//===- llvm/ADT/SparseMultiSet.h - Sparse multiset --------------*- C++ -*-===//
 
//
 
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 
// See https://llvm.org/LICENSE.txt for license information.
 
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
//
 
//===----------------------------------------------------------------------===//
 
///
 
/// \file
 
/// This file defines the SparseMultiSet class, which adds multiset behavior to
 
/// the SparseSet.
 
///
 
/// A sparse multiset holds a small number of objects identified by integer keys
 
/// from a moderately sized universe. The sparse multiset uses more memory than
 
/// other containers in order to provide faster operations. Any key can map to
 
/// multiple values. A SparseMultiSetNode class is provided, which serves as a
 
/// convenient base class for the contents of a SparseMultiSet.
 
///
 
//===----------------------------------------------------------------------===//
 
 
 
#ifndef LLVM_ADT_SPARSEMULTISET_H
 
#define LLVM_ADT_SPARSEMULTISET_H
 
 
 
#include "llvm/ADT/identity.h"
 
#include "llvm/ADT/SmallVector.h"
 
#include "llvm/ADT/SparseSet.h"
 
#include <cassert>
 
#include <cstdint>
 
#include <cstdlib>
 
#include <iterator>
 
#include <limits>
 
#include <utility>
 
 
 
namespace llvm {
 
 
 
/// Fast multiset implementation for objects that can be identified by small
 
/// unsigned keys.
 
///
 
/// SparseMultiSet allocates memory proportional to the size of the key
 
/// universe, so it is not recommended for building composite data structures.
 
/// It is useful for algorithms that require a single set with fast operations.
 
///
 
/// Compared to DenseSet and DenseMap, SparseMultiSet provides constant-time
 
/// fast clear() as fast as a vector.  The find(), insert(), and erase()
 
/// operations are all constant time, and typically faster than a hash table.
 
/// The iteration order doesn't depend on numerical key values, it only depends
 
/// on the order of insert() and erase() operations.  Iteration order is the
 
/// insertion order. Iteration is only provided over elements of equivalent
 
/// keys, but iterators are bidirectional.
 
///
 
/// Compared to BitVector, SparseMultiSet<unsigned> uses 8x-40x more memory, but
 
/// offers constant-time clear() and size() operations as well as fast iteration
 
/// independent on the size of the universe.
 
///
 
/// SparseMultiSet contains a dense vector holding all the objects and a sparse
 
/// array holding indexes into the dense vector.  Most of the memory is used by
 
/// the sparse array which is the size of the key universe. The SparseT template
 
/// parameter provides a space/speed tradeoff for sets holding many elements.
 
///
 
/// When SparseT is uint32_t, find() only touches up to 3 cache lines, but the
 
/// sparse array uses 4 x Universe bytes.
 
///
 
/// When SparseT is uint8_t (the default), find() touches up to 3+[N/256] cache
 
/// lines, but the sparse array is 4x smaller.  N is the number of elements in
 
/// the set.
 
///
 
/// For sets that may grow to thousands of elements, SparseT should be set to
 
/// uint16_t or uint32_t.
 
///
 
/// Multiset behavior is provided by providing doubly linked lists for values
 
/// that are inlined in the dense vector. SparseMultiSet is a good choice when
 
/// one desires a growable number of entries per key, as it will retain the
 
/// SparseSet algorithmic properties despite being growable. Thus, it is often a
 
/// better choice than a SparseSet of growable containers or a vector of
 
/// vectors. SparseMultiSet also keeps iterators valid after erasure (provided
 
/// the iterators don't point to the element erased), allowing for more
 
/// intuitive and fast removal.
 
///
 
/// @tparam ValueT      The type of objects in the set.
 
/// @tparam KeyFunctorT A functor that computes an unsigned index from KeyT.
 
/// @tparam SparseT     An unsigned integer type. See above.
 
