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
///
pub mod decode {
/// The error returned by [`decode()`](super::decode()).
pub type Error = gix_error::Exn<gix_error::ValidationError>;
}
/// Decode `data` as EWAH bitmap.
pub fn decode(data: &[u8]) -> Result<(Vec, &[u8]), decode::Error> {
use crate::decode;
use gix_error::{message, OptionExt};
let (num_bits, data) = decode::u32(data).ok_or_raise(|| message("eof reading amount of bits").into())?;
let (len, data) = decode::u32(data).ok_or_raise(|| message("eof reading chunk length").into())?;
let len = len as usize;
// NOTE: git does this by copying all bytes first, and then it will change the endianness in a separate loop.
// Maybe it's faster, but we can't do it without unsafe. Let's leave it to the optimizer and maybe
// one day somebody will find out that it's worth it to use unsafe here.
let (mut bits, data) = data
.split_at_checked(len * std::mem::size_of::<u64>())
.ok_or_raise(|| message("eof while reading bit data").into())?;
let mut buf = std::vec::Vec::<u64>::with_capacity(len);
for _ in 0..len {
let (bit_num, rest) = bits.split_at(std::mem::size_of::<u64>());
bits = rest;
buf.push(u64::from_be_bytes(bit_num.try_into().unwrap()));
}
let (rlw, data) = decode::u32(data).ok_or_raise(|| message("eof while reading run length width").into())?;
Ok((
Vec {
num_bits,
bits: buf,
rlw: rlw.into(),
},
data,
))
}
mod access {
use super::Vec;
impl Vec {
/// Create a bitmap from a sequence of bit values.
///
/// The resulting bitmap uses a literal-only EWAH representation.
///
/// Returns `None` if `bits.len()` exceeds `u32::MAX`.
pub fn from_bits(bits: &[bool]) -> Option<Self> {
let literal_words: std::vec::Vec<u64> = bits
.chunks(64)
.map(|chunk| {
chunk.iter().enumerate().fold(
0u64,
|word, (idx, bit)| {
if *bit {
word | (1u64 << idx)
} else {
word
}
},
)
})
.collect();
let num_bits = bits.len().try_into().ok()?;
Some(Vec {
num_bits,
bits: std::iter::once((literal_words.len() as u64) << (1 + RLW_RUNNING_BITS))
.chain(literal_words)
.collect(),
rlw: 0,
})
}
/// Write the bitmap as EWAH bytes to `out`.
///
/// These bytes can be parsed again with [`decode()`](super::decode()).
pub fn write_to(&self, out: &mut impl std::io::Write) -> std::io::Result<()> {
let len: u32 = self.bits.len().try_into().map_err(|_| {
std::io::Error::new(std::io::ErrorKind::InvalidInput, "bit word count exceeds u32::MAX")
})?;
let rlw: u32 = self.rlw.try_into().map_err(|_| {
std::io::Error::new(
std::io::ErrorKind::InvalidInput,
"run length word offset exceeds u32::MAX",
)
})?;
out.write_all(&self.num_bits.to_be_bytes())?;
out.write_all(&len.to_be_bytes())?;
for word in &self.bits {
out.write_all(&word.to_be_bytes())?;
}
out.write_all(&rlw.to_be_bytes())
}
/// Call `f(index)` for each bit that is true, given the index of the bit that identifies it uniquely within the bit array.
/// If `f` returns `None` the iteration will be stopped and `None` is returned.
///
/// The index is sequential like in any other vector.
pub fn for_each_set_bit(&self, mut f: impl FnMut(usize) -> Option<()>) -> Option<()> {
let num_bits = self.num_bits();
let mut index = 0usize;
let mut iter = self.bits.iter();
while let Some(word) = iter.next() {
if rlw_runbit_is_set(word) {
let len = usize::try_from(rlw_running_len_bits(word)).ok()?;
let end = index.checked_add(len)?;
if end > num_bits {
return None;
}
for _ in 0..len {
f(index)?;
index += 1;
}
} else {
let len = usize::try_from(rlw_running_len_bits(word)).ok()?;
let end = index.checked_add(len)?;
if end > num_bits {
return None;
}
index = end;
}
for _ in 0..rlw_literal_words(word) {
let word = iter.next()?;
let remaining = num_bits.checked_sub(index)?;
if remaining == 0 {
return None;
}
let bits_in_word = remaining.min(64);
if bits_in_word < 64 && (word >> bits_in_word) != 0 {
return None;
}
for bit_index in 0..bits_in_word {
if word & (1 << bit_index) != 0 {
f(index)?;
}
index += 1;
}
}
}
Some(())
}
/// The amount of bits we are currently holding.
pub fn num_bits(&self) -> usize {
self.num_bits.try_into().expect("we are not on 16 bit systems")
}
}
#[inline]
fn rlw_running_len_bits(w: &u64) -> u64 {
rlw_running_len(w) * 64
}
#[inline]
fn rlw_running_len(w: &u64) -> u64 {
(w >> 1) & RLW_LARGEST_RUNNING_COUNT
}
#[inline]
fn rlw_literal_words(w: &u64) -> u64 {
w >> (1 + RLW_RUNNING_BITS)
}
#[inline]
fn rlw_runbit_is_set(w: &u64) -> bool {
w & 1 == 1
}
const RLW_RUNNING_BITS: u64 = 4 * 8;
const RLW_LARGEST_RUNNING_COUNT: u64 = (1 << RLW_RUNNING_BITS) - 1;
}
/// A growable collection of u64 that are seen as stream of individual bits.
#[allow(dead_code)]
#[derive(Clone)]
pub struct Vec {
num_bits: u32,
bits: std::vec::Vec<u64>,
/// RLW is an offset into the `bits` buffer, so `1` translates into &bits\[1] essentially.
rlw: u64,
}