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
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
//! Varified
//! borrow from petgraph

use std::fmt;
use std::hash::Hash;

/// The default integer type for graph indices.
/// `u32` is the default to reduce the size of the graph's data and improve
/// performance in the common case.
///
/// Used for node and edge indices in `Graph` and `StableGraph`, used
/// for node indices in `Csr`.
pub type DefaultIx = u32;

/// Trait for the unsigned integer type used for node and edge indices.
///
/// Marked `unsafe` because: the trait must faithfully preserve
/// and convert index values.
pub unsafe trait IndexType: Copy + Default + Hash + Ord + fmt::Debug + 'static {
    fn new(x: usize) -> Self;
    fn index(&self) -> usize;
    fn max() -> Self;
}

unsafe impl IndexType for usize {
    #[inline(always)]
    fn new(x: usize) -> Self {
        x
    }
    #[inline(always)]
    fn index(&self) -> Self {
        *self
    }
    #[inline(always)]
    fn max() -> Self {
        ::std::usize::MAX
    }
}

unsafe impl IndexType for u32 {
    #[inline(always)]
    fn new(x: usize) -> Self {
        x as u32
    }
    #[inline(always)]
    fn index(&self) -> usize {
        *self as usize
    }
    #[inline(always)]
    fn max() -> Self {
        ::std::u32::MAX
    }
}

unsafe impl IndexType for u16 {
    #[inline(always)]
    fn new(x: usize) -> Self {
        x as u16
    }
    #[inline(always)]
    fn index(&self) -> usize {
        *self as usize
    }
    #[inline(always)]
    fn max() -> Self {
        ::std::u16::MAX
    }
}

unsafe impl IndexType for u8 {
    #[inline(always)]
    fn new(x: usize) -> Self {
        x as u8
    }
    #[inline(always)]
    fn index(&self) -> usize {
        *self as usize
    }
    #[inline(always)]
    fn max() -> Self {
        ::std::u8::MAX
    }
}

// `UnionFind<K>` is a disjoint-set data structure.

use std::cmp::Ordering;

/// `UnionFind<K>` is a disjoint-set data structure. It tracks set membership of *n* elements
/// indexed from *0* to *n - 1*. The scalar type is `K` which must be an unsigned integer type.
///
/// <http://en.wikipedia.org/wiki/Disjoint-set_data_structure>
///
/// Too awesome not to quote:
///
/// “The amortized time per operation is **O(α(n))** where **α(n)** is the
/// inverse of **f(x) = A(x, x)** with **A** being the extremely fast-growing Ackermann function.”
#[derive(Debug, Clone)]
pub struct UnionFind<K> {
    // For element at index *i*, store the index of its parent; the representative itself
    // stores its own index. This forms equivalence classes which are the disjoint sets, each
    // with a unique representative.
    parent: Vec<K>,
    // It is a balancing tree structure,
    // so the ranks are logarithmic in the size of the container -- a byte is more than enough.
    //
    // Rank is separated out both to save space and to save cache in when searching in the parent
    // vector.
    rank: Vec<u8>,
}

#[inline]
unsafe fn get_unchecked<K>(xs: &[K], index: usize) -> &K {
    debug_assert!(index < xs.len());
    xs.get_unchecked(index)
}

#[inline]
unsafe fn get_unchecked_mut<K>(xs: &mut [K], index: usize) -> &mut K {
    debug_assert!(index < xs.len());
    xs.get_unchecked_mut(index)
}

impl<K> UnionFind<K>
where
    K: IndexType,
{
    /// Create a new `UnionFind` of `n` disjoint sets.
    pub fn new(n: usize) -> Self {
        let rank = vec![0; n];
        let parent = (0..n).map(K::new).collect::<Vec<K>>();

        UnionFind { parent, rank }
    }

    /// Return the representative for `x`.
    ///
    /// **Panics** if `x` is out of bounds.
    pub fn find(&self, x: K) -> K {
        assert!(x.index() < self.parent.len());
        unsafe {
            let mut x = x;
            loop {
                // Use unchecked indexing because we can trust the internal set ids.
                let xparent = *get_unchecked(&self.parent, x.index());
                if xparent == x {
                    break;
                }
                x = xparent;
            }
            x
        }
    }

    /// Return the representative for `x`.
    ///
    /// Write back the found representative, flattening the internal
    /// datastructure in the process and quicken future lookups.
    ///
    /// **Panics** if `x` is out of bounds.
    pub fn find_mut(&mut self, x: K) -> K {
        assert!(x.index() < self.parent.len());
        unsafe { self.find_mut_recursive(x) }
    }

    unsafe fn find_mut_recursive(&mut self, mut x: K) -> K {
        let mut parent = *get_unchecked(&self.parent, x.index());
        while parent != x {
            let grandparent = *get_unchecked(&self.parent, parent.index());
            *get_unchecked_mut(&mut self.parent, x.index()) = grandparent;
            x = parent;
            parent = grandparent;
        }
        x
    }

    /// Returns `true` if the given elements belong to the same set, and returns
    /// `false` otherwise.
    pub fn equiv(&self, x: K, y: K) -> bool {
        self.find(x) == self.find(y)
    }

    /// Unify the two sets containing `x` and `y`.
    ///
    /// Return `false` if the sets were already the same, `true` if they were unified.
    ///
    /// **Panics** if `x` or `y` is out of bounds.
    pub fn union(&mut self, x: K, y: K) -> bool {
        if x == y {
            return false;
        }
        let xrep = self.find_mut(x);
        let yrep = self.find_mut(y);

        if xrep == yrep {
            return false;
        }

        let xrepu = xrep.index();
        let yrepu = yrep.index();
        let xrank = self.rank[xrepu];
        let yrank = self.rank[yrepu];

        // The rank corresponds roughly to the depth of the treeset, so put the
        // smaller set below the larger
        match xrank.cmp(&yrank) {
            Ordering::Less => self.parent[xrepu] = yrep,
            Ordering::Greater => self.parent[yrepu] = xrep,
            Ordering::Equal => {
                self.parent[yrepu] = xrep;
                self.rank[xrepu] += 1;
            }
        }
        true
    }

    /// Return a vector mapping each element to its representative.
    pub fn into_labeling(mut self) -> Vec<K> {
        // write in the labeling of each element
        unsafe {
            for ix in 0..self.parent.len() {
                let k = *get_unchecked(&self.parent, ix);
                let xrep = self.find_mut_recursive(k);
                *self.parent.get_unchecked_mut(ix) = xrep;
            }
        }
        self.parent
    }
}