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
use regex_automata::{dfa::Automaton, Anchored, Input};

use crate::{
    ext_slice::ByteSlice,
    unicode::fsm::{
        grapheme_break_fwd::GRAPHEME_BREAK_FWD,
        grapheme_break_rev::GRAPHEME_BREAK_REV,
        regional_indicator_rev::REGIONAL_INDICATOR_REV,
    },
    utf8,
};

/// An iterator over grapheme clusters in a byte string.
///
/// This iterator is typically constructed by
/// [`ByteSlice::graphemes`](trait.ByteSlice.html#method.graphemes).
///
/// Unicode defines a grapheme cluster as an *approximation* to a single user
/// visible character. A grapheme cluster, or just "grapheme," is made up of
/// one or more codepoints. For end user oriented tasks, one should generally
/// prefer using graphemes instead of [`Chars`](struct.Chars.html), which
/// always yields one codepoint at a time.
///
/// Since graphemes are made up of one or more codepoints, this iterator yields
/// `&str` elements. When invalid UTF-8 is encountered, replacement codepoints
/// are [substituted](index.html#handling-of-invalid-utf-8).
///
/// This iterator can be used in reverse. When reversed, exactly the same
/// set of grapheme clusters are yielded, but in reverse order.
///
/// This iterator only yields *extended* grapheme clusters, in accordance with
/// [UAX #29](https://www.unicode.org/reports/tr29/tr29-33.html#Grapheme_Cluster_Boundaries).
#[derive(Clone, Debug)]
pub struct Graphemes<'a> {
    bs: &'a [u8],
}

impl<'a> Graphemes<'a> {
    pub(crate) fn new(bs: &'a [u8]) -> Graphemes<'a> {
        Graphemes { bs }
    }

    /// View the underlying data as a subslice of the original data.
    ///
    /// The slice returned has the same lifetime as the original slice, and so
    /// the iterator can continue to be used while this exists.
    ///
    /// # Examples
    ///
    /// ```
    /// use bstr::ByteSlice;
    ///
    /// let mut it = b"abc".graphemes();
    ///
    /// assert_eq!(b"abc", it.as_bytes());
    /// it.next();
    /// assert_eq!(b"bc", it.as_bytes());
    /// it.next();
    /// it.next();
    /// assert_eq!(b"", it.as_bytes());
    /// ```
    #[inline]
    pub fn as_bytes(&self) -> &'a [u8] {
        self.bs
    }
}

impl<'a> Iterator for Graphemes<'a> {
    type Item = &'a str;

    #[inline]
    fn next(&mut self) -> Option<&'a str> {
        let (grapheme, size) = decode_grapheme(self.bs);
        if size == 0 {
            return None;
        }
        self.bs = &self.bs[size..];
        Some(grapheme)
    }
}

impl<'a> DoubleEndedIterator for Graphemes<'a> {
    #[inline]
    fn next_back(&mut self) -> Option<&'a str> {
        let (grapheme, size) = decode_last_grapheme(self.bs);
        if size == 0 {
            return None;
        }
        self.bs = &self.bs[..self.bs.len() - size];
        Some(grapheme)
    }
}

/// An iterator over grapheme clusters in a byte string and their byte index
/// positions.
///
/// This iterator is typically constructed by
/// [`ByteSlice::grapheme_indices`](trait.ByteSlice.html#method.grapheme_indices).
///
/// Unicode defines a grapheme cluster as an *approximation* to a single user
/// visible character. A grapheme cluster, or just "grapheme," is made up of
/// one or more codepoints. For end user oriented tasks, one should generally
/// prefer using graphemes instead of [`Chars`](struct.Chars.html), which
/// always yields one codepoint at a time.
///
/// Since graphemes are made up of one or more codepoints, this iterator
/// yields `&str` elements (along with their start and end byte offsets).
/// When invalid UTF-8 is encountered, replacement codepoints are
/// [substituted](index.html#handling-of-invalid-utf-8). Because of this, the
/// indices yielded by this iterator may not correspond to the length of the
/// grapheme cluster yielded with those indices. For example, when this
/// iterator encounters `\xFF` in the byte string, then it will yield a pair
/// of indices ranging over a single byte, but will provide an `&str`
/// equivalent to `"\u{FFFD}"`, which is three bytes in length. However, when
/// given only valid UTF-8, then all indices are in exact correspondence with
/// their paired grapheme cluster.
///
/// This iterator can be used in reverse. When reversed, exactly the same
/// set of grapheme clusters are yielded, but in reverse order.
///
/// This iterator only yields *extended* grapheme clusters, in accordance with
/// [UAX #29](https://www.unicode.org/reports/tr29/tr29-33.html#Grapheme_Cluster_Boundaries).
#[derive(Clone, Debug)]
pub struct GraphemeIndices<'a> {
    bs: &'a [u8],
    forward_index: usize,
    reverse_index: usize,
}

impl<'a> GraphemeIndices<'a> {
    pub(crate) fn new(bs: &'a [u8]) -> GraphemeIndices<'a> {
        GraphemeIndices { bs, forward_index: 0, reverse_index: bs.len() }
    }

