Datasets

What is a dataset?

A dataset wraps a SQL query (or a saved query, or a NoSQL find) with a name, optional column overrides, and a refresh policy. Once saved, the dataset becomes a reusable source you can point any chart or widget at without repeating the SQL.

Why bother?

• Reuse — define a complex JOIN once, use it in ten reports.
• Caching — heavy queries can be cached on a schedule, so reports load instantly.
• Column polish — rename ugly column names, hide internal IDs, reorder for presentation.
• Transforms — derived fields, deduplication, casts and filters without modifying the underlying query.

Anatomy of a dataset

• Source — the connection + query (or saved query) that produces the rows.
• Columns — auto-detected from the query result; you can rename, hide and reorder.
• Transforms — an optional pipeline of steps run after the query: rename, cast, derived, filter, deduplicate, sort, limit.
• Execution— “live” (run on every use) or “cached” (refresh on a schedule).

Live vs. cached

Live datasets re-run the query every time something reads them. Great for small queries or real-time dashboards.

Cached datasets refresh on a schedule and serve the cached result in between. Use them for expensive aggregates so your dashboard loads in milliseconds.