Files
servo/components/layout/flow/inline/text_run.rs
Martin Robinson 34578288c7 layout: Remove some newline-related code in shaping
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
2026-04-24 13:35:19 +02:00

666 lines
25 KiB
Rust
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at https://mozilla.org/MPL/2.0/. */
use std::mem;
use std::ops::Range;
use std::sync::Arc;
use app_units::Au;
use fonts::{FontContext, FontRef, GlyphStore, ShapingFlags, ShapingOptions};
use icu_locid::subtags::Language;
use icu_properties::{self, LineBreak};
use log::warn;
use malloc_size_of_derive::MallocSizeOf;
use servo_arc::Arc as ServoArc;
use servo_base::text::is_bidi_control;
use style::Zero;
use style::computed_values::text_rendering::T as TextRendering;
use style::computed_values::white_space_collapse::T as WhiteSpaceCollapse;
use style::computed_values::word_break::T as WordBreak;
use style::properties::ComputedValues;
use style::str::char_is_whitespace;
use style::values::computed::OverflowWrap;
use unicode_bidi::{BidiInfo, Level};
use unicode_script::Script;
use super::line_breaker::LineBreaker;
use super::{InlineFormattingContextLayout, SharedInlineStyles};
use crate::context::LayoutContext;
use crate::dom::WeakLayoutBox;
use crate::flow::inline::line::TextRunOffsets;
use crate::fragment_tree::BaseFragmentInfo;
// There are two reasons why we might want to break at the start:
//
// 1. The line breaker told us that a break was necessary between two separate
// instances of sending text to it.
// 2. We are following replaced content ie `have_deferred_soft_wrap_opportunity`.
//
// In both cases, we don't want to do this if the first character prevents a
// soft wrap opportunity.
#[derive(PartialEq)]
enum SegmentStartSoftWrapPolicy {
Force,
FollowLinebreaker,
}
/// A data structure which contains information used when shaping a [`TextRunSegment`].
#[derive(Clone, Debug, MallocSizeOf)]
pub(crate) struct FontAndScriptInfo {
/// The font used when shaping a [`TextRunSegment`].
pub font: FontRef,
/// The script used when shaping a [`TextRunSegment`].
pub script: Script,
/// The BiDi [`Level`] used when shaping a [`TextRunSegment`].
pub bidi_level: Level,
/// The [`Language`] used when shaping a [`TextRunSegment`].
pub language: Language,
/// Spacing to add between each letter. Corresponds to the CSS 2.1 `letter-spacing` property.
/// NB: You will probably want to set the `IGNORE_LIGATURES_SHAPING_FLAG` if this is non-null.
///
/// Letter spacing is not applied to all characters. Use [Self::letter_spacing_for_character] to
/// determine the amount of spacing to apply.
pub letter_spacing: Option<Au>,
/// Spacing to add between each word. Corresponds to the CSS 2.1 `word-spacing` property.
pub word_spacing: Option<Au>,
/// The [`TextRendering`] value from the original style.
pub text_rendering: TextRendering,
}
impl FontAndScriptInfo {
/// Creates a minimal [`FontAndScriptInfo`] for a single font, with generic language settings
/// and the default shaping configuration. This is only used to generate placeholders for
/// text carets on otherwise empty lines.
pub(crate) fn simple_for_font(font: FontRef) -> Self {
Self {
font,
script: Script::Common,
bidi_level: Level::ltr(),
language: Language::UND,
letter_spacing: None,
word_spacing: None,
text_rendering: TextRendering::Auto,
}
}
}
impl From<&FontAndScriptInfo> for ShapingOptions {
fn from(info: &FontAndScriptInfo) -> Self {
let mut flags = ShapingFlags::empty();
if info.bidi_level.is_rtl() {
flags.insert(ShapingFlags::RTL_FLAG);
}
// From https://www.w3.org/TR/css-text-3/#cursive-script:
// Cursive scripts do not admit gaps between their letters for either
// justification or letter-spacing.
let letter_spacing = info
.letter_spacing
.filter(|_| !is_cursive_script(info.script));
if letter_spacing.is_some() {
flags.insert(ShapingFlags::IGNORE_LIGATURES_SHAPING_FLAG);
};
if info.text_rendering == TextRendering::Optimizespeed {
flags.insert(ShapingFlags::IGNORE_LIGATURES_SHAPING_FLAG);
flags.insert(ShapingFlags::DISABLE_KERNING_SHAPING_FLAG)
}
Self {
letter_spacing,
word_spacing: info.word_spacing,
script: info.script,
language: info.language,
flags,
}
}
}
#[derive(Debug, MallocSizeOf)]
pub(crate) struct TextRunSegment {
/// Information about the font and language used in this text run. This is produced by
/// segmenting the inline formatting context's text content by font, script, and bidi level.
