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
use off64::u8;

// Two benefits of this over Vec<u8>:
// - Length type is u8, which is the type for `tail_page_count`.
// - No heap allocations; small enough to copy.
// How it works:
// - Page sizes can only be from `2^8` to `2^32` (see `{MIN,MAX}_PAGE_SIZE_POW2`).
// - There are only `32 - 8 = 24` page sizes.
// - It takes 5 bits to represent 24 distinct values.
// - There can be up to 24 entries, each taking 5 bits. It will take 5 bits to represent the length.
// - We can't fit 125 bits neatly, so we split up into two shards (64-bit ints), and use 4 bits in each int to represent the length of up to 12 entries covered by its int, using exactly `4 + 12 * 5 = 64` bits.
#[derive(Clone, Copy)]
pub struct TailPageSizes(u64, u64);

impl TailPageSizes {
  pub fn new() -> Self {
    Self(0, 0)
  }

  fn shard_get_len(shard: u64) -> u8 {
    u8!((shard & (0b1111 << 60)) >> 60)
  }

  fn shard_set_len(shard: &mut u64, len: u8) {
    assert!(len <= 12);
    *shard &= !(0b1111 << 60);
    *shard |= u64::from(len) << 60;
  }

  fn shard_get(shard: u64, idx: u8) -> u8 {
    assert!(idx < 12);
    let shift = idx * 5;
    u8!((shard & (0b11111 << shift)) >> shift)
  }

  fn shard_set(shard: &mut u64, idx: u8, val: u8) {
    assert!(val < (1 << 5));
    let shift = idx * 5;
    *shard &= !(0b11111 << shift);
    *shard |= u64::from(val) << shift;
  }

  pub fn len(&self) -> u8 {
    Self::shard_get_len(self.0) + Self::shard_get_len(self.1)
  }

  pub fn get(&self, idx: u8) -> Option<u8> {
    if idx >= self.len() {
      return None;
    };
    Some(if idx < 12 {
      Self::shard_get(self.0, idx)
    } else {
      Self::shard_get(self.1, idx - 12)
    })
  }

  pub fn push(&mut self, pow2: u8) {
    let idx = self.len();
    if idx < 12 {
      Self::shard_set(&mut self.0, idx, pow2);
      Self::shard_set_len(&mut self.0, idx + 1);
    } else {
      Self::shard_set(&mut self.1, idx - 12, pow2);
      Self::shard_set_len(&mut self.1, idx - 12 + 1);
    };
  }
}

impl IntoIterator for TailPageSizes {
  type IntoIter = TailPageSizesIter;
  type Item = (u8, u8);

  fn into_iter(self) -> Self::IntoIter {
    TailPageSizesIter(self, 0)
  }
}

// Convenient iterator that provides `u8` indices instead of `.enumerate()` and `u8!(idx)`.
pub struct TailPageSizesIter(TailPageSizes, u8);

impl Iterator for TailPageSizesIter {
  type Item = (u8, u8);

  fn next(&mut self) -> Option<Self::Item> {
    let idx = self.1;
    let entry = self.0.get(idx)?;
    self.1 += 1;
    Some((idx, entry))
  }
}

#[cfg(test)]
mod tests {
  use super::TailPageSizes;
  use itertools::Itertools;

  #[test]
  fn test_tail_page_sizes() {
    let mut tail = TailPageSizes::new();
    // Check iterator/state is correct when empty.
    assert_eq!(tail.into_iter().collect_vec(), vec![]);

    tail.push(5);
    assert_eq!(tail.into_iter().collect_vec(), vec![(0, 5)]);

    tail.push(3);
    assert_eq!(tail.into_iter().collect_vec(), vec![(0, 5), (1, 3)]);

    tail.push(8);
    assert_eq!(tail.into_iter().collect_vec(), vec![(0, 5), (1, 3), (2, 8)]);
    assert_eq!(tail.get(1), Some(3));
    assert_eq!(tail.get(3), None);

    tail.push(9);
    assert_eq!(tail.into_iter().collect_vec(), vec![
      (0, 5),
      (1, 3),
      (2, 8),
      (3, 9),
    ]);
    assert_eq!(tail.len(), 4);
    assert_eq!(tail.get(0), Some(5));
    assert_eq!(tail.get(4), None);

    tail.push(4);
    assert_eq!(tail.into_iter().collect_vec(), vec![
      (0, 5),
      (1, 3),
      (2, 8),
      (3, 9),
      (4, 4)
    ]);
    assert_eq!(tail.len(), 5);
    assert_eq!(tail.get(3), Some(9));
    assert_eq!(tail.get(5), None);

    tail.push(11);
    assert_eq!(tail.into_iter().collect_vec(), vec![
      (0, 5),
      (1, 3),
      (2, 8),
      (3, 9),
      (4, 4),
      (5, 11),
    ]);
    assert_eq!(tail.len(), 6);
    assert_eq!(tail.get(0), Some(5));
    assert_eq!(tail.get(4), Some(4));
    assert_eq!(tail.get(5), Some(11));
    assert_eq!(tail.get(6), None);

    tail.push(14);
    assert_eq!(tail.into_iter().collect_vec(), vec![
      (0, 5),
      (1, 3),
      (2, 8),
      (3, 9),
      (4, 4),
      (5, 11),
      (6, 14),
    ]);
    assert_eq!(tail.len(), 7);
    assert_eq!(tail.get(0), Some(5));
    assert_eq!(tail.get(4), Some(4));
    assert_eq!(tail.get(5), Some(11));
    assert_eq!(tail.get(6), Some(14));

    tail.push(2);
    assert_eq!(tail.into_iter().collect_vec(), vec![
      (0, 5),
      (1, 3),
      (2, 8),
      (3, 9),
      (4, 4),
      (5, 11),
      (6, 14),
      (7, 2),
    ]);
    assert_eq!(tail.len(), 8);
    assert_eq!(tail.get(0), Some(5));
    assert_eq!(tail.get(4), Some(4));
    assert_eq!(tail.get(5), Some(11));
    assert_eq!(tail.get(6), Some(14));
    assert_eq!(tail.get(7), Some(2));
    assert_eq!(tail.get(8), None);
  }
}