///
 
template<typename ValueT,
 
         typename KeyFunctorT = identity<unsigned>,
 
         typename SparseT = uint8_t>
 
class SparseMultiSet {
 
  static_assert(std::is_unsigned_v<SparseT>,
 
                "SparseT must be an unsigned integer type");
 
 
 
  /// The actual data that's stored, as a doubly-linked list implemented via
 
  /// indices into the DenseVector.  The doubly linked list is implemented
 
  /// circular in Prev indices, and INVALID-terminated in Next indices. This
 
  /// provides efficient access to list tails. These nodes can also be
 
  /// tombstones, in which case they are actually nodes in a single-linked
 
  /// freelist of recyclable slots.
 
  struct SMSNode {
 
    static constexpr unsigned INVALID = ~0U;
 
 
 
    ValueT Data;
 
    unsigned Prev;
 
    unsigned Next;
 
 
 
    SMSNode(ValueT D, unsigned P, unsigned N) : Data(D), Prev(P), Next(N) {}
 
 
 
    /// List tails have invalid Nexts.
 
    bool isTail() const {
 
      return Next == INVALID;
 
    }
 
 
 
    /// Whether this node is a tombstone node, and thus is in our freelist.
 
    bool isTombstone() const {
 
      return Prev == INVALID;
 
    }
 
 
 
    /// Since the list is circular in Prev, all non-tombstone nodes have a valid
 
    /// Prev.
 
    bool isValid() const { return Prev != INVALID; }
 
  };
 
 
 
  using KeyT = typename KeyFunctorT::argument_type;
 
  using DenseT = SmallVector<SMSNode, 8>;
 
  DenseT Dense;
 
  SparseT *Sparse = nullptr;
 
  unsigned Universe = 0;
 
  KeyFunctorT KeyIndexOf;
 
  SparseSetValFunctor<KeyT, ValueT, KeyFunctorT> ValIndexOf;
 
 
 
  /// We have a built-in recycler for reusing tombstone slots. This recycler
 
  /// puts a singly-linked free list into tombstone slots, allowing us quick
 
  /// erasure, iterator preservation, and dense size.
 
  unsigned FreelistIdx = SMSNode::INVALID;
 
  unsigned NumFree = 0;
 
 
 
  unsigned sparseIndex(const ValueT &Val) const {
 
    assert(ValIndexOf(Val) < Universe &&
 
           "Invalid key in set. Did object mutate?");
 
    return ValIndexOf(Val);
 
  }
 
  unsigned sparseIndex(const SMSNode &N) const { return sparseIndex(N.Data); }
 
 
 
  /// Whether the given entry is the head of the list. List heads's previous
 
  /// pointers are to the tail of the list, allowing for efficient access to the
 
  /// list tail. D must be a valid entry node.
 
  bool isHead(const SMSNode &D) const {
 
    assert(D.isValid() && "Invalid node for head");
 
    return Dense[D.Prev].isTail();
 
  }
 
 
 
  /// Whether the given entry is a singleton entry, i.e. the only entry with
 
  /// that key.
 
  bool isSingleton(const SMSNode &N) const {
 
    assert(N.isValid() && "Invalid node for singleton");
 
    // Is N its own predecessor?
 
    return &Dense[N.Prev] == &N;
 
  }
 
 
 
  /// Add in the given SMSNode. Uses a free entry in our freelist if
 
  /// available. Returns the index of the added node.
 
  unsigned addValue(const ValueT& V, unsigned Prev, unsigned Next) {
 
    if (NumFree == 0) {
 
      Dense.push_back(SMSNode(V, Prev, Next));
 
      return Dense.size() - 1;
 
    }
 
 
 
    // Peel off a free slot
 
    unsigned Idx = FreelistIdx;
 
    unsigned NextFree = Dense[Idx].Next;
 
    assert(Dense[Idx].isTombstone() && "Non-tombstone free?");
 
 
 
    Dense[Idx] = SMSNode(V, Prev, Next);
 
    FreelistIdx = NextFree;
 
    --NumFree;
 
    return Idx;
 
  }
 
 
 
  /// Make the current index a new tombstone. Pushes it onto the freelist.
 
  void makeTombstone(unsigned Idx) {
 
    Dense[Idx].Prev = SMSNode::INVALID;
 
    Dense[Idx].Next = FreelistIdx;
 
    FreelistIdx = Idx;
 
    ++NumFree;
 
  }
 
 
 
public:
 
  using value_type = ValueT;
 
  using reference = ValueT &;
 
  using const_reference = const ValueT &;
 
  using pointer = ValueT *;
 
  using const_pointer = const ValueT *;
 
  using size_type = unsigned;
 
 
 
  SparseMultiSet() = default;
 
  SparseMultiSet(const SparseMultiSet &) = delete;
 
  SparseMultiSet &operator=(const SparseMultiSet &) = delete;
 
  ~SparseMultiSet() { free(Sparse); }
 
 
 
  /// Set the universe size which determines the largest key the set can hold.
 