#[conditional_malloc_size_of]
pub info: Arc<FontAndScriptInfo>,
/// The range of bytes in the parent [`super::InlineFormattingContext`]'s text content.
pub byte_range: Range<usize>,
/// The range of characters in the parent [`super::InlineFormattingContext`]'s text content.
pub character_range: Range<usize>,
/// Whether or not the linebreaker said that we should allow a line break at the start of this
/// segment.
pub break_at_start: bool,
/// The shaped runs within this segment.
#[conditional_malloc_size_of]
pub runs: Vec<Arc<GlyphStore>>,
}
impl TextRunSegment {
fn new(
info: Arc<FontAndScriptInfo>,
byte_range: Range<usize>,
character_range: Range<usize>,
) -> Self {
Self {
info,
byte_range,
character_range,
runs: Vec::new(),
break_at_start: false,
}
}
/// Returns true if the new `Font`, `Script` and BiDi `Level` are compatible with this segment
/// or false otherwise.
fn is_compatible(
&self,
new_font: &Option<FontRef>,
new_script: Script,
new_bidi_level: Level,
) -> bool {
if self.info.bidi_level != new_bidi_level {
return false;
}
if new_font
.as_ref()
.is_some_and(|new_font| !Arc::ptr_eq(&self.info.font, new_font))
{
return false;
}
!script_is_specific(self.info.script) ||
!script_is_specific(new_script) ||
self.info.script == new_script
}
/// Update this segment to end at the given byte and character index. The update will only ever
/// make the Script specific and will not change it otherwise.
fn update(&mut self, next_byte_index: usize, next_character_index: usize, new_script: Script) {
if !script_is_specific(self.info.script) && script_is_specific(new_script) {
self.info = Arc::new(FontAndScriptInfo {
script: new_script,
..(*self.info).clone()
});
}
self.character_range.end = next_character_index;
self.byte_range.end = next_byte_index;
}
fn layout_into_line_items(
&self,
text_run: &TextRun,
mut soft_wrap_policy: SegmentStartSoftWrapPolicy,
ifc: &mut InlineFormattingContextLayout,
) {
if self.break_at_start && soft_wrap_policy == SegmentStartSoftWrapPolicy::FollowLinebreaker
{
soft_wrap_policy = SegmentStartSoftWrapPolicy::Force;
}
let mut character_range_start = self.character_range.start;
for (run_index, run) in self.runs.iter().enumerate() {
ifc.possibly_flush_deferred_forced_line_break();
let new_character_range_end = character_range_start + run.character_count();
let offsets = ifc
.ifc
.shared_selection
.clone()
.map(|shared_selection| TextRunOffsets {
shared_selection,
character_range: character_range_start..new_character_range_end,
});
// Break before each unbreakable run in this TextRun, except the first unless the
// linebreaker was set to break before the first run.
if run_index != 0 || soft_wrap_policy == SegmentStartSoftWrapPolicy::Force {
ifc.process_soft_wrap_opportunity();
}
ifc.push_glyph_store_to_unbreakable_segment(run.clone(), text_run, &self.info, offsets);
character_range_start = new_character_range_end;
}
}
fn shape_and_push_range(
&mut self,
range: &Range<usize>,
formatting_context_text: &str,
options: &ShapingOptions,
) {
self.runs.push(
self.info
.font
.shape_text(&formatting_context_text[range.clone()], options),
);
}
/// Shape the text of this [`TextRunSegment`], first finding "words" for the shaper by processing
/// the linebreaks found in the owning [`super::InlineFormattingContext`]. Linebreaks are filtered,
/// based on the style of the parent inline box.
fn shape_text(
&mut self,
parent_style: &ComputedValues,
formatting_context_text: &str,
linebreaker: &mut LineBreaker,
) {
let options: ShapingOptions = (&*self.info).into();
// Gather the linebreaks that apply to this segment from the inline formatting context's collection
// of line breaks. Also add a simulated break at the end of the segment in order to ensure the final
// piece of text is processed.
