Home | History | Annotate | Download | only in ADT
      1 //===--- StringMap.h - String Hash table map interface ----------*- C++ -*-===//
      2 //
      3 //                     The LLVM Compiler Infrastructure
      4 //
      5 // This file is distributed under the University of Illinois Open Source
      6 // License. See LICENSE.TXT for details.
      7 //
      8 //===----------------------------------------------------------------------===//
      9 //
     10 // This file defines the StringMap class.
     11 //
     12 //===----------------------------------------------------------------------===//
     13 
     14 #ifndef LLVM_ADT_STRINGMAP_H
     15 #define LLVM_ADT_STRINGMAP_H
     16 
     17 #include "llvm/ADT/StringRef.h"
     18 #include "llvm/Support/Allocator.h"
     19 #include <cstring>
     20 
     21 namespace llvm {
     22   template<typename ValueT>
     23   class StringMapConstIterator;
     24   template<typename ValueT>
     25   class StringMapIterator;
     26   template<typename ValueTy>
     27   class StringMapEntry;
     28 
     29 /// StringMapEntryInitializer - This datatype can be partially specialized for
     30 /// various datatypes in a stringmap to allow them to be initialized when an
     31 /// entry is default constructed for the map.
     32 template<typename ValueTy>
     33 class StringMapEntryInitializer {
     34 public:
     35   template <typename InitTy>
     36   static void Initialize(StringMapEntry<ValueTy> &T, InitTy InitVal) {
     37     T.second = InitVal;
     38   }
     39 };
     40 
     41 
     42 /// StringMapEntryBase - Shared base class of StringMapEntry instances.
     43 class StringMapEntryBase {
     44   unsigned StrLen;
     45 public:
     46   explicit StringMapEntryBase(unsigned Len) : StrLen(Len) {}
     47 
     48   unsigned getKeyLength() const { return StrLen; }
     49 };
     50 
     51 /// StringMapImpl - This is the base class of StringMap that is shared among
     52 /// all of its instantiations.
     53 class StringMapImpl {
     54 protected:
     55   // Array of NumBuckets pointers to entries, null pointers are holes.
     56   // TheTable[NumBuckets] contains a sentinel value for easy iteration. Followed
     57   // by an array of the actual hash values as unsigned integers.
     58   StringMapEntryBase **TheTable;
     59   unsigned NumBuckets;
     60   unsigned NumItems;
     61   unsigned NumTombstones;
     62   unsigned ItemSize;
     63 protected:
     64   explicit StringMapImpl(unsigned itemSize) : ItemSize(itemSize) {
     65     // Initialize the map with zero buckets to allocation.
     66     TheTable = 0;
     67     NumBuckets = 0;
     68     NumItems = 0;
     69     NumTombstones = 0;
     70   }
     71   StringMapImpl(unsigned InitSize, unsigned ItemSize);
     72   void RehashTable();
     73 
     74   /// LookupBucketFor - Look up the bucket that the specified string should end
     75   /// up in.  If it already exists as a key in the map, the Item pointer for the
     76   /// specified bucket will be non-null.  Otherwise, it will be null.  In either
     77   /// case, the FullHashValue field of the bucket will be set to the hash value
     78   /// of the string.
     79   unsigned LookupBucketFor(StringRef Key);
     80 
     81   /// FindKey - Look up the bucket that contains the specified key. If it exists
     82   /// in the map, return the bucket number of the key.  Otherwise return -1.
     83   /// This does not modify the map.
     84   int FindKey(StringRef Key) const;
     85 
     86   /// RemoveKey - Remove the specified StringMapEntry from the table, but do not
     87   /// delete it.  This aborts if the value isn't in the table.
     88   void RemoveKey(StringMapEntryBase *V);
     89 
     90   /// RemoveKey - Remove the StringMapEntry for the specified key from the
     91   /// table, returning it.  If the key is not in the table, this returns null.
     92   StringMapEntryBase *RemoveKey(StringRef Key);
     93 private:
     94   void init(unsigned Size);
     95 public:
     96   static StringMapEntryBase *getTombstoneVal() {
     97     return (StringMapEntryBase*)-1;
     98   }
     99 
    100   unsigned getNumBuckets() const { return NumBuckets; }
    101   unsigned getNumItems() const { return NumItems; }
    102 
    103   bool empty() const { return NumItems == 0; }
    104   unsigned size() const { return NumItems; }
    105 };
    106 
    107 /// StringMapEntry - This is used to represent one value that is inserted into
    108 /// a StringMap.  It contains the Value itself and the key: the string length
    109 /// and data.
