Interface CubeEngine

The engine's completions of Pure code written after a relation, e.g. ->filter(x | $x. after a node's input (user, 2026-10-16), or of code on its own, on the model as it is: undefined when the engine completes no code, as an engine without the service answers, and from then on; rejects with a CubeEngineError

Runs an execution lambda; rejects with a CubeEngineError

Runs an execution lambda and gives its rows as CSV text (PLAN §11.10): the engine's own, a header row first, every value as the engine writes it; rejects with a CubeEngineError

The engine's catalogue of Pure functions, its PCT documentation, for the expression editors' completions and help: read once, undefined when the engine doesn't serve it, and from then on

The databases and runtimes of a model, parsed once

An Extend column's expression as the engine's JSON (PLAN §11.7), located under the source id; rejects with a CubeEngineError placed in the text

Plans an execution lambda for its runtime's database, as a run would, without running it: some expressions type, then fail when planned (PLAN §11.7). Rejects with a CubeEngineError

The outline of an expression of one row, x | …, from the engine's JSON of it (EXPRESSION-EDITOR.md, round 2): its parts placed in the code when the JSON was parsed with its source information; none for anything but a lambda of one expression

The Pure text of a lambda, for display only: never parsed back

The schema of each table, in one engine call: one entry per key, failures included

A lambda as the engine receives it, pretty-printed, its numbers kept exact (PLAN §11.16): for a viewer to copy while debugging a run

The schema of each lambda, in one engine call: one entry per key, failures included