  /// The universe must be sized before any elements can be added.
 
  ///
 
  /// @param U Universe size. All object keys must be less than U.
 
  ///
 
  void setUniverse(unsigned U) {
 
    // It's not hard to resize the universe on a non-empty set, but it doesn't
 
    // seem like a likely use case, so we can add that code when we need it.
 
    assert(empty() && "Can only resize universe on an empty map");
 
    // Hysteresis prevents needless reallocations.
 
    if (U >= Universe/4 && U <= Universe)
 
      return;
 
    free(Sparse);
 
    // The Sparse array doesn't actually need to be initialized, so malloc
 
    // would be enough here, but that will cause tools like valgrind to
 
    // complain about branching on uninitialized data.
 
    Sparse = static_cast<SparseT*>(safe_calloc(U, sizeof(SparseT)));
 
    Universe = U;
 
  }
 
 
 
  /// Our iterators are iterators over the collection of objects that share a
 
  /// key.
 
  template <typename SMSPtrTy> class iterator_base {
 
    friend class SparseMultiSet;
 
 
 
  public:
 
    using iterator_category = std::bidirectional_iterator_tag;
 
    using value_type = ValueT;
 
    using difference_type = std::ptrdiff_t;
 
    using pointer = value_type *;
 
    using reference = value_type &;
 
 
 
  private:
 
    SMSPtrTy SMS;
 
    unsigned Idx;
 
    unsigned SparseIdx;
 
 
 
    iterator_base(SMSPtrTy P, unsigned I, unsigned SI)
 
      : SMS(P), Idx(I), SparseIdx(SI) {}
 
 
 
    /// Whether our iterator has fallen outside our dense vector.
 
    bool isEnd() const {
 
      if (Idx == SMSNode::INVALID)
 
        return true;
 
 
 
      assert(Idx < SMS->Dense.size() && "Out of range, non-INVALID Idx?");
 
      return false;
 
    }
 
 
 
    /// Whether our iterator is properly keyed, i.e. the SparseIdx is valid
 
    bool isKeyed() const { return SparseIdx < SMS->Universe; }
 
 
 
    unsigned Prev() const { return SMS->Dense[Idx].Prev; }
 
    unsigned Next() const { return SMS->Dense[Idx].Next; }
 
 
 
    void setPrev(unsigned P) { SMS->Dense[Idx].Prev = P; }
 
    void setNext(unsigned N) { SMS->Dense[Idx].Next = N; }
 
 
 
  public:
 
    reference operator*() const {
 
      assert(isKeyed() && SMS->sparseIndex(SMS->Dense[Idx].Data) == SparseIdx &&
 
             "Dereferencing iterator of invalid key or index");
 
 
 
      return SMS->Dense[Idx].Data;
 
    }
 
    pointer operator->() const { return &operator*(); }
 
 
 
    /// Comparison operators
 
    bool operator==(const iterator_base &RHS) const {
 
      // end compares equal
 
      if (SMS == RHS.SMS && Idx == RHS.Idx) {
 
        assert((isEnd() || SparseIdx == RHS.SparseIdx) &&
 
               "Same dense entry, but different keys?");
 
        return true;
 
      }
 
 
 
      return false;
 
    }
 
 
 
    bool operator!=(const iterator_base &RHS) const {
 
      return !operator==(RHS);
 
    }
 
 
 
    /// Increment and decrement operators
 
    iterator_base &operator--() { // predecrement - Back up
 
      assert(isKeyed() && "Decrementing an invalid iterator");
 
      assert((isEnd() || !SMS->isHead(SMS->Dense[Idx])) &&
 
             "Decrementing head of list");
 
 
 
      // If we're at the end, then issue a new find()
 
      if (isEnd())
 
        Idx = SMS->findIndex(SparseIdx).Prev();
 
      else
 
        Idx = Prev();
 
 
 
      return *this;
 
    }
 
    iterator_base &operator++() { // preincrement - Advance
 
      assert(!isEnd() && isKeyed() && "Incrementing an invalid/end iterator");
 
      Idx = Next();
 
      return *this;
 
    }
 
    iterator_base operator--(int) { // postdecrement
 
      iterator_base I(*this);
 
      --*this;
 
      return I;
 
