Analytical modelling has historically meant a separate tool, a separate connection and a separate place for the result to live. This release puts the whole loop in the browser: a modeller picks a database, a schema and a cube, composes the query by moving dimensions and measures into it, runs it against the live database, and reads the answer as a grid or a chart without leaving the page.
Composing the query
- Dimension and measure tree. The cube's dimensions and measures are presented as a searchable tree, and a query is composed by moving them into the editor rather than by writing the projection by hand.
- Aggregation per level. Each level in the query carries its own aggregate selection, so the granularity of a measure is set where it applies.
- The schema on screen. A schema viewer and selector sit in the editor toolbar, so the definition behind a cube is readable at the point the query is written, and the detail panel is scoped to the cube in hand.
- Compiled to the platform's SQL layer. The composed query is emitted as the platform's declarative SQL, so an analytical query is the same governed, portable artifact as any other statement the platform runs.
Running it and reading the answer
- Executed against the database. The editor runs the query on the selected connection and renders the result set in place, with a search over the returned grid and a cancel action for a query that is taking too long.
- Grid, chart or summary. A single toggle switches the result panel between a tabular grid, a chart and an information view of the execution.
- Execution statistics. A statistics drawer on the toolbar reports what the run actually cost, so a slow query is diagnosed rather than guessed at.
- Exported. A result is exported from the panel, and a saved query is promoted into a report.
Keeping what is worth keeping
- History and favourites. Executed queries are recorded to a history, and a query worth returning to is saved as a favourite and grouped under its own tab.
- Running queries in view. A dialog lists what is currently executing, so a long-running analytical query is visible instead of silent.
The editor is a new surface alongside the existing database workbench; cubes, schemas and reports defined previously are read as they are, and no existing analytical definition needs migrating to open it here.