#pragma once /* ** tarray.h ** Templated, automatically resizing array ** **--------------------------------------------------------------------------- ** Copyright 1998-2007 Randy Heit ** All rights reserved. ** ** Redistribution and use in source and binary forms, with or without ** modification, are permitted provided that the following conditions ** are met: ** ** 1. Redistributions of source code must retain the above copyright ** notice, this list of conditions and the following disclaimer. ** 2. Redistributions in binary form must reproduce the above copyright ** notice, this list of conditions and the following disclaimer in the ** documentation and/or other materials provided with the distribution. ** 3. The name of the author may not be used to endorse or promote products ** derived from this software without specific prior written permission. ** ** THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR ** IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES ** OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. ** IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, ** INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT ** NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, ** DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY ** THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT ** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF ** THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. **--------------------------------------------------------------------------- ** */ #include #include #include #include #include #include #if !defined(_WIN32) #include // for intptr_t #else #include // for mingw #endif template class TIterator : public std::iterator { public: typedef typename TIterator::value_type value_type; typedef typename TIterator::difference_type difference_type; typedef typename TIterator::pointer pointer; typedef typename TIterator::reference reference; typedef typename TIterator::iterator_category iterator_category; TIterator(T* ptr = nullptr) { m_ptr = ptr; } // Comparison operators bool operator==(const TIterator &other) const { return m_ptr == other.m_ptr; } bool operator!=(const TIterator &other) const { return m_ptr != other.m_ptr; } bool operator< (const TIterator &other) const { return m_ptr < other.m_ptr; } bool operator<=(const TIterator &other) const { return m_ptr <= other.m_ptr; } bool operator> (const TIterator &other) const { return m_ptr > other.m_ptr; } bool operator>=(const TIterator &other) const { return m_ptr >= other.m_ptr; } // Arithmetic operators TIterator &operator++() { ++m_ptr; return *this; } TIterator operator++(int) { pointer tmp = m_ptr; ++*this; return TIterator(tmp); } TIterator &operator--() { --m_ptr; return *this; } TIterator operator--(int) { pointer tmp = m_ptr; --*this; return TIterator(tmp); } TIterator &operator+=(difference_type offset) { m_ptr += offset; return *this; } TIterator operator+(difference_type offset) const { return TIterator(m_ptr + offset); } friend TIterator operator+(difference_type offset, const TIterator &other) { return TIterator(offset + other.m_ptr); } TIterator &operator-=(difference_type offset) { m_ptr -= offset; return *this; } TIterator operator-(difference_type offset) const { return TIterator(m_ptr - offset); } difference_type operator-(const TIterator &other) const { return m_ptr - other.m_ptr; } T &operator*() { return *m_ptr; } const T &operator*() const { return *m_ptr; } T* operator->() { return m_ptr; } protected: T* m_ptr; }; // TArray ------------------------------------------------------------------- // Must match TArray's layout. struct FArray { void *Array; unsigned int Count; unsigned int Most; }; // T is the type stored in the array. // TT is the type returned by operator(). template class TArray { public: typedef TIterator iterator; typedef TIterator const_iterator; typedef T value_type; iterator begin() { return &Array[0]; } const_iterator begin() const { return &Array[0]; } const_iterator cbegin() const { return &Array[0]; } iterator end() { return &Array[Count]; } const_iterator end() const { return &Array[Count]; } const_iterator cend() const { return &Array[Count]; } //////// // This is a dummy constructor that does nothing. The purpose of this // is so you can create a global TArray in the data segment