Mirrors/servo
1
0
Fork 0
mirror of https://github.com/servo/servo.git synced 2025-08-03 12:40:06 +01:00
Code Issues Projects Releases Packages Wiki Activity

Basic support for bidirectional text

Browse source
This commit is contained in:
Matt Brubeck 2015-06-26 08:19:29 -07:00
parent b386d7ae44
commit dfac8ce4a1
20 changed files with 309 additions and 107 deletions
Download patch file Download diff file Expand all files Collapse all files

2
components/layout/Cargo.toml
View file

@ -73,4 +73,4 @@ string_cache_plugin = "0.1"
euclid = "0.1"
serde = "0.4"
serde_macros = "0.4"
unicode-bidi = "0.2"

144
components/layout/inline.rs
View file

@ -28,10 +28,11 @@ use std::collections::VecDeque;
use std::fmt;
use std::mem;
use std::sync::Arc;
use std::u16;
use std::isize;
use style::computed_values::{display, overflow_x, position, text_align, text_justify};
use style::computed_values::{text_overflow, vertical_align, white_space};
use style::properties::ComputedValues;
use unicode_bidi;
use util::geometry::{Au, MAX_AU, ZERO_RECT};
use util::logical_geometry::{LogicalRect, LogicalSize, WritingMode};
use util::range::{Range, RangeIndex};
@ -66,7 +67,7 @@ static FONT_SUPERSCRIPT_OFFSET_RATIO: f32 = 0.34;
/// with a float or a horizontal wall of the containing block. The block-start
/// inline-start corner of the green zone is the same as that of the line, but
/// the green zone can be taller and wider than the line itself.
#[derive(RustcEncodable, Debug, Copy, Clone)]
#[derive(RustcEncodable, Debug, Clone)]
pub struct Line {
/// A range of line indices that describe line breaks.
///
@ -95,6 +96,11 @@ pub struct Line {
/// | 'I like' | 'truffles,' | '<img> yes' | 'I do.' |
pub range: Range<FragmentIndex>,
/// The bidirectional embedding level runs for this line, in visual order.
///
/// Can be set to `None` if the line is 100% left-to-right.
pub visual_runs: Option<Vec<(Range<FragmentIndex>, u8)>>,
/// The bounds are the exact position and extents of the line with respect
/// to the parent box.
///
@ -153,6 +159,23 @@ pub struct Line {
pub inline_metrics: InlineMetrics,
}
impl Line {
fn new(writing_mode: WritingMode,
minimum_block_size_above_baseline: Au,
minimum_depth_below_baseline: Au)
-> Line {
Line {
range: Range::empty(),
visual_runs: None,
bounds: LogicalRect::zero(writing_mode),
green_zone: LogicalSize::zero(writing_mode),
inline_metrics: InlineMetrics::new(minimum_block_size_above_baseline,
minimum_depth_below_baseline,
minimum_block_size_above_baseline),
}
}
}
int_range_index! {
#[derive(RustcEncodable)]
#[doc = "The index of a fragment in a flattened vector of DOM elements."]
@ -202,14 +225,9 @@ impl LineBreaker {
LineBreaker {
new_fragments: Vec::new(),
work_list: VecDeque::new(),
pending_line: Line {
range: Range::empty(),
bounds: LogicalRect::zero(float_context.writing_mode),
green_zone: LogicalSize::zero(float_context.writing_mode),
inline_metrics: InlineMetrics::new(minimum_block_size_above_baseline,
minimum_depth_below_baseline,
minimum_block_size_above_baseline),
},
pending_line: Line::new(float_context.writing_mode,
minimum_block_size_above_baseline,
minimum_depth_below_baseline),
floats: float_context,
lines: Vec::new(),
cur_b: Au(0),
@ -228,18 +246,10 @@ impl LineBreaker {
}
/// Reinitializes the pending line to blank data.
fn reset_line(&mut self) {
self.pending_line.range.reset(FragmentIndex(0), FragmentIndex(0));
self.pending_line.bounds = LogicalRect::new(self.floats.writing_mode,
Au(0),
self.cur_b,
Au(0),
Au(0));
self.pending_line.green_zone = LogicalSize::zero(self.floats.writing_mode);
self.pending_line.inline_metrics =
InlineMetrics::new(self.minimum_block_size_above_baseline,
self.minimum_depth_below_baseline,
self.minimum_block_size_above_baseline)
fn reset_line(&mut self) -> Line {
mem::replace(&mut self.pending_line, Line::new(self.floats.writing_mode,
self.minimum_block_size_above_baseline,
self.minimum_depth_below_baseline))
}
/// Reflows fragments for the given inline flow.
@ -260,10 +270,40 @@ impl LineBreaker {
// Do the reflow.
self.reflow_fragments(old_fragment_iter, flow, layout_context);
// Perform unicode bidirectional layout.
let para_level = flow.base.writing_mode.to_bidi_level();
// The text within a fragment is at a single bidi embedding level (because we split
// fragments on level run boundaries during flow construction), so we can build a level
// array with just one entry per fragment.
