[basiclines]

Direction

Make LEO a credible React alternative without moving application state into the component tree. Preserve LEOObject, LEOList and LEOElement and existing lifecycle/string-template entry points.

The research and staged acceptance criteria are in docs/react-direction.md, citing State of JS and React's own documentation. The 80% migration target and performance advantage remain unmeasured, not claimed.

Changes

• Replace destroy-and-rebuild rendering with keyed DOM reconciliation, stable input identity, controlled values, event replacement and callback refs.
• Add classic JSX/createElement, function props, children and fragments without a React dependency.
• Add synchronous model-owned setState with shallow patches and current-state updater functions; a multi-field update notifies after commit.
• Observe model/list references and release only the view's subscriptions; preserve shared data through disconnect/reconnect.
• Add shared-counter and searchable/keyed-list examples with model actions separate from view code.
• Update the generated README, migration boundaries, build compatibility and CI.

Evidence

52 unit tests pass. All 34 new unit tests fail under relevant reverted/mutated production behavior, then pass after exact restoration. Both built-example integration checks also fail with the original production implementation and pass after restoration.

Production bundles, generated docs and example bundles build successfully. Both demos were exercised in a real browser, including shared state after reconnection and keyed input identity, focus and selection.

Deliberate limits

This is the first foundation, not a production-readiness or drop-in React compatibility claim. SSR/hydration, async/error boundaries, concurrent scheduling, types and representative migration/performance studies remain follow-up milestones. Native DOM events differ from React synthetic events; string templates remain trusted HTML.

No npm release or automatic merge is included.