Add memchr implemenation based on rust-memchr to libstd

2 files changed, 402 insertions, 0 deletions

diff --git a/src/libstd/lib.rs b/src/libstd/lib.rs
index 0f8b2f6e17b..eba0c799cd2 100644
--- a/src/libstd/lib.rs
+++ b/src/libstd/lib.rs
@@ -248,6 +248,7 @@
 #![feature(link_args)]
 #![feature(linkage)]
 #![feature(macro_reexport)]
+#![feature(num_bits_bytes)]
 #![feature(on_unimplemented)]
 #![feature(oom)]
 #![feature(optin_builtin_traits)]
@@ -429,6 +430,7 @@ pub mod path;
 pub mod process;
 pub mod sync;
 pub mod time;
+mod memchr;
 
 #[macro_use]
 #[path = "sys/common/mod.rs"] mod sys_common;
diff --git a/src/libstd/memchr.rs b/src/libstd/memchr.rs
new file mode 100644
index 00000000000..8a2fc575759
--- /dev/null
+++ b/src/libstd/memchr.rs
@@ -0,0 +1,400 @@
+// Copyright 2015 The Rust Project Developers. See the COPYRIGHT
+// file at the top-level directory of this distribution and at
+// http://rust-lang.org/COPYRIGHT.
+//
+// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
+// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
+// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
+// option. This file may not be copied, modified, or distributed
+// except according to those terms.
+//
+// Original implementation taken from rust-memchr
+// Copyright 2015 Andrew Gallant, bluss and Nicolas Koch
+
+use libc::{c_void, c_int, size_t};
+
+
+/// A safe interface to `memchr`.
+///
+/// Returns the index corresponding to the first occurrence of `needle` in
+/// `haystack`, or `None` if one is not found.
+///
+/// memchr reduces to super-optimized machine code at around an order of
+/// magnitude faster than `haystack.iter().position(|&b| b == needle)`.
+/// (See benchmarks.)
+///
+/// # Example
+///
+/// This shows how to find the first position of a byte in a byte string.
+///
+/// ```rust
+/// use memchr::memchr;
+///
+/// let haystack = b"the quick brown fox";
+/// assert_eq!(memchr(b'k', haystack), Some(8));
+/// ```
+pub fn memchr(needle: u8, haystack: &[u8]) -> Option<usize> {
+    // libc memchr
+    #[cfg(any(not(target_os = "windows"),
+              not(any(target_pointer_width = "32",
+                      target_pointer_width = "64"))))]
+    fn memchr_specific(needle: u8, haystack: &[u8]) -> Option<usize> {
+        use libc::memchr as libc_memchr;
+
+        let p = unsafe {
+            libc_memchr(
+                haystack.as_ptr() as *const c_void,
+                needle as c_int,
+                haystack.len() as size_t)
+        };
+        if p.is_null() {
+            None
+        } else {
+            Some(p as usize - (haystack.as_ptr() as usize))
+        }
+    }
+
+    // use fallback on windows, since it's faster
+    #[cfg(all(target_os = "windows",
+              any(target_pointer_width = "32",
+                  target_pointer_width = "64")))]
+    fn memchr_specific(needle: u8, haystack: &[u8]) -> Option<usize> {
+        fallback::memchr(needle, haystack)
+    }
+
+    memchr_specific(needle, haystack)
+}
+
+/// A safe interface to `memrchr`.
+///
+/// Returns the index corresponding to the last occurrence of `needle` in
+/// `haystack`, or `None` if one is not found.
+///
+/// # Example
+///
+/// This shows how to find the last position of a byte in a byte string.
+///
+/// ```rust
+/// use memchr::memrchr;
+///
+/// let haystack = b"the quick brown fox";
+/// assert_eq!(memrchr(b'o', haystack), Some(17));
+/// ```
+pub fn memrchr(needle: u8, haystack: &[u8]) -> Option<usize> {
+
+    #[cfg(target_os = "linux")]
+    fn memrchr_specific(needle: u8, haystack: &[u8]) -> Option<usize> {
+        use libc::memrchr as libc_memrchr;
+
+        // GNU's memrchr() will - unlike memchr() - error if haystack is empty.
+        if haystack.is_empty() {return None}
+        let p = unsafe {
+            libc_memrchr(
+                haystack.as_ptr() as *const c_void,
+                needle as c_int,
+                haystack.len() as size_t)
+        };
+        if p.is_null() {
+            None
+        } else {
+            Some(p as usize - (haystack.as_ptr() as usize))
+        }
+    }
+
+    #[cfg(all(not(target_os = "linux"),
+              any(target_pointer_width = "32", target_pointer_width = "64")))]
+    fn memrchr_specific(needle: u8, haystack: &[u8]) -> Option<usize> {
+        fallback::memrchr(needle, haystack)
+    }
+
+    // For the rare case of neither 32 bit nor 64-bit platform.
