StringMap.h revision 9f1f00ab1b347953da0efb0e20352be5e0709276
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/Support/Allocator.h" 18#include <cstring> 19#include <string> 20 21namespace 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. 32template<typename ValueTy> 33class StringMapEntryInitializer { 34public: 35 template <typename InitTy> 36 static void Initialize(StringMapEntry<ValueTy> &T, InitTy InitVal) { 37 } 38}; 39 40 41/// StringMapEntryBase - Shared base class of StringMapEntry instances. 42class StringMapEntryBase { 43 unsigned StrLen; 44public: 45 explicit StringMapEntryBase(unsigned Len) : StrLen(Len) {} 46 47 unsigned getKeyLength() const { return StrLen; } 48}; 49 50/// StringMapImpl - This is the base class of StringMap that is shared among 51/// all of its instantiations. 52class StringMapImpl { 53public: 54 /// ItemBucket - The hash table consists of an array of these. If Item is 55 /// non-null, this is an extant entry, otherwise, it is a hole. 56 struct ItemBucket { 57 /// FullHashValue - This remembers the full hash value of the key for 58 /// easy scanning. 59 unsigned FullHashValue; 60 61 /// Item - This is a pointer to the actual item object. 62 StringMapEntryBase *Item; 63 }; 64 65protected: 66 ItemBucket *TheTable; 67 unsigned NumBuckets; 68 unsigned NumItems; 69 unsigned NumTombstones; 70 unsigned ItemSize; 71protected: 72 explicit StringMapImpl(unsigned itemSize) : ItemSize(itemSize) { 73 // Initialize the map with zero buckets to allocation. 74 TheTable = 0; 75 NumBuckets = 0; 76 NumItems = 0; 77 NumTombstones = 0; 78 } 79 StringMapImpl(unsigned InitSize, unsigned ItemSize); 80 void RehashTable(); 81 82 /// ShouldRehash - Return true if the table should be rehashed after a new 83 /// element was recently inserted. 84 bool ShouldRehash() const { 85 // If the hash table is now more than 3/4 full, or if fewer than 1/8 of 86 // the buckets are empty (meaning that many are filled with tombstones), 87 // grow the table. 88 return NumItems*4 > NumBuckets*3 || 89 NumBuckets-(NumItems+NumTombstones) < NumBuckets/8; 90 } 91 92 /// LookupBucketFor - Look up the bucket that the specified string should end 93 /// up in. If it already exists as a key in the map, the Item pointer for the 94 /// specified bucket will be non-null. Otherwise, it will be null. In either 95 /// case, the FullHashValue field of the bucket will be set to the hash value 96 /// of the string. 97 unsigned LookupBucketFor(const char *KeyStart, const char *KeyEnd); 98 99 /// FindKey - Look up the bucket that contains the specified key. If it exists 100 /// in the map, return the bucket number of the key. Otherwise return -1. 101 /// This does not modify the map. 102 int FindKey(const char *KeyStart, const char *KeyEnd) const; 103 104 /// RemoveKey - Remove the specified StringMapEntry from the table, but do not 105 /// delete it. This aborts if the value isn't in the table. 106 void RemoveKey(StringMapEntryBase *V); 107 108 /// RemoveKey - Remove the StringMapEntry for the specified key from the 109 /// table, returning it. If the key is not in the table, this returns null. 110 StringMapEntryBase *RemoveKey(const char *KeyStart, const char *KeyEnd); 111private: 112 void init(unsigned Size); 113public: 114 static StringMapEntryBase *getTombstoneVal() { 115 return (StringMapEntryBase*)-1; 116 } 117 118 unsigned getNumBuckets() const { return NumBuckets; } 119 unsigned getNumItems() const { return NumItems; } 120 121 bool empty() const { return NumItems == 0; } 122 unsigned size() const { return NumItems; } 123}; 124 125/// StringMapEntry - This is used to represent one value that is inserted into 126/// a StringMap. It contains the Value itself and the key: the string length 127/// and data. 128template<typename ValueTy> 129class StringMapEntry : public StringMapEntryBase { 130public: 131 ValueTy second; 132 133 explicit StringMapEntry(unsigned strLen) 134 : StringMapEntryBase(strLen), second() {} 135 StringMapEntry(unsigned strLen, const ValueTy &V) 136 : StringMapEntryBase(strLen), second(V) {} 137 138 const ValueTy &getValue() const { return second; } 139 ValueTy &getValue() { return second; } 140 141 void setValue(const ValueTy &V) { second = V; } 142 143 /// getKeyData - Return the start of the string data that is the key for this 144 /// value. The string data is always stored immediately after the 145 /// StringMapEntry object. 146 const char *getKeyData() const {return reinterpret_cast<const char*>(this+1);} 147 148 const char *first() const { return getKeyData(); } 149 150 /// Create - Create a StringMapEntry for the specified key and default 151 /// construct the value. 152 template<typename AllocatorTy, typename InitType> 153 static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd, 154 AllocatorTy &Allocator, 155 InitType InitVal) { 156 unsigned KeyLength = static_cast<unsigned>(KeyEnd-KeyStart); 157 158 // Okay, the item doesn't already exist, and 'Bucket' is the bucket to fill 159 // in. Allocate a new item with space for the string at the end and a null 160 // terminator. 161 162 unsigned AllocSize = static_cast<unsigned>(sizeof(StringMapEntry))+ 163 KeyLength+1; 164 unsigned Alignment = alignof<StringMapEntry>(); 165 166 StringMapEntry *NewItem = 167 static_cast<StringMapEntry*>(Allocator.Allocate(AllocSize,Alignment)); 168 169 // Default construct the value. 170 new (NewItem) StringMapEntry(KeyLength); 171 172 // Copy the string information. 173 char *StrBuffer = const_cast<char*>(NewItem->getKeyData()); 174 memcpy(StrBuffer, KeyStart, KeyLength); 175 StrBuffer[KeyLength] = 0; // Null terminate for convenience of clients. 176 177 // Initialize the value if the client wants to. 178 StringMapEntryInitializer<ValueTy>::Initialize(*NewItem, InitVal); 179 return NewItem; 180 } 181 182 template<typename AllocatorTy> 183 static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd, 184 AllocatorTy &Allocator) { 185 return Create(KeyStart, KeyEnd, Allocator, (void*)0); 186 } 187 188 189 /// Create - Create a StringMapEntry with normal malloc/free. 190 template<typename InitType> 191 static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd, 192 InitType InitVal) { 193 MallocAllocator A; 194 return Create(KeyStart, KeyEnd, A, InitVal); 195 } 196 197 static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd) { 198 return Create(KeyStart, KeyEnd, (void*)0); 199 } 200 201 /// GetStringMapEntryFromValue - Given a value that is known to be embedded 202 /// into a StringMapEntry, return the StringMapEntry itself. 203 static StringMapEntry &GetStringMapEntryFromValue(ValueTy &V) { 204 StringMapEntry *EPtr = 0; 205 char *Ptr = reinterpret_cast<char*>(&V) - 206 (reinterpret_cast<char*>(&EPtr->second) - 207 reinterpret_cast<char*>(EPtr)); 208 return *reinterpret_cast<StringMapEntry*>(Ptr); 209 } 210 static const StringMapEntry &GetStringMapEntryFromValue(const ValueTy &V) { 211 return GetStringMapEntryFromValue(const_cast<ValueTy&>(V)); 212 } 213 214 /// Destroy - Destroy this StringMapEntry, releasing memory back to the 215 /// specified allocator. 216 template<typename AllocatorTy> 217 void Destroy(AllocatorTy &Allocator) { 218 // Free memory referenced by the item. 219 this->~StringMapEntry(); 220 Allocator.Deallocate(this); 221 } 222 223 /// Destroy this object, releasing memory back to the malloc allocator. 