Files
aho_corasick
atty
beef
bitflags
bstr
byteorder
cfg_if
clap
clap_derive
codespan
codespan_reporting
crc_any
crypto_hash
csv
csv_core
debug_helper
filepath
fixed
fixed_macro
fixed_macro_impl
fixed_macro_types
fnv
foreign_types
foreign_types_shared
getrandom
glob
hashbrown
heck
hex
indexmap
itoa
lazy_static
libc
linked_hash_map
linked_hash_set
logos
logos_derive
lrl_test_compiler
maplit
memchr
memoffset
once_cell
openssl
openssl_sys
os_str_bytes
paste
pest
pest_derive
pest_generator
pest_meta
phf
phf_generator
phf_macros
phf_shared
ppv_lite86
proc_macro2
proc_macro_error
proc_macro_error_attr
proc_macro_hack
quote
rand
rand_chacha
rand_core
regex
regex_automata
regex_syntax
remove_dir_all
ring
rowan
rustc_hash
ryu
semver
semver_parser
serde
serde_derive
serde_json
siphasher
smallvec
smawk
smol_str
spin
stable_deref_trait
strsim
syn
taplo
tempfile
termcolor
text_size
textwrap
toml
triomphe
typenum
ucd_trie
unicode_linebreak
unicode_segmentation
unicode_width
unicode_xid
untrusted
utf8_ranges
vec_map
  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
// This is adapted from `fallback.rs` from rust-memchr. It's modified to return
// the 'inverse' query of memchr, e.g. finding the first byte not in the provided
// set. This is simple for the 1-byte case.

use core::cmp;
use core::usize;

#[cfg(target_pointer_width = "32")]
const USIZE_BYTES: usize = 4;

#[cfg(target_pointer_width = "64")]
const USIZE_BYTES: usize = 8;

// The number of bytes to loop at in one iteration of memchr/memrchr.
const LOOP_SIZE: usize = 2 * USIZE_BYTES;

/// Repeat the given byte into a word size number. That is, every 8 bits
/// is equivalent to the given byte. For example, if `b` is `\x4E` or
/// `01001110` in binary, then the returned value on a 32-bit system would be:
/// `01001110_01001110_01001110_01001110`.
#[inline(always)]
fn repeat_byte(b: u8) -> usize {
    (b as usize) * (usize::MAX / 255)
}

pub fn inv_memchr(n1: u8, haystack: &[u8]) -> Option<usize> {
    let vn1 = repeat_byte(n1);
    let confirm = |byte| byte != n1;
    let loop_size = cmp::min(LOOP_SIZE, haystack.len());
    let align = USIZE_BYTES - 1;
    let start_ptr = haystack.as_ptr();
    let end_ptr = haystack[haystack.len()..].as_ptr();
    let mut ptr = start_ptr;

    unsafe {
        if haystack.len() < USIZE_BYTES {
            return forward_search(start_ptr, end_ptr, ptr, confirm);
        }

        let chunk = read_unaligned_usize(ptr);
        if (chunk ^ vn1) != 0 {
            return forward_search(start_ptr, end_ptr, ptr, confirm);
        }

        ptr = ptr.add(USIZE_BYTES - (start_ptr as usize & align));
        debug_assert!(ptr > start_ptr);
        debug_assert!(end_ptr.sub(USIZE_BYTES) >= start_ptr);
        while loop_size == LOOP_SIZE && ptr <= end_ptr.sub(loop_size) {
            debug_assert_eq!(0, (ptr as usize) % USIZE_BYTES);

            let a = *(ptr as *const usize);
            let b = *(ptr.add(USIZE_BYTES) as *const usize);
            let eqa = (a ^ vn1) != 0;
            let eqb = (b ^ vn1) != 0;
            if eqa || eqb {
                break;
            }
            ptr = ptr.add(LOOP_SIZE);
        }
        forward_search(start_ptr, end_ptr, ptr, confirm)
    }
}

/// Return the last index not matching the byte `x` in `text`.
pub fn inv_memrchr(n1: u8, haystack: &[u8]) -> Option<usize> {
    let vn1 = repeat_byte(n1);
    let confirm = |byte| byte != n1;
    let loop_size = cmp::min(LOOP_SIZE, haystack.len());
    let align = USIZE_BYTES - 1;
    let start_ptr = haystack.as_ptr();
    let end_ptr = haystack[haystack.len()..].as_ptr();
    let mut ptr = end_ptr;

    unsafe {
        if haystack.len() < USIZE_BYTES {
            return reverse_search(start_ptr, end_ptr, ptr, confirm);
        }

        let chunk = read_unaligned_usize(ptr.sub(USIZE_BYTES));
        if (chunk ^ vn1) != 0 {
            return reverse_search(start_ptr, end_ptr, ptr, confirm);
        }

        ptr = (end_ptr as usize & !align) as *const u8;
        debug_assert!(start_ptr <= ptr && ptr <= end_ptr);
        while loop_size == LOOP_SIZE && ptr >= start_ptr.add(loop_size) {
            debug_assert_eq!(0, (ptr as usize) % USIZE_BYTES);

            let a = *(ptr.sub(2 * USIZE_BYTES) as *const usize);
            let b = *(ptr.sub(1 * USIZE_BYTES) as *const usize);
            let eqa = (a ^ vn1) != 0;
            let eqb = (b ^ vn1) != 0;
            if eqa || eqb {
                break;
            }
            ptr = ptr.sub(loop_size);
        }
        reverse_search(start_ptr, end_ptr, ptr, confirm)
    }
}

