fix(text-response): fix null handling of old solution evaluation breakdowns

[adi-herwana-nus]

Bug

Rollbar flagged a NoMethodError: undefined method 'index_by' for nil when rendering
the text response answer partial
(_text_response.json.jbuilder).

The solutionResults block reads the persisted auto-grading result:

last_attempt.auto_grading.result['evaluation_results'].index_by { |result| result['solution_id'] }

The evaluation_results key was only introduced in commit b6da95a4ad
("feat(text-response): implement spreadsheet formula autograding"). Answers graded
before that commit have an auto-grading result containing messages but no
evaluation_results key, so the lookup returns nil and .index_by raises, 500-ing the
answer view.

The guard on the surrounding condition only checked that result existed, not that it
contained evaluation_results:

) && last_attempt&.auto_grading&.result

Remediation

Tighten the guard so the entire solutionResults block is skipped when
evaluation_results is absent.

We are not backfilling existing DB records. Legacy answers simply render without a
grading breakdown; a user can regrade the answer to repopulate evaluation_results
and restore the breakdown on demand.

The recovery path above depends on a grader being able to re-run autograding, but the
Re-evaluate Answer button in
QuestionGrade.tsx
was only rendered for programming and autogradable rubric-based-response questions. We extended the button's render condition to also cover autogradable text-response questions.

Scope check

Audited every evaluation_results reference across app, lib, and spec. The jbuilder
is the only read-site of the persisted auto_grading.result['evaluation_results']
hash. All other references are producer-side (building the results in the grading
services) and cannot be nil in the same way. The sibling result['messages'] read is
safe — messages has existed since the original grading service, so every record has it.

Tests

Added regression coverage to
answers_controller_spec.rb
(the #show action renders this jbuilder end-to-end via render_views)

[Copilot]

Pull request overview

Fixes a runtime error when rendering text-response answer JSON for legacy autograding results that predate the introduction of evaluation_results, and ensures instructors can re-run autograding for affected text-response questions to repopulate the missing breakdown.

Changes:

• Guard solutionResults rendering in the text-response Jbuilder partial on presence of evaluation_results to avoid nil.index_by crashes for legacy records.
• Extend the grading UI to show the Re-evaluate Answer button for autogradable text-response questions (in addition to programming questions).
• Add controller-spec regression coverage for both legacy (no evaluation_results) and current (has evaluation_results) autograding result shapes when rendering #show.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
app/views/course/assessment/answer/text_responses/_text_response.json.jbuilder	Skips solutionResults block unless evaluation_results exists, preventing nil.index_by 500s for older persisted results.
client/app/bundles/course/assessment/submission/containers/QuestionGrade.tsx	Enables Re-evaluate Answer for autogradable text-response questions to allow regeneration of missing breakdown data.
spec/controllers/course/assessment/submission/answer/answers_controller_spec.rb	Adds end-to-end #show regression tests covering both legacy and current autograding result structures.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[cysjonathan]

Thanks!