    110 template<typename ValueTy>
    111 class StringMapEntry : public StringMapEntryBase {
    112 public:
    113   ValueTy second;
    114 
    115   explicit StringMapEntry(unsigned strLen)
    116     : StringMapEntryBase(strLen), second() {}
    117   StringMapEntry(unsigned strLen, const ValueTy &V)
    118     : StringMapEntryBase(strLen), second(V) {}
    119 
    120   StringRef getKey() const {
    121     return StringRef(getKeyData(), getKeyLength());
    122   }
    123 
    124   const ValueTy &getValue() const { return second; }
    125   ValueTy &getValue() { return second; }
    126 
    127   void setValue(const ValueTy &V) { second = V; }
    128 
    129   /// getKeyData - Return the start of the string data that is the key for this
    130   /// value.  The string data is always stored immediately after the
    131   /// StringMapEntry object.
    132   const char *getKeyData() const {return reinterpret_cast<const char*>(this+1);}
    133 
    134   StringRef first() const { return StringRef(getKeyData(), getKeyLength()); }
    135 
    136   /// Create - Create a StringMapEntry for the specified key and default
    137   /// construct the value.
    138   template<typename AllocatorTy, typename InitType>
    139   static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd,
    140                                 AllocatorTy &Allocator,
    141                                 InitType InitVal) {
    142     unsigned KeyLength = static_cast<unsigned>(KeyEnd-KeyStart);
    143 
    144     // Okay, the item doesn't already exist, and 'Bucket' is the bucket to fill
    145     // in.  Allocate a new item with space for the string at the end and a null
    146     // terminator.
    147 
    148     unsigned AllocSize = static_cast<unsigned>(sizeof(StringMapEntry))+
    149       KeyLength+1;
    150     unsigned Alignment = alignOf<StringMapEntry>();
    151 
    152     StringMapEntry *NewItem =
    153       static_cast<StringMapEntry*>(Allocator.Allocate(AllocSize,Alignment));
    154 
    155     // Default construct the value.
    156     new (NewItem) StringMapEntry(KeyLength);
    157 
    158     // Copy the string information.
    159     char *StrBuffer = const_cast<char*>(NewItem->getKeyData());
    160     memcpy(StrBuffer, KeyStart, KeyLength);
    161     StrBuffer[KeyLength] = 0;  // Null terminate for convenience of clients.
    162 
    163     // Initialize the value if the client wants to.
    164     StringMapEntryInitializer<ValueTy>::Initialize(*NewItem, InitVal);
    165     return NewItem;
    166   }
    167 
    168   template<typename AllocatorTy>
    169   static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd,
    170                                 AllocatorTy &Allocator) {
    171     return Create(KeyStart, KeyEnd, Allocator, 0);
    172   }
    173 
    174   /// Create - Create a StringMapEntry with normal malloc/free.
    175   template<typename InitType>
    176   static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd,
    177                                 InitType InitVal) {
    178     MallocAllocator A;
    179     return Create(KeyStart, KeyEnd, A, InitVal);
    180   }
    181 
    182   static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd) {
    183     return Create(KeyStart, KeyEnd, ValueTy());
    184   }
    185 
    186   /// GetStringMapEntryFromValue - Given a value that is known to be embedded
    187   /// into a StringMapEntry, return the StringMapEntry itself.
    188   static StringMapEntry &GetStringMapEntryFromValue(ValueTy &V) {
    189     StringMapEntry *EPtr = 0;
    190     char *Ptr = reinterpret_cast<char*>(&V) -
    191                   (reinterpret_cast<char*>(&EPtr->second) -
    192                    reinterpret_cast<char*>(EPtr));
    193     return *reinterpret_cast<StringMapEntry*>(Ptr);
    194   }
    195   static const StringMapEntry &GetStringMapEntryFromValue(const ValueTy &V) {
    196     return GetStringMapEntryFromValue(const_cast<ValueTy&>(V));
    197   }
    198 
    199   /// GetStringMapEntryFromKeyData - Given key data that is known to be embedded
    200   /// into a StringMapEntry, return the StringMapEntry itself.
    201   static StringMapEntry &GetStringMapEntryFromKeyData(const char *KeyData) {
    202     char *Ptr = const_cast<char*>(KeyData) - sizeof(StringMapEntry<ValueTy>);
    203     return *reinterpret_cast<StringMapEntry*>(Ptr);
    204   }
    205 
    206   /// Destroy - Destroy this StringMapEntry, releasing memory back to the
    207   /// specified allocator.
    208   template<typename AllocatorTy>
    209   void Destroy(AllocatorTy &Allocator) {
    210     // Free memory referenced by the item.
    211     this->~StringMapEntry();
    212     Allocator.Deallocate(this);
    213   }
    214 
    215   /// Destroy this object, releasing memory back to the malloc allocator.
    216   void Destroy() {
    217     MallocAllocator A;
    218     Destroy(A);
    219   }
    220 };
    221 
    222 
    223 /// StringMap - This is an unconventional map that is specialized for handling
    224 /// keys that are "strings", which are basically ranges of bytes. This does some
    225 /// funky memory allocation and hashing things to make it extremely efficient,
    226 /// storing the string data *after* the value in the map.
