StringMap.h revision ddcdcc88631c6bd4ad43d9198b98bc9a829be036
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 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 T.second = InitVal; 38 } 39}; 40 41 42/// StringMapEntryBase - Shared base class of StringMapEntry instances. 43class StringMapEntryBase { 44 unsigned StrLen; 45public: 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. 53class StringMapImpl { 54public: 55 /// ItemBucket - The hash table consists of an array of these. If Item is 56 /// non-null, this is an extant entry, otherwise, it is a hole. 57 struct ItemBucket { 58 /// FullHashValue - This remembers the full hash value of the key for 59 /// easy scanning. 60 unsigned FullHashValue; 61 62 /// Item - This is a pointer to the actual item object. 63 StringMapEntryBase *Item; 64 }; 65 66protected: 67 ItemBucket *TheTable; 68 unsigned NumBuckets; 69 unsigned NumItems; 70 unsigned NumTombstones; 71 unsigned ItemSize; 72protected: 73 explicit StringMapImpl(unsigned itemSize) : ItemSize(itemSize) { 74 // Initialize the map with zero buckets to allocation. 75 TheTable = 0; 76 NumBuckets = 0; 77 NumItems = 0; 78 NumTombstones = 0; 79 } 80 StringMapImpl(unsigned InitSize, unsigned ItemSize); 81 void RehashTable(); 82 83 /// LookupBucketFor - Look up the bucket that the specified string should end 84 /// up in. If it already exists as a key in the map, the Item pointer for the 85 /// specified bucket will be non-null. Otherwise, it will be null. In either 86 /// case, the FullHashValue field of the bucket will be set to the hash value 87 /// of the string. 88 unsigned LookupBucketFor(StringRef Key); 89 90 /// FindKey - Look up the bucket that contains the specified key. If it exists 91 /// in the map, return the bucket number of the key. Otherwise return -1. 92 /// This does not modify the map. 93 int FindKey(StringRef Key) const; 94 95 /// RemoveKey - Remove the specified StringMapEntry from the table, but do not 96 /// delete it. This aborts if the value isn't in the table. 97 void RemoveKey(StringMapEntryBase *V); 98 99 /// RemoveKey - Remove the StringMapEntry for the specified key from the 100 /// table, returning it. If the key is not in the table, this returns null. 101 StringMapEntryBase *RemoveKey(StringRef Key); 102private: 103 void init(unsigned Size); 104public: 105 static StringMapEntryBase *getTombstoneVal() { 106 return (StringMapEntryBase*)-1; 107 } 108 109 unsigned getNumBuckets() const { return NumBuckets; } 110 unsigned getNumItems() const { return NumItems; } 111 112 bool empty() const { return NumItems == 0; } 113 unsigned size() const { return NumItems; } 114}; 115 116/// StringMapEntry - This is used to represent one value that is inserted into 117/// a StringMap. It contains the Value itself and the key: the string length 118/// and data. 119template<typename ValueTy> 120class StringMapEntry : public StringMapEntryBase { 121public: 122 ValueTy second; 123 124 explicit StringMapEntry(unsigned strLen) 125 : StringMapEntryBase(strLen), second() {} 126 StringMapEntry(unsigned strLen, const ValueTy &V) 127 : StringMapEntryBase(strLen), second(V) {} 128 129 StringRef getKey() const { 130 return StringRef(getKeyData(), getKeyLength()); 131 } 132 133 const ValueTy &getValue() const { return second; } 134 ValueTy &getValue() { return second; } 135 136 void setValue(const ValueTy &V) { second = V; } 137 138 /// getKeyData - Return the start of the string data that is the key for this 139 /// value. The string data is always stored immediately after the 140 /// StringMapEntry object. 141 const char *getKeyData() const {return reinterpret_cast<const char*>(this+1);} 142 143 const char *first() const { return getKeyData(); } 144 145 /// Create - Create a StringMapEntry for the specified key and default 146 /// construct the value. 147 template<typename AllocatorTy, typename InitType> 148 static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd, 149 AllocatorTy &Allocator, 150 InitType InitVal) { 151 unsigned KeyLength = static_cast<unsigned>(KeyEnd-KeyStart); 152 153 // Okay, the item doesn't already exist, and 'Bucket' is the bucket to fill 154 // in. Allocate a new item with space for the string at the end and a null 155 // terminator. 156 157 unsigned AllocSize = static_cast<unsigned>(sizeof(StringMapEntry))+ 158 KeyLength+1; 159 unsigned Alignment = alignOf<StringMapEntry>(); 160 161 StringMapEntry *NewItem = 162 static_cast<StringMapEntry*>(Allocator.Allocate(AllocSize,Alignment)); 163 164 // Default construct the value. 165 new (NewItem) StringMapEntry(KeyLength); 166 167 // Copy the string information. 168 char *StrBuffer = const_cast<char*>(NewItem->getKeyData()); 169 memcpy(StrBuffer, KeyStart, KeyLength); 170 StrBuffer[KeyLength] = 0; // Null terminate for convenience of clients. 171 172 // Initialize the value if the client wants to. 173 StringMapEntryInitializer<ValueTy>::Initialize(*NewItem, InitVal); 174 return NewItem; 175 } 176 177 template<typename AllocatorTy> 178 static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd, 179 AllocatorTy &Allocator) { 180 return Create(KeyStart, KeyEnd, Allocator, 0); 181 } 182 183 184 /// Create - Create a StringMapEntry with normal malloc/free. 185 template<typename InitType> 186 static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd, 187 InitType InitVal) { 188 MallocAllocator A; 189 return Create(KeyStart, KeyEnd, A, InitVal); 190 } 191 192 static StringMapEntry *Create(const char *KeyStart, const char *KeyEnd) { 193 return Create(KeyStart, KeyEnd, ValueTy()); 194 } 195 196 /// GetStringMapEntryFromValue - Given a value that is known to be embedded 197 /// into a StringMapEntry, return the StringMapEntry itself. 198 static StringMapEntry &GetStringMapEntryFromValue(ValueTy &V) { 199 StringMapEntry *EPtr = 0; 200 char *Ptr = reinterpret_cast<char*>(&V) - 201 (reinterpret_cast<char*>(&EPtr->second) - 202 reinterpret_cast<char*>(EPtr)); 203 return *reinterpret_cast<StringMapEntry*>(Ptr); 204 } 205 static const StringMapEntry &GetStringMapEntryFromValue(const ValueTy &V) { 206 return GetStringMapEntryFromValue(const_cast<ValueTy&>(V)); 207 } 208 209 /// GetStringMapEntryFromKeyData - Given key data that is known to be embedded 210 /// into a StringMapEntry, return the StringMapEntry itself. 211 static StringMapEntry &GetStringMapEntryFromKeyData(const char *KeyData) { 212 char *Ptr = const_cast<char*>(KeyData) - sizeof(StringMapEntry<ValueTy>); 213 return *reinterpret_cast<StringMapEntry*>(Ptr); 214 } 215 216 217 /// Destroy - Destroy this StringMapEntry, releasing memory back to the 218 /// specified allocator. 219 template<typename AllocatorTy> 220 void Destroy(AllocatorTy &Allocator) { 221 // Free memory referenced by the item. 222 this->~StringMapEntry(); 223 Allocator.Deallocate(this); 224 } 225 226 /// Destroy this object, releasing memory back to the malloc allocator. 227 void Destroy() { 228 MallocAllocator A; 229 Destroy(A); 230 } 231}; 232 233 234/// StringMap - This is an unconventional map that is specialized for handling 235/// keys that are "strings", which are basically ranges of bytes. This does some 236/// funky memory allocation and hashing things to make it extremely efficient, 237/// storing the string data *after* the value in the map. 238template<typename ValueTy, typename AllocatorTy = MallocAllocator> 239class StringMap : public StringMapImpl { 240 AllocatorTy Allocator; 241 typedef StringMapEntry<ValueTy> MapEntryTy; 242public: 243 StringMap() : StringMapImpl(static_cast<unsigned>(sizeof(MapEntryTy))) {} 244 explicit StringMap(unsigned InitialSize) 245 : StringMapImpl(InitialSize, static_cast<unsigned>(sizeof(MapEntryTy))) {} 246 247 explicit StringMap(AllocatorTy A) 248 : StringMapImpl(static_cast<unsigned>(sizeof(MapEntryTy))), Allocator(A) {} 249 250 explicit StringMap(const StringMap &RHS) 251 : StringMapImpl(static_cast<unsigned>(sizeof(MapEntryTy))) { 252 assert(RHS.empty() && 253 "Copy ctor from non-empty stringmap not implemented yet!"); 254 (void)RHS; 255 } 256 void operator=(const StringMap &RHS) { 257 assert(RHS.empty() && 258 "assignment from non-empty stringmap not implemented yet!"); 259 (void)RHS; 260 clear(); 261 } 262 263 typedef typename ReferenceAdder<AllocatorTy>::result AllocatorRefTy; 264 typedef typename ReferenceAdder<const AllocatorTy>::result AllocatorCRefTy; 265 AllocatorRefTy getAllocator() { return Allocator; } 266 AllocatorCRefTy