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
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
/*!
Crate `utf8-ranges` converts ranges of Unicode scalar values to equivalent
ranges of UTF-8 bytes. This is useful for constructing byte based automatons
that need to embed UTF-8 decoding.

See the documentation on the `Utf8Sequences` iterator for more details and
an example.

# Wait, what is this?

This is simplest to explain with an example. Let's say you wanted to test
whether a particular byte sequence was a Cyrillic character. One possible
scalar value range is `[0400-04FF]`. The set of allowed bytes for this
range can be expressed as a sequence of byte ranges:

```ignore
[D0-D3][80-BF]
```

This is simple enough: simply encode the boundaries, `0400` encodes to
`D0 80` and `04FF` encodes to `D3 BF`, and create ranges from each
corresponding pair of bytes: `D0` to `D3` and `80` to `BF`.

However, what if you wanted to add the Cyrillic Supplementary characters to
your range? Your range might then become `[0400-052F]`. The same procedure
as above doesn't quite work because `052F` encodes to `D4 AF`. The byte ranges
you'd get from the previous transformation would be `[D0-D4][80-AF]`. However,
this isn't quite correct because this range doesn't capture many characters,
for example, `04FF` (because its last byte, `BF` isn't in the range `80-AF`).

Instead, you need multiple sequences of byte ranges:

```ignore
[D0-D3][80-BF]  # matches codepoints 0400-04FF
[D4][80-AF]     # matches codepoints 0500-052F
```

This gets even more complicated if you want bigger ranges, particularly if
they naively contain surrogate codepoints. For example, the sequence of byte
ranges for the basic multilingual plane (`[0000-FFFF]`) look like this:

```ignore
[0-7F]
[C2-DF][80-BF]
[E0][A0-BF][80-BF]
[E1-EC][80-BF][80-BF]
[ED][80-9F][80-BF]
[EE-EF][80-BF][80-BF]
```

Note that the byte ranges above will *not* match any erroneous encoding of
UTF-8, including encodings of surrogate codepoints.

And, of course, for all of Unicode (`[000000-10FFFF]`):

```ignore
[0-7F]
[C2-DF][80-BF]
[E0][A0-BF][80-BF]
[E1-EC][80-BF][80-BF]
[ED][80-9F][80-BF]
[EE-EF][80-BF][80-BF]
[F0][90-BF][80-BF][80-BF]
[F1-F3][80-BF][80-BF][80-BF]
[F4][80-8F][80-BF][80-BF]
```

This crate automates the process of creating these byte ranges from ranges of
Unicode scalar values.

# Why would I ever use this?

You probably won't ever need this. In 99% of cases, you just decode the byte
sequence into a Unicode scalar value and compare scalar values directly.
However, this explicit decoding step isn't always possible. For example, the
construction of some finite state machines may benefit from converting ranges
of scalar values into UTF-8 decoder automata (e.g., for character classes in
regular expressions).

# Lineage

I got the idea and general implementation strategy from Russ Cox in his
[article on regexps](https://swtch.com/~rsc/regexp/regexp3.html) and RE2.
Russ Cox got it from Ken Thompson's `grep` (no source, folk lore?).
I also got the idea from
[Lucene](https://github.com/apache/lucene-solr/blob/trunk/lucene/core/src/java/org/apache/lucene/util/automaton/UTF32ToUTF8.java),
which uses it for executing automata on their term index.
*/

#![deny(missing_docs)]

#[cfg(test)] extern crate quickcheck;

use std::char;
use std::fmt;
use std::slice;

use char_utf8::encode_utf8;

const MAX_UTF8_BYTES: usize = 4;

mod char_utf8;

/// Utf8Sequence represents a sequence of byte ranges.
///
/// To match a Utf8Sequence, a candidate byte sequence must match each
/// successive range.
///
/// For example, if there are two ranges, `[C2-DF][80-BF]`, then the byte
/// sequence `\xDD\x61` would not match because `0x61 < 0x80`.
#[derive(Copy, Clone, Eq, PartialEq)]
pub enum Utf8Sequence {
    /// One byte range.
    One(Utf8Range),
    /// Two successive byte ranges.
    Two([Utf8Range; 2]),
    /// Three successive byte ranges.
    Three([Utf8Range; 3]),
    /// Four successive byte ranges.
    Four([Utf8Range; 4]),
}