that gets // used by code before startup without worrying about the constructor // resetting it after it's already been used. You MUST NOT use it for // heap- or stack-allocated TArrays. enum ENoInit { NoInit }; TArray (ENoInit dummy) { } //////// TArray () { Most = 0; Count = 0; Array = NULL; } explicit TArray (size_t max, bool reserve = false) { Most = (unsigned)max; Count = (unsigned)(reserve? max : 0); Array = (T *)malloc (sizeof(T)*max); if (reserve && Count > 0) { ConstructEmpty(0, Count - 1); } } TArray (const TArray &other) { DoCopy (other); } TArray (TArray &&other) { Array = other.Array; other.Array = NULL; Most = other.Most; other.Most = 0; Count = other.Count; other.Count = 0; } TArray &operator= (const TArray &other) { if (&other != this) { if (Array != NULL) { if (Count > 0) { DoDelete (0, Count-1); } free (Array); } DoCopy (other); } return *this; } TArray &operator= (TArray &&other) { if (Array) { if (Count > 0) { DoDelete (0, Count-1); } free (Array); } Array = other.Array; other.Array = NULL; Most = other.Most; other.Most = 0; Count = other.Count; other.Count = 0; return *this; } ~TArray () { if (Array) { if (Count > 0) { DoDelete (0, Count-1); } free (Array); Array = NULL; Count = 0; Most = 0; } } // Check equality of two arrays bool operator==(const TArray &other) const { if (Count != other.Count) { return false; } for (unsigned int i = 0; i < Count; ++i) { if (Array[i] != other.Array[i]) { return false; } } return true; } // Return a reference to an element. // Note that the asserts must let the element after the end pass because this gets frequently used as a sentinel pointer. T &operator[] (size_t index) const { assert(index <= Count); return Array[index]; } // Returns the value of an element TT operator() (size_t index) const { assert(index <= Count); return Array[index]; } // Returns a reference to the last element T &Last() const { return Array[Count-1]; } // returns address of first element T *Data() const { return &Array[0]; } unsigned int Find(const T& item) const { unsigned int i; for(i = 0;i < Count;++i) { if(Array[i] == item) break; } return i; } template unsigned int FindEx(Func compare) const { unsigned int i; for (i = 0; i < Count; ++i) { if (compare(Array[i])) break; } return i; } unsigned int Push (const T &item) { Grow (1); ::new((void*)&Array[Count]) T(item); return Count++; } unsigned int Push(T &&item) { Grow(1); ::new((void*)&Array[Count]) T(std::move(item)); return Count++; } unsigned Append(const TArray &item) { unsigned start = Count; Grow(item.Size()); Count += item.Size(); for (unsigned i = 0; i < item.Size(); i++) { new(&Array[start + i]) T(item[i]); } return start; } unsigned Append(TArray &&item) { unsigned start = Count; Grow(item.Size()); Count += item.Size(); for (unsigned i = 0; i < item.Size(); i++) { new(&Array[start + i]) T(std::move(item[i])); } return start; } bool Pop () { if (Count > 0) { Array[--Count].~T(); return true; } return false; } bool Pop (T &item) { if (Count > 0) { item = Array[--Count]; Array[Count].~T(); return true; } return false; } void Delete (unsigned int index) { if (index < Count) { Array[index].~T(); if (index < --Count) { memmove (&Array[index], &Array[index+1], sizeof(T)*(Count - index)); } } } void Delete (unsigned int index, int deletecount) { if (index + deletecount > Count) { deletecount = Count - index; } if (deletecount > 0) { for (int i = 0; i < deletecount; i++) { Array[index + i].~T(); } Count -= deletecount; if (index < Count) { memmove (&Array[index], &Array[index+deletecount], sizeof(T)*(Count - index)); } } } // Inserts an item into the array, shifting elements as needed void Insert (unsigned int index, const T &item) { if (index >= Count) { // Inserting somewhere past the end of the array, so we can // just add it without moving things. Resize (index + 1); ::new ((void *)&Array[index]) T(item); } else { // Inserting somewhere in the middle of the array, // so make room for it Resize (Count + 1); // Now move items from the index and onward out of the way memmove (&Array[index+1], &Array[index], sizeof(T)*(Count - index - 1)); // And put the new element in ::new ((void *)&Array[index]) T(item); } } void ShrinkToFit () { if (Most > Count) { Most = Count; if (Most == 0) { if (Array != NULL) { free (Array); Array = NULL; } } else { DoResize (); } } } // Grow