xutf 1.4.0

Permissive UTF-8/16/32 transcoding, comparison and BOM detection with SIMD ASCII fast paths
Documentation 
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
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222

//! Differential validation of the generated category/script tables against
//! the `unicode-properties` and `unicode-script` crates.
//!
//! The oracle crates may lag our UCD snapshot (see [`crate::UNICODE_VERSION`]),
//! so codepoints the oracle reports as `Unassigned` / `Unknown` are excluded:
//! newly-encoded characters diverge by version, not by bug. Everything the
//! oracle *does* assign must match exactly.

use std::{fmt::Write as _, string::String, vec::Vec};

use unicode_properties::UnicodeGeneralCategory;
use unicode_script::UnicodeScript;

use crate::ucd::{
	GeneralCategory as GC, GeneralCategoryGroup as GCG, Script, general_category,
	general_category_group, script,
};

fn oracle_category(c: char) -> GC {
	use unicode_properties::GeneralCategory as O;
	match c.general_category() {
		O::UppercaseLetter => GC::UppercaseLetter,
		O::LowercaseLetter => GC::LowercaseLetter,
		O::TitlecaseLetter => GC::TitlecaseLetter,
		O::ModifierLetter => GC::ModifierLetter,
		O::OtherLetter => GC::OtherLetter,
		O::NonspacingMark => GC::NonspacingMark,
		O::SpacingMark => GC::SpacingMark,
		O::EnclosingMark => GC::EnclosingMark,
		O::DecimalNumber => GC::DecimalNumber,
		O::LetterNumber => GC::LetterNumber,
		O::OtherNumber => GC::OtherNumber,
		O::ConnectorPunctuation => GC::ConnectorPunctuation,
		O::DashPunctuation => GC::DashPunctuation,
		O::OpenPunctuation => GC::OpenPunctuation,
		O::ClosePunctuation => GC::ClosePunctuation,
		O::InitialPunctuation => GC::InitialPunctuation,
		O::FinalPunctuation => GC::FinalPunctuation,
		O::OtherPunctuation => GC::OtherPunctuation,
		O::MathSymbol => GC::MathSymbol,
		O::CurrencySymbol => GC::CurrencySymbol,
		O::ModifierSymbol => GC::ModifierSymbol,
		O::OtherSymbol => GC::OtherSymbol,
		O::SpaceSeparator => GC::SpaceSeparator,
		O::LineSeparator => GC::LineSeparator,
		O::ParagraphSeparator => GC::ParagraphSeparator,
		O::Control => GC::Control,
		O::Format => GC::Format,
		O::Surrogate => GC::Surrogate,
		O::PrivateUse => GC::PrivateUse,
		O::Unassigned => GC::Unassigned,
	}
}

/// Renders contiguous diffs as ranges for triage, then fails when any exist.
fn assert_no_diffs<A, B>(diffs: &[(u32, A, B)])
where
	A: Copy + core::fmt::Debug + PartialEq,
	B: Copy + core::fmt::Debug + PartialEq,
{
	let mut msg = String::new();
	let mut groups = 0usize;
	let mut i = 0usize;
	while i < diffs.len() {
		let (start, ours, theirs) = diffs[i];
		let mut end = start;
		while i + 1 < diffs.len() {
			let (cp, o, t) = diffs[i + 1];
			if cp != end + 1 || o != ours || t != theirs {
				break;
			}
			end = cp;
			i += 1;
		}
		groups += 1;
		if groups <= 40 {
			let _ = writeln!(msg, "U+{start:04X}..=U+{end:04X}: ours {ours:?}, oracle {theirs:?}");
		}
		i += 1;
	}
	assert!(diffs.is_empty(), "{} diffs in {groups} ranges:\n{msg}", diffs.len());
}

/// Every scalar the oracle assigns must have an identical category.
/// When compiled against UCD 16 (default), UCD 17 additions in the oracle
/// (ours Unassigned) are skipped.
#[test]
fn category_matches_unicode_properties() {
	let mut diffs: Vec<(u32, GC, GC)> = Vec::new();
	for cp in (0..0x11_0000u32).filter_map(char::from_u32) {
		let ours = general_category(cp as u32);
		let theirs = oracle_category(cp);
		if ours != theirs {
			#[cfg(not(feature = "ucd-17"))]
			if ours == GC::Unassigned
				|| (cp as u32 == 0x0295 && ours == GC::LowercaseLetter && theirs == GC::OtherLetter)
			{
				continue;
			}
			if theirs != GC::Unassigned {
				diffs.push((cp as u32, ours, theirs));
			}
		}
	}
	assert_no_diffs(&diffs);
}

