slint/internal/core/software_renderer.rs
2022-11-21 15:26:35 +01:00

1425 lines
55 KiB
Rust

// Copyright © SixtyFPS GmbH <info@slint-ui.com>
// SPDX-License-Identifier: GPL-3.0-only OR LicenseRef-Slint-commercial
//! This module contains the [`SoftwareRenderer`] and related types.
#![warn(missing_docs)]
mod draw_functions;
mod fonts;
use crate::api::Window;
use crate::graphics::{IntRect, PixelFormat, SharedImageBuffer};
use crate::item_rendering::ItemRenderer;
use crate::items::{ImageFit, Item, ItemRc};
use crate::lengths::{
LogicalLength, LogicalPoint, LogicalRect, LogicalSize, LogicalVector, PhysicalPx, PointLengths,
RectLengths, ScaleFactor, SizeLengths,
};
use crate::renderer::Renderer;
use crate::textlayout::{FontMetrics as _, TextParagraphLayout};
use crate::window::{WindowAdapter, WindowInner};
use crate::{Brush, Color, Coord, ImageInner, StaticTextures};
use alloc::rc::{Rc, Weak};
use alloc::{vec, vec::Vec};
use core::cell::{Cell, RefCell};
use core::pin::Pin;
use euclid::num::Zero;
pub use draw_functions::{PremultipliedRgbaColor, Rgb565Pixel, TargetPixel};
type PhysicalLength = euclid::Length<i16, PhysicalPx>;
type PhysicalRect = euclid::Rect<i16, PhysicalPx>;
type PhysicalSize = euclid::Size2D<i16, PhysicalPx>;
type PhysicalPoint = euclid::Point2D<i16, PhysicalPx>;
type DirtyRegion = PhysicalRect;
/// This trait defines a bi-directional interface between Slint and your code to send lines to your screen, when using
/// the [`SoftwareRenderer::render_by_line`] function.
///
/// * Through the associated `TargetPixel` type Slint knows how to create and manipulate pixels without having to know
/// the exact device-specific binary representation and operations for blending.
/// * Through the `process_line` function Slint notifies you when a line can be rendered and provides a callback that
/// you can invoke to fill a slice of pixels for the given line.
///
/// See the [`render_by_line`](SoftwareRenderer::render_by_line) documentation for an example.
pub trait LineBufferProvider {
/// The pixel type of the buffer
type TargetPixel: TargetPixel;
/// Called once per line, you will have to call the render_fn back with the buffer.
///
/// The `line` is the y position of the line to be drawn.
/// The `range` is the range within the line that is going to be rendered (eg, within the dirty region)
/// The `render_fn` function should be called to render the line, passing the buffer
/// corresponding to the specified line and range.
fn process_line(
&mut self,
line: usize,
range: core::ops::Range<usize>,
render_fn: impl FnOnce(&mut [Self::TargetPixel]),
);
}
/// A Renderer that do the rendering in software
///
/// The renderer can remember what items needs to be redrawn from the previous iteration.
///
/// There are two kind of possible rendering
/// 1. Using [`render()`](Self::render()) to render the window in a buffer
/// 2. Using [`render_by_line()`](Self::render()) to render the window line by line. This
/// is only useful if the device does not have enough memory to render the whole window
/// in one single buffer
///
/// ### `MAX_BUFFER_AGE`
///
/// The `MAX_BUFFER_AGE` parameter specifies how many buffers are being re-used.
/// This means that the buffer passed to the render functions still contains a rendering of
/// the window that was refreshed as least that amount of frame ago.
/// It will impact how much of the screen needs to be redrawn.
///
/// Typical value can be:
/// - **0:** No attempt at tracking dirty items will be made. The full screen is always redrawn.
/// - **1:** Only redraw the parts that have changed since the previous call to render.
/// This is assuming that the same buffer is passed on every call to render.
/// - **2:** Redraw the part that have changed during the two last frames.
/// This is assuming double buffering and swapping of the buffers.
pub struct SoftwareRenderer<const MAX_BUFFER_AGE: usize> {
partial_cache: RefCell<crate::item_rendering::PartialRenderingCache>,
/// This is the area which we are going to redraw in the next frame, no matter if the items are dirty or not
force_dirty: Cell<crate::item_rendering::DirtyRegion>,
/// This is the area which was dirty on the previous frames, in case we do double buffering
///
/// We really only need MAX_BUFFER_AGE - 1 but that's not allowed because we cannot do operations with
/// generic parameters
prev_frame_dirty: [Cell<DirtyRegion>; MAX_BUFFER_AGE],
window: Weak<dyn crate::window::WindowAdapter>,
}
impl<const MAX_BUFFER_AGE: usize> SoftwareRenderer<MAX_BUFFER_AGE> {
/// Create a new Renderer for a given window.
///
/// The `window` parameter can be coming from [`Rc::new_cyclic()`](alloc::rc::Rc::new_cyclic())
/// since the `WindowAdapter` most likely own the Renderer
pub fn new(window: Weak<dyn crate::window::WindowAdapter>) -> Self {
Self {
window: window.clone(),
partial_cache: Default::default(),
force_dirty: Default::default(),
prev_frame_dirty: [DirtyRegion::default(); MAX_BUFFER_AGE].map(|x| x.into()),
}
}
/// Internal function to apply a dirty region depending on the dirty_tracking_policy.
/// Returns the region to actually draw.
fn apply_dirty_region(
&self,
dirty_region: DirtyRegion,
screen_size: PhysicalSize,
) -> DirtyRegion {
if MAX_BUFFER_AGE == 0 {
PhysicalRect { origin: euclid::point2(0, 0), size: screen_size }
} else if MAX_BUFFER_AGE == 1 {
dirty_region
} else if MAX_BUFFER_AGE == 2 {
dirty_region.union(&self.prev_frame_dirty[0].replace(dirty_region))
} else {
let mut prev = dirty_region;
let mut union = dirty_region;
for x in self.prev_frame_dirty.iter().skip(1) {
prev = x.replace(prev);
union = union.union(&prev);
}
union
}
.intersection(&PhysicalRect { origin: euclid::point2(0, 0), size: screen_size })
.unwrap_or_default()
}
/// Render the window to the given frame buffer.
