Trait RasterBackend 

pub trait RasterBackend {
    type Device: RenderDevice;

    // Required methods
    fn new_target(&self, w: u32, h: u32, clear: Color) -> Self::Device;
    fn new_target_with_backdrop(&self, base: &Pixmap) -> Self::Device;
    fn snapshot(&self, d: &Self::Device) -> Pixmap;
    fn finish(&self, d: Self::Device) -> Pixmap;
}

The factory that lets the engine rasterize offscreen: soft masks, transparency groups, pattern cells and the fill+stroke knockout buffer all need a second target.

Required Associated Types

type Device: RenderDevice

The device this backend produces.

Required Methods

fn new_target(&self, w: u32, h: u32, clear: Color) -> Self::Device

A fresh target of the given size, every pixel set to clear.

Both axes must be at most MAX_TARGET_DIMENSION: vello_cpu sizes its scenes, pixmaps and masks with u16. The engine enforces the bound before calling.

fn new_target_with_backdrop(&self, base: &Pixmap) -> Self::Device

A fresh target seeded with base’s pixels.

This is a non-isolated transparency group’s backdrop (ISO 32000 §11.4.6) and the fill+stroke knockout buffer’s saved destination.

fn snapshot(&self, d: &Self::Device) -> Pixmap

The device’s current pixels, without consuming it.

Only valid with every RenderDevice::push_layer matched by a RenderDevice::pop — vello_cpu asserts it. On a retained-scene backend this costs a full rasterization of the scene so far, not a rectangle copy, which is why the engine renders each group into its own target rather than snapshotting sub-rectangles of a shared one.

fn finish(&self, d: Self::Device) -> Pixmap

Consume the device, yielding its premultiplied RGBA8 pixels.

Dyn Compatibility

This trait is dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors