GS SQL Query Editor

Purpose-built SQL for temporal data. One table. Every dimension of time.

GS SQL is a purpose-built query language for temporal data. The STREAM table virtually projects across millions of independent time-series structures. Query with familiar SQL syntax extended for temporal range parameters, roll-up statistics, and time zone-aware date functions.

Catalogs — Many Schemas, One Query Language

Tables in GS SQL come from catalogs. The built-in system catalog projects every stream, sample, event, notification, and job. User catalogs project your component templates as tables — one catalog per logical schema. Reference any of them by qualified name (system.stream, employees.salaries) or use the unqualified name when the default catalog is unambiguous. One language reaches them all.

Entity diagram showing tables from multiple catalogs in different colors — system tables and several user catalogs, each with their own template tables and foreign-key relationships — An org with multiple catalogs — each color groups the tables that belong to one catalog. Every table is reachable from the same query language.

TDQ™ — Temporal Deep Query

The core system layer. Query the STREAM table, SAMPLE data, ACTIVEEVENT, NOTIFICATION, and JOB tables directly. Full access to every column, every roll-up statistic, every metadata field.

One Table. Every Stream.

The system.stream table virtually projects across every stream in your org — every component, every property, every relationship. Combine metadata columns (_component_name, _name, _last_value) with the _sample() accessor to pull metadata and historical samples in a single query. Roll-up statistics like CycleId='hour', Stat='avg' are pre-computed and free.

GS SQL editor running a system.stream query against the solar demo org — pulling component name, stream name, last value, and a 7-day hourly-average sample array for the inverters — A single `system.stream` query returning metadata and a 7-day hourly-average sample for every inverter stream.

Temporal Range Parameters

The range() parameter inside _sample() defines the temporal window. Use range(last=N) for count-based windows, range(sd=…, ed=…) for time bounds, and add CycleId='hour' | 'day' | 'month' with Stat='avg' | 'max' | 'min' | 'sum' | 'last' | 'first' to project rollups at any granularity. One language. Every dimension of time.

GS SQL editor showing 24 rows of hourly-average AC output for Inverter 1 over the last day — a clean solar production curve — Hourly-average AC output for Inverter 1 over the last 24 hours — `range(sd=-1d, ed=now), CycleId='hour', Stat='avg'`.

See How the Engine Runs Your Query

Check the Show trees box in the query editor and three trees print before the results — the parsed AST, the optimized Pushdown plan, and the Physical Plan with the operators that will execute (IndexScannerOp, JoinOp, FilterOp, ProjectionOp, SortOp, and others). No black box.

Show-trees output showing the AST, Pushdown, and Physical Plan trees for the collapsed-schema employees query — **Show trees** output for the collapsed-schema query above — AST, Pushdown, and Physical Plan.

TEQ™ — Temporal Entity Query

Query your semantic model. Use component template IDs as table names and stream IDs as column names. Entity-level results with temporal extensions built in.

Same Question. Fewer Joins.

Two imports of MySQL's employees sample database. On the left, the original normalized schema. On the right, the AI-collapsed version — salaries and titles are no longer child tables, but temporal streams hanging directly off employees, accessed via the range(last=20) temporal accessor. Identical results, no FK joins on the temporal children.