///
/// The renderer uses a cache internally and will only render the part of the window
/// which are dirty. The `extra_draw_region` is an extra regin which will also
/// be rendered. (eg: the previous dirty region in case of double buffering)
///
/// returns the dirty region for this frame (not including the extra_draw_region)
pub fn render(&self, buffer: &mut [impl TargetPixel], buffer_stride: usize) {
let window = self.window.upgrade().expect("render() called on a destroyed Window");
let window_inner = WindowInner::from_pub(window.window());
let factor = ScaleFactor::new(window_inner.scale_factor());
let (size, background) = if let Some(window_item) =
window_inner.window_item().as_ref().map(|item| item.as_pin_ref())
{
(
(LogicalSize::from_lengths(window_item.width(), window_item.height()).cast()
* factor)
.cast(),
window_item.background(),
)
} else {
(
euclid::size2(buffer_stride as _, (buffer.len() / buffer_stride) as _),
Brush::default(),
)
};
let buffer_renderer = SceneBuilder::new(
size,
factor,
window_inner,
RenderToBuffer { buffer, stride: buffer_stride },
);
let mut renderer = crate::item_rendering::PartialRenderer::new(
&self.partial_cache,
self.force_dirty.take(),
buffer_renderer,
);
window_inner.draw_contents(|components| {
for (component, origin) in components {
renderer.compute_dirty_regions(component, *origin);
}
let dirty_region = (renderer.dirty_region.to_rect().cast() * factor).round_out().cast();
let to_draw = self.apply_dirty_region(dirty_region, size);
renderer.combine_clip(
(to_draw.cast() / factor).cast(),
LogicalLength::zero(),
LogicalLength::zero(),
);
if !background.is_transparent() {
// FIXME: gradient
renderer.actual_renderer.processor.process_rectangle(to_draw, background.color());
}
for (component, origin) in components {
crate::item_rendering::render_component_items(component, &mut renderer, *origin);
}
});
}
/// Render the window, line by line, into the line buffer provided by the `line_callback` function.
///
/// The renderer uses a cache internally and will only render the part of the window
/// which are dirty, depending on the dirty tracking policy set in [`SoftwareRenderer::new`]
///
/// The line callback will be called for each line and should provide a buffer to draw into.
///
/// As an example, let's imagine we want to render into a plain buffer.
/// (You wouldn't normally use `render_by_line` for that because the [`Self::render`] would
/// then be more efficient)
///
/// ```rust
/// # use i_slint_core::software_renderer::{LineBufferProvider, SoftwareRenderer, Rgb565Pixel};
/// # fn xxx<'a>(the_frame_buffer: &'a mut [Rgb565Pixel], display_width: usize, renderer: &SoftwareRenderer<0>) {
/// struct FrameBuffer<'a>{ frame_buffer: &'a mut [Rgb565Pixel], stride: usize }
/// impl<'a> LineBufferProvider for FrameBuffer<'a> {
/// type TargetPixel = Rgb565Pixel;
/// fn process_line(
/// &mut self,
/// line: usize,
/// range: core::ops::Range<usize>,
/// render_fn: impl FnOnce(&mut [Self::TargetPixel]),
/// ) {
/// let line_begin = line * self.stride;
/// render_fn(&mut self.frame_buffer[line_begin..][range]);
/// // The line has been rendered and there could be code here to
/// // send the pixel to the display
/// }
/// }
/// renderer.render_by_line(FrameBuffer{ frame_buffer: the_frame_buffer, stride: display_width });
/// # }
/// ```
pub fn render_by_line(&self, line_buffer: impl LineBufferProvider) {
let window = self.window.upgrade().expect("render() called on a destroyed Window");
let window_inner = WindowInner::from_pub(window.window());
let component_rc = window_inner.component();
let component = crate::component::ComponentRc::borrow_pin(&component_rc);
if let Some(window_item) = crate::items::ItemRef::downcast_pin::<crate::items::WindowItem>(
component.as_ref().get_item_ref(0),
) {
let factor = ScaleFactor::new(window_inner.scale_factor());
let size = LogicalSize::from_lengths(window_item.width(), window_item.height()).cast()
* factor;
render_window_frame_by_line(
window_inner,
window_item.background(),
size.cast(),
&self,
line_buffer,
);
}
}
#[doc(hidden)]
pub fn default_font_size() -> LogicalLength {
self::fonts::DEFAULT_FONT_SIZE
}
}
#[doc(hidden)]
impl<const MAX_BUFFER_AGE: usize> Renderer for SoftwareRenderer<MAX_BUFFER_AGE> {
fn text_size(
&self,
font_request: crate::graphics::FontRequest,
text: &str,
max_width: Option<LogicalLength>,
scale_factor: ScaleFactor,
) -> LogicalSize {
fonts::text_size(font_request, text, max_width, scale_factor)
}
fn text_input_byte_offset_for_position(
&self,
_text_input: Pin<&crate::items::TextInput>,
_pos: LogicalPoint,
) -> usize {
0
}
fn text_input_cursor_rect_for_byte_offset(
&self,
_text_input: Pin<&crate::items::TextInput>,
_byte_offset: usize,
) -> LogicalRect {
Default::default()
}
fn free_graphics_resources(
&self,
items: &mut dyn Iterator<Item = Pin<crate::items::ItemRef<'_>>>,
) {
for item in items {
let cache_entry =
item.cached_rendering_data_offset().release(&mut self.partial_cache.borrow_mut());
drop(cache_entry);
}
}
fn mark_dirty_region(&self, region: crate::item_rendering::DirtyRegion) {
self.force_dirty.set(self.force_dirty.get().union(&region))
}
fn register_bitmap_font(&self, font_data: &'static crate::graphics::BitmapFont) {
fonts::register_bitmap_font(font_data);
}
}
fn render_window_frame_by_line<const MAX_BUFFER_AGE: usize>(
window: &WindowInner,
background: Brush,
size: PhysicalSize,
renderer: &SoftwareRenderer<MAX_BUFFER_AGE>,
mut line_buffer: impl LineBufferProvider,
) {
let mut scene = prepare_scene(window, size, renderer);
let dirty_region = scene.dirty_region;
debug_assert!(scene.current_line >= dirty_region.origin.y_length());