    }
 
    iterator_base operator++(int) { // postincrement
 
      iterator_base I(*this);
 
      ++*this;
 
      return I;
 
    }
 
  };
 
 
 
  using iterator = iterator_base<SparseMultiSet *>;
 
  using const_iterator = iterator_base<const SparseMultiSet *>;
 
 
 
  // Convenience types
 
  using RangePair = std::pair<iterator, iterator>;
 
 
 
  /// Returns an iterator past this container. Note that such an iterator cannot
 
  /// be decremented, but will compare equal to other end iterators.
 
  iterator end() { return iterator(this, SMSNode::INVALID, SMSNode::INVALID); }
 
  const_iterator end() const {
 
    return const_iterator(this, SMSNode::INVALID, SMSNode::INVALID);
 
  }
 
 
 
  /// Returns true if the set is empty.
 
  ///
 
  /// This is not the same as BitVector::empty().
 
  ///
 
  bool empty() const { return size() == 0; }
 
 
 
  /// Returns the number of elements in the set.
 
  ///
 
  /// This is not the same as BitVector::size() which returns the size of the
 
  /// universe.
 
  ///
 
  size_type size() const {
 
    assert(NumFree <= Dense.size() && "Out-of-bounds free entries");
 
    return Dense.size() - NumFree;
 
  }
 
 
 
  /// Clears the set.  This is a very fast constant time operation.
 
  ///
 
  void clear() {
 
    // Sparse does not need to be cleared, see find().
 
    Dense.clear();
 
    NumFree = 0;
 
    FreelistIdx = SMSNode::INVALID;
 
  }
 
 
 
  /// Find an element by its index.
 
  ///
 
  /// @param   Idx A valid index to find.
 
  /// @returns An iterator to the element identified by key, or end().
 
  ///
 
  iterator findIndex(unsigned Idx) {
 
    assert(Idx < Universe && "Key out of range");
 
    const unsigned Stride = std::numeric_limits<SparseT>::max() + 1u;
 
    for (unsigned i = Sparse[Idx], e = Dense.size(); i < e; i += Stride) {
 
      const unsigned FoundIdx = sparseIndex(Dense[i]);
 
      // Check that we're pointing at the correct entry and that it is the head
 
      // of a valid list.
 
      if (Idx == FoundIdx && Dense[i].isValid() && isHead(Dense[i]))
 
        return iterator(this, i, Idx);
 
      // Stride is 0 when SparseT >= unsigned.  We don't need to loop.
 
      if (!Stride)
 
        break;
 
    }
 
    return end();
 
  }
 
 
 
  /// Find an element by its key.
 
  ///
 
  /// @param   Key A valid key to find.
 
  /// @returns An iterator to the element identified by key, or end().
 
  ///
 
  iterator find(const KeyT &Key) {
 
    return findIndex(KeyIndexOf(Key));
 
  }
 
 
 
  const_iterator find(const KeyT &Key) const {
 
    iterator I = const_cast<SparseMultiSet*>(this)->findIndex(KeyIndexOf(Key));
 
    return const_iterator(I.SMS, I.Idx, KeyIndexOf(Key));
 
  }
 
 
 
  /// Returns the number of elements identified by Key. This will be linear in
 
  /// the number of elements of that key.
 
  size_type count(const KeyT &Key) const {
 
    unsigned Ret = 0;
 
    for (const_iterator It = find(Key); It != end(); ++It)
 
      ++Ret;
 
 
 
    return Ret;
 
  }
 
 
 
  /// Returns true if this set contains an element identified by Key.
 
  bool contains(const KeyT &Key) const {
 
    return find(Key) != end();
 
  }
 
 
 
  /// Return the head and tail of the subset's list, otherwise returns end().
 
  iterator getHead(const KeyT &Key) { return find(Key); }
 
  iterator getTail(const KeyT &Key) {
 
    iterator I = find(Key);
 
    if (I != end())
 
      I = iterator(this, I.Prev(), KeyIndexOf(Key));
 
    return I;
 
  }
 
 
 
  /// The bounds of the range of items sharing Key K. First member is the head
 
  /// of the list, and the second member is a decrementable end iterator for
 
  /// that key.
 