let levels: Vec<u8> = self.new_fragments.iter().map(|fragment| match fragment.specific {
SpecificFragmentInfo::ScannedText(ref info) => info.run.bidi_level,
_ => para_level
}).collect();
let mut lines = mem::replace(&mut self.lines, Vec::new());
// If everything is LTR, don't bother with reordering.
let has_rtl = levels.iter().cloned().any(unicode_bidi::is_rtl);
if has_rtl {
// Compute and store the visual ordering of the fragments within the line.
for line in &mut lines {
let range = line.range.begin().to_usize()..line.range.end().to_usize();
let runs = unicode_bidi::visual_runs(range, &levels);
line.visual_runs = Some(runs.iter().map(|run| {
let start = FragmentIndex(run.start as isize);
let len = FragmentIndex(run.len() as isize);
(Range::new(start, len), levels[run.start])
}).collect());
}
}
// Place the fragments back into the flow.
old_fragments.fragments = mem::replace(&mut self.new_fragments, vec![]);
flow.fragments = old_fragments;
flow.lines = mem::replace(&mut self.lines, Vec::new());
flow.lines = lines;
}
/// Reflows the given fragments, which have been plucked out of the inline flow.
@ -350,7 +390,7 @@ impl LineBreaker {
fn flush_current_line(&mut self) {
debug!("LineBreaker: flushing line {}: {:?}", self.lines.len(), self.pending_line);
self.strip_trailing_whitespace_from_pending_line_if_necessary();
self.lines.push(self.pending_line);
self.lines.push(self.pending_line.clone());
self.cur_b = self.pending_line.bounds.start.b + self.pending_line.bounds.size.block;
self.reset_line();
}
@ -612,7 +652,7 @@ impl LineBreaker {
line_flush_mode: LineFlushMode) {
let indentation = self.indentation_for_pending_fragment();
if self.pending_line_is_empty() {
assert!(self.new_fragments.len() <= (u16::MAX as usize));
debug_assert!(self.new_fragments.len() <= (isize::MAX as usize));
self.pending_line.range.reset(FragmentIndex(self.new_fragments.len() as isize),
FragmentIndex(0));
}
@ -922,18 +962,47 @@ impl InlineFlow {
text_align::T::left | text_align::T::right => unreachable!()
}
for fragment_index in line.range.each_index() {
let fragment = fragments.get_mut(fragment_index.to_usize());
inline_start_position_for_fragment = inline_start_position_for_fragment +
fragment.margin.inline_start;
fragment.border_box = LogicalRect::new(fragment.style.writing_mode,
inline_start_position_for_fragment,
fragment.border_box.start.b,
fragment.border_box.size.inline,
fragment.border_box.size.block);
fragment.update_late_computed_inline_position_if_necessary();
inline_start_position_for_fragment = inline_start_position_for_fragment +
fragment.border_box.size.inline + fragment.margin.inline_end;
// Lay out the fragments in visual order.
let run_count = match line.visual_runs {
Some(ref runs) => runs.len(),
None => 1
};
for run_idx in 0..run_count {
let (range, level) = match line.visual_runs {
Some(ref runs) if is_ltr => runs[run_idx],
Some(ref runs) => runs[run_count - run_idx - 1], // reverse order for RTL runs
None => (line.range, 0)
};
// If the bidi embedding direction is opposite the layout direction, lay out this
// run in reverse order.
let reverse = unicode_bidi::is_ltr(level) != is_ltr;
let fragment_indices = if reverse {
(range.end().get() - 1..range.begin().get() - 1).step_by(-1)
} else {
(range.begin().get()..range.end().get()).step_by(1)
};
for fragment_index in fragment_indices {
let fragment = fragments.get_mut(fragment_index as usize);
inline_start_position_for_fragment = inline_start_position_for_fragment +
fragment.margin.inline_start;
let border_start = if fragment.style.writing_mode.is_bidi_ltr() == is_ltr {
inline_start_position_for_fragment
} else {
line.green_zone.inline - inline_start_position_for_fragment
- fragment.margin.inline_end
- fragment.border_box.size.inline
};
fragment.border_box = LogicalRect::new(fragment.style.writing_mode,
border_start,
fragment.border_box.start.b,
fragment.border_box.size.inline,
fragment.border_box.size.block);
fragment.update_late_computed_inline_position_if_necessary();
inline_start_position_for_fragment = inline_start_position_for_fragment +
fragment.border_box.size.inline + fragment.margin.inline_end;
}
}
}
@ -1303,6 +1372,7 @@ impl Flow for InlineFlow {
self.minimum_depth_below_baseline);
scanner.scan_for_lines(self, layout_context);
// Now, go through each line and lay out the fragments inside.
let mut line_distance_from_flow_block_start = Au(0);
let line_count = self.lines.len();