// FIXME gradient
let background_color = background.color().into();
while scene.current_line < dirty_region.origin.y_length() + dirty_region.size.height_length() {
line_buffer.process_line(
scene.current_line.get() as usize,
dirty_region.min_x() as usize..dirty_region.max_x() as usize,
|line_buffer| {
let offset = dirty_region.min_x() as usize;
TargetPixel::blend_slice(line_buffer, background_color);
for span in scene.items[0..scene.current_items_index].iter().rev() {
debug_assert!(scene.current_line >= span.pos.y_length());
debug_assert!(
scene.current_line < span.pos.y_length() + span.size.height_length(),
);
match span.command {
SceneCommand::Rectangle { color } => {
TargetPixel::blend_slice(
&mut line_buffer[span.pos.x as usize - offset
..(span.pos.x_length() + span.size.width_length()).get()
as usize
- offset],
color,
);
}
SceneCommand::Texture { texture_index } => {
let texture = &scene.textures[texture_index as usize];
draw_functions::draw_texture_line(
&PhysicalRect {
origin: span.pos - euclid::vec2(offset as i16, 0),
size: span.size,
},
scene.current_line,
texture,
line_buffer,
);
}
SceneCommand::SharedBuffer { shared_buffer_index } => {
let texture =
scene.shared_buffers[shared_buffer_index as usize].as_texture();
draw_functions::draw_texture_line(
&PhysicalRect {
origin: span.pos - euclid::vec2(offset as i16, 0),
size: span.size,
},
scene.current_line,
&texture,
line_buffer,
);
}
SceneCommand::RoundedRectangle { rectangle_index } => {
let rr = &scene.rounded_rectangles[rectangle_index as usize];
draw_functions::draw_rounded_rectangle_line(
&PhysicalRect {
origin: span.pos - euclid::vec2(offset as i16, 0),
size: span.size,
},
scene.current_line,
rr,
line_buffer,
);
}
}
}
},
);
if scene.current_line < dirty_region.origin.y_length() + dirty_region.size.height_length() {
scene.next_line();
}
}
}
struct Scene {
/// the next line to be processed
current_line: PhysicalLength,
/// The items are sorted like so:
/// - `items[future_items_index..]` are the items that have `y > current_line`.
/// They must be sorted by `y` (top to bottom), then by `z` (front to back)
/// - `items[..current_items_index]` are the items that overlap with the current_line,
/// sorted by z (front to back)
items: Vec<SceneItem>,
future_items_index: usize,
current_items_index: usize,
textures: Vec<SceneTexture<'static>>,
rounded_rectangles: Vec<RoundedRectangle>,
shared_buffers: Vec<SharedBufferCommand>,
dirty_region: DirtyRegion,
}
impl Scene {
pub fn new(
mut items: Vec<SceneItem>,
textures: Vec<SceneTexture<'static>>,
rounded_rectangles: Vec<RoundedRectangle>,
shared_buffers: Vec<SharedBufferCommand>,
dirty_region: DirtyRegion,
) -> Self {
let current_line = dirty_region.origin.y_length();
items.retain(|i| i.pos.y_length() + i.size.height_length() > current_line);
items.sort_unstable_by(|a, b| compare_scene_item(a, b));
let current_items_index = items.partition_point(|i| i.pos.y_length() <= current_line);
items[..current_items_index].sort_unstable_by(|a, b| b.z.cmp(&a.z));
Self {
items,
current_line,
current_items_index,
future_items_index: current_items_index,
textures,
rounded_rectangles,
shared_buffers,
dirty_region,
}
}
/// Updates `current_items_index` and `future_items_index` to match the invariant
pub fn next_line(&mut self) {
self.current_line += PhysicalLength::new(1);
// The items array is split in part:
// 1. [0..i] are the items that have already been processed, that are on this line
// 2. [j..current_items_index] are the items from the previous line that might still be
// valid on this line
// 3. [tmp1, tmp2] is a buffer where we swap items so we can make room for the items in [0..i]
// 4. [future_items_index..] are the items which might get processed now
// 5. [current_items_index..tmp1], [tmp2..future_items_index] and [i..j] is garbage
//
// At each step, we selecting the item with the higher z from the list 2 or 3 or 4 and take it from
// that list. Then we add it to the list [0..i] if it needs more processing. If needed,
// we move the first item from list 2. to list 3. to make some room
let (mut i, mut j, mut tmp1, mut tmp2) =
(0, 0, self.current_items_index, self.current_items_index);
'outer: loop {
let future_next_z = self
.items
.get(self.future_items_index)
.filter(|i| i.pos.y_length() <= self.current_line)
.map(|i| i.z);
let item = loop {
if tmp1 != tmp2 {
if future_next_z.map_or(true, |z| self.items[tmp1].z > z) {
let idx = tmp1;
tmp1 += 1;
if tmp1 == tmp2 {
tmp1 = self.current_items_index;
tmp2 = self.current_items_index;
}
break self.items[idx];
}
} else if j < self.current_items_index {
let item = &self.items[j];
if item.pos.y_length() + item.size.height_length() <= self.current_line {
j += 1;
continue;
}
if future_next_z.map_or(true, |z| item.z > z) {
j += 1;
break *item;
}
}
if future_next_z.is_some() {
self.future_items_index += 1;
break self.items[self.future_items_index - 1];
}
break 'outer;
};
if i != j {
// there is room
} else if j >= self.current_items_index && tmp1 == tmp2 {
// the current_items list is empty
j += 1
} else if self.items[j].pos.y_length() + self.items[j].size.height_length()
<= self.current_line
{
// next item in the current_items array is no longer in this line
j += 1;
} else if tmp2 < self.future_items_index && j < self.current_items_index {
// move the next item in current_items
let to_move = self.items[j];
self.items[tmp2] = to_move;
j += 1;
tmp2 += 1;
} else {
debug_assert!(tmp1 >= self.current_items_index);
let sort_begin = i;
// merge sort doesn't work because we don't have enough tmp space, just bring all items and use a normal sort.