    /// View the underlying data as a subslice of the original data.
    ///
    /// The slice returned has the same lifetime as the original slice, and so
    /// the iterator can continue to be used while this exists.
    ///
    /// # Examples
    ///
    /// ```
    /// use bstr::ByteSlice;
    ///
    /// let mut it = b"abc".grapheme_indices();
    ///
    /// assert_eq!(b"abc", it.as_bytes());
    /// it.next();
    /// assert_eq!(b"bc", it.as_bytes());
    /// it.next();
    /// it.next();
    /// assert_eq!(b"", it.as_bytes());
    /// ```
    #[inline]
    pub fn as_bytes(&self) -> &'a [u8] {
        self.bs
    }
}

impl<'a> Iterator for GraphemeIndices<'a> {
    type Item = (usize, usize, &'a str);

    #[inline]
    fn next(&mut self) -> Option<(usize, usize, &'a str)> {
        let index = self.forward_index;
        let (grapheme, size) = decode_grapheme(self.bs);
        if size == 0 {
            return None;
        }
        self.bs = &self.bs[size..];
        self.forward_index += size;
        Some((index, index + size, grapheme))
    }
}

impl<'a> DoubleEndedIterator for GraphemeIndices<'a> {
    #[inline]
    fn next_back(&mut self) -> Option<(usize, usize, &'a str)> {
        let (grapheme, size) = decode_last_grapheme(self.bs);
        if size == 0 {
            return None;
        }
        self.bs = &self.bs[..self.bs.len() - size];
        self.reverse_index -= size;
        Some((self.reverse_index, self.reverse_index + size, grapheme))
    }
}

/// Decode a grapheme from the given byte string.
///
/// This returns the resulting grapheme (which may be a Unicode replacement
/// codepoint if invalid UTF-8 was found), along with the number of bytes
/// decoded in the byte string. The number of bytes decoded may not be the
/// same as the length of grapheme in the case where invalid UTF-8 is found.
pub fn decode_grapheme(bs: &[u8]) -> (&str, usize) {
    if bs.is_empty() {
        ("", 0)
    } else if bs.len() >= 2
        && bs[0].is_ascii()
        && bs[1].is_ascii()
        && !bs[0].is_ascii_whitespace()
    {
        // FIXME: It is somewhat sad that we have to special case this, but it
        // leads to a significant speed up in predominantly ASCII text. The
        // issue here is that the DFA has a bit of overhead, and running it for
        // every byte in mostly ASCII text results in a bit slowdown. We should
        // re-litigate this once regex-automata 0.3 is out, but it might be
        // hard to avoid the special case. A DFA is always going to at least
        // require some memory access.

        // Safe because all ASCII bytes are valid UTF-8.
        let grapheme = unsafe { bs[..1].to_str_unchecked() };
        (grapheme, 1)
    } else if let Some(hm) = {
        let input = Input::new(bs).anchored(Anchored::Yes);
        GRAPHEME_BREAK_FWD.try_search_fwd(&input).unwrap()
    } {
        // Safe because a match can only occur for valid UTF-8.
        let grapheme = unsafe { bs[..hm.offset()].to_str_unchecked() };
        (grapheme, grapheme.len())
    } else {
        const INVALID: &'static str = "\u{FFFD}";
        // No match on non-empty bytes implies we found invalid UTF-8.
        let (_, size) = utf8::decode_lossy(bs);
        (INVALID, size)
    }
}

fn decode_last_grapheme(bs: &[u8]) -> (&str, usize) {
    if bs.is_empty() {
        ("", 0)
    } else if let Some(hm) = {
        let input = Input::new(bs).anchored(Anchored::Yes);
        GRAPHEME_BREAK_REV.try_search_rev(&input).unwrap()
    } {
        let start = adjust_rev_for_regional_indicator(bs, hm.offset());
        // Safe because a match can only occur for valid UTF-8.
        let grapheme = unsafe { bs[start..].to_str_unchecked() };
        (grapheme, grapheme.len())
    } else {
        const INVALID: &'static str = "\u{FFFD}";
        // No match on non-empty bytes implies we found invalid UTF-8.
        let (_, size) = utf8::decode_last_lossy(bs);
        (INVALID, size)
    }
}