/// Same contract for Script, keyed by UCD long name (our variants are the
/// long names with underscores removed).
#[test]
fn script_matches_unicode_script() {
	let mut diffs: Vec<(u32, Script, unicode_script::Script)> = Vec::new();
	for cp in (0..0x11_0000u32).filter_map(char::from_u32) {
		let ours = script(cp as u32);
		let theirs = cp.script();
		if theirs == unicode_script::Script::Unknown {
			continue;
		}
		#[cfg(not(feature = "ucd-17"))]
		if ours == Script::Unknown {
			continue;
		}
		if format!("{ours:?}") != theirs.full_name().replace('_', "") {
			diffs.push((cp as u32, ours, theirs));
		}
	}
	assert_no_diffs(&diffs);
}

/// The pi-repo call sites in one place: tokenizer scanner classes
/// (`\p{L}`/`\p{N}`/`\p{M}`/`\p{Han}`), digit-class splits, and the
/// permissive out-of-range/surrogate conventions.
#[test]
fn call_site_semantics() {
	// \p{Han} includes Nl 〇 and Lm 々 (regex-syntax parity).
	assert_eq!(script('中' as u32), Script::Han);
	assert_eq!(script('〇' as u32), Script::Han);
	assert_eq!(script('々' as u32), Script::Han);
	assert_eq!(script('a' as u32), Script::Latin);
	assert_eq!(script(0x0300), Script::Inherited); // combining grave
	assert_eq!(script(' ' as u32), Script::Common);

	assert_eq!(general_category('A' as u32), GC::UppercaseLetter);
	assert_eq!(general_category('Dž' as u32), GC::TitlecaseLetter);
	assert_eq!(general_category('0' as u32), GC::DecimalNumber);
	assert_eq!(general_category(0x0660), GC::DecimalNumber); // Arabic-Indic zero
	assert_eq!(general_category('½' as u32), GC::OtherNumber);
	assert_eq!(general_category_group('中' as u32), GCG::Letter);
	assert_eq!(general_category_group(0x0300), GCG::Mark);
	assert_eq!(general_category_group(' ' as u32), GCG::Separator);
	assert_eq!(general_category_group('!' as u32), GCG::Punctuation);
	assert_eq!(general_category_group('+' as u32), GCG::Symbol);
	assert_eq!(general_category_group('\t' as u32), GCG::Other);

	// Permissive u32 inputs `char`-based crates cannot classify.
	assert_eq!(general_category(0xd800), GC::Surrogate);
	assert_eq!(script(0xd800), Script::Unknown);
	assert_eq!(general_category(0x11_0000), GC::Unassigned);
	assert_eq!(script(0x11_0000), Script::Unknown);
	assert_eq!(general_category(u32::MAX), GC::Unassigned);
}

/// The [`crate::Ucd`] methods agree with the free functions for `u32`
/// (permissive inputs included) and `char`. Scoped import and qualified
/// paths: the oracle crates' traits also add these method names to `char`
/// in this module.
#[test]
fn trait_matches_free_functions() {
	use crate::ucd::Ucd;
	for cp in [0x41u32, 0x0300, 0x4e2d, 0xd800, 0x11_0000, u32::MAX] {
		assert_eq!(Ucd::general_category(cp), general_category(cp));
		assert_eq!(Ucd::general_category_group(cp), general_category_group(cp));
		assert_eq!(Ucd::script(cp), script(cp));
	}
	assert_eq!(Ucd::script('中'), Script::Han);
	assert_eq!(Ucd::general_category('A'), GC::UppercaseLetter);
	assert_eq!(Ucd::general_category_group('0'), GCG::Number);
}

/// Group derivation covers all 30 categories with the UAX #44 first-letter
/// rule.
#[test]
fn groups_are_exhaustive() {
	use GCG as G;
	let by_group: [(&[GC], G); 7] = [
		(
			&[
				GC::UppercaseLetter,
				GC::LowercaseLetter,
				GC::TitlecaseLetter,
				GC::ModifierLetter,
				GC::OtherLetter,
			],
			G::Letter,
		),
		(&[GC::NonspacingMark, GC::SpacingMark, GC::EnclosingMark], G::Mark),
		(&[GC::DecimalNumber, GC::LetterNumber, GC::OtherNumber], G::Number),
		(
			&[
				GC::ConnectorPunctuation,
				GC::DashPunctuation,
				GC::OpenPunctuation,
				GC::ClosePunctuation,
				GC::InitialPunctuation,
				GC::FinalPunctuation,
				GC::OtherPunctuation,
			],
			G::Punctuation,
		),
		(&[GC::MathSymbol, GC::CurrencySymbol, GC::ModifierSymbol, GC::OtherSymbol], G::Symbol),
		(&[GC::SpaceSeparator, GC::LineSeparator, GC::ParagraphSeparator], G::Separator),
		(&[GC::Control, GC::Format, GC::Surrogate, GC::PrivateUse, GC::Unassigned], G::Other),
	];
	let mut seen = 0usize;
	for (cats, group) in by_group {
		for &cat in cats {
			assert_eq!(cat.group(), group, "{cat:?}");
			seen += 1;
		}
	}
	assert_eq!(seen, 30);
}