2
components/layout/lib.rs
View file

@ -14,6 +14,7 @@
#![feature(plugin)]
#![feature(raw)]
#![feature(slice_chars)]
#![feature(step_by)]
#![feature(str_char)]
#![feature(unsafe_no_drop_flag)]
@ -59,6 +60,7 @@ extern crate serde;
extern crate smallvec;
extern crate string_cache;
extern crate style;
extern crate unicode_bidi;
extern crate url;
// Listed first because of macro definitions

134
components/layout/text.rs
View file

@ -10,7 +10,7 @@ use fragment::{Fragment, SpecificFragmentInfo, ScannedTextFragmentInfo, Unscanne
use inline::InlineFragments;
use gfx::font::{DISABLE_KERNING_SHAPING_FLAG, FontMetrics, IGNORE_LIGATURES_SHAPING_FLAG};
use gfx::font::{RunMetrics, ShapingFlags, ShapingOptions};
use gfx::font::{RTL_FLAG, RunMetrics, ShapingFlags, ShapingOptions};
use gfx::font_context::FontContext;
use gfx::text::glyph::CharIndex;
use gfx::text::text_run::TextRun;
@ -23,11 +23,39 @@ use style::computed_values::{line_height, text_orientation, text_rendering, text
use style::computed_values::{white_space};
use style::properties::ComputedValues;
use style::properties::style_structs::Font as FontStyle;
use unicode_bidi::{is_rtl, process_text};
use util::geometry::Au;
use util::linked_list::split_off_head;
use util::logical_geometry::{LogicalSize, WritingMode};
use util::range::{Range, RangeIndex};
/// Returns the concatenated text of a list of unscanned text fragments.
fn text(fragments: &LinkedList<Fragment>) -> String {
// FIXME: Some of this work is later duplicated in split_first_fragment_at_newline_if_necessary
// and transform_text. This code should be refactored so that the all the scanning for
// newlines is done in a single pass.
let mut text = String::new();
for fragment in fragments {
match fragment.specific {
SpecificFragmentInfo::UnscannedText(ref info) => {
match fragment.white_space() {
white_space::T::normal | white_space::T::nowrap => {
text.push_str(&info.text.replace("\n", " "));
}
white_space::T::pre => {
text.push_str(&info.text);
}
}
}
_ => {}
}
}
text
}
/// A stack-allocated object for scanning an inline flow into `TextRun`-containing `TextFragment`s.
pub struct TextRunScanner {
pub clump: LinkedList<Fragment>,
@ -45,11 +73,26 @@ impl TextRunScanner {
mut fragments: LinkedList<Fragment>)
-> InlineFragments {
debug!("TextRunScanner: scanning {} fragments for text runs...", fragments.len());
debug_assert!(!fragments.is_empty());
// Calculate bidi embedding levels, so we can split bidirectional fragments for reordering.
let text = text(&fragments);
let para_level = fragments.front().unwrap().style.writing_mode.to_bidi_level();
let bidi_info = process_text(&text, Some(para_level));
// Optimization: If all the text is LTR, don't bother splitting on bidi levels.
let bidi_levels = if bidi_info.levels.iter().cloned().any(is_rtl) {
Some(&bidi_info.levels[..])
} else {
None
};
// FIXME(pcwalton): We want to be sure not to allocate multiple times, since this is a
// performance-critical spot, but this may overestimate and allocate too much memory.
let mut new_fragments = Vec::with_capacity(fragments.len());
let mut last_whitespace = true;
let mut paragraph_bytes_processed = 0;
while !fragments.is_empty() {
// Create a clump.
split_first_fragment_at_newline_if_necessary(&mut fragments);
@ -66,6 +109,8 @@ impl TextRunScanner {
