fix(quickcheck): omit empty categories from ObservationSummary debug rendering

[bobzhang]

In the snapshot-first testing workflow, a report's Debug rendering is effectively the API surface — and every classes-only coverage snapshot (the common case when using classify) currently carries a noisy labels: {} field, e.g. in a downstream project (moonbit-community/toml-parser#122):

observations={ labels: {}, classes: { "contains-table": 570, ... } }

ObservationSummary now renders via a custom Debug impl that omits empty categories, so the above reads observations={ classes: {...} } (symmetrically, labels-only summaries drop classes: {}). The text failure_report path already skipped empty categories; this aligns the Debug path.

Test plan: new classes-only rendering test; one existing labels-only snapshot updated (drops , classes: {}); moon test quickcheck 68/68; moon check --deny-warn, moon fmt, moon info clean (no interface change — the type is private).

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR improves snapshot stability/readability for QuickCheck reports by changing ObservationSummary’s Debug rendering to omit empty observation categories (labels/classes), matching the existing failure_report output behavior and reducing noise in snapshot-first workflows.

Changes:

• Replace derive(@debug.Debug) on ObservationSummary with a custom @debug.Debug::to_repr that conditionally includes labels / classes only when non-empty.
• Update an existing snapshot expectation to drop the now-omitted empty classes: {} field.
• Add a new snapshot test intended to assert classes-only rendering omits the empty labels field.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
quickcheck/observation.mbt	Adds a custom Debug impl for ObservationSummary that omits empty categories for snapshot-friendly output.
quickcheck/driver_test.mbt	Updates one snapshot and adds a new snapshot test for the new classes-only rendering behavior.

---

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

[bobzhang]

Not flaky as-is (the driver's default seed=37 makes the run deterministic), but the robustness point stands — the count would drift if Int generation or the default seed ever changed, and this test only exercises the rendering shape. Fixed in 1f9fa57: the classification is now unconditional (classify(true, "observed")), so the count always equals the test count.

[coveralls]

Coverage Report for CI Build 5815

Coverage increased (+0.003%) to 90.323%

Details

• Coverage increased (+0.003%) from the base build.
• Patch coverage: 5 of 5 lines across 1 file are fully covered (100%).
• No coverage regressions found.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

No coverage regressions found.

---

Coverage Stats

Relevant Lines:	17970
Covered Lines:	16231
Line Coverage:	90.32%
Coverage Strength:	169206.38 hits per line

---

💛 - Coveralls

[bobzhang]

@debug.Repr(Map[
         .. if !self.labels.is_empty() { [ ("labels", Repr(self.labels)]},
          .. if !self.class.is_empty() { ...}
         ])

[bobzhang]

Done — the record is now built with conditional spreads in the literal. One deliberate deviation from the snippet: kept Repr::record(...) instead of Repr(Map(...)), since rendering the summary as a Map would quote the keys ({ "labels": ... }) and churn every observation snapshot; record keeps the unquoted field style.