1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
|
use typst_library::layout::{Dir, Em};
use unicode_bidi::{BidiInfo, Level as BidiLevel};
use super::*;
/// A representation in which children are already layouted and text is already
/// preshaped.
///
/// In many cases, we can directly reuse these results when constructing a line.
/// Only when a line break falls onto a text index that is not safe-to-break per
/// rustybuzz, we have to reshape that portion.
pub struct Preparation<'a> {
/// The full text.
pub text: &'a str,
/// Configuration for inline layout.
pub config: &'a Config,
/// Bidirectional text embedding levels.
///
/// This is `None` if all text directions are uniform (all the base
/// direction).
pub bidi: Option<BidiInfo<'a>>,
/// Text runs, spacing and layouted elements.
pub items: Vec<(Range, Item<'a>)>,
/// Maps from byte indices to item indices.
pub indices: Vec<usize>,
/// The span mapper.
pub spans: SpanMapper,
}
impl<'a> Preparation<'a> {
/// Get the item that contains the given `text_offset`.
pub fn get(&self, offset: usize) -> &(Range, Item<'a>) {
let idx = self.indices.get(offset).copied().unwrap_or(0);
&self.items[idx]
}
/// Iterate over the items that intersect the given `sliced` range.
pub fn slice(&self, sliced: Range) -> impl Iterator<Item = &(Range, Item<'a>)> {
// Usually, we don't want empty-range items at the start of the line
// (because they will be part of the previous line), but for the first
// line, we need to keep them.
let start = match sliced.start {
0 => 0,
n => self.indices.get(n).copied().unwrap_or(0),
};
self.items[start..].iter().take_while(move |(range, _)| {
range.start < sliced.end || range.end <= sliced.end
})
}
}
/// Performs BiDi analysis and then prepares further layout by building a
/// representation on which we can do line breaking without layouting each and
/// every line from scratch.
#[typst_macros::time]
pub fn prepare<'a>(
engine: &mut Engine,
config: &'a Config,
text: &'a str,
segments: Vec<Segment<'a>>,
spans: SpanMapper,
) -> SourceResult<Preparation<'a>> {
let default_level = match config.dir {
Dir::RTL => BidiLevel::rtl(),
_ => BidiLevel::ltr(),
};
let bidi = BidiInfo::new(text, Some(default_level));
let is_bidi = bidi
.levels
.iter()
.any(|level| level.is_ltr() != default_level.is_ltr());
let mut cursor = 0;
let mut items = Vec::with_capacity(segments.len());
// Shape the text to finalize the items.
for segment in segments {
let len = segment.textual_len();
let end = cursor + len;
let range = cursor..end;
match segment {
Segment::Text(_, styles) => {
shape_range(&mut items, engine, text, &bidi, range, styles);
}
Segment::Item(item) => items.push((range, item)),
}
cursor = end;
}
// Build the mapping from byte to item indices.
let mut indices = Vec::with_capacity(text.len());
for (i, (range, _)) in items.iter().enumerate() {
indices.extend(range.clone().map(|_| i));
}
if config.cjk_latin_spacing {
add_cjk_latin_spacing(&mut items);
}
Ok(Preparation {
config,
text,
bidi: is_bidi.then_some(bidi),
items,
indices,
spans,
})
}
/// Add some spacing between Han characters and western characters. See
/// Requirements for Chinese Text Layout, Section 3.2.2 Mixed Text Composition
/// in Horizontal Written Mode
fn add_cjk_latin_spacing(items: &mut [(Range, Item)]) {
let mut items = items
.iter_mut()
.filter(|(_, x)| !matches!(x, Item::Tag(_)))
.peekable();
let mut prev: Option<&ShapedGlyph> = None;
while let Some((_, item)) = items.next() {
let Some(text) = item.text_mut() else {
prev = None;
continue;
};
// Since we only call this function in [`prepare`], we can assume that
// the Cow is owned, and `to_mut` can be called without overhead.
debug_assert!(matches!(text.glyphs, std::borrow::Cow::Owned(_)));
let mut glyphs = text.glyphs.to_mut().iter_mut().peekable();
while let Some(glyph) = glyphs.next() {
let next = glyphs.peek().map(|n| n as _).or_else(|| {
items
.peek()
.and_then(|(_, i)| i.text())
.and_then(|shaped| shaped.glyphs.first())
});
// Case 1: CJ followed by a Latin character
if glyph.is_cj_script() && next.is_some_and(|g| g.is_letter_or_number()) {
// The spacing is default to 1/4 em, and can be shrunk to 1/8 em.
glyph.x_advance += Em::new(0.25);
glyph.adjustability.shrinkability.1 += Em::new(0.125);
text.width += Em::new(0.25).at(glyph.size);
}
// Case 2: Latin followed by a CJ character
if glyph.is_cj_script() && prev.is_some_and(|g| g.is_letter_or_number()) {
glyph.x_advance += Em::new(0.25);
glyph.x_offset += Em::new(0.25);
glyph.adjustability.shrinkability.0 += Em::new(0.125);
text.width += Em::new(0.25).at(glyph.size);
}
prev = Some(glyph);
}
}
}
|