+    #[cfg(all(not(target_os = "linux"),
+              not(target_pointer_width = "32"),
+              not(target_pointer_width = "64")))]
+    fn memrchr_specific(needle: u8, haystack: &[u8]) -> Option<usize> {
+        haystack.iter().rposition(|&b| b == needle)
+    }
+
+    memrchr_specific(needle, haystack)
+}
+
+#[allow(dead_code)]
+mod fallback {
+    use cmp;
+    use usize;
+
+    const LO_U64: u64 = 0x0101010101010101;
+    const HI_U64: u64 = 0x8080808080808080;
+
+    // use truncation
+    const LO_USIZE: usize = LO_U64 as usize;
+    const HI_USIZE: usize = HI_U64 as usize;
+
+    /// Return `true` if `x` contains any zero byte.
+    ///
+    /// From *Matters Computational*, J. Arndt
+    ///
+    /// "The idea is to subtract one from each of the bytes and then look for
+    /// bytes where the borrow propagated all the way to the most significant
+    /// bit."
+    #[inline]
+    fn contains_zero_byte(x: usize) -> bool {
+        x.wrapping_sub(LO_USIZE) & !x & HI_USIZE != 0
+    }
+
+    #[cfg(target_pointer_width = "32")]
+    #[inline]
+    fn repeat_byte(b: u8) -> usize {
+        let mut rep = (b as usize) << 8 | b as usize;
+        rep = rep << 16 | rep;
+        rep
+    }
+
+    #[cfg(target_pointer_width = "64")]
+    #[inline]
+    fn repeat_byte(b: u8) -> usize {
+        let mut rep = (b as usize) << 8 | b as usize;
+        rep = rep << 16 | rep;
+        rep = rep << 32 | rep;
+        rep
+    }
+
+    /// Return the first index matching the byte `a` in `text`.
+    pub fn memchr(x: u8, text: &[u8]) -> Option<usize> {
+        // Scan for a single byte value by reading two `usize` words at a time.
+        //
+        // Split `text` in three parts
+        // - unaligned inital part, before the first word aligned address in text
+        // - body, scan by 2 words at a time
+        // - the last remaining part, < 2 word size
+        let len = text.len();
+        let ptr = text.as_ptr();
+
+        // search up to an aligned boundary
+        let align = (ptr as usize) & (usize::BYTES- 1);
+        let mut offset;
+        if align > 0 {
+            offset = cmp::min(usize::BYTES - align, len);
+            if let Some(index) = text[..offset].iter().position(|elt| *elt == x) {
+                return Some(index);
+            }
+        } else {
+            offset = 0;
+        }
+
+        // search the body of the text
+        let repeated_x = repeat_byte(x);
+
+        if len >= 2 * usize::BYTES {
+            while offset <= len - 2 * usize::BYTES {
+                unsafe {
+                    let u = *(ptr.offset(offset as isize) as *const usize);
+                    let v = *(ptr.offset((offset + usize::BYTES) as isize) as *const usize);
+
+                    // break if there is a matching byte
+                    let zu = contains_zero_byte(u ^ repeated_x);
+                    let zv = contains_zero_byte(v ^ repeated_x);
+                    if zu || zv {
+                        break;
+                    }
+                }
+                offset += usize::BYTES * 2;
+            }
+        }
+
+        // find the byte after the point the body loop stopped
+        text[offset..].iter().position(|elt| *elt == x).map(|i| offset + i)
+    }
+
+    /// Return the last index matching the byte `a` in `text`.
+    pub fn memrchr(x: u8, text: &[u8]) -> Option<usize> {
+        // Scan for a single byte value by reading two `usize` words at a time.
