characterminator 0.1.2

find and eliminate characters outside an allowed set, per file type and per file, so text costs fewer tokens
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
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
//! The `.ctrm-sets` grammar: one line declares one named set.
//!
//! `<name> <member>...`, where a member is a literal character, `U+XXXX`,
//! `U+XXXX-U+YYYY`, or the name of another set (V25). The grammar lives
//! with its parser, and builtin preset data files are written in the same
//! grammar (V22), so this is the only reader of either.
//!
//! The unit is a LINE, not a file: `--set <line>` is the flag twin of a
//! `.ctrm-sets` line (`src/rules:V18`), and both must reach the same
//! parser. Reading files and ordering them by precedence belongs to
//! `src/rules`.

use super::CharRange;
use std::fmt;

/// What a set is built from.
#[derive(Debug, Clone, PartialEq, Eq)]
pub(crate) enum SetMember {
    /// A single character, written as itself.
    Literal(char),
    /// A span, written `U+XXXX` or `U+XXXX-U+YYYY`.
    Range(CharRange),
    /// The name of another set, to be composed in (V25).
    Named(String),
    /// A member granted only at ONE fidelity, written `<family>:<member>`
    /// (V41). The notation is `src/fix:V28`'s, because a labelled member
    /// here and a labelled class member there name the same families.
    Labelled {
        family: String,
        member: Box<SetMember>,
    },
}

/// One parsed line: a name and the members granted to it.
///
/// Members stay UNRESOLVED here. A line can name a set declared later in
/// the same file, or in a file loaded after it, so resolving at parse time
/// would make a definition's meaning depend on the order lines are read.
#[derive(Debug, Clone, PartialEq, Eq)]
pub(crate) struct SetDefinition {
    pub name: String,
    pub members: Vec<SetMember>,
}

/// Why a line could not be read.
#[derive(Debug, Clone, PartialEq, Eq)]
pub(crate) enum ParseError {
    /// A name with no members: an empty set is almost always a typo.
    NoMembers { name: String },
    /// A name that could never be referenced as a member, because it would
    /// be read as something else (V25).
    UnusableName { name: String },
    /// A `U+XXXX` token that is not a code point.
    BadCodePoint { token: String },
    /// A range whose ends are the wrong way round.
    ReversedRange { token: String },
    /// A `<family>:<member>` token whose two halves do not both hold
    /// (V41): no family, no member, or a label on a label.
    BadLabel { token: String },
}

/// The prefix that marks a code point, per V22.
const CODE_POINT: &str = "U+";

/// The marker for a whole-line comment.
const COMMENT: char = '\u{0023}';

/// Why a token that is not a name cannot declare a set (V25).
const UNUSABLE: &str = "cannot name a set: a name is 2 or more \
                        characters and is never written U+XXXX";

/// Why a labelled member cannot be read (V41).
const LABEL: &str = "is not `<family>:<member>`: both halves are required \
                     and a label takes no second label";

/// What separates a fidelity label from the member it governs (V41).
const LABEL_MARK: char = '\u{003A}';

impl ParseError {
    /// The offending token, whichever kind of fault this is.
    ///
    /// Every message is built as token-then-reason so that a line naming
    /// several bad tokens reports them in one recognisable shape.
    fn token(&self) -> &str {
        match self {
            Self::NoMembers { name } | Self::UnusableName { name } => name,
            Self::BadCodePoint { token }
            | Self::ReversedRange { token }
            | Self::BadLabel { token } => token,
        }
    }