Array to be large enough to hold amount more entries without // further growing. void Grow (unsigned int amount) { if (Count + amount > Most) { const unsigned int choicea = Count + amount; const unsigned int choiceb = Most = (Most >= 16) ? Most + Most / 2 : 16; Most = (choicea > choiceb ? choicea : choiceb); DoResize (); } } // Resize Array so that it has exactly amount entries in use. void Resize (unsigned int amount) { if (Count < amount) { // Adding new entries Grow (amount - Count); ConstructEmpty(Count, amount - 1); } else if (Count != amount) { // Deleting old entries DoDelete (amount, Count - 1); } Count = amount; } // Ensures that the array has at most amount entries. // Useful in cases where the initial allocation may be larger than the final result. // Resize would create a lot of unneeded code in those cases. void Clamp(unsigned int amount) { if (Count > amount) { // Deleting old entries DoDelete(amount, Count - 1); Count = amount; } } void Alloc(unsigned int amount) { // first destroys all content and then rebuilds the array. if (Count > 0) DoDelete(0, Count - 1); Count = 0; Resize(amount); ShrinkToFit(); } // Reserves amount entries at the end of the array, but does nothing // with them. unsigned int Reserve (size_t amount) { Grow ((unsigned)amount); unsigned int place = Count; Count += (unsigned)amount; if (Count > 0) ConstructEmpty(place, Count - 1); return place; } unsigned int Size () const { return Count; } unsigned int Max () const { return Most; } void Clear () { if (Count > 0) { DoDelete (0, Count-1); Count = 0; } } void Reset() { Clear(); Most = 0; if (Array != nullptr) { free(Array); Array = nullptr; } } void Swap(TArray &other) { std::swap(Array, other.Array); std::swap(Count, other.Count); std::swap(Most, other.Most); } private: T *Array; unsigned int Count; unsigned int Most; void DoCopy (const TArray &other) { Most = Count = other.Count; if (Count != 0) { Array = (T *)malloc (sizeof(T)*Most); for (unsigned int i = 0; i < Count; ++i) { ::new(&Array[i]) T(other.Array[i]); } } else { Array = NULL; } } void DoResize () { size_t allocsize = sizeof(T)*Most; Array = (T *)realloc (Array, allocsize); } void DoDelete (unsigned int first, unsigned int last) { assert (last != ~0u); for (unsigned int i = first; i <= last; ++i) { Array[i].~T(); } } void ConstructEmpty(unsigned int first, unsigned int last) { assert(last != ~0u); for (unsigned int i = first; i <= last; ++i) { ::new(&Array[i]) T; } } }; // TDeletingArray ----------------------------------------------------------- // An array that deletes its elements when it gets deleted. template class TDeletingArray : public TArray { public: TDeletingArray() : TArray() {} TDeletingArray(TDeletingArray &&other) : TArray(std::move(other)) {} TDeletingArray &operator=(TDeletingArray &&other) { TArray::operator=(std::move(other)); return *this; } ~TDeletingArray () { for (unsigned int i = 0; i < TArray::Size(); ++i) { if ((*this)[i] != NULL) delete (*this)[i]; } } void DeleteAndClear() { for (unsigned int i = 0; i < TArray::Size(); ++i) { if ((*this)[i] != NULL) delete (*this)[i]; } this->Clear(); } }; // This is only used for exposing the sector's Lines array to ZScript. // Unlike TArrayView, its members are public as needed by the map loader. template class TStaticPointedArray { public: typedef TIterator iterator; typedef TIterator const_iterator; iterator begin() { return &Array[0]; } const_iterator begin() const { return &Array[0]; } const_iterator cbegin() const { return &Array[0]; } iterator end() { return &Array[Count]; } const_iterator end() const { return &Array[Count]; } const_iterator cend() const { return &Array[Count]; } void Init(T *ptr, unsigned cnt) { Array = ptr; Count = cnt; } // Return a reference to an element T &operator[] (size_t index) const { return Array[index]; } T &At(size_t index) const { return Array[index]; } unsigned int Size() const { return Count; } // Some code needs to access these directly so they cannot be private. T *Array; unsigned int Count; }; // TAutoGrowArray ----------------------------------------------------------- // An array with accessors that automatically grow the array as needed. // It can still be used as a normal TArray if needed. ACS uses this for // world and global arrays. template class TAutoGrowArray : public TArray { public: T