1 // Copyright 2015 The Rust Project Developers. See the COPYRIGHT
2 // file at the top-level directory of this distribution and at
3 // http://rust-lang.org/COPYRIGHT.
5 // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
6 // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
7 // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
8 // option. This file may not be copied, modified, or distributed
9 // except according to those terms.
11 // Original implementation taken from rust-memchr
12 // Copyright 2015 Andrew Gallant, bluss and Nicolas Koch
19 const LO_U64: u64 = 0x0101010101010101;
20 const HI_U64: u64 = 0x8080808080808080;
23 const LO_USIZE: usize = LO_U64 as usize;
24 const HI_USIZE: usize = HI_U64 as usize;
26 /// Return `true` if `x` contains any zero byte.
28 /// From *Matters Computational*, J. Arndt
30 /// "The idea is to subtract one from each of the bytes and then look for
31 /// bytes where the borrow propagated all the way to the most significant
34 fn contains_zero_byte(x: usize) -> bool {
35 x.wrapping_sub(LO_USIZE) & !x & HI_USIZE != 0
38 #[cfg(target_pointer_width = "32")]
40 fn repeat_byte(b: u8) -> usize {
41 let mut rep = (b as usize) << 8 | b as usize;
42 rep = rep << 16 | rep;
46 #[cfg(target_pointer_width = "64")]
48 fn repeat_byte(b: u8) -> usize {
49 let mut rep = (b as usize) << 8 | b as usize;
50 rep = rep << 16 | rep;
51 rep = rep << 32 | rep;
55 /// Return the first index matching the byte `a` in `text`.
56 pub fn memchr(x: u8, text: &[u8]) -> Option<usize> {
57 // Scan for a single byte value by reading two `usize` words at a time.
59 // Split `text` in three parts
60 // - unaligned initial part, before the first word aligned address in text
61 // - body, scan by 2 words at a time
62 // - the last remaining part, < 2 word size
64 let ptr = text.as_ptr();
65 let usize_bytes = mem::size_of::<usize>();
67 // search up to an aligned boundary
68 let mut offset = ptr.align_offset(usize_bytes);
70 offset = cmp::min(offset, len);
71 if let Some(index) = text[..offset].iter().position(|elt| *elt == x) {
76 // search the body of the text
77 let repeated_x = repeat_byte(x);
79 if len >= 2 * usize_bytes {
80 while offset <= len - 2 * usize_bytes {
82 let u = *(ptr.offset(offset as isize) as *const usize);
83 let v = *(ptr.offset((offset + usize_bytes) as isize) as *const usize);
85 // break if there is a matching byte
86 let zu = contains_zero_byte(u ^ repeated_x);
87 let zv = contains_zero_byte(v ^ repeated_x);
92 offset += usize_bytes * 2;
96 // find the byte after the point the body loop stopped
97 text[offset..].iter().position(|elt| *elt == x).map(|i| offset + i)
100 /// Return the last index matching the byte `a` in `text`.
101 pub fn memrchr(x: u8, text: &[u8]) -> Option<usize> {
102 // Scan for a single byte value by reading two `usize` words at a time.
104 // Split `text` in three parts
105 // - unaligned tail, after the last word aligned address in text
106 // - body, scan by 2 words at a time
107 // - the first remaining bytes, < 2 word size
108 let len = text.len();
109 let ptr = text.as_ptr();
110 let usize_bytes = mem::size_of::<usize>();
112 // search to an aligned boundary
113 let end_align = (ptr as usize + len) & (usize_bytes - 1);
116 offset = if end_align >= len { 0 } else { len - end_align };
117 if let Some(index) = text[offset..].iter().rposition(|elt| *elt == x) {
118 return Some(offset + index);
124 // search the body of the text
125 let repeated_x = repeat_byte(x);
127 while offset >= 2 * usize_bytes {
129 let u = *(ptr.offset(offset as isize - 2 * usize_bytes as isize) as *const usize);
130 let v = *(ptr.offset(offset as isize - usize_bytes as isize) as *const usize);
132 // break if there is a matching byte
133 let zu = contains_zero_byte(u ^ repeated_x);
134 let zv = contains_zero_byte(v ^ repeated_x);
139 offset -= 2 * usize_bytes;
142 // find the byte before the point the body loop stopped
143 text[..offset].iter().rposition(|elt| *elt == x)
146 // test fallback implementations on all platforms
149 assert_eq!(Some(0), memchr(b'a', b"a"));
154 assert_eq!(Some(0), memchr(b'a', b"aaaa"));
159 assert_eq!(Some(4), memchr(b'z', b"aaaaz"));
164 assert_eq!(Some(4), memchr(b'\x00', b"aaaa\x00"));
168 fn matches_past_nul() {
169 assert_eq!(Some(5), memchr(b'z', b"aaaa\x00z"));
173 fn no_match_empty() {
174 assert_eq!(None, memchr(b'a', b""));
179 assert_eq!(None, memchr(b'a', b"xyz"));
183 fn matches_one_reversed() {
184 assert_eq!(Some(0), memrchr(b'a', b"a"));
188 fn matches_begin_reversed() {
189 assert_eq!(Some(3), memrchr(b'a', b"aaaa"));
193 fn matches_end_reversed() {
194 assert_eq!(Some(0), memrchr(b'z', b"zaaaa"));
198 fn matches_nul_reversed() {
199 assert_eq!(Some(4), memrchr(b'\x00', b"aaaa\x00"));
203 fn matches_past_nul_reversed() {
204 assert_eq!(Some(0), memrchr(b'z', b"z\x00aaaa"));
208 fn no_match_empty_reversed() {
209 assert_eq!(None, memrchr(b'a', b""));
213 fn no_match_reversed() {
214 assert_eq!(None, memrchr(b'a', b"xyz"));
218 fn each_alignment_reversed() {
219 let mut data = [1u8; 64];
224 assert_eq!(Some(pos - start), memrchr(needle, &data[start..]));