  RangePair equal_range(const KeyT &K) {
 
    iterator B = find(K);
 
    iterator E = iterator(this, SMSNode::INVALID, B.SparseIdx);
 
    return std::make_pair(B, E);
 
  }
 
 
 
  /// Insert a new element at the tail of the subset list. Returns an iterator
 
  /// to the newly added entry.
 
  iterator insert(const ValueT &Val) {
 
    unsigned Idx = sparseIndex(Val);
 
    iterator I = findIndex(Idx);
 
 
 
    unsigned NodeIdx = addValue(Val, SMSNode::INVALID, SMSNode::INVALID);
 
 
 
    if (I == end()) {
 
      // Make a singleton list
 
      Sparse[Idx] = NodeIdx;
 
      Dense[NodeIdx].Prev = NodeIdx;
 
      return iterator(this, NodeIdx, Idx);
 
    }
 
 
 
    // Stick it at the end.
 
    unsigned HeadIdx = I.Idx;
 
    unsigned TailIdx = I.Prev();
 
    Dense[TailIdx].Next = NodeIdx;
 
    Dense[HeadIdx].Prev = NodeIdx;
 
    Dense[NodeIdx].Prev = TailIdx;
 
 
 
    return iterator(this, NodeIdx, Idx);
 
  }
 
 
 
  /// Erases an existing element identified by a valid iterator.
 
  ///
 
  /// This invalidates iterators pointing at the same entry, but erase() returns
 
  /// an iterator pointing to the next element in the subset's list. This makes
 
  /// it possible to erase selected elements while iterating over the subset:
 
  ///
 
  ///   tie(I, E) = Set.equal_range(Key);
 
  ///   while (I != E)
 
  ///     if (test(*I))
 
  ///       I = Set.erase(I);
 
  ///     else
 
  ///       ++I;
 
  ///
 
  /// Note that if the last element in the subset list is erased, this will
 
  /// return an end iterator which can be decremented to get the new tail (if it
 
  /// exists):
 
  ///
 
  ///  tie(B, I) = Set.equal_range(Key);
 
  ///  for (bool isBegin = B == I; !isBegin; /* empty */) {
 
  ///    isBegin = (--I) == B;
 
  ///    if (test(I))
 
  ///      break;
 
  ///    I = erase(I);
 
  ///  }
 
  iterator erase(iterator I) {
 
    assert(I.isKeyed() && !I.isEnd() && !Dense[I.Idx].isTombstone() &&
 
           "erasing invalid/end/tombstone iterator");
 
 
 
    // First, unlink the node from its list. Then swap the node out with the
 
    // dense vector's last entry
 
    iterator NextI = unlink(Dense[I.Idx]);
 
 
 
    // Put in a tombstone.
 
    makeTombstone(I.Idx);
 
 
 
    return NextI;
 
  }
 
 
 
  /// Erase all elements with the given key. This invalidates all
 
  /// iterators of that key.
 
  void eraseAll(const KeyT &K) {
 
    for (iterator I = find(K); I != end(); /* empty */)
 
      I = erase(I);
 
  }
 
 
 
private:
 
  /// Unlink the node from its list. Returns the next node in the list.
 
  iterator unlink(const SMSNode &N) {
 
    if (isSingleton(N)) {
 
      // Singleton is already unlinked
 
      assert(N.Next == SMSNode::INVALID && "Singleton has next?");
 
      return iterator(this, SMSNode::INVALID, ValIndexOf(N.Data));
 
    }
 
 
 
    if (isHead(N)) {
 
      // If we're the head, then update the sparse array and our next.
 
      Sparse[sparseIndex(N)] = N.Next;
 
      Dense[N.Next].Prev = N.Prev;
 
      return iterator(this, N.Next, ValIndexOf(N.Data));
 
    }
 
 
 
    if (N.isTail()) {
 
      // If we're the tail, then update our head and our previous.
 
      findIndex(sparseIndex(N)).setPrev(N.Prev);
 
      Dense[N.Prev].Next = N.Next;
 
 
 
      // Give back an end iterator that can be decremented
 
      iterator I(this, N.Prev, ValIndexOf(N.Data));
 
      return ++I;
 
    }
 
 
 
    // Otherwise, just drop us
 
    Dense[N.Next].Prev = N.Prev;
 
    Dense[N.Prev].Next = N.Next;
 
    return iterator(this, N.Next, ValIndexOf(N.Data));
 
  }
 
};
 
 
 
} // end namespace llvm
 
 
 
#endif // LLVM_ADT_SPARSEMULTISET_H