    /// What is wrong with it.
    const fn reason(&self) -> &'static str {
        match self {
            Self::NoMembers { .. } => "grants nothing",
            Self::UnusableName { .. } => UNUSABLE,
            Self::BadCodePoint { .. } => "is not a code point",
            Self::ReversedRange { .. } => "ends below where it starts",
            Self::BadLabel { .. } => LABEL,
        }
    }
}

impl fmt::Display for ParseError {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        write!(f, "'{}' {}", self.token(), self.reason())
    }
}

impl std::error::Error for ParseError {}

/// Read one line of the `.ctrm-sets` grammar.
///
/// `Ok(None)` is a line that declares nothing: blank, or a comment. Only a
/// WHOLE-line comment is recognised, because `#` is an ordinary character
/// and so a legal literal member; a trailing comment could not be told from
/// a grant of `#`.
///
/// # Errors
///
/// Returns a [`ParseError`] naming the offending token.
pub(crate) fn parse_line(
    line: &str,
) -> Result<Option<SetDefinition>, ParseError> {
    let content = line.trim();
    if content.is_empty() || content.starts_with(COMMENT) {
        return Ok(None);
    }
    let mut tokens = content.split_whitespace();
    let Some(name) = tokens.next() else {
        return Ok(None);
    };
    check_name(name)?;
    let members = tokens.map(member).collect::<Result<Vec<_>, _>>()?;
    definition(name, members)
}

/// Pair a name with its members, rejecting a set that grants nothing.
fn definition(
    name: &str,
    members: Vec<SetMember>,
) -> Result<Option<SetDefinition>, ParseError> {
    if members.is_empty() {
        return Err(ParseError::NoMembers {
            name: name.to_owned(),
        });
    }
    Ok(Some(SetDefinition {
        name: name.to_owned(),
        members,
    }))
}

/// A name must be a token that reads back as a NAME when used as a member.
///
/// A one-character name, or one spelled `U+XXXX`, would be read as a
/// literal or a code point everywhere it was referenced, so the set could
/// never be composed into another. Declaring it is a trap rather than a
/// grant, and this fails closed instead.
fn check_name(name: &str) -> Result<(), ParseError> {
    match member(name)? {
        SetMember::Named(_) => Ok(()),
        SetMember::Literal(_)
        | SetMember::Range(_)
        | SetMember::Labelled { .. } => Err(ParseError::UnusableName {
            name: name.to_owned(),
        }),
    }
}

/// Classify one whitespace-separated token.
fn member(token: &str) -> Result<SetMember, ParseError> {
    if let Some(digits) = token.strip_prefix(CODE_POINT) {
        return code_form(token, digits);
    }
    let mut chars = token.chars();
    match (chars.next(), chars.next()) {
        (Some(only), None) => Ok(SetMember::Literal(only)),
        _ => unlabelled_or_labelled(token),
    }
}

/// A token of two or more characters: a set name, or a member carrying a
/// fidelity label (V41).
///
/// The label is recognised before the name, so `emoji:U+2705` is one
/// labelled member rather than a set called `emoji:U+2705`. A name may
/// therefore not contain the mark, which costs nothing: every name in the
/// grammar so far is a word.
fn unlabelled_or_labelled(token: &str) -> Result<SetMember, ParseError> {
    match token.split_once(LABEL_MARK) {
        None => Ok(SetMember::Named(token.to_owned())),
        Some((family, rest)) => labelled(token, family, rest),
    }
}

/// The two halves of a `<family>:<member>` token, both required.
fn labelled(
    token: &str,
    family: &str,
    rest: &str,
) -> Result<SetMember, ParseError> {
    let bad = || ParseError::BadLabel {
        token: token.to_owned(),
    };
    if family.is_empty() || rest.is_empty() {
        return Err(bad());
    }
    match member(rest)? {
        // A label on a label says a member belongs to two families at
        // once, which V41 has no meaning for.
        SetMember::Labelled { .. } => Err(bad()),
        inner => Ok(SetMember::Labelled {
            family: family.to_owned(),
            member: Box::new(inner),
        }),
    }
}

/// A token that began `U+`: either one code point or a span.
fn code_form(token: &str, digits: &str) -> Result<SetMember, ParseError> {
    match digits.split_once('\u{002D}') {
        None => code_point(token, digits)
            .map(|point| SetMember::Range(CharRange::single(point))),
        Some((low, high)) => span(token, low, high),
    }
}