while j < self.current_items_index {
let item = self.items[j];
if item.pos.y_length() + item.size.height_length() > self.current_line {
self.items[i] = item;
i += 1;
}
j += 1;
}
self.items.copy_within(tmp1..tmp2, i);
i += tmp2 - tmp1;
debug_assert!(i < self.future_items_index);
self.items[i] = item;
i += 1;
while self.future_items_index < self.items.len() {
let item = self.items[self.future_items_index];
if item.pos.y_length() > self.current_line {
break;
}
self.items[i] = item;
i += 1;
self.future_items_index += 1;
}
self.items[sort_begin..i].sort_unstable_by(|a, b| b.z.cmp(&a.z));
break;
}
self.items[i] = item;
i += 1;
}
self.current_items_index = i;
// check that current items are properly sorted
debug_assert!(self.items[0..self.current_items_index].windows(2).all(|x| x[0].z >= x[1].z));
}
}
#[derive(Clone, Copy, Debug)]
struct SceneItem {
pos: PhysicalPoint,
size: PhysicalSize,
// this is the order of the item from which it is in the item tree
z: u16,
command: SceneCommand,
}
fn compare_scene_item(a: &SceneItem, b: &SceneItem) -> core::cmp::Ordering {
// First, order by line (top to bottom)
match a.pos.y.partial_cmp(&b.pos.y) {
None | Some(core::cmp::Ordering::Equal) => {}
Some(ord) => return ord,
}
// Then by the reverse z (front to back)
match a.z.partial_cmp(&b.z) {
None | Some(core::cmp::Ordering::Equal) => {}
Some(ord) => return ord.reverse(),
}
// anything else, we don't care
core::cmp::Ordering::Equal
}
#[derive(Clone, Copy, Debug)]
#[repr(u8)]
enum SceneCommand {
Rectangle {
color: PremultipliedRgbaColor,
},
/// texture_index is an index in the Scene::textures array
Texture {
texture_index: u16,
},
/// shared_buffer_index is an index in Scene::shared_buffers
SharedBuffer {
shared_buffer_index: u16,
},
/// rectangle_index is an index in the Scene::rounded_rectangle array
RoundedRectangle {
rectangle_index: u16,
},
}
struct SceneTexture<'a> {
data: &'a [u8],
format: PixelFormat,
/// bytes between two lines in the source
stride: u16,
source_size: PhysicalSize,
color: Color,
alpha: u8,
}
struct SharedBufferCommand {
buffer: SharedImageBuffer,
/// The source rectangle that is mapped into this command span
source_rect: PhysicalRect,
colorize: Color,
alpha: u8,
}
impl SharedBufferCommand {
fn as_texture(&self) -> SceneTexture<'_> {
let begin = self.buffer.width() as usize * self.source_rect.min_y() as usize
+ self.source_rect.min_x() as usize;
match &self.buffer {
SharedImageBuffer::RGB8(b) => SceneTexture {
data: &b.as_bytes()[begin * 3..],
stride: 3 * b.stride() as u16,
format: PixelFormat::Rgb,
source_size: self.source_rect.size,
color: self.colorize,
alpha: self.alpha,
},
SharedImageBuffer::RGBA8(b) => SceneTexture {
data: &b.as_bytes()[begin * 4..],
stride: 4 * b.stride() as u16,
format: PixelFormat::Rgba,
source_size: self.source_rect.size,
color: self.colorize,
alpha: self.alpha,
},
SharedImageBuffer::RGBA8Premultiplied(b) => SceneTexture {
data: &b.as_bytes()[begin * 4..],
stride: 4 * b.stride() as u16,
format: PixelFormat::RgbaPremultiplied,
source_size: self.source_rect.size,
color: self.colorize,
alpha: self.alpha,
},
}
}
}
#[derive(Debug)]
struct RoundedRectangle {
radius: PhysicalLength,
/// the border's width
width: PhysicalLength,
border_color: PremultipliedRgbaColor,
inner_color: PremultipliedRgbaColor,
/// The clips is the amount of pixels of the rounded rectangle that is clipped away.