224 void Destroy() { 225 MallocAllocator A; 226 Destroy(A); 227 } 228}; 229 230 231/// StringMap - This is an unconventional map that is specialized for handling 232/// keys that are "strings", which are basically ranges of bytes. This does some 233/// funky memory allocation and hashing things to make it extremely efficient, 234/// storing the string data *after* the value in the map. 235template<typename ValueTy, typename AllocatorTy = MallocAllocator> 236class StringMap : public StringMapImpl { 237 AllocatorTy Allocator; 238 typedef StringMapEntry<ValueTy> MapEntryTy; 239public: 240 StringMap() : StringMapImpl(static_cast<unsigned>(sizeof(MapEntryTy))) {} 241 explicit StringMap(unsigned InitialSize) 242 : StringMapImpl(InitialSize, static_cast<unsigned>(sizeof(MapEntryTy))) {} 243 244 AllocatorTy &getAllocator() { return Allocator; } 245 const AllocatorTy &getAllocator() const { return Allocator; } 246 247 typedef const char* key_type; 248 typedef ValueTy mapped_type; 249 typedef StringMapEntry<ValueTy> value_type; 250 typedef size_t size_type; 251 252 typedef StringMapConstIterator<ValueTy> const_iterator; 253 typedef StringMapIterator<ValueTy> iterator; 254 255 iterator begin() { 256 return iterator(TheTable, NumBuckets == 0); 257 } 258 iterator end() { 259 return iterator(TheTable+NumBuckets, true); 260 } 261 const_iterator begin() const { 262 return const_iterator(TheTable, NumBuckets == 0); 263 } 264 const_iterator end() const { 265 return const_iterator(TheTable+NumBuckets, true); 266 } 267 268 iterator find(const char *KeyStart, const char *KeyEnd) { 269 int Bucket = FindKey(KeyStart, KeyEnd); 270 if (Bucket == -1) return end(); 271 return iterator(TheTable+Bucket); 272 } 273 iterator find(const char *Key) { 274 return find(Key, Key + strlen(Key)); 275 } 276 iterator find(const std::string &Key) { 277 const char* key_start = (Key.empty() ? NULL : &Key[0]); 278 return find(key_start, key_start + Key.size()); 279 } 280 281 const_iterator find(const char *KeyStart, const char *KeyEnd) const { 282 int Bucket = FindKey(KeyStart, KeyEnd); 283 if (Bucket == -1) return end(); 284 return const_iterator(TheTable+Bucket); 285 } 286 const_iterator find(const char *Key) const { 287 return find(Key, Key + strlen(Key)); 288 } 289 const_iterator find(const std::string &Key) const { 290 const char* key_start = (Key.empty() ? NULL : &Key[0]); 291 return find(key_start, key_start + Key.size()); 292 } 293 294 ValueTy& operator[](const char *Key) { 295 value_type& entry = GetOrCreateValue(Key, Key + strlen(Key)); 296 return entry.getValue(); 297 } 298 ValueTy& operator[](const std::string &Key) { 299 const char* key_start = (Key.empty() ? NULL : &Key[0]); 300 value_type& entry = GetOrCreateValue(key_start, key_start + Key.size()); 301 return entry.getValue(); 302 } 303 304 size_type count(const char *KeyStart, const char *KeyEnd) const { 305 return find(KeyStart, KeyEnd) == end() ? 0 : 1; 306 } 307 size_type count(const char *Key) const { 308 return count(Key, Key + strlen(Key)); 309 } 310 size_type count(const std::string &Key) const { 311 const char* key_start = (Key.empty() ? NULL : &Key[0]); 312 return count(key_start, key_start + Key.size()); 313 } 314 315 /// insert - Insert the specified key/value pair into the map. If the key 316 /// already exists in the map, return false and ignore the request, otherwise 317 /// insert it and return true. 318 bool insert(MapEntryTy *KeyValue) { 319 unsigned BucketNo = 320 LookupBucketFor(KeyValue->getKeyData(), 321 KeyValue->getKeyData()+KeyValue->getKeyLength()); 322 ItemBucket &Bucket = TheTable[BucketNo]; 323 if (Bucket.Item && Bucket.Item != getTombstoneVal()) 324 return false; // Already exists in map. 325 326 if (Bucket.Item == getTombstoneVal()) 327 --NumTombstones; 328 Bucket.Item = KeyValue; 329 ++NumItems; 330 331 if (ShouldRehash()) 332 RehashTable(); 333 return true; 334 } 335 336 // clear - Empties out the StringMap 337 void clear() { 338 if (empty()) return; 339 340 // Zap all values, resetting the keys back to non-present (not tombstone), 341 // which is safe because we're removing all elements. 