#[inline(always)]
unsafe fn forward_search<F: Fn(u8) -> bool>(
    start_ptr: *const u8,
    end_ptr: *const u8,
    mut ptr: *const u8,
    confirm: F,
) -> Option<usize> {
    debug_assert!(start_ptr <= ptr);
    debug_assert!(ptr <= end_ptr);

    while ptr < end_ptr {
        if confirm(*ptr) {
            return Some(sub(ptr, start_ptr));
        }
        ptr = ptr.offset(1);
    }
    None
}

#[inline(always)]
unsafe fn reverse_search<F: Fn(u8) -> bool>(
    start_ptr: *const u8,
    end_ptr: *const u8,
    mut ptr: *const u8,
    confirm: F,
) -> Option<usize> {
    debug_assert!(start_ptr <= ptr);
    debug_assert!(ptr <= end_ptr);

    while ptr > start_ptr {
        ptr = ptr.offset(-1);
        if confirm(*ptr) {
            return Some(sub(ptr, start_ptr));
        }
    }
    None
}

unsafe fn read_unaligned_usize(ptr: *const u8) -> usize {
    (ptr as *const usize).read_unaligned()
}

/// Subtract `b` from `a` and return the difference. `a` should be greater than
/// or equal to `b`.
fn sub(a: *const u8, b: *const u8) -> usize {
    debug_assert!(a >= b);
    (a as usize) - (b as usize)
}

/// Safe wrapper around `forward_search`
#[inline]
pub(crate) fn forward_search_bytes<F: Fn(u8) -> bool>(
    s: &[u8],
    confirm: F,
) -> Option<usize> {
    unsafe {
        let start = s.as_ptr();
        let end = start.add(s.len());
        forward_search(start, end, start, confirm)
    }
}

/// Safe wrapper around `reverse_search`
#[inline]
pub(crate) fn reverse_search_bytes<F: Fn(u8) -> bool>(
    s: &[u8],
    confirm: F,
) -> Option<usize> {
    unsafe {
        let start = s.as_ptr();
        let end = start.add(s.len());
        reverse_search(start, end, end, confirm)
    }
}

#[cfg(test)]
mod tests {
    use super::{inv_memchr, inv_memrchr};
    // search string, search byte, inv_memchr result, inv_memrchr result.
    // these are expanded into a much larger set of tests in build_tests
    const TESTS: &[(&[u8], u8, usize, usize)] = &[
        (b"z", b'a', 0, 0),
        (b"zz", b'a', 0, 1),
        (b"aza", b'a', 1, 1),
        (b"zaz", b'a', 0, 2),
        (b"zza", b'a', 0, 1),
        (b"zaa", b'a', 0, 0),
        (b"zzz", b'a', 0, 2),
    ];

    type TestCase = (Vec<u8>, u8, Option<(usize, usize)>);

    fn build_tests() -> Vec<TestCase> {
        let mut result = vec![];
        for &(search, byte, fwd_pos, rev_pos) in TESTS {
            result.push((search.to_vec(), byte, Some((fwd_pos, rev_pos))));
            for i in 1..515 {
                // add a bunch of copies of the search byte to the end.
                let mut suffixed: Vec<u8> = search.into();
                suffixed.extend(std::iter::repeat(byte).take(i));
                result.push((suffixed, byte, Some((fwd_pos, rev_pos))));

                // add a bunch of copies of the search byte to the start.
                let mut prefixed: Vec<u8> =
                    std::iter::repeat(byte).take(i).collect();
                prefixed.extend(search);
                result.push((
                    prefixed,
                    byte,
                    Some((fwd_pos + i, rev_pos + i)),
                ));

                // add a bunch of copies of the search byte to both ends.
                let mut surrounded: Vec<u8> =
                    std::iter::repeat(byte).take(i).collect();
                surrounded.extend(search);
                surrounded.extend(std::iter::repeat(byte).take(i));
                result.push((
                    surrounded,
                    byte,
                    Some((fwd_pos + i, rev_pos + i)),
                ));
            }
        }

        // build non-matching tests for several sizes
        for i in 0..515 {
            result.push((
                std::iter::repeat(b'\0').take(i).collect(),
                b'\0',
                None,
            ));
        }

        result
    }

    #[test]
    fn test_inv_memchr() {
        use {ByteSlice, B};
        for (search, byte, matching) in build_tests() {
            assert_eq!(
                inv_memchr(byte, &search),
                matching.map(|m| m.0),
                "inv_memchr when searching for {:?} in {:?}",
                byte as char,
                // better printing
                B(&search).as_bstr(),
            );
            assert_eq!(
                inv_memrchr(byte, &search),
                matching.map(|m| m.1),
                "inv_memrchr when searching for {:?} in {:?}",
                byte as char,
                // better printing
                B(&search).as_bstr(),
            );
            // Test a rather large number off offsets for potential alignment issues
            for offset in 1..130 {
                if offset >= search.len() {
                    break;
                }
                // If this would cause us to shift the results off the end, skip
                // it so that we don't have to recompute them.
                if let Some((f, r)) = matching {
                    if offset > f || offset > r {
                        break;
                    }
                }
                let realigned = &search[offset..];

                let forward_pos = matching.map(|m| m.0 - offset);
                let reverse_pos = matching.map(|m| m.1 - offset);

                assert_eq!(
                    inv_memchr(byte, &realigned),
                    forward_pos,
                    "inv_memchr when searching (realigned by {}) for {:?} in {:?}",
                    offset,
                    byte as char,
                    realigned.as_bstr(),
                );
                assert_eq!(
                    inv_memrchr(byte, &realigned),
                    reverse_pos,
                    "inv_memrchr when searching (realigned by {}) for {:?} in {:?}",
                    offset,
                    byte as char,
                    realigned.as_bstr(),
                );
            }
        }
    }
}