/// For example, if left_clip > width, then the left border will not be visible, and
/// if left_clip > radius, then no radius will be seen in the left side
left_clip: PhysicalLength,
right_clip: PhysicalLength,
top_clip: PhysicalLength,
bottom_clip: PhysicalLength,
}
fn prepare_scene<const MAX_BUFFER_AGE: usize>(
window: &WindowInner,
size: PhysicalSize,
software_renderer: &SoftwareRenderer<MAX_BUFFER_AGE>,
) -> Scene {
let factor = ScaleFactor::new(window.scale_factor());
let prepare_scene = SceneBuilder::new(size, factor, window, PrepareScene::default());
let mut renderer = crate::item_rendering::PartialRenderer::new(
&software_renderer.partial_cache,
software_renderer.force_dirty.take(),
prepare_scene,
);
let mut dirty_region = PhysicalRect::default();
window.draw_contents(|components| {
for (component, origin) in components {
renderer.compute_dirty_regions(component, *origin);
}
dirty_region = (renderer.dirty_region.to_rect().cast() * factor).round_out().cast();
dirty_region = software_renderer.apply_dirty_region(dirty_region, size);
renderer.combine_clip(
(dirty_region.cast() / factor).cast(),
LogicalLength::zero(),
LogicalLength::zero(),
);
for (component, origin) in components {
crate::item_rendering::render_component_items(component, &mut renderer, *origin);
}
});
let prepare_scene = renderer.into_inner();
Scene::new(
prepare_scene.processor.items,
prepare_scene.processor.textures,
prepare_scene.processor.rounded_rectangles,
prepare_scene.processor.shared_buffers,
dirty_region,
)
}
trait ProcessScene {
fn process_texture(&mut self, geometry: PhysicalRect, texture: SceneTexture<'static>);
fn process_rectangle(&mut self, geometry: PhysicalRect, color: Color);
fn process_rounded_rectangle(&mut self, geometry: PhysicalRect, data: RoundedRectangle);
fn process_shared_image_buffer(&mut self, geometry: PhysicalRect, buffer: SharedBufferCommand);
}
struct RenderToBuffer<'a, TargetPixel> {
buffer: &'a mut [TargetPixel],
stride: usize,
}
impl<'a, T: TargetPixel> ProcessScene for RenderToBuffer<'a, T> {
fn process_texture(&mut self, geometry: PhysicalRect, texture: SceneTexture<'static>) {
for line in geometry.min_y()..geometry.max_y() {
draw_functions::draw_texture_line(
&geometry,
PhysicalLength::new(line),
&texture,
&mut self.buffer[line as usize * self.stride..],
);
}
}
fn process_shared_image_buffer(&mut self, geometry: PhysicalRect, buffer: SharedBufferCommand) {
let texture = buffer.as_texture();
for line in geometry.min_y()..geometry.max_y() {
draw_functions::draw_texture_line(
&geometry,
PhysicalLength::new(line),
&texture,
&mut self.buffer[line as usize * self.stride..],
);
}
}
fn process_rectangle(&mut self, geometry: PhysicalRect, color: Color) {
let color = PremultipliedRgbaColor::from(color);
for line in geometry.min_y()..geometry.max_y() {
let begin = line as usize * self.stride + geometry.origin.x as usize;
TargetPixel::blend_slice(
&mut self.buffer[begin..begin + geometry.width() as usize],
color,
);
}
}
fn process_rounded_rectangle(&mut self, geometry: PhysicalRect, rr: RoundedRectangle) {
for line in geometry.min_y()..geometry.max_y() {
draw_functions::draw_rounded_rectangle_line(
&geometry,
PhysicalLength::new(line),
&rr,
&mut self.buffer[line as usize * self.stride..],
);
}
}
}
#[derive(Default)]
struct PrepareScene {
items: Vec<SceneItem>,
textures: Vec<SceneTexture<'static>>,
rounded_rectangles: Vec<RoundedRectangle>,
shared_buffers: Vec<SharedBufferCommand>,
}
impl ProcessScene for PrepareScene {
fn process_texture(&mut self, geometry: PhysicalRect, texture: SceneTexture<'static>) {
let size = geometry.size;
if !size.is_empty() {
let texture_index = self.textures.len() as u16;
self.textures.push(texture);
self.items.push(SceneItem {
pos: geometry.origin,
size,
z: self.items.len() as u16,
command: SceneCommand::Texture { texture_index },
});
}
}
fn process_shared_image_buffer(&mut self, geometry: PhysicalRect, buffer: SharedBufferCommand) {
let size = geometry.size;
if !size.is_empty() {
let shared_buffer_index = self.shared_buffers.len() as u16;
self.shared_buffers.push(buffer);
self.items.push(SceneItem {
pos: geometry.origin,
size,
z: self.items.len() as u16,
command: SceneCommand::SharedBuffer { shared_buffer_index },
});
}
}
fn process_rectangle(&mut self, geometry: PhysicalRect, color: Color) {
let size = geometry.size;
if !size.is_empty() {
let z = self.items.len() as u16;
let pos = geometry.origin;
let color = PremultipliedRgbaColor::from(color);
self.items.push(SceneItem { pos, size, z, command: SceneCommand::Rectangle { color } });
}
}
fn process_rounded_rectangle(&mut self, geometry: PhysicalRect, data: RoundedRectangle) {
let size = geometry.size;
if !size.is_empty() {
let rectangle_index = self.rounded_rectangles.len() as u16;
self.rounded_rectangles.push(data);
self.items.push(SceneItem {
pos: geometry.origin,
size,
z: self.items.len() as u16,
command: SceneCommand::RoundedRectangle { rectangle_index },
});
}
}
}
struct SceneBuilder<'a, T> {
processor: T,
state_stack: Vec<RenderState>,
current_state: RenderState,
scale_factor: ScaleFactor,
window: &'a WindowInner,
}
impl<'a, T: ProcessScene> SceneBuilder<'a, T> {
fn new(
size: PhysicalSize,
scale_factor: ScaleFactor,