    227 template<typename ValueTy, typename AllocatorTy = MallocAllocator>
    228 class StringMap : public StringMapImpl {
    229   AllocatorTy Allocator;
    230 public:
    231   typedef StringMapEntry<ValueTy> MapEntryTy;
    232 
    233   StringMap() : StringMapImpl(static_cast<unsigned>(sizeof(MapEntryTy))) {}
    234   explicit StringMap(unsigned InitialSize)
    235     : StringMapImpl(InitialSize, static_cast<unsigned>(sizeof(MapEntryTy))) {}
    236 
    237   explicit StringMap(AllocatorTy A)
    238     : StringMapImpl(static_cast<unsigned>(sizeof(MapEntryTy))), Allocator(A) {}
    239 
    240   StringMap(unsigned InitialSize, AllocatorTy A)
    241     : StringMapImpl(InitialSize, static_cast<unsigned>(sizeof(MapEntryTy))),
    242       Allocator(A) {}
    243 
    244   StringMap(const StringMap &RHS)
    245     : StringMapImpl(static_cast<unsigned>(sizeof(MapEntryTy))) {
    246     assert(RHS.empty() &&
    247            "Copy ctor from non-empty stringmap not implemented yet!");
    248     (void)RHS;
    249   }
    250   void operator=(const StringMap &RHS) {
    251     assert(RHS.empty() &&
    252            "assignment from non-empty stringmap not implemented yet!");
    253     (void)RHS;
    254     clear();
    255   }
    256 
    257   typedef typename ReferenceAdder<AllocatorTy>::result AllocatorRefTy;
    258   typedef typename ReferenceAdder<const AllocatorTy>::result AllocatorCRefTy;
    259   AllocatorRefTy getAllocator() { return Allocator; }
    260   AllocatorCRefTy getAllocator() const { return Allocator; }
    261 
    262   typedef const char* key_type;
    263   typedef ValueTy mapped_type;
    264   typedef StringMapEntry<ValueTy> value_type;
    265   typedef size_t size_type;
    266 
    267   typedef StringMapConstIterator<ValueTy> const_iterator;
    268   typedef StringMapIterator<ValueTy> iterator;
    269 
    270   iterator begin() {
    271     return iterator(TheTable, NumBuckets == 0);
    272   }
    273   iterator end() {
    274     return iterator(TheTable+NumBuckets, true);
    275   }
    276   const_iterator begin() const {
    277     return const_iterator(TheTable, NumBuckets == 0);
    278   }
    279   const_iterator end() const {
    280     return const_iterator(TheTable+NumBuckets, true);
    281   }
    282 
    283   iterator find(StringRef Key) {
    284     int Bucket = FindKey(Key);
    285     if (Bucket == -1) return end();
    286     return iterator(TheTable+Bucket, true);
    287   }
    288 
    289   const_iterator find(StringRef Key) const {
    290     int Bucket = FindKey(Key);
    291     if (Bucket == -1) return end();
    292     return const_iterator(TheTable+Bucket, true);
    293   }
    294 
    295   /// lookup - Return the entry for the specified key, or a default
    296   /// constructed value if no such entry exists.
    297   ValueTy lookup(StringRef Key) const {
    298     const_iterator it = find(Key);
    299     if (it != end())
    300       return it->second;
    301     return ValueTy();
    302   }
    303 
    304   ValueTy &operator[](StringRef Key) {
    305     return GetOrCreateValue(Key).getValue();
    306   }
    307 
    308   size_type count(StringRef Key) const {
    309     return find(Key) == end() ? 0 : 1;
    310   }
    311 
    312   /// insert - Insert the specified key/value pair into the map.  If the key
    313   /// already exists in the map, return false and ignore the request, otherwise
    314   /// insert it and return true.
    315   bool insert(MapEntryTy *KeyValue) {
    316     unsigned BucketNo = LookupBucketFor(KeyValue->getKey());
    317     StringMapEntryBase *&Bucket = TheTable[BucketNo];
    318     if (Bucket && Bucket != getTombstoneVal())
    319       return false;  // Already exists in map.
    320 
    321     if (Bucket == getTombstoneVal())
    322       --NumTombstones;
    323     Bucket = KeyValue;
    324     ++NumItems;
    325     assert(NumItems + NumTombstones <= NumBuckets);
    326 
    327     RehashTable();
    328     return true;
    329   }
    330 
    331   // clear - Empties out the StringMap
    332   void clear() {
    333     if (empty()) return;
    334 
    335     // Zap all values, resetting the keys back to non-present (not tombstone),
    336     // which is safe because we're removing all elements.