impl Utf8Sequence {
    /// Creates a new UTF-8 sequence from the encoded bytes of a scalar value
    /// range.
    ///
    /// This assumes that `start` and `end` have the same length.
    fn from_encoded_range(start: &[u8], end: &[u8]) -> Self {
        assert_eq!(start.len(), end.len());
        match start.len() {
            2 => Utf8Sequence::Two([
                Utf8Range::new(start[0], end[0]),
                Utf8Range::new(start[1], end[1]),
            ]),
            3 => Utf8Sequence::Three([
                Utf8Range::new(start[0], end[0]),
                Utf8Range::new(start[1], end[1]),
                Utf8Range::new(start[2], end[2]),
            ]),
            4 => Utf8Sequence::Four([
                Utf8Range::new(start[0], end[0]),
                Utf8Range::new(start[1], end[1]),
                Utf8Range::new(start[2], end[2]),
                Utf8Range::new(start[3], end[3]),
            ]),
            n => unreachable!("invalid encoded length: {}", n),
        }
    }

    /// Returns the underlying sequence of byte ranges as a slice.
    pub fn as_slice(&self) -> &[Utf8Range] {
        use self::Utf8Sequence::*;
        match *self {
            One(ref r) => unsafe { slice::from_raw_parts(r, 1) },
            Two(ref r) => &r[..],
            Three(ref r) => &r[..],
            Four(ref r) => &r[..],
        }
    }

    /// Returns the number of byte ranges in this sequence.
    ///
    /// The length is guaranteed to be in the closed interval `[1, 4]`.
    pub fn len(&self) -> usize {
        self.as_slice().len()
    }

    /// Returns true if and only if a prefix of `bytes` matches this sequence
    /// of byte ranges.
    pub fn matches(&self, bytes: &[u8]) -> bool {
        if bytes.len() < self.len() {
            return false;
        }
        for (&b, r) in bytes.iter().zip(self) {
            if !r.matches(b) {
                return false;
            }
        }
        true
    }
}

impl<'a> IntoIterator for &'a Utf8Sequence {
    type IntoIter = slice::Iter<'a, Utf8Range>;
    type Item = &'a Utf8Range;

    fn into_iter(self) -> Self::IntoIter {
        self.as_slice().into_iter()
    }
}

impl fmt::Debug for Utf8Sequence {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        use self::Utf8Sequence::*;
        match *self {
            One(ref r) => write!(f, "{:?}", r),
            Two(ref r) => write!(f, "{:?}{:?}", r[0], r[1]),
            Three(ref r) => write!(f, "{:?}{:?}{:?}", r[0], r[1], r[2]),
            Four(ref r) => write!(f, "{:?}{:?}{:?}{:?}",
                                  r[0], r[1], r[2], r[3]),
        }
    }
}

/// A single inclusive range of UTF-8 bytes.
#[derive(Clone, Copy, PartialEq, Eq)]
pub struct Utf8Range {
    /// Start of byte range (inclusive).
    pub start: u8,
    /// End of byte range (inclusive).
    pub end: u8,
}

impl Utf8Range {
    fn new(start: u8, end: u8) -> Self {
        Utf8Range { start: start, end: end }
    }

    /// Returns true if and only if the given byte is in this range.
    pub fn matches(&self, b: u8) -> bool {
        self.start <= b && b <= self.end
    }
}

impl fmt::Debug for Utf8Range {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        if self.start == self.end {
            write!(f, "[{:X}]", self.start)
        } else {
            write!(f, "[{:X}-{:X}]", self.start, self.end)
        }
    }
}