window: &'a WindowInner,
processor: T,
) -> Self {
Self {
processor,
state_stack: vec![],
current_state: RenderState {
alpha: 1.,
offset: LogicalPoint::default(),
clip: LogicalRect::new(
LogicalPoint::default(),
(size.cast() / scale_factor).cast(),
),
},
scale_factor,
window,
}
}
fn should_draw(&self, rect: &LogicalRect) -> bool {
!rect.size.is_empty()
&& self.current_state.alpha > 0.01
&& self.current_state.clip.intersects(rect)
}
fn draw_image_impl(
&mut self,
geom: LogicalRect,
source: &crate::graphics::Image,
mut source_rect: IntRect,
image_fit: ImageFit,
colorize: Color,
) {
let alpha_u16 = self.current_state.alpha_u8() as u16;
let image_inner: &ImageInner = source.into();
let size: euclid::default::Size2D<u32> = source_rect.size.cast();
let phys_size = geom.size_length().cast() * self.scale_factor;
let source_to_target_x = phys_size.width / (size.width as f32);
let source_to_target_y = phys_size.height / (size.height as f32);
let mut image_fit_offset = euclid::Vector2D::default();
let (source_to_target_x, source_to_target_y) = match image_fit {
ImageFit::Fill => (source_to_target_x, source_to_target_y),
ImageFit::Cover => {
let ratio = f32::max(source_to_target_x, source_to_target_y);
if size.width as f32 > phys_size.width / ratio {
let diff = (size.width as f32 - phys_size.width / ratio) as i32;
source_rect.origin.x += diff / 2;
source_rect.size.width -= diff;
}
if size.height as f32 > phys_size.height / ratio {
let diff = (size.height as f32 - phys_size.height / ratio) as i32;
source_rect.origin.y += diff / 2;
source_rect.size.height -= diff;
}
(ratio, ratio)
}
ImageFit::Contain => {
let ratio = f32::min(source_to_target_x, source_to_target_y);
if (size.width as f32) < phys_size.width / ratio {
image_fit_offset.x = (phys_size.width - size.width as f32 * ratio) / 2.;
}
if (size.height as f32) < phys_size.height / ratio {
image_fit_offset.y = (phys_size.height - size.height as f32 * ratio) / 2.;
}
(ratio, ratio)
}
};
let offset =
self.current_state.offset.to_vector().cast() * self.scale_factor + image_fit_offset;
let renderer_clip_in_source_rect_space = (self.current_state.clip.cast()
* self.scale_factor)
.scale(1. / source_to_target_x, 1. / source_to_target_y);
match image_inner {
ImageInner::None => (),
ImageInner::StaticTextures(StaticTextures { data, textures, .. }) => {
for t in textures.as_slice() {
if let Some(clipped_relative_source_rect) =
t.rect.intersection(&source_rect).and_then(|clipped_source_rect| {
let relative_clipped_source_rect = clipped_source_rect
.translate(-source_rect.origin.to_vector())
.cast();
euclid::Rect::<_, PhysicalPx>::from_untyped(
&relative_clipped_source_rect,
)
.intersection(&renderer_clip_in_source_rect_space)
})
{
let target_rect = clipped_relative_source_rect
.scale(source_to_target_x, source_to_target_y)
.translate(offset)
.round();
let actual_x = clipped_relative_source_rect.origin.x as usize
+ source_rect.origin.x as usize
- t.rect.origin.x as usize;
let actual_y = clipped_relative_source_rect.origin.y as usize
+ source_rect.origin.y as usize
- t.rect.origin.y as usize;
let stride = t.rect.width() as u16 * t.format.bpp() as u16;
let t_alpha_u16 = t.color.alpha() as u16;
self.processor.process_texture(
target_rect.cast(),
SceneTexture {
data: &data.as_slice()[(t.index
+ (stride as usize) * actual_y
+ (t.format.bpp()) * actual_x)..],
stride,
source_size: clipped_relative_source_rect.size.ceil().cast(),
format: t.format,
color: if colorize.alpha() > 0 { colorize } else { t.color },
alpha: if colorize.alpha() > 0 {
(((alpha_u16 * colorize.alpha() as u16) / 255
* t_alpha_u16 as u16)
/ 255) as u8
} else {
((alpha_u16 * t_alpha_u16) / 255) as u8
},
},
);
}
}
}
_ => {
let img_src_size = source.size();
if let Some(buffer) = image_inner.render_to_buffer(Some(
crate::graphics::fit_size(image_fit, phys_size, img_src_size).cast(),
)) {
if let Some(clipped_relative_source_rect) = renderer_clip_in_source_rect_space
.intersection(&euclid::rect(
0.,
0.,
source_rect.width() as f32,
source_rect.height() as f32,
))
{
let target_rect = clipped_relative_source_rect
.scale(source_to_target_x, source_to_target_y)
.translate(offset)
.round();
let buf_size = buffer.size().cast::<f32>();
self.processor.process_shared_image_buffer(
target_rect.cast(),
SharedBufferCommand {
buffer,
source_rect: clipped_relative_source_rect
.translate(
euclid::Point2D::from_untyped(source_rect.origin.cast())
.to_vector(),
)
.scale(
buf_size.width / img_src_size.width as f32,
buf_size.height / img_src_size.height as f32,
)
.cast(),
colorize,
alpha: ((self.current_state.alpha_u8() as u16
* colorize.alpha() as u16)
/ 255) as u8,
},
);
}
} else {
unimplemented!("The image cannot be rendered")
}
}
};
}
/// Returns the color of the brush, mixed with the current_state's alpha
fn alpha_color(&self, brush: &Brush) -> Color {
let mut color = brush.color();
if self.current_state.alpha < 1.0 {
color = Color::from_argb_u8(
(color.alpha() as f32 * self.current_state.alpha) as u8,
color.red(),
color.green(),
color.blue(),
);
}
color
}
}
#[derive(Clone, Copy)]
struct RenderState {
alpha: f32,
offset: LogicalPoint,
clip: LogicalRect,
}
impl RenderState {
/// Converts the f32 alpha value to an u8 based representation.
