Frame-QL: An Introduction
Frame-QL is a storage-independent query language for declaring a governed analytical result. The requester names the analytical quantities to be returned, declares the final anchor at which the result frame should exist, and states conditions that shape the request. The requester does not prescribe the relational operations or physical commands used to produce the result. This storage independence rests on the published Theory of Data, Version 7.1: a theory of analytical data whose identity and laws are defined independently of physical storage. A measure family is a uniquely governed analytical family such as Revenue. A measure is that family at one anchor: $$\boxed{Measure = MeasureFamily @ Anchor}$$ Thus Revenue@CustomerMonth is a measure. A Frame-QL statement such as: FROM retail_manifold SELECT revenue AT {customer, cal.month} asks the governed environment for Revenue at the customer-month output anchor without requiring the requester to manufacture the path through physical data. Frame-QL uses two anchor spellings for different structural responsibilities. Query-level AT is mandatory and declares the one final output anchor of the frame. Expression-local @ {…} pins the anchor at which a subexpression is to be established for use by an enclosing operation. Conceptually, an expression such as revenue @ {order} denotes Revenue at Order for that use, but @ does not replace AT as an output declaration. This paper develops the central idea of Frame-QL: the output frame is the query. It explains why rows, tables, joins, and GROUP BY are not primitives of the request interface; how a governed family name and an anchored measure differ from a physical column; how ambiguity becomes clarification rather than silent choice; and why Frame-QL provides a natural analytical-intent boundary for AI-assisted analytics.ToD defines analytical identity and lawful transformation. The Frame-QL specification defines how a request expresses that meaning, through its versioned grammar and semantics. Implementations separately demonstrate their coverage. Language design need not wait for implementation, and implementation behavior cannot redefine the analytical target.