let range = self.byte_range.clone();
let linebreaks = linebreaker.advance_to_linebreaks_in_range(self.byte_range.clone());
let linebreak_iter = linebreaks.iter().chain(std::iter::once(&range.end));
self.runs.clear();
self.runs.reserve(linebreaks.len());
self.break_at_start = false;
let text_style = parent_style.get_inherited_text().clone();
let can_break_anywhere = text_style.word_break == WordBreak::BreakAll ||
text_style.overflow_wrap == OverflowWrap::Anywhere ||
text_style.overflow_wrap == OverflowWrap::BreakWord;
let mut last_slice = self.byte_range.start..self.byte_range.start;
for break_index in linebreak_iter {
let mut options = options;
if *break_index == self.byte_range.start {
self.break_at_start = true;
continue;
}
// Extend the slice to the next UAX#14 line break opportunity.
let mut slice = last_slice.end..*break_index;
let word = &formatting_context_text[slice.clone()];
// Split off any trailing whitespace into a separate glyph run.
let mut whitespace = slice.end..slice.end;
let rev_char_indices = word.char_indices().rev().peekable();
let mut ends_with_whitespace = false;
if let Some((first_white_space_index, first_white_space_character)) = rev_char_indices
.take_while(|&(_, character)| char_is_whitespace(character))
.last()
{
ends_with_whitespace = true;
whitespace.start = slice.start + first_white_space_index;
// If line breaking for a piece of text that has `white-space-collapse:
// break-spaces` there is a line break opportunity *after* every preserved space,
// but not before. This means that we should not split off the first whitespace.
//
// An exception to this is if the style tells us that we can break in the middle of words.
if text_style.white_space_collapse == WhiteSpaceCollapse::BreakSpaces &&
!can_break_anywhere
{
whitespace.start += first_white_space_character.len_utf8();
options
.flags
.insert(ShapingFlags::ENDS_WITH_WHITESPACE_SHAPING_FLAG);
}
slice.end = whitespace.start;
}
// If there's no whitespace and `word-break` is set to `keep-all`, try increasing the slice.
// TODO: This should only happen for CJK text.
if !ends_with_whitespace &&
*break_index != self.byte_range.end &&
text_style.word_break == WordBreak::KeepAll &&
!can_break_anywhere
{
continue;
}
// Only advance the last slice if we are not going to try to expand the slice.
last_slice = slice.start..*break_index;
// Push the non-whitespace part of the range.
if !slice.is_empty() {
self.shape_and_push_range(&slice, formatting_context_text, &options);
}
if whitespace.is_empty() {
continue;
}
options.flags.insert(
ShapingFlags::IS_WHITESPACE_SHAPING_FLAG |
ShapingFlags::ENDS_WITH_WHITESPACE_SHAPING_FLAG,
);
// If `white-space-collapse: break-spaces` is active, insert a line breaking opportunity
// between each white space character in the white space that we trimmed off.
if text_style.white_space_collapse == WhiteSpaceCollapse::BreakSpaces {
let start_index = whitespace.start;
for (index, character) in formatting_context_text[whitespace].char_indices() {
let index = start_index + index;
self.shape_and_push_range(
&(index..index + character.len_utf8()),
formatting_context_text,
&options,
);
}
continue;
}
self.shape_and_push_range(&whitespace, formatting_context_text, &options);
}
}
}
/// A single item in a [`TextRun`].
#[derive(Debug, MallocSizeOf)]
pub(crate) enum TextRunItem {
/// A hard line break i.e. a "\n" as other types line breaks are normalized to "\n".
LineBreak { character_index: usize },
/// Any other text for which a font can be matched. We store a `Box` here as [`TextRunSegment`]
/// is quite a bit larger than the other enum variants.
TextSegment(Box<TextRunSegment>),
}
/// A single [`TextRun`] for the box tree. These are all descendants of
/// [`super::InlineBox`] or the root of the [`super::InlineFormattingContext`]. During
/// box tree construction, text is split into [`TextRun`]s based on their font, script,
/// etc. When these are created text is already shaped.