fn alpha_u8(&self) -> u8 {
(self.alpha * 255.) as u8
}
}
impl<'a, T: ProcessScene> crate::item_rendering::ItemRenderer for SceneBuilder<'a, T> {
fn draw_rectangle(&mut self, rect: Pin<&crate::items::Rectangle>, _: &ItemRc) {
let geom = LogicalRect::new(LogicalPoint::default(), rect.geometry().size_length());
if self.should_draw(&geom) {
let geom = match geom.intersection(&self.current_state.clip) {
Some(geom) => geom,
None => return,
};
// FIXME: gradients
let color = self.alpha_color(&rect.background());
if color.alpha() == 0 {
return;
}
self.processor.process_rectangle(
(geom.translate(self.current_state.offset.to_vector()).cast() * self.scale_factor)
.round()
.cast(),
color,
);
}
}
fn draw_border_rectangle(&mut self, rect: Pin<&crate::items::BorderRectangle>, _: &ItemRc) {
let geom = LogicalRect::new(LogicalPoint::default(), rect.geometry().size_length());
if self.should_draw(&geom) {
let border = rect.border_width();
let radius = rect.border_radius();
// FIXME: gradients
let color = self.alpha_color(&rect.background());
let border_color = if border.get() as f32 > 0.01 {
self.alpha_color(&rect.border_color())
} else {
Color::default()
};
if radius.get() > 0 as _ {
let radius = radius
.min(geom.width_length() / 2 as Coord)
.min(geom.height_length() / 2 as Coord);
if let Some(clipped) = geom.intersection(&self.current_state.clip) {
let geom2 = geom.cast() * self.scale_factor;
let clipped2 = clipped.cast() * self.scale_factor;
// Add a small value to make sure that the clip is always positive despite floating point shenanigans
const E: f32 = 0.00001;
self.processor.process_rounded_rectangle(
(clipped.translate(self.current_state.offset.to_vector()).cast()
* self.scale_factor)
.round()
.cast(),
RoundedRectangle {
radius: (radius.cast() * self.scale_factor).cast(),
width: (border.cast() * self.scale_factor).cast(),
border_color: border_color.into(),
inner_color: color.into(),
top_clip: PhysicalLength::new(
(clipped2.min_y() - geom2.min_y() + E) as _,
),
bottom_clip: PhysicalLength::new(
(geom2.max_y() - clipped2.max_y() + E) as _,
),
left_clip: PhysicalLength::new(
(clipped2.min_x() - geom2.min_x() + E) as _,
),
right_clip: PhysicalLength::new(
(geom2.max_x() - clipped2.max_x() + E) as _,
),
},
);
}
return;
}
if color.alpha() > 0 {
if let Some(r) = geom
.inflate(-border.get(), -border.get())
.intersection(&self.current_state.clip)
{
self.processor.process_rectangle(
(r.translate(self.current_state.offset.to_vector()).cast()
* self.scale_factor)
.round()
.cast(),
color,
);
}
}
// FIXME: gradients
if border_color.alpha() > 0 {
let mut add_border = |r: LogicalRect| {
if let Some(r) = r.intersection(&self.current_state.clip) {
self.processor.process_rectangle(
(r.translate(self.current_state.offset.to_vector()).cast()
* self.scale_factor)
.round()
.cast(),
border_color,
);
}
};
let b = border.get();
add_border(euclid::rect(0 as _, 0 as _, geom.width(), b));
add_border(euclid::rect(0 as _, geom.height() - b, geom.width(), b));
add_border(euclid::rect(0 as _, b, b, geom.height() - b - b));
add_border(euclid::rect(geom.width() - b, b, b, geom.height() - b - b));
}
}
}
fn draw_image(&mut self, image: Pin<&crate::items::ImageItem>, _: &ItemRc) {
let geom =
LogicalRect::new(LogicalPoint::default(), image.as_ref().geometry().size_length());
if self.should_draw(&geom) {
let source = image.source();
self.draw_image_impl(
geom,
&source,
euclid::Rect::new(Default::default(), source.size().cast()),
image.image_fit(),
Default::default(),
);
}
}
fn draw_clipped_image(&mut self, image: Pin<&crate::items::ClippedImage>, _: &ItemRc) {
let geom = LogicalRect::new(LogicalPoint::default(), image.geometry().size_length());
if self.should_draw(&geom) {
let source = image.source();
let source_clip_x = image.source_clip_x();
let source_clip_y = image.source_clip_y();
let source_size = source.size();
let mut source_clip_width = image.source_clip_width();
// when the source_clip size is empty, make it full
if source_clip_width == 0 {
source_clip_width = source_size.width as i32 - source_clip_x;
}
let mut source_clip_height = image.source_clip_height();
if source_clip_height == 0 {
source_clip_height = source_size.height as i32 - source_clip_y;
}
self.draw_image_impl(
geom,
&source,
euclid::rect(source_clip_x, source_clip_y, source_clip_width, source_clip_height),
image.image_fit(),
image.colorize().color(),
);
}
}
fn draw_text(&mut self, text: Pin<&crate::items::Text>, _: &ItemRc) {
let string = text.text();
if string.trim().is_empty() {
return;
}
let geom = LogicalRect::new(LogicalPoint::default(), text.geometry().size_length());
if !self.should_draw(&geom) {
return;
}
let font_request = text.font_request(self.window);
let font = fonts::match_font(&font_request, self.scale_factor);
let layout = fonts::text_layout_for_font(&font, &font_request, self.scale_factor);
let color = self.alpha_color(&text.color());
let max_size = (geom.size.cast() * self.scale_factor).cast();
let paragraph = TextParagraphLayout {
string: &string,
layout,
max_width: max_size.width_length(),
max_height: max_size.height_length(),
horizontal_alignment: text.horizontal_alignment(),
vertical_alignment: text.vertical_alignment(),
wrap: text.wrap(),
overflow: text.overflow(),
single_line: false,
};
// Clip glyphs not only against the global clip but also against the Text's geometry to avoid drawing outside
// of its boundaries (that breaks partial rendering and the cast to usize for the item relative coordinate below).
