Interface SceneRaycastParams

Canvas-pixel cursor position relative to view.htmlElement — the format you'd compute with e.clientX/Y - canvas.getBoundingClientRect(). Required when view is the only spatial input.

Optional pre-filter on candidate object IDs. Returning false drops the object before any per-mesh / per-triangle work — composes with the visible/pickable filter (both must accept).

World-from-pick affine transform. The raycaster treats the input as a matrix that maps a canonical local ray (origin (0,0,0), direction (0,0,1)) into world space:

• origin = matrix · (0,0,0,1) = the matrix's translation column.
• direction = matrix · (0,0,1,0) = the matrix's third basis column.

Equivalent to passing the camera's inverseViewMatrix for a "shoot a ray straight ahead from the camera" query, except that GL view space looks down -Z, so for that case the caller should negate the resulting direction or pre-flip the third column. The raycaster doesn't second-guess the caller's convention — it just reads the matrix at face value.

Pre-built world-space ray. dir need not be normalised; tHit in the result is reported in dir-multiples either way.

Maximum parametric distance along the ray. Defaults to Infinity.

Minimum parametric distance along the ray. Defaults to 0.

Active view, required when canvasPos is used (the raycaster unprojects through this view's camera) and used to filter non-pickable objects when visiblePickableOnly is enabled.

When true (and view is provided), candidate objects must also satisfy view.objects[id].visible and view.objects[id].pickable !== false. Defaults to true when view is given, false otherwise — a programmatic ray pick without a view inspects the entire BVH.