Mirrors/servo
1
0
Fork 0
mirror of https://github.com/servo/servo.git synced 2025-08-06 06:00:15 +01:00
Code Issues Projects Releases Packages Wiki Activity

Use byte indices instead of char indices for text runs

Browse source 
Replace character indices with UTF-8 byte offsets throughout the code dealing
with text shaping and breaking.  This eliminates a lot of complexity when
converting from one to the other, and interoperates better with the rest of
the Rust ecosystem.
This commit is contained in:
Matt Brubeck 2016-04-27 11:22:02 -07:00
parent dba878dfb2
commit 659305fe0a
15 changed files with 259 additions and 437 deletions
Download patch file Download diff file Expand all files Collapse all files

149
components/gfx/text/shaping/harfbuzz.rs
View file

@ -35,7 +35,7 @@ use harfbuzz::{hb_position_t, hb_tag_t};
use libc::{c_char, c_int, c_uint, c_void};
use platform::font::FontTable;
use std::{char, cmp, ptr};
use text::glyph::{CharIndex, GlyphData, GlyphId, GlyphStore};
use text::glyph::{ByteIndex, GlyphData, GlyphId, GlyphStore};
use text::shaping::ShaperMethods;
use text::util::{fixed_to_float, float_to_fixed, is_bidi_control};
@ -45,8 +45,7 @@ macro_rules! hb_tag {
);
}
static NO_GLYPH: i32 = -1;
static CONTINUATION_BYTE: i32 = -2;
const NO_GLYPH: i32 = -1;
static KERN: u32 = hb_tag!('k', 'e', 'r', 'n');
static LIGA: u32 = hb_tag!('l', 'i', 'g', 'a');
@ -258,44 +257,18 @@ impl Shaper {
let glyph_data = ShapedGlyphData::new(buffer);
let glyph_count = glyph_data.len();
let byte_max = text.len();
let char_max = text.chars().count();
// GlyphStore records are indexed by character, not byte offset.
// so, we must be careful to increment this when saving glyph entries.
let (mut char_idx, char_step) = if options.flags.contains(RTL_FLAG) {
(CharIndex(char_max as isize - 1), CharIndex(-1))
} else {
(CharIndex(0), CharIndex(1))
};
debug!("Shaped text[char count={}], got back {} glyph info records.",
char_max,
debug!("Shaped text[byte count={}], got back {} glyph info records.",
byte_max,
glyph_count);
if char_max != glyph_count {
debug!("NOTE: Since these are not equal, we probably have been given some complex \
glyphs.");
}
// make map of what chars have glyphs
let mut byte_to_glyph: Vec<i32>;
// fast path: all chars are single-byte.
if byte_max == char_max {
byte_to_glyph = vec![NO_GLYPH; byte_max];
} else {
byte_to_glyph = vec![CONTINUATION_BYTE; byte_max];
for (i, _) in text.char_indices() {
byte_to_glyph[i] = NO_GLYPH;
}
}
let mut byte_to_glyph = vec![NO_GLYPH; byte_max];
debug!("(glyph idx) -> (text byte offset)");
for i in 0..glyph_data.len() {
// loc refers to a *byte* offset within the utf8 string.
let loc = glyph_data.byte_offset_of_glyph(i) as usize;
if loc < byte_max {
assert!(byte_to_glyph[loc] != CONTINUATION_BYTE);
byte_to_glyph[loc] = i as i32;
} else {
debug!("ERROR: tried to set out of range byte_to_glyph: idx={}, glyph idx={}",
@ -312,10 +285,7 @@ impl Shaper {
}
let mut glyph_span = 0..0;
// This span contains first byte of first char, to last byte of last char in range.
// So, char_byte_span.end points to first byte of last+1 char, if it's less than byte_max.
let mut char_byte_span;
let mut byte_range = 0..0;
let mut y_pos = Au(0);
@ -325,106 +295,62 @@ impl Shaper {
while glyph_span.start < glyph_count {
debug!("Processing glyph at idx={}", glyph_span.start);
glyph_span.end = glyph_span.start;
byte_range.end = glyph_data.byte_offset_of_glyph(glyph_span.start) as usize;
let char_byte_start = glyph_data.byte_offset_of_glyph(glyph_span.start) as usize;
char_byte_span = char_byte_start..char_byte_start;
let mut glyph_spans_multiple_characters = false;
// find a range of chars corresponding to this glyph, plus
// any trailing chars that do not have associated glyphs.
while char_byte_span.end < byte_max {
let ch = text[char_byte_span.end..].chars().next().unwrap();
char_byte_span.end += ch.len_utf8();
debug!("Processing char byte span: off={}, len={} for glyph idx={}",
char_byte_span.start, char_byte_span.len(), glyph_span.start);
while char_byte_span.end != byte_max &&
byte_to_glyph[char_byte_span.end] == NO_GLYPH {
debug!("Extending char byte span to include byte offset={} with no associated \
glyph", char_byte_span.end);
let ch = text[char_byte_span.end..].chars().next().unwrap();
char_byte_span.end += ch.len_utf8();
glyph_spans_multiple_characters = true;
while byte_range.end < byte_max {
byte_range.end += 1;
// Extend the byte range to include any following byte without its own glyph.
while byte_range.end < byte_max && byte_to_glyph[byte_range.end] == NO_GLYPH {
byte_range.end += 1;
}
// extend glyph range to max glyph index covered by char_span,
// in cases where one char made several glyphs and left some unassociated chars.
// Extend the glyph range to include all glyphs covered by bytes processed so far.
let mut max_glyph_idx = glyph_span.end;