+        //
+        // Split `text` in three parts
+        // - unaligned tail, after the last word aligned address in text
+        // - body, scan by 2 words at a time
+        // - the first remaining bytes, < 2 word size
+        let len = text.len();
+        let ptr = text.as_ptr();
+
+        // search to an aligned boundary
+        let end_align = (ptr as usize + len) & (usize::BYTES - 1);
+        let mut offset;
+        if end_align > 0 {
+            offset = len - cmp::min(usize::BYTES - end_align, len);
+            if let Some(index) = text[offset..].iter().rposition(|elt| *elt == x) {
+                return Some(offset + index);
+            }
+        } else {
+            offset = len;
+        }
+
+        // search the body of the text
+        let repeated_x = repeat_byte(x);
+
+        while offset >= 2 * usize::BYTES {
+            unsafe {
+                let u = *(ptr.offset(offset as isize - 2 * usize::BYTES as isize) as *const usize);
+                let v = *(ptr.offset(offset as isize - usize::BYTES as isize) as *const usize);
+
+                // break if there is a matching byte
+                let zu = contains_zero_byte(u ^ repeated_x);
+                let zv = contains_zero_byte(v ^ repeated_x);
+                if zu || zv {
+                    break;
+                }
+            }
+            offset -= 2 * usize::BYTES;
+        }
+
+        // find the byte before the point the body loop stopped
+        text[..offset].iter().rposition(|elt| *elt == x)
+    }
+
+    // test fallback implementations on all plattforms
+    #[test]
+    fn matches_one() {
+        assert_eq!(Some(0), memchr(b'a', b"a"));
+    }
+
+    #[test]
+    fn matches_begin() {
+        assert_eq!(Some(0), memchr(b'a', b"aaaa"));
+    }
+
+    #[test]
+    fn matches_end() {
+        assert_eq!(Some(4), memchr(b'z', b"aaaaz"));
+    }
+
+    #[test]
+    fn matches_nul() {
+        assert_eq!(Some(4), memchr(b'\x00', b"aaaa\x00"));
+    }
+
+    #[test]
+    fn matches_past_nul() {
+        assert_eq!(Some(5), memchr(b'z', b"aaaa\x00z"));
+    }
+
+    #[test]
+    fn no_match_empty() {
+        assert_eq!(None, memchr(b'a', b""));
+    }
+
+    #[test]
+    fn no_match() {
+        assert_eq!(None, memchr(b'a', b"xyz"));
+    }
+
+    #[test]
+    fn matches_one_reversed() {
+        assert_eq!(Some(0), memrchr(b'a', b"a"));
+    }
+
+    #[test]
+    fn matches_begin_reversed() {
+        assert_eq!(Some(3), memrchr(b'a', b"aaaa"));
+    }
+
+    #[test]
+    fn matches_end_reversed() {
+        assert_eq!(Some(0), memrchr(b'z', b"zaaaa"));
+    }
+
+    #[test]
+    fn matches_nul_reversed() {
+        assert_eq!(Some(4), memrchr(b'\x00', b"aaaa\x00"));
+    }
+
+    #[test]
+    fn matches_past_nul_reversed() {
+        assert_eq!(Some(0), memrchr(b'z', b"z\x00aaaa"));
+    }
+
+    #[test]
+    fn no_match_empty_reversed() {
+        assert_eq!(None, memrchr(b'a', b""));
+    }
+
+    #[test]
+    fn no_match_reversed() {
+        assert_eq!(None, memrchr(b'a', b"xyz"));
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    // test the implementations for the current plattform
+    use super::{memchr, memrchr};
+
+    #[test]
+    fn matches_one() {
+        assert_eq!(Some(0), memchr(b'a', b"a"));
+    }
+
+    #[test]
+    fn matches_begin() {
+        assert_eq!(Some(0), memchr(b'a', b"aaaa"));
+    }
+
+    #[test]
+    fn matches_end() {
+        assert_eq!(Some(4), memchr(b'z', b"aaaaz"));
+    }
+
+    #[test]
+    fn matches_nul() {
+        assert_eq!(Some(4), memchr(b'\x00', b"aaaa\x00"));
+    }
+
+    #[test]
+    fn matches_past_nul() {
+        assert_eq!(Some(5), memchr(b'z', b"aaaa\x00z"));
+    }
+
+    #[test]
+    fn no_match_empty() {
+        assert_eq!(None, memchr(b'a', b""));
+    }
+
+    #[test]
+    fn no_match() {
+        assert_eq!(None, memchr(b'a', b"xyz"));
+    }
+
+    #[test]
+    fn matches_one_reversed() {
+        assert_eq!(Some(0), memrchr(b'a', b"a"));
+    }
+
+    #[test]
+    fn matches_begin_reversed() {
+        assert_eq!(Some(3), memrchr(b'a', b"aaaa"));
+    }
+
+    #[test]
+    fn matches_end_reversed() {
+        assert_eq!(Some(0), memrchr(b'z', b"zaaaa"));
+    }
+
+    #[test]
+    fn matches_nul_reversed() {
+        assert_eq!(Some(4), memrchr(b'\x00', b"aaaa\x00"));
+    }
+
+    #[test]
+    fn matches_past_nul_reversed() {
+        assert_eq!(Some(0), memrchr(b'z', b"z\x00aaaa"));
+    }
+
+    #[test]
+    fn no_match_empty_reversed() {
+        assert_eq!(None, memrchr(b'a', b""));
+    }
+
+    #[test]
+    fn no_match_reversed() {
+        assert_eq!(None, memrchr(b'a', b"xyz"));
+    }
+}