/// `U+XXXX-U+YYYY`: the high end repeats the prefix, per V22.
fn span(token: &str, low: &str, high: &str) -> Result<SetMember, ParseError> {
    let upper = high.strip_prefix(CODE_POINT).ok_or_else(|| {
        ParseError::BadCodePoint {
            token: token.to_owned(),
        }
    })?;
    let start = code_point(token, low)?;
    let end = code_point(token, upper)?;
    CharRange::new(start, end)
        .map(SetMember::Range)
        .ok_or_else(|| ParseError::ReversedRange {
            token: token.to_owned(),
        })
}

/// Hexadecimal digits to a character, or the token named as bad.
fn code_point(token: &str, digits: &str) -> Result<char, ParseError> {
    hex(digits).ok_or_else(|| ParseError::BadCodePoint {
        token: token.to_owned(),
    })
}

/// `U+XXXX`, or a sequence `U+XXXX+U+YYYY...`, as the text it spells.
///
/// The ONE decoder of the code-point spelling every grammar here shares:
/// a set member, a map token (`src/fix`) and a line of the vendored emoji
/// sequences (`src/lint/hazard:V63`). `None` for anything else, so each caller
/// names the refusal in its own grammar's words.
pub(crate) fn code_points(token: &str) -> Option<String> {
    token
        .strip_prefix(CODE_POINT)?
        .split("+U+")
        .map(hex)
        .collect()
}

/// Hexadecimal digits to a character.
///
/// The digits are checked before conversion because `from_str_radix` also
/// accepts a leading sign, which would read `U++41` as a letter.
fn hex(digits: &str) -> Option<char> {
    if digits.is_empty() || !digits.bytes().all(|b| b.is_ascii_hexdigit()) {
        return None;
    }
    char::from_u32(u32::from_str_radix(digits, 16).ok()?)
}

#[cfg(test)]
mod tests {
    use super::{CharRange, ParseError, SetMember, parse_line};

    fn members(line: &str) -> Result<Vec<SetMember>, ParseError> {
        parse_line(line)
            .map(|parsed| parsed.map(|set| set.members).unwrap_or_default())
    }

    #[test]
    fn blank_and_comment_declare_nothing() {
        assert_eq!(parse_line("   "), Ok(None));
        assert_eq!(parse_line("# a note"), Ok(None));
        assert_eq!(parse_line("  # indented note"), Ok(None));
    }

    #[test]
    fn a_single_character_is_a_literal() {
        let one = Ok(vec![SetMember::Literal('\u{00A9}')]);
        assert_eq!(members("legal \u{00A9}"), one);
    }

    #[test]
    fn a_two_character_token_is_a_set_name() {
        let named = Ok(vec![SetMember::Named("box".to_owned())]);
        assert_eq!(members("spec box"), named);
    }

    #[test]
    fn a_code_point_is_a_single_range() {
        let point = SetMember::Range(CharRange::single('\u{2014}'));
        assert_eq!(members("dash U+2014"), Ok(vec![point]));
    }

    #[test]
    fn a_span_is_a_range() {
        let range = CharRange {
            start: '\u{2500}',
            end: '\u{257F}',
        };
        let span = SetMember::Range(range);
        assert_eq!(members("box U+2500-U+257F"), Ok(vec![span]));
    }

    #[test]
    fn the_name_itself_is_not_a_member() {
        assert_eq!(
            parse_line("ab cd").map(|set| set.map(|s| s.name)),
            Ok(Some("ab".to_owned()))
        );
    }

    #[test]
    fn a_set_granting_nothing_is_rejected() {
        let err = Err(ParseError::NoMembers {
            name: "lonely".to_owned(),
        });
        assert_eq!(parse_line("lonely"), err);
    }

    #[test]
    fn a_one_character_name_is_rejected() {
        let err = Err(ParseError::UnusableName {
            name: "x".to_owned(),
        });
        assert_eq!(parse_line("x U+2014"), err);
    }

