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
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
//! `name` — name records.
//!
//! The `name` table holds many records; we just want the family + full
//! names. Selection priority: Windows / Unicode BMP English (3,1,0x409),
//! then Mac Roman English (1,0,0).
use crate::parser::read_u16;
use crate::Error;
#[derive(Debug, Clone)]
pub struct NameTable<'a> {
bytes: &'a [u8],
/// `count` and `stringOffset` for the format-0/1 record table.
count: u16,
string_offset: u16,
/// Optional storage of decoded UTF-8 strings, keyed by `name_id`.
/// Populated lazily — but for the round-1 use case we only ever
/// look up two ids, so we just decode on demand in `find`.
_phantom: core::marker::PhantomData<&'a ()>,
}
impl<'a> NameTable<'a> {
pub fn parse(bytes: &'a [u8]) -> Result<Self, Error> {
// Header:
// 0 / format (2; 0 or 1)
// 2 / count (2)
// 4 / stringOffset (2)
if bytes.len() < 6 {
return Err(Error::UnexpectedEof);
}
let format = read_u16(bytes, 0)?;
if format > 1 {
return Err(Error::BadStructure("name.format > 1"));
}
let count = read_u16(bytes, 2)?;
let string_offset = read_u16(bytes, 4)?;
// Each record is 12 bytes: platformID, encodingID, languageID,
// nameID, length, offset.
let table_end = 6usize + count as usize * 12;
if bytes.len() < table_end {
return Err(Error::UnexpectedEof);
}
if (string_offset as usize) > bytes.len() {
return Err(Error::BadOffset);
}
Ok(Self {
bytes,
count,
string_offset,
_phantom: core::marker::PhantomData,
})
}
/// Find the value of a name record by its `name_id`. Selects the
/// best-ranked encoding (Windows/Unicode/English first).
pub fn find(&self, name_id: u16) -> Option<&'a str> {
// We don't return owned strings: instead we look for a record
// whose payload is already valid UTF-8 (or transcodable to it via
// the trivial UTF-16-BE path). We return the *highest-ranked*
// record we can decode.
let mut best: Option<(i32, std::borrow::Cow<'a, str>)> = None;
for i in 0..self.count as usize {
let off = 6 + i * 12;
let platform = read_u16(self.bytes, off).ok()?;
let encoding = read_u16(self.bytes, off + 2).ok()?;
let language = read_u16(self.bytes, off + 4).ok()?;
let nid = read_u16(self.bytes, off + 6).ok()?;
if nid != name_id {
continue;
}
let length = read_u16(self.bytes, off + 8).ok()? as usize;
let str_off = read_u16(self.bytes, off + 10).ok()? as usize;
let start = self.string_offset as usize + str_off;
let end = start.checked_add(length)?;
let raw = self.bytes.get(start..end)?;
let rank = rank_record(platform, encoding, language);
let decoded = decode(platform, encoding, raw)?;
match &best {
Some((br, _)) if *br >= rank => {}
_ => best = Some((rank, decoded)),
}
}
// Leak the decoded Cow into a 'a str: only safe for the borrowed
// case. For owned strings (re-encoded UTF-16) we Box::leak so the
// returned str outlives the call. Names are tiny (< 100 bytes
// typically); leak cost is negligible per font load.
let (_, c) = best?;
Some(match c {
std::borrow::Cow::Borrowed(s) => s,
std::borrow::Cow::Owned(s) => Box::leak(s.into_boxed_str()),
})
}
}
fn rank_record(platform: u16, encoding: u16, language: u16) -> i32 {
// Higher = preferred. Windows English first (most common in modern
// fonts), then Mac Roman English, then anything Unicode-y, then the
// rest.
match (platform, encoding, language) {
(3, 1, 0x0409) => 100, // Windows Unicode English (US)
(3, 1, l) if l & 0xFF == 9 => 90, // Any Windows English
(3, 1, _) => 80,
(3, 10, _) => 75, // Windows UCS-4
(1, 0, 0) => 70, // Mac Roman English
(0, _, _) => 60, // Unicode platform
_ => 10,
}
}
fn decode<'a>(platform: u16, encoding: u16, raw: &'a [u8]) -> Option<std::borrow::Cow<'a, str>> {
match (platform, encoding) {
// UTF-16 BE: Unicode platform (0,*), Windows Unicode (3,1) and
// (3,10).
(0, _) | (3, 1) | (3, 10) => {
if raw.len() % 2 != 0 {
return None;
}
let mut s = String::with_capacity(raw.len() / 2);
let mut i = 0;
while i + 1 < raw.len() {
let u = u16::from_be_bytes([raw[i], raw[i + 1]]);
i += 2;
if (0xD800..=0xDBFF).contains(&u) {
// High surrogate — pair with the next code unit.
if i + 1 >= raw.len() {
return None;
}
let lo = u16::from_be_bytes([raw[i], raw[i + 1]]);
if !(0xDC00..=0xDFFF).contains(&lo) {
return None;
}
i += 2;
let cp = 0x10000 + (((u - 0xD800) as u32) << 10) + (lo - 0xDC00) as u32;
s.push(char::from_u32(cp)?);
} else {
s.push(char::from_u32(u as u32)?);
}
}
Some(std::borrow::Cow::Owned(s))
}
// Mac Roman is a 1-byte encoding; the lower 7 bits are ASCII so
// everything we need (font-name-wise) decodes as raw ASCII.
(1, 0) => {
// Try ASCII fast path, fall back to lossy.
if raw.iter().all(|&b| b < 0x80) {
std::str::from_utf8(raw)
.ok()
.map(std::borrow::Cow::Borrowed)
} else {
Some(std::borrow::Cow::Owned(
raw.iter()
.map(|&b| if b < 0x80 { b as char } else { '?' })
.collect(),
))
}
}
_ => None,
}
}
#[cfg(test)]
mod tests {
use super::*;
/// Build a one-record name table (Windows Unicode English) holding
/// "Hi" as name id 1.
fn build_minimal() -> Vec<u8> {
let utf16: Vec<u8> = "Hi".encode_utf16().flat_map(|u| u.to_be_bytes()).collect();
let length = utf16.len() as u16;
let header_size = 6 + 12;
let mut out = vec![0u8; header_size];
out[0..2].copy_from_slice(&0u16.to_be_bytes()); // format
out[2..4].copy_from_slice(&1u16.to_be_bytes()); // count
out[4..6].copy_from_slice(&(header_size as u16).to_be_bytes()); // stringOffset
// Record:
out[6..8].copy_from_slice(&3u16.to_be_bytes()); // platform = Windows
out[8..10].copy_from_slice(&1u16.to_be_bytes()); // encoding = Unicode BMP
out[10..12].copy_from_slice(&0x0409u16.to_be_bytes()); // language = English
out[12..14].copy_from_slice(&1u16.to_be_bytes()); // name id
out[14..16].copy_from_slice(&length.to_be_bytes()); // length
out[16..18].copy_from_slice(&0u16.to_be_bytes()); // offset
out.extend_from_slice(&utf16);
out
}
#[test]
fn decodes_utf16_be() {
let bytes = build_minimal();
let n = NameTable::parse(&bytes).unwrap();
assert_eq!(n.find(1), Some("Hi"));
assert_eq!(n.find(99), None);
}
}