ShapeUtils.triangulateShape should accept Vector2Like instead of relying on Vector2.equals

[atox996]

Description

Currently, ShapeUtils.triangulateShape internally calls removeDupEndPts, which relies on Vector2.equals. This introduces a strong coupling to Vector2 instances, forcing users to instantiate potentially large numbers of Vector2 objects just to perform triangulation.

Ideally, triangulateShape should accept inputs adhering to the Vector2Like type ({ readonly x: number; readonly y: number; }) and avoid depending on Vector2.equals. This would allow users to pass plain objects without creating Vector2 instances, improving performance and reducing memory overhead in high-vertex-count scenarios.

Proposed solution

• Allow triangulateShape to accept Vector2Like objects.
• Replace Vector2.equals checks in removeDupEndPts with a simple value comparison (v1.x === v2.x && v1.y === v2.y).

Reproduction steps

none

Code

const contour: Vector2Like[] = [];
const holes: Vector2Like[][] = [];
ShapeUtils.triangulateShape(contour, holes)

Live example

none

Screenshots

No response

Version

r181

Device

No response

Browser

No response

OS

No response

[Mugen87]

This would allow users to pass plain objects without creating Vector2 instances, improving performance and reducing memory overhead in high-vertex-count scenarios.

Can you please share a performance test that demonstrates the performance differences between POJOs and instances of Vector2? At the end, both are objects and treated as such by the JS engine.