Proposal: HostOps.nodeRect — node geometry readback (root-space border box)

Aperta
#618 1 commento 0 reazioni 0 assegnatari Vedi su GitHub

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
5/5
Tempo stimato
Più di una settimana
Idoneità per principianti
35/100
Tipo di issue
Funzionalità
Chiarezza
Abbastanza chiara
Stato di attività
Attiva
Stack tecnologico
react, rust, typescript
Ambito
api, frontend

Direzione di ricerca

Read contracts/spec/ui-ops.ts and inspect the generated bindings and engine/runtime-cabi/src/compat/ui/mod.rs; the issue says the prototype targeted the removed engine/wasm crate. First settle the result shape and auxiliary-surface semantics with maintainers, then port the core method, C ABI export, and framework/web binding. Done means tests cover nested geometry, missing or stale IDs, and agreement with the drawn position.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

Proposal: HostOps.nodeRect — node geometry readback

Summary

Add a read-only op that returns a node's border box in root space, so renderers can anchor UI (popovers, menus, tooltips, floating buttons) without probing the framebuffer.

Problem

The core computes layout, but the host-visible op surface has no way to read a node's rectangle:

  • ui_node_layout returns parent-relative values and is not part of HostOps, so every caller has to re-walk the ancestor chain.
  • There is no op that returns a node's box in root space.

A renderer that wants to place a popover next to its trigger therefore has to reconstruct the box from hit tests.

Measured cost in a real renderer

Our React renderer for PocketJS (compat layer over this core) anchors Menu/Select/tooltip panels to their triggers today by scanning for a seed point with hitTestBounds, then four binary searches for the edges:

Scene Hit-test queries per measure Queries per popover open (8 frames)
bare node 1,839 —
labelled node 3,040 —
clipped node 449 —
popover open — 22,322

That is per interaction, on a 480×272 viewport, with two frames of latency before the panel can be placed.

Proposal

interface NodeRect {
  readonly x: number;
  readonly y: number;
  readonly w: number;
  readonly h: number;
}
interface HostOps {
  nodeRect?(id: number): NodeRect | null;
}

Semantics:

  • Root-space border box in logical pixels: the node's parent-relative layout rect summed up the parent chain.
  • Untransformed (transforms are not applied) and unclipped (overflow clips and occlusion are not applied); a painted/clipped rect can be a later op.
  • Margin excluded.
  • null for id 0, a stale id or a destroyed node.
  • Optional member: hosts without the op omit it, and callers feature-detect. Old hosts keep working.

A working implementation exists in a local branch (see below): a core method that sums the parent chain, a C ABI export writing four f32s to an output buffer, and the framework/web binding. Tests: a hand-computed nested rect, the none/stale cases, and a pixel-free assertion that the box matches the drawn position.

Feedback we would like before opening a ported PR:

  1. Result shape. The op needs to return four numbers. Options we see: (a) a new rect result kind in contracts/spec/ui-ops.ts with the generated bindings extended for it, (b) an out-buffer call with the existing integer success result — cheaper to wire, less ergonomic. We are happy to do either; (a) touches the generator chain and the guests, so we would rather ask first.
  2. Untransformed/unclipped as stated, or should it intersect with clipping ancestors?
  3. Should the same op serve auxiliary surfaces (rect in that surface's space, as hit tests do)?

Status of the local implementation

The prototype predates the runtime-drivers refactor (#603): it added ui_node_rect to the old engine/wasm crate, which no longer exists. The core method and the C ABI export port cleanly to engine/runtime-cabi/src/compat/ui/mod.rs; what needs deciding is the descriptor/result-kind shape above. We have not opened that PR yet because we would rather agree on the API than hand you a PR that needs reworking in all the generators.

Happy to post the prototype diff on request, and to rebase the ported version the moment the shape is settled.

Lingua principale
TypeScript
Stelle
1.6k
Fork
88
Metriche di merge delle PR
Nessuna PR unita negli ultimi 30g

Preparare l'ambiente

Non abbiamo ancora controllato i file di configurazione di questo progetto. Parti dal suo README e consulta la nostra guida al primo contributo per i passaggi generali.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di pocket-stack/pocketjs

Tutte le issue di pocket-stack/pocketjs

Issue simili

Altre issue su TypeScript

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.