    337     for (unsigned I = 0, E = NumBuckets; I != E; ++I) {
    338       StringMapEntryBase *&Bucket = TheTable[I];
    339       if (Bucket && Bucket != getTombstoneVal()) {
    340         static_cast<MapEntryTy*>(Bucket)->Destroy(Allocator);
    341       }
    342       Bucket = 0;
    343     }
    344 
    345     NumItems = 0;
    346     NumTombstones = 0;
    347   }
    348 
    349   /// GetOrCreateValue - Look up the specified key in the table.  If a value
    350   /// exists, return it.  Otherwise, default construct a value, insert it, and
    351   /// return.
    352   template <typename InitTy>
    353   MapEntryTy &GetOrCreateValue(StringRef Key, InitTy Val) {
    354     unsigned BucketNo = LookupBucketFor(Key);
    355     StringMapEntryBase *&Bucket = TheTable[BucketNo];
    356     if (Bucket && Bucket != getTombstoneVal())
    357       return *static_cast<MapEntryTy*>(Bucket);
    358 
    359     MapEntryTy *NewItem =
    360       MapEntryTy::Create(Key.begin(), Key.end(), Allocator, Val);
    361 
    362     if (Bucket == getTombstoneVal())
    363       --NumTombstones;
    364     ++NumItems;
    365     assert(NumItems + NumTombstones <= NumBuckets);
    366 
    367     // Fill in the bucket for the hash table.  The FullHashValue was already
    368     // filled in by LookupBucketFor.
    369     Bucket = NewItem;
    370 
    371     RehashTable();
    372     return *NewItem;
    373   }
    374 
    375   MapEntryTy &GetOrCreateValue(StringRef Key) {
    376     return GetOrCreateValue(Key, ValueTy());
    377   }
    378 
    379   /// remove - Remove the specified key/value pair from the map, but do not
    380   /// erase it.  This aborts if the key is not in the map.
    381   void remove(MapEntryTy *KeyValue) {
    382     RemoveKey(KeyValue);
    383   }
    384 
    385   void erase(iterator I) {
    386     MapEntryTy &V = *I;
    387     remove(&V);
    388     V.Destroy(Allocator);
    389   }
    390 
    391   bool erase(StringRef Key) {
    392     iterator I = find(Key);
    393     if (I == end()) return false;
    394     erase(I);
    395     return true;
    396   }
    397 
    398   ~StringMap() {
    399     clear();
    400     free(TheTable);
    401   }
    402 };
    403 
    404 
    405 template<typename ValueTy>
    406 class StringMapConstIterator {
    407 protected:
    408   StringMapEntryBase **Ptr;
    409 public:
    410   typedef StringMapEntry<ValueTy> value_type;
    411 
    412   explicit StringMapConstIterator(StringMapEntryBase **Bucket,
    413                                   bool NoAdvance = false)
    414   : Ptr(Bucket) {
    415     if (!NoAdvance) AdvancePastEmptyBuckets();
    416   }
    417 
    418   const value_type &operator*() const {
    419     return *static_cast<StringMapEntry<ValueTy>*>(*Ptr);
    420   }
    421   const value_type *operator->() const {
    422     return static_cast<StringMapEntry<ValueTy>*>(*Ptr);
    423   }
    424 
    425   bool operator==(const StringMapConstIterator &RHS) const {
    426     return Ptr == RHS.Ptr;
    427   }
    428   bool operator!=(const StringMapConstIterator &RHS) const {
    429     return Ptr != RHS.Ptr;
    430   }
    431 
    432   inline StringMapConstIterator& operator++() {   // Preincrement
    433     ++Ptr;
    434     AdvancePastEmptyBuckets();
    435     return *this;
    436   }
    437   StringMapConstIterator operator++(int) {        // Postincrement
    438     StringMapConstIterator tmp = *this; ++*this; return tmp;
    439   }
    440 
    441 private:
    442   void AdvancePastEmptyBuckets() {
    443     while (*Ptr == 0 || *Ptr == StringMapImpl::getTombstoneVal())
    444       ++Ptr;
    445   }
    446 };
    447 
    448 template<typename ValueTy>
    449 class StringMapIterator : public StringMapConstIterator<ValueTy> {
    450 public:
    451   explicit StringMapIterator(StringMapEntryBase **Bucket,
    452                              bool NoAdvance = false)
    453     : StringMapConstIterator<ValueTy>(Bucket, NoAdvance) {
    454   }
    455   StringMapEntry<ValueTy> &operator*() const {
    456     return *static_cast<StringMapEntry<ValueTy>*>(*this->Ptr);
    457   }
    458   StringMapEntry<ValueTy> *operator->() const {
    459     return static_cast<StringMapEntry<ValueTy>*>(*this->Ptr);
    460   }
    461 };
    462 
    463 }
    464 
    465 #endif
    466