    #[test]
    fn a_name_spelled_as_a_code_point_is_rejected() {
        let err = Err(ParseError::UnusableName {
            name: "U+2014".to_owned(),
        });
        assert_eq!(parse_line("U+2014 U+2015"), err);
    }

    #[test]
    fn a_surrogate_is_not_a_code_point() {
        let err = Err(ParseError::BadCodePoint {
            token: "U+D800".to_owned(),
        });
        assert_eq!(parse_line("bad U+D800"), err);
    }

    #[test]
    fn nonsense_digits_are_rejected() {
        let err = Err(ParseError::BadCodePoint {
            token: "U+ZZZZ".to_owned(),
        });
        assert_eq!(parse_line("bad U+ZZZZ"), err);
    }

    #[test]
    fn a_signed_code_point_is_rejected() {
        let err = Err(ParseError::BadCodePoint {
            token: "U++41".to_owned(),
        });
        assert_eq!(parse_line("bad U++41"), err);
    }

    #[test]
    fn a_span_without_the_second_prefix_is_rejected() {
        let err = Err(ParseError::BadCodePoint {
            token: "U+0041-005A".to_owned(),
        });
        assert_eq!(parse_line("bad U+0041-005A"), err);
    }

    #[test]
    fn a_reversed_span_is_rejected() {
        let err = Err(ParseError::ReversedRange {
            token: "U+005A-U+0041".to_owned(),
        });
        assert_eq!(parse_line("bad U+005A-U+0041"), err);
    }

    #[test]
    fn a_hyphen_alone_is_a_literal() {
        let dash = Ok(vec![SetMember::Literal('\u{002D}')]);
        assert_eq!(members("punct \u{002D}"), dash);
    }

    #[test]
    fn several_members_keep_their_order() {
        let parsed = members("mix \u{00A9} U+2014 box");
        assert_eq!(
            parsed,
            Ok(vec![
                SetMember::Literal('\u{00A9}'),
                SetMember::Range(CharRange::single('\u{2014}')),
                SetMember::Named("box".to_owned()),
            ])
        );
    }
    #[test]
    fn a_labelled_member_carries_its_family() {
        let labelled = SetMember::Labelled {
            family: "emoji".to_owned(),
            member: Box::new(SetMember::Range(CharRange::single('\u{2705}'))),
        };
        assert_eq!(members("marks emoji:U+2705"), Ok(vec![labelled]));
    }

    #[test]
    fn a_label_may_govern_a_set_name() {
        let labelled = SetMember::Labelled {
            family: "emoji".to_owned(),
            member: Box::new(SetMember::Named("box".to_owned())),
        };
        assert_eq!(members("wide emoji:box"), Ok(vec![labelled]));
    }

    #[test]
    fn a_label_with_a_missing_half_is_rejected() {
        let no_family = Err(ParseError::BadLabel {
            token: ":U+2705".to_owned(),
        });
        assert_eq!(parse_line("marks :U+2705"), no_family);
        let no_member = Err(ParseError::BadLabel {
            token: "emoji:".to_owned(),
        });
        assert_eq!(parse_line("marks emoji:"), no_member);
    }

    /// A member belonging to two families at once is a thing V41 has no
    /// meaning for, so it is refused rather than read as one of them.
    #[test]
    fn a_label_on_a_label_is_rejected() {
        let err = Err(ParseError::BadLabel {
            token: "text:emoji:U+2705".to_owned(),
        });
        assert_eq!(parse_line("marks text:emoji:U+2705"), err);
    }

    /// The label is recognised BEFORE the name, so a name carrying the
    /// mark reads as a labelled member and cannot name a set (V25). Every
    /// name in this grammar is a word, so nothing is lost.
    #[test]
    fn a_name_may_not_carry_the_label_mark() {
        let err = Err(ParseError::UnusableName {
            name: "a:b".to_owned(),
        });
        assert_eq!(parse_line("a:b U+2705"), err);
    }
}