mirror of
https://github.com/apache/superset.git
synced 2026-07-18 12:45:44 +00:00
52a3ab0d71
Two opt-in tests (SUPERSET_PERF_VALIDATION=1 to run, skipped otherwise) under the existing PerfValidationTests class: * test_av_sc001_activity_endpoint_p95_under_1500ms — seeds dense history on the birth_names dashboard (~12 chart edits, 5 dataset edits, 1 dashboard edit), then hits /api/v1/dashboard/<uuid>/activity/ 50 times after warmup and asserts p95 < 1500ms per SC-AV-001. * test_av_sc003_save_path_p95_unaffected_by_activity_view — under the same fixture, runs 50 dashboard saves and re-measures the SC-004 version-capture overhead (50ms p95 budget). Confirms the activity- view branch hasn't added a new save-path coupling that regresses sc-103156's save-path budget. PASSES on the current branch — read- only feature, no new save-path code paths. Seed fixture differs from the spec's "25 charts × 3 dataset windows each" target — the birth_names fixture has ~12 charts on a single dataset, so we approximate by editing many charts plus the dataset multiple times. A bespoke multi-dataset fixture would let us measure against the spec's exact scale; that's future work. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>