/// An iterator over ranges of matching UTF-8 byte sequences.
///
/// The iteration represents an alternation of comprehensive byte sequences
/// that match precisely the set of UTF-8 encoded scalar values.
///
/// A byte sequence corresponds to one of the scalar values in the range given
/// if and only if it completely matches exactly one of the sequences of byte
/// ranges produced by this iterator.
///
/// Each sequence of byte ranges matches a unique set of bytes. That is, no two
/// sequences will match the same bytes.
///
/// # Example
///
/// This shows how to match an arbitrary byte sequence against a range of
/// scalar values.
///
/// ```rust
/// use utf8_ranges::{Utf8Sequences, Utf8Sequence};
///
/// fn matches(seqs: &[Utf8Sequence], bytes: &[u8]) -> bool {
///     for range in seqs {
///         if range.matches(bytes) {
///             return true;
///         }
///     }
///     false
/// }
///
/// // Test the basic multilingual plane.
/// let seqs: Vec<_> = Utf8Sequences::new('\u{0}', '\u{FFFF}').collect();
///
/// // UTF-8 encoding of 'a'.
/// assert!(matches(&seqs, &[0x61]));
/// // UTF-8 encoding of '☃' (`\u{2603}`).
/// assert!(matches(&seqs, &[0xE2, 0x98, 0x83]));
/// // UTF-8 encoding of `\u{10348}` (outside the BMP).
/// assert!(!matches(&seqs, &[0xF0, 0x90, 0x8D, 0x88]));
/// // Tries to match against a UTF-8 encoding of a surrogate codepoint,
/// // which is invalid UTF-8, and therefore fails, despite the fact that
/// // the corresponding codepoint (0xD800) falls in the range given.
/// assert!(!matches(&seqs, &[0xED, 0xA0, 0x80]));
/// // And fails against plain old invalid UTF-8.
/// assert!(!matches(&seqs, &[0xFF, 0xFF]));
/// ```
///
/// If this example seems circuitous, that's because it is! It's meant to be
/// illustrative. In practice, you could just try to decode your byte sequence
/// and compare it with the scalar value range directly. However, this is not
/// always possible (for example, in a byte based automaton).
pub struct Utf8Sequences {
    range_stack: Vec<ScalarRange>,
}

impl Utf8Sequences {
    /// Create a new iterator over UTF-8 byte ranges for the scalar value range
    /// given.
    pub fn new(start: char, end: char) -> Self {
        let mut it = Utf8Sequences { range_stack: vec![] };
        it.push(start as u32, end as u32);
        it
    }

    /// reset resets the scalar value range.
    /// Any existing state is cleared, but resources may be reused.
    ///
    /// N.B. Benchmarks say that this method is dubious.
    #[doc(hidden)]
    pub fn reset(&mut self, start: char, end: char) {
        self.range_stack.clear();
        self.push(start as u32, end as u32);
    }

    fn push(&mut self, start: u32, end: u32) {
        self.range_stack.push(ScalarRange { start: start, end: end });
    }
}

struct ScalarRange {
    start: u32,
    end: u32,
}

impl fmt::Debug for ScalarRange {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        write!(f, "ScalarRange({:X}, {:X})", self.start, self.end)
    }
}

impl Iterator for Utf8Sequences {
    type Item = Utf8Sequence;

    fn next(&mut self) -> Option<Self::Item> {
    'TOP:
        while let Some(mut r) = self.range_stack.pop() {
        'INNER:
            loop {
                if let Some((r1, r2)) = r.split() {
                    self.push(r2.start, r2.end);
                    r.start = r1.start;
                    r.end = r1.end;
                    continue 'INNER;
                }
                if !r.is_valid() {
                    continue 'TOP;
                }
                for i in 1..MAX_UTF8_BYTES {
                    let max = max_scalar_value(i);
                    if r.start <= max && max < r.end {
                        self.push(max + 1, r.end);
                        r.end = max;
                        continue 'INNER;
                    }
                }
                if let Some(ascii_range) = r.as_ascii() {
                    return Some(Utf8Sequence::One(ascii_range));
                }
                for i in 1..MAX_UTF8_BYTES {
                    let m = (1 << (6 * i)) - 1;
                    if (r.start & !m) != (r.end & !m) {
                        if (r.start & m) != 0 {
                            self.push((r.start | m) + 1, r.end);
                            r.end = r.start | m;
                            continue 'INNER;
                        }
                        if (r.end & m) != m {
                            self.push(r.end & !m, r.end);
                            r.end = (r.end & !m) - 1;
                            continue 'INNER;
                        }
                    }
                }
                let mut start = [0; MAX_UTF8_BYTES];
                let mut end = [0; MAX_UTF8_BYTES];
                let n = r.encode(&mut start, &mut end);
                return Some(Utf8Sequence::from_encoded_range(
                    &start[0..n], &end[0..n]));
            }
        }
        None
    }
}

impl ScalarRange {
    /// split splits this range if it overlaps with a surrogate codepoint.
    ///
    /// Either or both ranges may be invalid.
    fn split(&self) -> Option<(ScalarRange, ScalarRange)> {
        if self.start < 0xE000 && self.end > 0xD7FF {
            Some((ScalarRange {
                start: self.start,
                end: 0xD7FF,
            }, ScalarRange {
                start: 0xE000,
                end: self.end,
            }))
        } else {
            None
        }
    }