// FIXME: we should allow drawing outside of the Text element's boundaries.
let physical_clip = if let Some(logical_clip) = self.current_state.clip.intersection(&geom)
{
logical_clip.cast() * self.scale_factor
} else {
return; // This should have been caught earlier already
};
let offset = self.current_state.offset.to_vector().cast() * self.scale_factor;
paragraph.layout_lines(|glyphs, line_x, line_y| {
let baseline_y = line_y + font.ascent();
while let Some(positioned_glyph) = glyphs.next() {
let src_rect = PhysicalRect::new(
PhysicalPoint::from_lengths(
line_x + positioned_glyph.x + positioned_glyph.platform_glyph.x(),
baseline_y
- positioned_glyph.platform_glyph.y()
- positioned_glyph.platform_glyph.height(),
),
positioned_glyph.platform_glyph.size(),
)
.cast();
if let Some(clipped_src) = src_rect.intersection(&physical_clip) {
let geometry = clipped_src.translate(offset).round();
let origin = (geometry.origin - offset.round()).cast::<usize>();
let actual_x = origin.x - src_rect.origin.x as usize;
let actual_y = origin.y - src_rect.origin.y as usize;
let stride = positioned_glyph.platform_glyph.width().get() as u16;
let geometry = geometry.cast();
self.processor.process_texture(
geometry,
SceneTexture {
data: &positioned_glyph.platform_glyph.data().as_slice()
[actual_x + actual_y * stride as usize..],
stride,
source_size: geometry.size,
format: PixelFormat::AlphaMap,
color,
alpha: self.current_state.alpha_u8(),
},
);
}
}
});
}
fn draw_text_input(&mut self, text_input: Pin<&crate::items::TextInput>, _: &ItemRc) {
text_input.geometry();
// TODO
}
#[cfg(feature = "std")]
fn draw_path(&mut self, path: Pin<&crate::items::Path>, _: &ItemRc) {
path.geometry();
// TODO
}
fn draw_box_shadow(&mut self, box_shadow: Pin<&crate::items::BoxShadow>, _: &ItemRc) {
box_shadow.geometry();
// TODO
}
fn combine_clip(
&mut self,
other: LogicalRect,
_radius: LogicalLength,
_border_width: LogicalLength,
) -> bool {
match self.current_state.clip.intersection(&other) {
Some(r) => {
self.current_state.clip = r;
true
}
None => {
self.current_state.clip = LogicalRect::default();
false
}
}
// TODO: handle radius and border
}
fn get_current_clip(&self) -> LogicalRect {
self.current_state.clip
}
fn translate(&mut self, distance: LogicalVector) {
self.current_state.offset += distance;
self.current_state.clip = self.current_state.clip.translate(-distance)
}
fn rotate(&mut self, _angle_in_degrees: f32) {
todo!()
}
fn apply_opacity(&mut self, opacity: f32) {
self.current_state.alpha *= opacity;
}
fn save_state(&mut self) {
self.state_stack.push(self.current_state);
}
fn restore_state(&mut self) {
self.current_state = self.state_stack.pop().unwrap();
}
fn scale_factor(&self) -> f32 {
self.scale_factor.0
}
fn draw_cached_pixmap(
&mut self,
_: &ItemRc,
_update_fn: &dyn Fn(&mut dyn FnMut(u32, u32, &[u8])),
) {
todo!()
}
fn draw_string(&mut self, _string: &str, _color: Color) {
todo!()
}
fn window(&self) -> &crate::api::Window {
unreachable!("this backend don't query the window")
}
fn as_any(&mut self) -> Option<&mut dyn core::any::Any> {
None
}
}
/// This is a minimal adaptor for a Window that doesn't have any other feature than rendering
/// using the software renderer.
///
/// The [`MAX_BUFFER_AGE`](SoftwareRenderer#max_buffer_age) generic parameter is forwarded to
/// the [`SoftwareRenderer`]
pub struct MinimalSoftwareWindow<const MAX_BUFFER_AGE: usize> {
window: Window,
renderer: SoftwareRenderer<MAX_BUFFER_AGE>,
needs_redraw: Cell<bool>,
}
impl<const MAX_BUFFER_AGE: usize> MinimalSoftwareWindow<MAX_BUFFER_AGE> {
/// Instantiate a new MinimalWindowAdaptor
pub fn new() -> Rc<Self> {
Rc::new_cyclic(|w: &Weak<Self>| Self {
window: Window::new(w.clone()),
renderer: SoftwareRenderer::new(w.clone()),
needs_redraw: Default::default(),
})
}
/// If the window needs to be redrawn, the callback will be called with the
/// [renderer](SoftwareRenderer) that should be used to do the drawing.
///
/// [`SoftwareRenderer::render()`] or [`SoftwareRenderer::render_by_line()`] should be called
/// in that callback.
///
/// Return true if something was redrawn.
pub fn draw_if_needed(
&self,
render_callback: impl FnOnce(&SoftwareRenderer<MAX_BUFFER_AGE>),
) -> bool {
if self.needs_redraw.replace(false) {
render_callback(&self.renderer);
true
} else {
false
}
}
}
impl<const MAX_BUFFER_AGE: usize> crate::window::WindowAdapterSealed
for MinimalSoftwareWindow<MAX_BUFFER_AGE>
{
fn request_redraw(&self) {
self.needs_redraw.set(true);
}
fn renderer(&self) -> &dyn Renderer {
&self.renderer
}
fn register_root_component(&self, window_item: Pin<&crate::items::WindowItem>) {
let default_font_size_prop =
crate::items::WindowItem::FIELD_OFFSETS.default_font_size.apply_pin(window_item);
if default_font_size_prop.get().get() <= 0 as Coord {
default_font_size_prop.set(SoftwareRenderer::<MAX_BUFFER_AGE>::default_font_size());
}
}
}
impl<const MAX_BUFFER_AGE: usize> WindowAdapter for MinimalSoftwareWindow<MAX_BUFFER_AGE> {
fn window(&self) -> &Window {
&self.window
}
}
impl<const MAX_BUFFER_AGE: usize> core::ops::Deref for MinimalSoftwareWindow<MAX_BUFFER_AGE> {
type Target = Window;
fn deref(&self) -> &Self::Target {
&self.window
}
}