// Flush that clump to the list of fragments we're building up.
last_whitespace = self.flush_clump_to_list(font_context,
&mut new_fragments,
&mut paragraph_bytes_processed,
bidi_levels,
last_whitespace);
}
@ -84,6 +129,8 @@ impl TextRunScanner {
fn flush_clump_to_list(&mut self,
font_context: &mut FontContext,
out_fragments: &mut Vec<Fragment>,
paragraph_bytes_processed: &mut usize,
bidi_levels: Option<&[u8]>,
mut last_whitespace: bool)
-> bool {
debug!("TextRunScanner: flushing {} fragments in range", self.clump.len());
@ -137,47 +184,51 @@ impl TextRunScanner {
};
let (mut start_position, mut end_position) = (0, 0);
for character in text.chars() {
// Search for the first font in this font group that contains a glyph for this
// character.
for font_index in 0..fontgroup.fonts.len() {
if font_index < fontgroup.fonts.len() - 1 &&
fontgroup.fonts
.get(font_index)
.unwrap()
.borrow()
.glyph_index(character)
.is_none() {
continue
let mut font_index = 0;
while font_index < fontgroup.fonts.len() - 1 {
if fontgroup.fonts.get(font_index).unwrap().borrow()
.glyph_index(character)
.is_some() {
break
}
// We found the font we want to use. Now, if necessary, flush the mapping
// we were building up.
if run_info.font_index != font_index {
if run_info.text.len() > 0 {
mapping.flush(&mut mappings,
&mut run_info,
&**text,
compression,
text_transform,
&mut last_whitespace,
&mut start_position,
end_position);
run_info_list.push(run_info);
run_info = RunInfo::new();
mapping = RunMapping::new(&run_info_list[..],
&run_info,
fragment_index);
}
run_info.font_index = font_index
}
// Consume this character.
end_position += character.len_utf8();
break
font_index += 1;
}
let bidi_level = match bidi_levels {
Some(levels) => levels[*paragraph_bytes_processed],
None => 0
};
// Now, if necessary, flush the mapping we were building up.
if run_info.font_index != font_index || run_info.bidi_level != bidi_level {
if end_position > start_position {
mapping.flush(&mut mappings,
&mut run_info,
&**text,
compression,
text_transform,
&mut last_whitespace,
&mut start_position,
end_position);
}
if run_info.text.len() > 0 {
run_info_list.push(run_info);
run_info = RunInfo::new();
mapping = RunMapping::new(&run_info_list[..],
&run_info,
fragment_index);
}
run_info.font_index = font_index;
run_info.bidi_level = bidi_level;
}
// Consume this character.
end_position += character.len_utf8();
*paragraph_bytes_processed += character.len_utf8();
}
// If the mapping is zero-length, don't flush it.
@ -220,8 +271,12 @@ impl TextRunScanner {
// FIXME(https://github.com/rust-lang/rust/issues/23338)
run_info_list.into_iter().map(|run_info| {
let mut options = options;
if is_rtl(run_info.bidi_level) {
options.flags.insert(RTL_FLAG);
}
let mut font = fontgroup.fonts.get(run_info.font_index).unwrap().borrow_mut();
Arc::new(box TextRun::new(&mut *font, run_info.text, &options))
Arc::new(box TextRun::new(&mut *font, run_info.text, &options, run_info.bidi_level))
}).collect::<Vec<_>>()
};
@ -364,6 +419,8 @@ struct RunInfo {
font_index: usize,
/// A cached copy of the number of Unicode characters in the text run.
character_length: usize,
/// The bidirection embedding level of this text run.
bidi_level: u8,
}
impl RunInfo {
@ -372,6 +429,7 @@ impl RunInfo {
text: String::new(),
font_index: 0,
character_length: 0,
bidi_level: 0,
}
}
}