///
/// <https://www.w3.org/TR/css-display-3/#css-text-run>
#[derive(Debug, MallocSizeOf)]
pub(crate) struct TextRun {
/// The [`BaseFragmentInfo`] for this [`TextRun`]. Usually this comes from the
/// original text node in the DOM for the text.
pub base_fragment_info: BaseFragmentInfo,
/// A weak reference to the parent of this layout box. This becomes valid as soon
/// as the *parent* of this box is added to the tree.
pub parent_box: Option<WeakLayoutBox>,
/// The [`crate::SharedStyle`] from this [`TextRun`]s parent element. This is
/// shared so that incremental layout can simply update the parent element and
/// this [`TextRun`] will be updated automatically.
pub inline_styles: SharedInlineStyles,
/// The range of text in [`super::InlineFormattingContext::text_content`] of the
/// [`super::InlineFormattingContext`] that owns this [`TextRun`]. These are UTF-8 offsets.
pub text_range: Range<usize>,
/// The range of characters in this text in [`super::InlineFormattingContext::text_content`]
/// of the [`super::InlineFormattingContext`] that owns this [`TextRun`]. These are *not*
/// UTF-8 offsets.
pub character_range: Range<usize>,
/// The [`TextRunItem`]s of this text run. This is produced by segmenting the incoming text
/// by things such as font and script as well as separating out hard line breaks.
/// segments, and shaped.
pub items: Vec<TextRunItem>,
}
impl TextRun {
pub(crate) fn new(
base_fragment_info: BaseFragmentInfo,
inline_styles: SharedInlineStyles,
text_range: Range<usize>,
character_range: Range<usize>,
) -> Self {
Self {
base_fragment_info,
parent_box: None,
inline_styles,
text_range,
character_range,
items: Vec::new(),
}
}
pub(super) fn segment_and_shape(
&mut self,
formatting_context_text: &str,
layout_context: &LayoutContext,
linebreaker: &mut LineBreaker,
bidi_info: &BidiInfo,
) {
let parent_style = self.inline_styles.style.borrow().clone();
self.items = self.segment_text_by_font(
layout_context,
formatting_context_text,
bidi_info,
&parent_style,
);
for item in self.items.iter_mut() {
if let TextRunItem::TextSegment(text_segment) = item {
text_segment.shape_text(&parent_style, formatting_context_text, linebreaker)
}
}
}
/// Take the [`TextRun`]'s text and turn it into [`TextRunSegment`]s. Each segment has a matched
/// font and script. Fonts may differ when glyphs are found in fallback fonts.
/// [`super::InlineFormattingContext`].
fn segment_text_by_font(
&mut self,
layout_context: &LayoutContext,
formatting_context_text: &str,
bidi_info: &BidiInfo,
parent_style: &ServoArc<ComputedValues>,
) -> Vec<TextRunItem> {
let font_style = parent_style.clone_font();
let language = font_style._x_lang.0.parse().unwrap_or(Language::UND);
let font_size = font_style.font_size.computed_size().into();
let font_group = layout_context.font_context.font_group(font_style);
let inherited_text_style = parent_style.get_inherited_text().clone();
let word_spacing = Some(inherited_text_style.word_spacing.to_used_value(font_size));
let letter_spacing = inherited_text_style
.letter_spacing
.0
.to_used_value(font_size);
let letter_spacing = if !letter_spacing.is_zero() {
Some(letter_spacing)
} else {
None
};
let text_rendering = inherited_text_style.text_rendering;
let mut current: Option<TextRunSegment> = None;
let mut results = Vec::new();
let finish_current_segment =
|current: &mut Option<TextRunSegment>, results: &mut Vec<TextRunItem>| {
if let Some(current) = current.take() {
results.push(TextRunItem::TextSegment(Box::new(current)));
}
};
let text_run_text = &formatting_context_text[self.text_range.clone()];
let char_iterator = TwoCharsAtATimeIterator::new(text_run_text.chars());
// The next bytes index of the character within the entire inline formatting context's text.
let mut next_byte_index = self.text_range.start;
for (relative_character_index, (character, next_character)) in char_iterator.enumerate() {
// The current character index within the entire inline formatting context's text.
let current_character_index = self.character_range.start + relative_character_index;
let current_byte_index = next_byte_index;
next_byte_index += character.len_utf8();
if character == '\n' {
finish_current_segment(&mut current, &mut results);
results.push(TextRunItem::LineBreak {
character_index: current_character_index,
});
continue;
}
let (font, script, bidi_level) = if character_cannot_change_font(character) {
(None, Script::Common, bidi_info.levels[current_byte_index])
} else {
(
font_group.find_by_codepoint(
&layout_context.font_context,
character,
next_character,
language,
),
Script::from(character),
bidi_info.levels[current_byte_index],
)
};
// If the existing segment is compatible with the character, just merge the character into it.