    /// is_valid returns true if and only if start <= end.
    fn is_valid(&self) -> bool {
        self.start <= self.end
    }

    /// as_ascii returns this range as a Utf8Range if and only if all scalar
    /// values in this range can be encoded as a single byte.
    fn as_ascii(&self) -> Option<Utf8Range> {
        if self.is_ascii() {
            Some(Utf8Range::new(self.start as u8, self.end as u8))
        } else {
            None
        }
    }

    /// is_ascii returns true if the range is ASCII only (i.e., takes a single
    /// byte to encode any scalar value).
    fn is_ascii(&self) -> bool {
        self.is_valid() && self.end <= 0x7f
    }

    /// encode writes the UTF-8 encoding of the start and end of this range
    /// to the corresponding destination slices.
    ///
    /// The slices should have room for at least `MAX_UTF8_BYTES`.
    fn encode(&self, start: &mut [u8], end: &mut [u8]) -> usize {
        let cs = char::from_u32(self.start).unwrap();
        let ce = char::from_u32(self.end).unwrap();
        let n = encode_utf8(cs, start).unwrap();
        let m = encode_utf8(ce, end).unwrap();
        assert_eq!(n, m);
        n
    }
}

fn max_scalar_value(nbytes: usize) -> u32 {
    match nbytes {
        1 => 0x007F,
        2 => 0x07FF,
        3 => 0xFFFF,
        4 => 0x10FFFF,
        _ => unreachable!("invalid UTF-8 byte sequence size"),
    }
}

#[cfg(test)]
mod tests {
    use quickcheck::{TestResult, quickcheck};

    use char_utf8::encode_utf8;
    use {MAX_UTF8_BYTES, Utf8Range, Utf8Sequences};

    fn rutf8(s: u8, e: u8) -> Utf8Range {
        Utf8Range::new(s, e)
    }

    fn never_accepts_surrogate_codepoints(start: char, end: char) {
        let mut buf = [0; MAX_UTF8_BYTES];
        for cp in 0xD800..0xE000 {
            let c = unsafe { ::std::mem::transmute(cp) };
            let n = encode_utf8(c, &mut buf).unwrap();
            for r in Utf8Sequences::new(start, end) {
                if r.matches(&buf[0..n]) {
                    panic!("Sequence ({:X}, {:X}) contains range {:?}, \
                            which matches surrogate code point {:X} \
                            with encoded bytes {:?}",
                           start as u32, end as u32, r, cp, &buf[0..n]);
                }
            }
        }
    }

    #[test]
    fn codepoints_no_surrogates() {
        never_accepts_surrogate_codepoints('\u{0}', '\u{FFFF}');
        never_accepts_surrogate_codepoints('\u{0}', '\u{10FFFF}');
        never_accepts_surrogate_codepoints('\u{0}', '\u{10FFFE}');
        never_accepts_surrogate_codepoints('\u{80}', '\u{10FFFF}');
        never_accepts_surrogate_codepoints('\u{D7FF}', '\u{E000}');
    }

    #[test]
    fn qc_codepoints_no_surrogate() {
        fn p(s: char, e: char) -> TestResult {
            if s > e {
                return TestResult::discard();
            }
            never_accepts_surrogate_codepoints(s, e);
            TestResult::passed()
        }
        quickcheck(p as fn(char, char) -> TestResult);
    }

    #[test]
    fn bmp() {
        use Utf8Sequence::*;

        let seqs = Utf8Sequences::new('\u{0}', '\u{FFFF}')
                                 .collect::<Vec<_>>();
        assert_eq!(seqs, vec![
            One(rutf8(0x0, 0x7F)),
            Two([rutf8(0xC2, 0xDF), rutf8(0x80, 0xBF)]),
            Three([rutf8(0xE0, 0xE0), rutf8(0xA0, 0xBF), rutf8(0x80, 0xBF)]),
            Three([rutf8(0xE1, 0xEC), rutf8(0x80, 0xBF), rutf8(0x80, 0xBF)]),
            Three([rutf8(0xED, 0xED), rutf8(0x80, 0x9F), rutf8(0x80, 0xBF)]),
            Three([rutf8(0xEE, 0xEF), rutf8(0x80, 0xBF), rutf8(0x80, 0xBF)]),
        ]);
    }

    #[test]
    fn scratch() {
        for range in Utf8Sequences::new('\u{0}', '\u{FFFF}') {
            println!("{:?}", range);
        }
    }
}