Explanation

Why this crate exists

Splitting eval into abstraction (engine-exec) and engines (…-tree-walk, …-bytecode) is what makes the backend swappable: a kit registers an engine under one operation name and the same language definition runs on the interpreter or the VM. The abstraction owns the parts that must not diverge — handler/journal keys, native dispatch, pattern matching — so conformance between backends is structural rather than coincidental.

Where it sits

This crate belongs to the engine-* family. Plugins, kits and hosts reach it through the kit's operation slot, so the engine can be swapped without touching a language definition.

See the reference for the API and the how-to guides for tasks.