if let Some(current) = current.as_mut() {
if current.is_compatible(&font, script, bidi_level) {
current.update(next_byte_index, current_character_index + 1, script);
continue;
}
}
let Some(font) = font.or_else(|| font_group.first(&layout_context.font_context)) else {
continue;
};
let info = FontAndScriptInfo {
font,
script,
bidi_level,
language,
word_spacing,
letter_spacing,
text_rendering,
};
finish_current_segment(&mut current, &mut results);
assert!(current.is_none());
current = Some(TextRunSegment::new(
Arc::new(info),
current_byte_index..next_byte_index,
current_character_index..current_character_index + 1,
));
}
finish_current_segment(&mut current, &mut results);
results
}
pub(super) fn layout_into_line_items(&self, ifc: &mut InlineFormattingContextLayout) {
if self.text_range.is_empty() {
return;
}
// If we are following replaced content, we should have a soft wrap opportunity, unless the
// first character of this `TextRun` prevents that soft wrap opportunity. If we see such a
// character it should also override the LineBreaker's indication to break at the start.
let have_deferred_soft_wrap_opportunity =
mem::replace(&mut ifc.have_deferred_soft_wrap_opportunity, false);
let mut soft_wrap_policy = match have_deferred_soft_wrap_opportunity {
true => SegmentStartSoftWrapPolicy::Force,
false => SegmentStartSoftWrapPolicy::FollowLinebreaker,
};
for item in self.items.iter() {
match item {
// If this whitespace forces a line break, queue up a hard line break the next time we
// see any content. We don't line break immediately, because we'd like to finish processing
// any ongoing inline boxes before ending the line.
TextRunItem::LineBreak { character_index } => {
ifc.possibly_flush_deferred_forced_line_break();
ifc.defer_forced_line_break_at_character_offset(*character_index);
},
TextRunItem::TextSegment(segment) => {
segment.layout_into_line_items(self, soft_wrap_policy, ifc)
},
}
soft_wrap_policy = SegmentStartSoftWrapPolicy::FollowLinebreaker;
}
}
}
/// From <https://www.w3.org/TR/css-text-3/#cursive-script>:
/// Cursive scripts do not admit gaps between their letters for either justification
/// or letter-spacing. The following Unicode scripts are included: Arabic, Hanifi
/// Rohingya, Mandaic, Mongolian, NKo, Phags Pa, Syriac
fn is_cursive_script(script: Script) -> bool {
matches!(
script,
Script::Arabic |
Script::Hanifi_Rohingya |
Script::Mandaic |
Script::Mongolian |
Script::Nko |
Script::Phags_Pa |
Script::Syriac
)
}
/// Whether or not this character should be able to change the font during segmentation. Certain
/// character are not rendered at all, so it doesn't matter what font we use to render them. They
/// should just be added to the current segment.
fn character_cannot_change_font(character: char) -> bool {
if character.is_control() {
return true;
}
if character == '\u{00A0}' {
return true;
}
if is_bidi_control(character) {
return false;
}
matches!(
icu_properties::maps::line_break().get(character),
LineBreak::CombiningMark |
LineBreak::Glue |
LineBreak::ZWSpace |
LineBreak::WordJoiner |
LineBreak::ZWJ
)
}
pub(super) fn get_font_for_first_font_for_style(
style: &ComputedValues,
font_context: &FontContext,
) -> Option<FontRef> {
let font = font_context
.font_group(style.clone_font())
.first(font_context);
if font.is_none() {
warn!("Could not find font for style: {:?}", style.clone_font());
}
font
}
pub(crate) struct TwoCharsAtATimeIterator<InputIterator> {
/// The input character iterator.
iterator: InputIterator,
/// The first character to produce in the next run of the iterator.
next_character: Option<char>,
}
impl<InputIterator> TwoCharsAtATimeIterator<InputIterator> {
fn new(iterator: InputIterator) -> Self {
Self {
iterator,
next_character: None,
}
}
}
impl<InputIterator> Iterator for TwoCharsAtATimeIterator<InputIterator>
where
InputIterator: Iterator<Item = char>,
{
type Item = (char, Option<char>);
fn next(&mut self) -> Option<Self::Item> {
// If the iterator isn't initialized do that now.
if self.next_character.is_none() {
self.next_character = self.iterator.next();
}
let character = self.next_character?;
self.next_character = self.iterator.next();
Some((character, self.next_character))
}
}
fn script_is_specific(script: Script) -> bool {
script != Script::Common && script != Script::Inherited
}