342 for (ItemBucket *I = TheTable, *E = TheTable+NumBuckets; I != E; ++I) { 343 if (I->Item && I->Item != getTombstoneVal()) { 344 static_cast<MapEntryTy*>(I->Item)->Destroy(Allocator); 345 I->Item = 0; 346 } 347 } 348 } 349 350 /// GetOrCreateValue - Look up the specified key in the table. If a value 351 /// exists, return it. Otherwise, default construct a value, insert it, and 352 /// return. 353 template <typename InitTy> 354 StringMapEntry<ValueTy> &GetOrCreateValue(const char *KeyStart, 355 const char *KeyEnd, 356 InitTy Val) { 357 unsigned BucketNo = LookupBucketFor(KeyStart, KeyEnd); 358 ItemBucket &Bucket = TheTable[BucketNo]; 359 if (Bucket.Item && Bucket.Item != getTombstoneVal()) 360 return *static_cast<MapEntryTy*>(Bucket.Item); 361 362 MapEntryTy *NewItem = MapEntryTy::Create(KeyStart, KeyEnd, Allocator, Val); 363 364 if (Bucket.Item == getTombstoneVal()) 365 --NumTombstones; 366 ++NumItems; 367 368 // Fill in the bucket for the hash table. The FullHashValue was already 369 // filled in by LookupBucketFor. 370 Bucket.Item = NewItem; 371 372 if (ShouldRehash()) 373 RehashTable(); 374 return *NewItem; 375 } 376 377 StringMapEntry<ValueTy> &GetOrCreateValue(const char *KeyStart, 378 const char *KeyEnd) { 379 return GetOrCreateValue(KeyStart, KeyEnd, (void*)0); 380 } 381 382 /// remove - Remove the specified key/value pair from the map, but do not 383 /// erase it. This aborts if the key is not in the map. 384 void remove(MapEntryTy *KeyValue) { 385 RemoveKey(KeyValue); 386 } 387 388 void erase(iterator I) { 389 MapEntryTy &V = *I; 390 remove(&V); 391 V.Destroy(Allocator); 392 } 393 394 bool erase(const char *Key) { 395 iterator I = find(Key); 396 if (I == end()) return false; 397 erase(I); 398 return true; 399 } 400 401 bool erase(const std::string &Key) { 402 iterator I = find(Key); 403 if (I == end()) return false; 404 erase(I); 405 return true; 406 } 407 408 ~StringMap() { 409 clear(); 410 free(TheTable); 411 } 412private: 413 StringMap(const StringMap &); // FIXME: Implement. 414 void operator=(const StringMap &); // FIXME: Implement. 415}; 416 417 418template<typename ValueTy> 419class StringMapConstIterator { 420protected: 421 StringMapImpl::ItemBucket *Ptr; 422public: 423 typedef StringMapEntry<ValueTy> value_type; 424 425 explicit StringMapConstIterator(StringMapImpl::ItemBucket *Bucket, 426 bool NoAdvance = false) 427 : Ptr(Bucket) { 428 if (!NoAdvance) AdvancePastEmptyBuckets(); 429 } 430 431 const value_type &operator*() const { 432 return *static_cast<StringMapEntry<ValueTy>*>(Ptr->Item); 433 } 434 const value_type *operator->() const { 435 return static_cast<StringMapEntry<ValueTy>*>(Ptr->Item); 436 } 437 438 bool operator==(const StringMapConstIterator &RHS) const { 439 return Ptr == RHS.Ptr; 440 } 441 bool operator!=(const StringMapConstIterator &RHS) const { 442 return Ptr != RHS.Ptr; 443 } 444 445 inline StringMapConstIterator& operator++() { // Preincrement 446 ++Ptr; 447 AdvancePastEmptyBuckets(); 448 return *this; 449 } 450 StringMapConstIterator operator++(int) { // Postincrement 451 StringMapConstIterator tmp = *this; ++*this; return tmp; 452 } 453 454private: 455 void AdvancePastEmptyBuckets() { 456 while (Ptr->Item == 0 || Ptr->Item == StringMapImpl::getTombstoneVal()) 457 ++Ptr; 458 } 459}; 460 461template<typename ValueTy> 462class StringMapIterator : public StringMapConstIterator<ValueTy> { 463public: 464 explicit StringMapIterator(StringMapImpl::ItemBucket *Bucket, 465 bool NoAdvance = false) 466 : StringMapConstIterator<ValueTy>(Bucket, NoAdvance) { 467 } 468 StringMapEntry<ValueTy> &operator*() const { 469 return *static_cast<StringMapEntry<ValueTy>*>(this->Ptr->Item); 470 } 471 StringMapEntry<ValueTy> *operator->() const { 472 return static_cast<StringMapEntry<ValueTy>*>(this->Ptr->Item); 473 } 474}; 475 476} 477 478#endif 479