for i in char_byte_span.clone() {
if byte_to_glyph[i] > NO_GLYPH {
max_glyph_idx = cmp::max(byte_to_glyph[i] as usize + 1, max_glyph_idx);
for glyph_idx in &byte_to_glyph[byte_range.clone()] {
if *glyph_idx != NO_GLYPH {
max_glyph_idx = cmp::max(*glyph_idx as usize + 1, max_glyph_idx);
}
}
if max_glyph_idx > glyph_span.end {
glyph_span.end = max_glyph_idx;
debug!("Extended glyph span (off={}, len={}) to cover char byte span's max \
glyph index",
glyph_span.start, glyph_span.len());
debug!("Extended glyph span to {:?}", glyph_span);
}
// if there's just one glyph, then we don't need further checks.
if glyph_span.len() == 1 { break; }
// if no glyphs were found yet, extend the char byte range more.
if glyph_span.len() == 0 { continue; }
debug!("Complex (multi-glyph to multi-char) association found. This case \
probably doesn't work.");
// If byte_range now includes all the byte offsets found in glyph_span, then we
// have found a contiguous "cluster" and can stop extending it.
let mut all_glyphs_are_within_cluster: bool = true;
for j in glyph_span.clone() {
let loc = glyph_data.byte_offset_of_glyph(j);
if !char_byte_span.contains(loc as usize) {
if !byte_range.contains(loc as usize) {
all_glyphs_are_within_cluster = false;
break
}
}
debug!("All glyphs within char_byte_span cluster?: {}",
all_glyphs_are_within_cluster);
// found a valid range; stop extending char_span.
if all_glyphs_are_within_cluster {
break
}
// Otherwise, the bytes we have seen so far correspond to a non-contiguous set of
// glyphs. Keep extending byte_range until we fill in all the holes in the glyph
// span or reach the end of the text.
}
// character/glyph clump must contain characters.
assert!(char_byte_span.len() > 0);
// character/glyph clump must contain glyphs.
assert!(byte_range.len() > 0);
assert!(glyph_span.len() > 0);
// now char_span is a ligature clump, formed by the glyphs in glyph_span.
// we need to find the chars that correspond to actual glyphs (char_extended_span),
//and set glyph info for those and empty infos for the chars that are continuations.
// Now byte_range is the ligature clump formed by the glyphs in glyph_span.
// We will save these glyphs to the glyph store at the index of the first byte.
let byte_idx = ByteIndex(byte_range.start as isize);
// a simple example:
// chars: 'f' 't' 't'
// glyphs: 'ftt' '' ''
// cgmap: t f f
// gspan: [-]
// cspan: [-]
// covsp: [---------------]
let mut covered_byte_span = char_byte_span.clone();
// extend, clipping at end of text range.
while covered_byte_span.end < byte_max &&
byte_to_glyph[covered_byte_span.end] == NO_GLYPH {
let ch = text[covered_byte_span.end..].chars().next().unwrap();
covered_byte_span.end += ch.len_utf8();
}
if covered_byte_span.start >= byte_max {
// oops, out of range. clip and forget this clump.
glyph_span.start = glyph_span.end;
char_byte_span.start = char_byte_span.end;
}
// fast path: 1-to-1 mapping of single char and single glyph.
if glyph_span.len() == 1 && !glyph_spans_multiple_characters {
if glyph_span.len() == 1 {
// Fast path: 1-to-1 mapping of byte offset to single glyph.
//
// TODO(Issue #214): cluster ranges need to be computed before
// shaping, and then consulted here.
// for now, just pretend that every character is a cluster start.
@ -433,7 +359,7 @@ impl Shaper {
//
// NB: When we acquire the ability to handle ligatures that cross word boundaries,
// we'll need to do something special to handle `word-spacing` properly.
let character = text[char_byte_span.clone()].chars().next().unwrap();
let character = text[byte_range.clone()].chars().next().unwrap();
if is_bidi_control(character) {
// Don't add any glyphs for bidi control chars
} else if character == '\t' {
@ -449,7 +375,7 @@ impl Shaper {
Default::default(),
true,
true);
glyphs.add_glyph_for_char_index(char_idx, character, &data);
glyphs.add_glyph_for_byte_index(byte_idx, character, &data);
} else {
let shape = glyph_data.entry_for_glyph(glyph_span.start, &mut y_pos);
let advance = self.advance_for_shaped_glyph(shape.advance, character, options);
@ -458,7 +384,7 @@ impl Shaper {
shape.offset,
true,
true);
glyphs.add_glyph_for_char_index(char_idx, character, &data);
glyphs.add_glyph_for_byte_index(byte_idx, character, &data);
}
} else {
// collect all glyphs to be assigned to the first character.
@ -473,15 +399,12 @@ impl Shaper {
glyph_i > glyph_span.start));
// all but first are ligature continuations
}
// now add the detailed glyph entry.
glyphs.add_glyphs_for_char_index(char_idx, &datas);
glyphs.add_glyphs_for_byte_index(byte_idx, &datas);
}
// shift up our working spans past things we just handled.
glyph_span.start = glyph_span.end;
char_byte_span.start = char_byte_span.end;
char_idx = char_idx + char_step;
byte_range.start = byte_range.end;
}
// this must be called after adding all glyph data; it sorts the