rapidhash 3.1.0

A rust port of rapidhash: an extremely fast, high quality, platform-independent hashing algorithm.
Documentation 
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

//! Internal module for reading unaligned bytes from a slice into `u64` and `u32` values.

/// Hacky const-friendly memory-safe unaligned bytes to u64. Compiler can't seem to remove the
/// bounds check, and so we have an unsafe version behind the `unsafe` feature flag.
#[cfg(not(feature = "unsafe"))]
#[inline(always)]
pub(crate) const fn read_u64(slice: &[u8], offset: usize) -> u64 {
    // equivalent to slice[offset..offset+8].try_into().unwrap(), but const-friendly
    let maybe_buf = slice.split_at(offset).1.first_chunk::<8>();
    let buf = match maybe_buf {
        Some(buf) => *buf,
        None => panic!("read_u64: slice too short"),
    };
    u64::from_le_bytes(buf)
}

/// Hacky const-friendly memory-safe unaligned bytes to u64. Compiler can't seem to remove the
/// bounds check, and so we have an unsafe version behind the `unsafe` feature flag.
#[cfg(not(feature = "unsafe"))]
#[inline(always)]
pub(crate) const fn read_u32(slice: &[u8], offset: usize) -> u32 {
    // equivalent to slice[offset..offset+4].try_into().unwrap(), but const-friendly
    let maybe_buf = slice.split_at(offset).1.first_chunk::<4>();
    let buf = match maybe_buf {
        Some(buf) => *buf,
        None => panic!("read_u32: slice too short"),
    };
    u32::from_le_bytes(buf)
}

/// Unsafe but const-friendly unaligned bytes to u64. The compiler can't seem to remove the bounds
/// checks for small integers because we do some funky bit shifting in the indexing.
///
/// SAFETY: `slice` must be at least `offset+8` bytes long, which we guarantee in this rapidhash
/// implementation.
#[cfg(feature = "unsafe")]
#[inline(always)]
pub(crate) const fn read_u64(slice: &[u8], offset: usize) -> u64 {
    debug_assert!(offset as isize >= 0);
    debug_assert!(slice.len() >= 8 + offset);
    let val = unsafe { core::ptr::read_unaligned(slice.as_ptr().offset(offset as isize) as *const u64) };
    val.to_le()  // swap bytes on big-endian systems to get the same u64 value
}

/// Unsafe but const-friendly unaligned bytes to u32. The compiler can't seem to remove the bounds
/// checks for small integers because we do some funky bit shifting in the indexing.
///
/// SAFETY: `slice` must be at least `offset+8` bytes long, which we guarantee in this rapidhash
/// implementation.
#[cfg(feature = "unsafe")]
#[inline(always)]
pub(crate) const fn read_u32(slice: &[u8], offset: usize) -> u32 {
    debug_assert!(offset as isize >= 0);
    debug_assert!(slice.len() >= 4 + offset);
    let val = unsafe { core::ptr::read_unaligned(slice.as_ptr().offset(offset as isize) as *const u32) };
    val.to_le()  // swap bytes on big-endian systems to get the same u64 value
}

/// Only used in rapidhash V1
#[inline(always)]
pub(crate) const fn read_u32_combined(slice: &[u8], offset_top: usize, offset_bot: usize) -> u64 {
    debug_assert!(slice.len() >= 4 + offset_top && slice.len() >= 4 + offset_bot);
    let top = read_u32(slice, offset_top) as u64;
    let bot = read_u32(slice, offset_bot) as u64;
    (top << 32) | bot
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_read_u32() {
        let bytes = &[23, 145, 3, 34];

        let split_result = bytes.split_at(0).1;
        assert_eq!(split_result.len(), 4);
        let maybe_buf = split_result.first_chunk::<4>();
        assert_eq!(maybe_buf, Some(&[23, 145, 3, 34]));

        assert_eq!(read_u32(bytes, 0), 570659095);

        let bytes = &[24, 54, 3, 23, 145, 3, 34];
        assert_eq!(read_u32(bytes, 3), 570659095);

        assert_eq!(read_u32(&[0, 0, 0, 0], 0), 0);
        assert_eq!(read_u32(&[1, 0, 0, 0], 0), 1);
        assert_eq!(read_u32(&[12, 0, 0, 0], 0), 12);
        assert_eq!(read_u32(&[0, 10, 0, 0], 0), 2560);
    }

    #[test]
    fn test_read_u64() {
        let bytes = [23, 145, 3, 34, 0, 0, 0, 0, 0, 0, 0].as_slice();
        assert_eq!(read_u64(bytes, 0), 570659095);

        let bytes = [1, 2, 3, 23, 145, 3, 34, 0, 0, 0, 0, 0, 0, 0].as_slice();
        assert_eq!(read_u64(bytes, 3), 570659095);

        let bytes = [0, 0, 0, 0, 0, 0, 0, 0].as_slice();
        assert_eq!(read_u64(bytes, 0), 0);
    }

    #[cfg(feature = "std")]
    #[test]
    fn test_u32_to_u128_delta() {
        fn formula(len: u64) -> u64 {
            (len & 24) >> (len >> 3)
        }

        fn formula2(len: u64) -> u64 {
            match len {
                8.. => 4,
                _ => 0,
            }
        }

        let inputs: std::vec::Vec<u64> = (4..=16).collect();
        let outputs: std::vec::Vec<u64> = inputs.iter().map(|&x| formula(x)).collect();
        let expected = std::vec![0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, 4];
        assert_eq!(outputs, expected);
        assert_eq!(outputs, inputs.iter().map(|&x| formula2(x)).collect::<Vec<u64>>());
    }

    #[test]
    #[should_panic]
    #[cfg(any(test, not(feature = "unsafe")))]
    fn test_read_u32_to_short_panics() {
        let bytes = [23, 145, 0].as_slice();
        assert_eq!(read_u32(bytes, 0), 0);
    }

    #[test]
    #[should_panic]
    #[cfg(any(test, not(feature = "unsafe")))]
    fn test_read_u64_to_short_panics() {
        let bytes = [23, 145, 0].as_slice();
        assert_eq!(read_u64(bytes, 0), 0);
    }
}