GetVal (unsigned int index) { if (index >= this->Size()) { return 0; } return (*this)[index]; } void SetVal (unsigned int index, T val) { if ((int)index < 0) return; // These always result in an out of memory condition. if (index >= this->Size()) { this->Resize (index + 1); } (*this)[index] = val; } }; // TMap --------------------------------------------------------------------- // An associative array, similar in concept to the STL extension // class hash_map. It is implemented using Lua's table algorithm: /* ** Hash uses a mix of chained scatter table with Brent's variation. ** A main invariant of these tables is that, if an element is not ** in its main position (i.e. the `original' position that its hash gives ** to it), then the colliding element is in its own main position. ** Hence even when the load factor reaches 100%, performance remains good. */ /****************************************************************************** * Copyright (C) 1994-2006 Lua.org, PUC-Rio. All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining * a copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sublicense, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ******************************************************************************/ typedef unsigned int hash_t; template struct THashTraits { // Returns the hash value for a key. hash_t Hash(const KT key) { return (hash_t)(intptr_t)key; } hash_t Hash(double key) { hash_t keyhash[2]; memcpy(&keyhash, &key, sizeof(keyhash)); return keyhash[0] ^ keyhash[1]; } // Compares two keys, returning zero if they are the same. int Compare(const KT left, const KT right) { return left != right; } }; template<> struct THashTraits { // Use all bits when hashing singles instead of converting them to ints. hash_t Hash(float key) { hash_t keyhash; memcpy(&keyhash, &key, sizeof(keyhash)); return keyhash; } int Compare(float left, float right) { return left != right; } }; template<> struct THashTraits { // Use all bits when hashing doubles instead of converting them to ints. hash_t Hash(double key) { hash_t keyhash[2]; memcpy(&keyhash, &key, sizeof(keyhash)); return keyhash[0] ^ keyhash[1]; } int Compare(double left, double right) { return left != right; } }; template struct TValueTraits { // Initializes a value for TMap. If a regular constructor isn't // good enough, you can override it. void Init(VT &value) { ::new(&value) VT; } }; // Must match layout of TMap struct FMap { void *Nodes; void *LastFree; hash_t Size; hash_t NumUsed; }; template class TMapIterator; template class TMapConstIterator; template, class ValueTraits=TValueTraits > class TMap { template friend class TMapIterator; template friend class TMapConstIterator; public: typedef class TMap MyType; typedef class TMapIterator Iterator; typedef class TMapConstIterator ConstIterator; typedef struct { const KT Key; VT Value; } Pair; typedef const Pair ConstPair; TMap() { NumUsed = 0; SetNodeVector(1); } TMap(hash_t size) { NumUsed = 0; SetNodeVector(size); } ~TMap() { ClearNodeVector(); } TMap(const TMap &o) { NumUsed = 0; SetNodeVector(o.CountUsed()); CopyNodes(o.Nodes, o.Size); } TMap &operator= (const TMap &o) { NumUsed = 0; ClearNodeVector(); SetNodeVector(o.CountUsed()); CopyNodes(o.Nodes, o.Size); return *this; } //======================================================================= // // TransferFrom // // Moves the contents from one TMap to another, leaving the TMap moved // from empty. // //======================================================================= void TransferFrom(TMap &o) { // Clear all our nodes. NumUsed = 0; ClearNodeVector(); // Copy all of o's nodes. Nodes = o.Nodes; LastFree = o.LastFree; Size = o.Size; NumUsed = o.NumUsed; // Tell o it doesn't have any nodes. o.Nodes = NULL; o.Size = 0; o.LastFree = NULL; o.NumUsed = 0; // Leave o functional with one empty node. o.SetNodeVector(1); } //======================================================================= // // Clear // // Empties out the table and resizes it with room for count entries. // //======================================================================= void Clear(hash_t count=1) { ClearNodeVector(); SetNodeVector(count); } //======================================================================= // // CountUsed // // Returns the number of entries in use in the