[Perf] Canvas layer composition is quadratic: splitRenderLayers re-copies the node array per node

Abierto Apto para principiantes
#275 0 comentarios 0 reacciones 0 asignados Ver en GitHub

Los mantenedores suelen responder en 1 día

Nadie ha tomado este issue todavía.

Evaluación

Dificultad
2/5
Tiempo estimado
1-3 horas
Aptitud para principiantes
78/100
Tipo de issue
Error
Claridad
Bien especificado
Estado de actividad
Activo
Stack tecnológico
typescript

Línea de trabajo

Lee packages/charts-core/src/canvas.ts alrededor de splitRenderLayers (líneas 639–646) e inspecciona cómo agrupa nodos consecutivos por renderer. El issue incluye una reproducción con una marca dot de canvas y resultados de benchmarks; úsalos para comprobar el coste antes y después del cambio. El trabajo estará terminado cuando la composición de capas evite copiar arrays repetidamente y la reproducción deje de escalar cuadráticamente.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

splitRenderLayers in packages/charts-core/src/canvas.ts builds each layer by re-copying its node array once per node, so composing layers is quadratic in the number of consecutive nodes on one renderer:

https://github.com/TanStack/charts/blob/b236dae2d57f1921b9e0296b6ecf023bf012bb5f/packages/charts-core/src/canvas.ts#L639-L646

const append = (layer: ChartRenderLayer) => {
  const previous = layers.at(-1)
  if (previous?.renderer === layer.renderer) {
    previous.nodes = [...previous.nodes, ...layer.nodes]
  } else {
    layers.push(layer)
  }
}

It runs whenever a chart mixes renderers, for example one mark with renderer: canvasChartRenderer on the default SVG renderer. The function also recurses into every group to check for a custom renderer, including the default focus layer groups, which hold one node per point. So a canvas dot mark with n points costs about n²/2 element copies per render, on the very marks that were moved to canvas to make them cheap.

Reproduction

@tanstack/charts 1.5.0, Node 24 (vitest, jsdom). Median of 5 calls of the layered renderer's prerender:

import { createChartScene, defineChart, dot } from '@tanstack/charts'
import { canvasChartRenderer, createLayeredChartRenderer } from '@tanstack/charts/canvas'
import { scaleLinear } from '@tanstack/charts/scales/linear'
import { createSvgChartRenderer } from '@tanstack/charts/svg/renderer'

const layered = createLayeredChartRenderer(createSvgChartRenderer())
for (const n of [2000, 4000, 8000, 16000, 32000]) {
  const data = Array.from({ length: n }, (_, i) => ({ x: i, y: (i * 7919) % 1000 }))
  const chart = defineChart({
    marks: [dot(data, { x: (d) => d.x, y: (d) => d.y, renderer: canvasChartRenderer })],
    scales: { x: { scale: scaleLinear().domain([0, n]) }, y: { scale: scaleLinear().domain([0, 1000]) } },
  })
  const scene = createChartScene(chart, { width: 800, height: 400 })
  const t0 = performance.now()
  layered.prerender(scene, { ariaLabel: 'repro' })
  console.log(n, performance.now() - t0)
}
points now with the fix below
2,000 2.5 ms 0.4 ms
4,000 8.2 ms 0.2 ms
8,000 23.6 ms 0.3 ms
16,000 206 ms 0.4 ms
32,000 1,390 ms 1.0 ms

In a real app (a Vue Chart, a scatter chart of about 7,000 points with its points on canvas), wheel zoom spent about 300 ms per render in this function: about 51 million element copies per zoom step. That made the canvas version roughly ten times slower than plain SVG. With the fix, frames are about three times shorter than with SVG.

Suggested fix

Append in place. Every array append receives is a fresh literal built in this function, so mutating it is safe:

type BuildingLayer = Omit<ChartRenderLayer, 'nodes'> & { nodes: SceneNode[] }

const layers: BuildingLayer[] = []
const append = (layer: BuildingLayer) => {
  const previous = layers.at(-1)
  if (previous?.renderer === layer.renderer) {
    for (const node of layer.nodes) previous.nodes.push(node)
  } else {
    layers.push(layer)
  }
}

(A loop rather than push(...layer.nodes), which can exceed the argument limit for very large layers.)


Found and measured with an AI coding agent (Claude Code); the numbers above are from real runs.

Lenguaje dominante
TypeScript
Estrellas
870
Forks
57
Merge medio
11 h 58 min
PR fusionados (30 d)
18

Preparar el entorno

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Más de TanStack/charts

Todos los issues de TanStack/charts

Issues similares

Más issues de TypeScript

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.