Saved queries with typed, filterable and parameterized columns turn ad-hoc SQL into shared, reliable datasets.
The gap between a clever query and a reusable dataset is bigger than it looks. Someone writes SQL that answers a question, pastes it into a dashboard, and three months later nobody remembers what the columns mean or whether the filters are still correct.
Views close that gap. In QXX, a view is a saved query exposed as a first-class dataset: its columns are typed, its filters are declared, and its inputs are parameterized. Consumers pick a symbol, a time range or a threshold without ever touching the underlying SQL.
Because views are typed and searchable, they appear in the catalog like any other dataset — filterable, sortable and documented. A quant can publish a signal as a view; a product team can subscribe to it; neither has to coordinate on schema by hand.
And because a view is just SQL over the same surface, it works in both modes: query it one-shot for history, or run it with emit changes to stream results live. One definition, reused everywhere.