/// Return the correct offset for the next grapheme decoded at the end of the
/// given byte string, where `i` is the initial guess. In particular,
/// `&bs[i..]` represents the candidate grapheme.
///
/// `i` is returned by this function in all cases except when `&bs[i..]` is
/// a pair of regional indicator codepoints. In that case, if an odd number of
/// additional regional indicator codepoints precedes `i`, then `i` is
/// adjusted such that it points to only a single regional indicator.
///
/// This "fixing" is necessary to handle the requirement that a break cannot
/// occur between regional indicators where it would cause an odd number of
/// regional indicators to exist before the break from the *start* of the
/// string. A reverse regex cannot detect this case easily without look-around.
fn adjust_rev_for_regional_indicator(mut bs: &[u8], i: usize) -> usize {
    // All regional indicators use a 4 byte encoding, and we only care about
    // the case where we found a pair of regional indicators.
    if bs.len() - i != 8 {
        return i;
    }
    // Count all contiguous occurrences of regional indicators. If there's an
    // even number of them, then we can accept the pair we found. Otherwise,
    // we can only take one of them.
    //
    // FIXME: This is quadratic in the worst case, e.g., a string of just
    // regional indicator codepoints. A fix probably requires refactoring this
    // code a bit such that we don't rescan regional indicators.
    let mut count = 0;
    while let Some(hm) = {
        let input = Input::new(bs).anchored(Anchored::Yes);
        REGIONAL_INDICATOR_REV.try_search_rev(&input).unwrap()
    } {
        bs = &bs[..hm.offset()];
        count += 1;
    }
    if count % 2 == 0 {
        i
    } else {
        i + 4
    }
}

#[cfg(all(test, feature = "std"))]
mod tests {
    use alloc::{
        string::{String, ToString},
        vec,
        vec::Vec,
    };

    #[cfg(not(miri))]
    use ucd_parse::GraphemeClusterBreakTest;

    use crate::tests::LOSSY_TESTS;

    use super::*;

    #[test]
    #[cfg(not(miri))]
    fn forward_ucd() {
        for (i, test) in ucdtests().into_iter().enumerate() {
            let given = test.grapheme_clusters.concat();
            let got: Vec<String> = Graphemes::new(given.as_bytes())
                .map(|cluster| cluster.to_string())
                .collect();
            assert_eq!(
                test.grapheme_clusters,
                got,
                "\ngrapheme forward break test {} failed:\n\
                 given:    {:?}\n\
                 expected: {:?}\n\
                 got:      {:?}\n",
                i,
                uniescape(&given),
                uniescape_vec(&test.grapheme_clusters),
                uniescape_vec(&got),
            );
        }
    }

    #[test]
    #[cfg(not(miri))]
    fn reverse_ucd() {
        for (i, test) in ucdtests().into_iter().enumerate() {
            let given = test.grapheme_clusters.concat();
            let mut got: Vec<String> = Graphemes::new(given.as_bytes())
                .rev()
                .map(|cluster| cluster.to_string())
                .collect();
            got.reverse();
            assert_eq!(
                test.grapheme_clusters,
                got,
                "\n\ngrapheme reverse break test {} failed:\n\
                 given:    {:?}\n\
                 expected: {:?}\n\
                 got:      {:?}\n",
                i,
                uniescape(&given),
                uniescape_vec(&test.grapheme_clusters),
                uniescape_vec(&got),
            );
        }
    }

    #[test]
    fn forward_lossy() {
        for &(expected, input) in LOSSY_TESTS {
            let got = Graphemes::new(input.as_bytes()).collect::<String>();
            assert_eq!(expected, got);
        }
    }

    #[test]
    fn reverse_lossy() {
        for &(expected, input) in LOSSY_TESTS {
            let expected: String = expected.chars().rev().collect();
            let got =
                Graphemes::new(input.as_bytes()).rev().collect::<String>();
            assert_eq!(expected, got);
        }
    }

    #[cfg(not(miri))]
    fn uniescape(s: &str) -> String {
        s.chars().flat_map(|c| c.escape_unicode()).collect::<String>()
    }

    #[cfg(not(miri))]
    fn uniescape_vec(strs: &[String]) -> Vec<String> {
        strs.iter().map(|s| uniescape(s)).collect()
    }

    /// Return all of the UCD for grapheme breaks.
    #[cfg(not(miri))]
    fn ucdtests() -> Vec<GraphemeClusterBreakTest> {
        const TESTDATA: &'static str =
            include_str!("data/GraphemeBreakTest.txt");

        let mut tests = vec![];
        for mut line in TESTDATA.lines() {
            line = line.trim();
            if line.starts_with("#") || line.contains("surrogate") {
                continue;
            }
            tests.push(line.parse().unwrap());
        }
        tests
    }
}