table. // //======================================================================= hash_t CountUsed() const { #ifdef _DEBUG hash_t used = 0; hash_t ct = Size; for (Node *n = Nodes; ct-- > 0; ++n) { if (!n->IsNil()) { ++used; } } assert (used == NumUsed); #endif return NumUsed; } //======================================================================= // // operator[] // // Returns a reference to the value associated with a particular key, // creating the pair if the key isn't already in the table. // //======================================================================= VT &operator[] (const KT key) { return GetNode(key)->Pair.Value; } const VT &operator[] (const KT key) const { return GetNode(key)->Pair.Value; } //======================================================================= // // CheckKey // // Returns a pointer to the value associated with a particular key, or // NULL if the key isn't in the table. // //======================================================================= VT *CheckKey (const KT key) { Node *n = FindKey(key); return n != NULL ? &n->Pair.Value : NULL; } const VT *CheckKey (const KT key) const { const Node *n = FindKey(key); return n != NULL ? &n->Pair.Value : NULL; } //======================================================================= // // Insert // // Adds a key/value pair to the table if key isn't in the table, or // replaces the value for the existing pair if the key is in the table. // // This is functionally equivalent to (*this)[key] = value; but can be // slightly faster if the pair needs to be created because it doesn't run // the constructor on the value part twice. // //======================================================================= VT &Insert(const KT key, const VT &value) { Node *n = FindKey(key); if (n != NULL) { n->Pair.Value = value; } else { n = NewKey(key); ::new(&n->Pair.Value) VT(value); } return n->Pair.Value; } VT &Insert(const KT key, VT &&value) { Node *n = FindKey(key); if (n != NULL) { n->Pair.Value = value; } else { n = NewKey(key); ::new(&n->Pair.Value) VT(value); } return n->Pair.Value; } VT &InsertNew(const KT key) { Node *n = FindKey(key); if (n != NULL) { n->Pair.Value.~VT(); } else { n = NewKey(key); } ::new(&n->Pair.Value) VT; return n->Pair.Value; } //======================================================================= // // Remove // // Removes the key/value pair for a particular key if it is in the table. // //======================================================================= void Remove(const KT key) { DelKey(key); } void Swap(MyType &other) { std::swap(Nodes, other.Nodes); std::swap(LastFree, other.LastFree); std::swap(Size, other.Size); std::swap(NumUsed, other.NumUsed); } protected: struct IPair // This must be the same as Pair above, but with a { // non-const Key. KT Key; VT Value; }; struct Node { Node *Next; IPair Pair; void SetNil() { Next = (Node *)1; } bool IsNil() const { return Next == (Node *)1; } }; /* This is used instead of memcpy, because Node is likely to be small, * such that the time spent calling a function would eclipse the time * spent copying. */ struct NodeSizedStruct { unsigned char Pads[sizeof(Node)]; }; Node *Nodes; Node *LastFree; /* any free position is before this position */ hash_t Size; /* must be a power of 2 */ hash_t NumUsed; const Node *MainPosition(const KT k) const { HashTraits Traits; return &Nodes[Traits.Hash(k) & (Size - 1)]; } Node *MainPosition(const KT k) { HashTraits Traits; return &Nodes[Traits.Hash(k) & (Size - 1)]; } void SetNodeVector(hash_t size) { // Round size up to nearest power of 2 for (Size = 1; Size < size; Size <<= 1) { } Nodes = (Node *)malloc(Size * sizeof(Node)); LastFree = &Nodes[Size]; /* all positions are free */ for (hash_t i = 0; i < Size; ++i) { Nodes[i].SetNil(); } } void ClearNodeVector() { for (hash_t i = 0; i < Size; ++i) { if (!Nodes[i].IsNil()) { Nodes[i].~Node(); } } free(Nodes); Nodes = NULL; Size = 0; LastFree = NULL; NumUsed = 0; } void Resize(hash_t nhsize) { hash_t i, oldhsize = Size; Node *nold = Nodes; /* create new hash part with appropriate size */ SetNodeVector(nhsize); /* re-insert elements from hash part */ NumUsed = 0; for (i = 0; i < oldhsize; ++i) { if (!nold[i].IsNil()) { Node *n = NewKey(nold[i].Pair.Key); ::new(&n->Pair.Value) VT(std::move(nold[i].Pair.Value)); nold[i].