profile-dataset

To check dataset quality and calculate null rates, profile a dataset by its id to return a structured, multi-layer profile of row/column shape, freshness, and per-column metrics. It defaults to profiling all columns but allows narrowing via a focus list.

Dataset profiling reveals coverage, null/fill rates, cardinality, and ranges by generating a three-layer profile: shape, per-column coverage, and quality flags. This structured output is suitable for downstream workflows like data triage and match reporting.

To profile specific columns instead of an entire dataset, supply a dataset id along with a focus list to narrow the scope. The profiler will return per-column stats and quality flags restricted to your selected columns using bundled stats and a sample.

When dataset statistics are missing or stale, the profiler applies gated recalc prompts and graceful sample-only fallbacks. It supports tiered workflows, triggering Tier-2 config recalculation or Tier-3 custom NQL escalation only when necessary for complex measures.

Yes, you can use dataset profiling for downstream data triage workflows. The structured output surfaces explicit stats, inferred shapes, and quality flags that directly support rosetta stone mappings, match reporting, and data triage tasks.

Bundled statistics provide a Tier-1 profile but may lack complex measures. Limitations are addressed through gated Tier-2 recalculation for config updates and Tier-3 custom NQL escalation, with graceful sample-only fallbacks when recalculation is not possible.