Vizualizáció ellenőrzése

Visualization Verification

Use this skill when testing and verifying charts, metric transformations, linked dashboards, and data exports.

Mathematical & Source-to-Pixel Verification

1. Verify Calculations Against Small, Hand-Checkable Fixtures:
   • Test every aggregation, rate/ratio denominator, currency conversion, moving average, and percentile calculation against deterministic fixtures where the exact expected numbers are hand-computed.
   • Verify that displayed chart values, tooltip readouts, KPI summary cards, and the tabular/export view agree to the exact unit and rounding rule.
2. Test Adversarial Data Shapes:
   • Empty & single-point datasets: [] and [singlePoint] (ensure scales do not produce NaN / Infinity or division-by-zero when min === max).
   • All-zero, negative, and mixed-sign values: Verify zero baselines and negative bar/area rendering below the zero axis.
   • Missing data (null / gaps) & sparse timestamps: Verify that missing buckets render as gaps rather than false zeros, and that out-of-order timestamps are sorted or bucketed deterministically.
   • Extreme outliers & long labels: Verify that a $1000\times$ spike or a 90-character category name does not clip axes, overlap labels, or break layout.

Cross-View Synchronization & Temporal Edge Cases

• Linked Filter & Brush Consistency:
  • Verify that selecting a time window, clicking a category bar, or clearing a filter updates every linked chart, summary KPI, and detail table to the exact same subset, and that rapid filter toggles never leave widgets showing mismatched time ranges.
• Timezone, DST & Boundary Cutoffs:
  • Test bucketing across UTC vs. local/workspace timezone boundaries, month-end rollovers, and Daylight Saving Time transitions so events at 23:59:59 or 00:00:00 land in the right daily/monthly bucket.

Accessibility & Performance Verification

• Verify keyboard-only navigation across controls/data points, screen-reader summary/table parity, and non-color differentiation.
• Benchmark rendering and interaction (brush/zoom/filter latency and DOM node count) at the target dataset scale, and document the verified data volume and known limits in the handoff report.

Checklist

• Do chart points, tooltips, KPI cards, and table/export rows match hand-verified reference calculations?
• Have empty, single-point (min === max), negative, null-gap, and extreme-outlier datasets been tested without NaN scale failures?
• Do linked views stay synchronized across filter changes, resets, and timezone/date boundaries?
• Have keyboard/screen-reader accessibility and target-scale rendering performance been verified and documented?