~Node(); } } free(nold); } void Rehash() { Resize (Size << 1); } Node *GetFreePos() { while (LastFree-- > Nodes) { if (LastFree->IsNil()) { return LastFree; } } return NULL; /* could not find a free place */ } /* ** Inserts a new key into a hash table; first, check whether key's main ** position is free. If not, check whether colliding node is in its main ** position or not: if it is not, move colliding node to an empty place and ** put new key in its main position; otherwise (colliding node is in its main ** position), new key goes to an empty position. ** ** The Value field is left unconstructed. */ Node *NewKey(const KT key) { Node *mp = MainPosition(key); if (!mp->IsNil()) { Node *othern; Node *n = GetFreePos(); /* get a free place */ if (n == NULL) /* cannot find a free place? */ { Rehash(); /* grow table */ return NewKey(key); /* re-insert key into grown table */ } othern = MainPosition(mp->Pair.Key); if (othern != mp) /* is colliding node out of its main position? */ { /* yes; move colliding node into free position */ while (othern->Next != mp) /* find previous */ { othern = othern->Next; } othern->Next = n; /* redo the chain with 'n' in place of 'mp' */ CopyNode(n, mp); /* copy colliding node into free pos. (mp->Next also goes) */ mp->Next = NULL; /* now 'mp' is free */ } else /* colliding node is in its own main position */ { /* new node will go into free position */ n->Next = mp->Next; /* chain new position */ mp->Next = n; mp = n; } } else { mp->Next = NULL; } ++NumUsed; ::new(&mp->Pair.Key) KT(key); return mp; } void DelKey(const KT key) { Node *mp = MainPosition(key), **mpp; HashTraits Traits; if (mp->IsNil()) { /* the key is definitely not present, because there is nothing at its main position */ } else if (!Traits.Compare(mp->Pair.Key, key)) /* the key is in its main position */ { if (mp->Next != NULL) /* move next node to its main position */ { Node *n = mp->Next; mp->~Node(); /* deconstruct old node */ CopyNode(mp, n); /* copy next node */ n->SetNil(); /* next node is now nil */ } else { mp->~Node(); mp->SetNil(); /* there is no chain, so main position is nil */ } --NumUsed; } else /* the key is either not present or not in its main position */ { for (mpp = &mp->Next, mp = *mpp; mp != NULL && Traits.Compare(mp->Pair.Key, key); mpp = &mp->Next, mp = *mpp) { } /* look for the key */ if (mp != NULL) /* found it */ { *mpp = mp->Next; /* rechain so this node is skipped */ mp->~Node(); mp->SetNil(); /* because this node is now nil */ --NumUsed; } } } Node *FindKey(const KT key) { HashTraits Traits; Node *n = MainPosition(key); while (n != NULL && !n->IsNil() && Traits.Compare(n->Pair.Key, key)) { n = n->Next; } return n == NULL || n->IsNil() ? NULL : n; } const Node *FindKey(const KT key) const { HashTraits Traits; const Node *n = MainPosition(key); while (n != NULL && !n->IsNil() && Traits.Compare(n->Pair.Key, key)) { n = n->Next; } return n == NULL || n->IsNil() ? NULL : n; } Node *GetNode(const KT key) { Node *n = FindKey(key); if (n != NULL) { return n; } n = NewKey(key); ValueTraits traits; traits.Init(n->Pair.Value); return n; } /* Perform a bit-wise copy of the node. Used when relocating a node in the table. */ void CopyNode(Node *dst, const Node *src) { *(NodeSizedStruct *)dst = *(const NodeSizedStruct *)src; } /* Copy all nodes in the node vector to this table. */ void CopyNodes(const Node *nodes, hash_t numnodes) { for (; numnodes-- > 0; ++nodes) { if (!nodes->IsNil()) { Node *n = NewKey(nodes->Pair.Key); ::new(&n->Pair.Value) VT(nodes->Pair.Value); } } } }; // TMapIterator ------------------------------------------------------------- // A class to iterate over all the pairs in a TMap. template > class TMapIterator { public: TMapIterator(MapType &map) : Map(map), Position(0) { } //======================================================================= // // NextPair // // Returns false if there are no more entries in the table. Otherwise, it // returns true, and pair is filled with a pointer to the pair in the // table. // //======================================================================= bool NextPair(typename MapType::Pair *&pair) { if (Position >= Map.Size) { return false; } do { if (!Map.Nodes[Position].IsNil()) { pair = reinterpret_cast(&Map.Nodes[Position].Pair); Position += 1; return true; } } while (++Position < Map.Size); return false; } //======================================================================= // // Reset // // Restarts the iteration so you can do it all over again. // //======================================================================= void Reset() { Position = 0; } protected: MapType ⤅ hash_t Position; }; // TMapConstIterator -------------------------------------------------------- // Exactly the same as TMapIterator, but it works with a const TMap. template > class TMapConstIterator { public: TMapConstIterator(const MapType &map) : Map(map), Position(0) { } bool NextPair(typename MapType::ConstPair *&pair) { if (Position >= Map.Size) { return false; } do { if (!Map.Nodes[Position].IsNil()) { pair = reinterpret_cast(&Map.Nodes[Position].Pair); Position += 1; return true; } } while (++Position < Map.Size); return false; } protected: const MapType ⤅ hash_t Position; }; //========================================================================== // // an array to hold a small number of unique entries // //========================================================================== template class UniqueList { TArray Array; public: T * Get(T * t) { for (unsigned i = 0; i bytes; unsigned size; public: void Resize(unsigned elem) { bytes.Resize((elem + 7) / 8); size = elem; } BitArray() : size(0) { } BitArray(unsigned elem) : bytes((elem + 7) / 8, true) { } BitArray(const BitArray & arr) : bytes(arr.bytes) { size = arr.size; } BitArray &operator=(const BitArray & arr) { bytes = arr.bytes; size = arr.size; return *this; } BitArray(BitArray && arr) : bytes(std::move(arr.bytes)) { size = arr.size; arr.size = 0; } BitArray &operator=(BitArray && arr) { bytes = std::move(arr.bytes); size = arr.size; arr.size = 0; return *this; } bool operator[](size_t index) const { return !!(bytes[index >> 3] & (1 << (index & 7))); } void Set(size_t index, bool set = true) { if (!set) Clear(index); else bytes[index >> 3] |= (1 << (index & 7)); } void Clear(size_t index) { bytes[index >> 3] &= ~(1 << (index & 7)); } unsigned Size() const { return size; } void Zero() { memset(&bytes[0], 0, bytes.Size()); } }; template class FixedBitArray { uint8_t bytes[(size + 7) / 8]; public: FixedBitArray() = default; FixedBitArray(bool set) { memset(bytes, set ? -1 : 0, sizeof(bytes)); } bool operator[](size_t index) const { return !!(bytes[index >> 3] & (1 << (index & 7))); } void Set(size_t index, bool set = true) { if (!set) Clear(index); else bytes[index >> 3] |= (1 << (index & 7)); } void Clear(size_t index) { bytes[index >> 3] &= ~(1 << (index & 7)); } constexpr unsigned Size() const { return size; } void Zero() { memset(&bytes[0], 0, sizeof(bytes)); } void SetAll(bool on) { memset(&bytes[0], on ? -1 : 0, sizeof(bytes)); } }; // A wrapper to externally stored data. // I would have expected something for this in the stl, but std::span is only in C++20. template class TArrayView { public: typedef TIterator iterator; typedef TIterator const_iterator; iterator begin() { return &Array[0]; } const_iterator begin() const { return &Array[0]; } const_iterator cbegin() const { return &Array[0]; } iterator end() { return &Array[Count]; } const_iterator end() const { return &Array[Count]; } const_iterator cend() const { return &Array[Count]; } //////// TArrayView() = default; // intended to keep this type trivial. TArrayView(T *data, unsigned count = 0) { Count = count; Array = data; } TArrayView(const TArrayView &other) = default; TArrayView &operator= (const TArrayView &other) = default; // Check equality of two arrays bool operator==(const TArrayView &other) const { if (Count != other.Count) { return false; } for (unsigned int i = 0; i < Count; ++i) { if (Array[i] != other.Array[i]) { return false; } } return true; } // Return a reference to an element T &operator[] (size_t index) const { return Array[index]; } // Returns a reference to the last element T &Last() const { return Array[Count - 1]; } // returns address of first element T *Data() const { return &Array[0]; } unsigned Size() const { return Count; } unsigned int Find(const T& item) const { unsigned int i; for (i = 0; i < Count; ++i) { if (Array[i] == item) break; } return i; } void Set(T *data, unsigned count) { Array = data; Count = count; } void Clear() { Count = 0; Array = nullptr; } private: T *Array; unsigned int Count; };