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
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
use crate::{Doubling, Fragment, Growth, SplitVec};

/// Equivalent to [`Doubling`] strategy except for the following:
///
/// * enables zero-cost (no-ops) `append` operation:
///   * we can append standard vectors, vectors of vectors, split vectors, etc., any data that implements `IntoFragments` trait,
///   * by simply accepting it as a whole fragment,
///   * according to benchmarks documented in the crate definition:
///     * `SplitVec<_, Recursive>` is infinitely faster than other growth strategies or standard vector :)
///     * since its time complexity is independent of size of the data to be appended.
/// * at the expense of providing slower random-access performance:
///   * random access time complexity of `Doubling` strategy is constant time;
///   * that of `Recursive` strategy is linear in the number of fragments;
///   * according to benchmarks documented in the crate definition:
///     * `SplitVec<_, Doubling>` or standard vector are around 4 to 7 times faster than `SplitVec<_, Recursive>`,
///     * and 1.5 times faster when the elements get very large (16 x `u64`).
///
/// Note that other operations such as serial access are equivalent to `Doubling` strategy.
///
/// # Examples
///
/// ```
/// use orx_split_vec::*;
///
/// // SplitVec<usize, Recursive>
/// let mut vec = SplitVec::with_recursive_growth();
///
/// vec.push('a');
/// assert_eq!(vec, &['a']);
///
/// vec.append(vec!['b', 'c']);
/// assert_eq!(vec, &['a', 'b', 'c']);
///
/// vec.append(vec![vec!['d'], vec!['e', 'f']]);
/// assert_eq!(vec, &['a', 'b', 'c', 'd', 'e', 'f']);
///
/// let other_split_vec: SplitVec<_> = vec!['g', 'h'].into();
/// vec.append(other_split_vec);
/// assert_eq!(vec, &['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h']);
/// ```
#[derive(Debug, Default, Clone, PartialEq)]
pub struct Recursive;

impl Growth for Recursive {
    #[inline(always)]
    fn new_fragment_capacity<T>(&self, fragments: &[Fragment<T>]) -> usize {
        Doubling.new_fragment_capacity(fragments)
    }
}

impl<T> SplitVec<T, Recursive> {
    /// Strategy which allows to create a fragment with double the capacity
    /// of the prior fragment every time the split vector needs to expand.
    ///
    /// Notice that this is similar to the `Doubling` growth strategy.
    /// However, `Recursive` and `Doubling` strategies have the two following important differences in terms of performance:
    ///
    /// * Random access by indices is much faster with `Doubling`.
    /// * Recursive strategy enables copy-free `append` method which merges another vector to this vector in constant time.
    ///
    /// All other operations are expected to have similar complexity.
    ///
    /// ## Random Access
    ///
    /// * `Doubling` strategy provides a constant time access by random indices.
    /// * `Recursive` strategy provides a random access time complexity that is linear in the number of fragments.
    /// Note that this is significantly faster than the linear-in-number-of-elements complexity of linked lists;
    /// however, significantly slower than the `Doubling` strategy's constant time.
    ///
    /// ## Append
    ///
    /// * `Recursive` strategy provides `append` operation which allows merging two vectors in constant time without copies.
    ///
    /// `SplitVec::append` method should not be confused with `std::vec::Vec::append` method:
    /// * The split vector version consumes the vector to be appended.
    /// It takes advantage of its split nature and appends the other vector simply by owning its pointer.
    /// In other words, the other vector is appended to this vector with no cost and no copies.
    /// * The standard vector version mutates the vector to be appended,
    /// moving all its element to the first vector leaving the latter empty.
    /// This operation is carried out by memory copies.
    ///
    /// # Examples
    ///
    /// ```
    /// use orx_split_vec::*;
    ///
    /// // SplitVec<usize, Doubling>
    /// let mut vec = SplitVec::with_recursive_growth();
    ///
    /// assert_eq!(1, vec.fragments().len());
    /// assert_eq!(Some(4), vec.fragments().first().map(|f| f.capacity()));
    /// assert_eq!(Some(0), vec.fragments().first().map(|f| f.len()));
    ///
    /// // fill the first 5 fragments
    /// let expected_fragment_capacities = vec![4, 8, 16, 32];
    /// let num_items: usize = expected_fragment_capacities.iter().sum();
    /// for i in 0..num_items {
    ///     vec.push(i);
    /// }
    ///
    /// assert_eq!(
    ///     expected_fragment_capacities,
    ///     vec.fragments()
    ///     .iter()
    ///     .map(|f| f.capacity())
    ///     .collect::<Vec<_>>()
    /// );
    /// assert_eq!(
    ///     expected_fragment_capacities,
    ///     vec.fragments().iter().map(|f| f.len()).collect::<Vec<_>>()
    /// );
    ///
    /// // create the 6-th fragment doubling the capacity
    /// vec.push(42);
    /// assert_eq!(
    ///     vec.fragments().len(),
    ///     expected_fragment_capacities.len() + 1
    /// );
    ///
    /// assert_eq!(vec.fragments().last().map(|f| f.capacity()), Some(32 * 2));
    /// assert_eq!(vec.fragments().last().map(|f| f.len()), Some(1));
    /// ```
    pub fn with_recursive_growth() -> Self {
        SplitVec::with_doubling_growth().into()
    }
}

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

    #[test]
    fn get_fragment_and_inner_indices() {
        let growth = Recursive;

        let vecs = vec![
            vec![0, 1, 2, 3],
            vec![4, 5],
            vec![6, 7, 8],
            vec![9],
            vec![10, 11, 12, 13, 14],
        ];
        let mut fragments: Vec<Fragment<_>> = vecs.clone().into_iter().map(|x| x.into()).collect();
        let len = fragments.iter().map(|x| x.len()).sum();

        let mut index = 0;
        for (f, vec) in vecs.iter().enumerate() {
            for (i, _) in vec.iter().enumerate() {
                let maybe_fi = growth.get_fragment_and_inner_indices(len, &fragments, index);
                assert_eq!(maybe_fi, Some((f, i)));

                let ptr = unsafe { growth.get_ptr_mut(&mut fragments, index) }.expect("is-some");
                assert_eq!(unsafe { *ptr }, index);

                unsafe { *ptr = 10 * index };
                assert_eq!(unsafe { *ptr }, 10 * index);

                index += 1;
            }
        }
    }

    #[test]
    fn get_fragment_and_inner_indices_exhaustive() {
        let growth = Recursive;

        let mut fragments: Vec<Fragment<_>> = vec![];

        let lengths = [30, 52, 14, 1, 7, 3, 79, 248, 147, 530];
        let mut index = 0;
        for _ in 0..100 {
            for &len in &lengths {
                let mut vec = Vec::with_capacity(len);
                for _ in 0..len {
                    vec.push(index);
                    index += 1;
                }
                fragments.push(vec.into());
            }
        }

        let total_len = fragments.iter().map(|x| x.len()).sum();

        let mut index = 0;
        let mut f = 0;
        for _ in 0..100 {
            for &len in &lengths {
                for i in 0..len {
                    let maybe_fi =
                        growth.get_fragment_and_inner_indices(total_len, &fragments, index);

                    assert_eq!(maybe_fi, Some((f, i)));

                    let ptr =
                        unsafe { growth.get_ptr_mut(&mut fragments, index) }.expect("is-some");
                    assert_eq!(unsafe { *ptr }, index);

                    unsafe { *ptr = 10 * index };
                    assert_eq!(unsafe { *ptr }, 10 * index);

                    index += 1;
                }
                f += 1;
            }
        }
    }
}