration.coerceValue(…, BaseType.FLOAT), the same check fromXML applies, so a default like 1,234 that Kolibri accepts no longer drops the answers.

- Replace the loose regex, which accepted `e`, `-`, `1e2e3` and rejected `1e-5`, `1E5` - Compare numeric answers by value, so `21` and `21.0` are duplicates Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Float coercion dropped every answer on an invalid value and merged `21` with `21.0`, so headless validation reported an incomplete question the editor showed no error for. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

rtibblesbot · 2026-10-01T17:13:17Z

Outside this PR's scope — tracked under #6103:

[QTI] A Text entry answer of NULL reopens empty and makes the question incomplete #6272 [QTI] A Text entry answer of NULL reopens empty and makes the question incomplete

@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

AlexVelezLl

Code changes make sense, and manual QA checks out! LGTM!

rtibblesbot marked this pull request as ready for review October 1, 2026 06:19

rtibblesbot requested review from a team and rtibbles October 1, 2026 06:19

rtibblesbot force-pushed the issue-6150-4c7ff5 branch 2 times, most recently from 640b7be to 11627c0 Compare October 1, 2026 06:53

AlexVelezLl self-assigned this Oct 1, 2026

AlexVelezLl requested changes Oct 1, 2026

View reviewed changes

rtibblesbot force-pushed the issue-6150-4c7ff5 branch from 11627c0 to c820430 Compare October 1, 2026 14:39

rtibblesbot force-pushed the issue-6150-4c7ff5 branch 2 times, most recently from c820430 to e26c7ea Compare October 1, 2026 16:04

rtibblesbot and others added 2 commits October 1, 2026 09:32

fix(qti): flag every repeated Text entry and Numeric answer

2354019

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

rtibblesbot force-pushed the issue-6150-4c7ff5 branch from e26c7ea to 2354019 Compare October 1, 2026 17:02

AlexVelezLl approved these changes Oct 2, 2026

View reviewed changes

AlexVelezLl merged commit 3332334 into learningequality:unstable Oct 2, 2026
18 checks passed

rtibblesbot deleted the issue-6150-4c7ff5 branch October 2, 2026 17:10

Navigation Menu

Uh oh!

Validate QTI numeric answers the same way inline and headlessly - #6258

Validate QTI numeric answers the same way inline and headlessly#6258
AlexVelezLl merged 3 commits into
learningequality:unstablefrom
rtibblesbot:issue-6150-4c7ff5

rtibblesbot commented Oct 1, 2026 •

edited

Loading

🟡 Waiting for feedback

Uh oh!

AlexVelezLl Oct 1, 2026

Uh oh!

rtibblesbot Oct 1, 2026

Uh oh!

AlexVelezLl Oct 1, 2026

Uh oh!

rtibblesbot Oct 1, 2026

Uh oh!

rtibblesbot Oct 1, 2026

Uh oh!

rtibblesbot Oct 1, 2026

Uh oh!

rtibblesbot commented Oct 1, 2026

Uh oh!

AlexVelezLl left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

rtibblesbot commented Oct 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

References

Reviewer guidance

QA steps

Evidence

Numeric repeated answers

Text entry repeated answers

AI usage

Deviations from the issue spec

🟡 Waiting for feedback

Uh oh!

AlexVelezLl Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

rtibblesbot Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

AlexVelezLl Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

rtibblesbot Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

rtibblesbot Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

rtibblesbot Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

rtibblesbot commented Oct 1, 2026

Uh oh!

AlexVelezLl left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

rtibblesbot commented Oct 1, 2026 •

edited

Loading