getAllocator() const { return Allocator; } 267 268 typedef const char* key_type; 269 typedef ValueTy mapped_type; 270 typedef StringMapEntry<ValueTy> value_type; 271 typedef size_t size_type; 272 273 typedef StringMapConstIterator<ValueTy> const_iterator; 274 typedef StringMapIterator<ValueTy> iterator; 275 276 iterator begin() { 277 return iterator(TheTable, NumBuckets == 0); 278 } 279 iterator end() { 280 return iterator(TheTable+NumBuckets, true); 281 } 282 const_iterator begin() const { 283 return const_iterator(TheTable, NumBuckets == 0); 284 } 285 const_iterator end() const { 286 return const_iterator(TheTable+NumBuckets, true); 287 } 288 289 iterator find(StringRef Key) { 290 int Bucket = FindKey(Key); 291 if (Bucket == -1) return end(); 292 return iterator(TheTable+Bucket); 293 } 294 295 const_iterator find(StringRef Key) const { 296 int Bucket = FindKey(Key); 297 if (Bucket == -1) return end(); 298 return const_iterator(TheTable+Bucket); 299 } 300 301 /// lookup - Return the entry for the specified key, or a default 302 /// constructed value if no such entry exists. 303 ValueTy lookup(StringRef Key) const { 304 const_iterator it = find(Key); 305 if (it != end()) 306 return it->second; 307 return ValueTy(); 308 } 309 310 ValueTy& operator[](StringRef Key) { 311 return GetOrCreateValue(Key).getValue(); 312 } 313 314 size_type count(StringRef Key) const { 315 return find(Key) == end() ? 0 : 1; 316 } 317 318 /// insert - Insert the specified key/value pair into the map. If the key 319 /// already exists in the map, return false and ignore the request, otherwise 320 /// insert it and return true. 321 bool insert(MapEntryTy *KeyValue) { 322 unsigned BucketNo = LookupBucketFor(KeyValue->getKey()); 323 ItemBucket &Bucket = TheTable[BucketNo]; 324 if (Bucket.Item && Bucket.Item != getTombstoneVal()) 325 return false; // Already exists in map. 326 327 if (Bucket.Item == getTombstoneVal()) 328 --NumTombstones; 329 Bucket.Item = KeyValue; 330 ++NumItems; 331 assert(NumItems + NumTombstones <= NumBuckets); 332 333 RehashTable(); 334 return true; 335 } 336 337 // clear - Empties out the StringMap 338 void clear() { 339 if (empty()) return; 340 341 // Zap all values, resetting the keys back to non-present (not tombstone), 342 // which is safe because we're removing all elements. 343 for (ItemBucket *I = TheTable, *E = TheTable+NumBuckets; I != E; ++I) { 344 if (I->Item && I->Item != getTombstoneVal()) { 345 static_cast<MapEntryTy*>(I->Item)->Destroy(Allocator); 346 I->Item = 0; 347 } 348 } 349 350 NumItems = 0; 351 NumTombstones = 0; 352 } 353 354 /// GetOrCreateValue - Look up the specified key in the table. If a value 355 /// exists, return it. Otherwise, default construct a value, insert it, and 356 /// return. 357 template <typename InitTy> 358 StringMapEntry<ValueTy> &GetOrCreateValue(StringRef Key, 359 InitTy Val) { 360 unsigned BucketNo = LookupBucketFor(Key); 361 ItemBucket &Bucket = TheTable[BucketNo]; 362 if (Bucket.Item && Bucket.Item != getTombstoneVal()) 363 return *static_cast<MapEntryTy*>(Bucket.Item); 364 365 MapEntryTy *NewItem = 366 MapEntryTy::Create(Key.begin(), Key.end(), Allocator, Val); 367 368 if (Bucket.Item == getTombstoneVal()) 369 --NumTombstones; 370 ++NumItems; 371 assert(NumItems + NumTombstones <= NumBuckets); 372 373 // Fill in the bucket for the hash table. The FullHashValue was already 374 // filled in by LookupBucketFor. 375 Bucket.Item = NewItem; 376 377 RehashTable(); 378 return *NewItem; 379 } 380 381 StringMapEntry<ValueTy> &GetOrCreateValue(StringRef Key) { 382 return GetOrCreateValue(Key, ValueTy()); 383 } 384 385 template <typename InitTy> 386 StringMapEntry<ValueTy> &GetOrCreateValue(const char *KeyStart, 387 const char *KeyEnd, 388 InitTy Val) { 389 return GetOrCreateValue(StringRef(KeyStart, KeyEnd - KeyStart), Val); 390 } 391 392 StringMapEntry<ValueTy> &GetOrCreateValue(const char *KeyStart, 393 const char *KeyEnd) { 394 return GetOrCreateValue(StringRef(KeyStart, KeyEnd - KeyStart)); 395 } 396 397 /// remove - Remove the specified key/value pair from the map, but do not 398 /// erase it. This aborts if the key is not in the map. 399 void remove(MapEntryTy *KeyValue) { 400 RemoveKey(KeyValue); 401 } 402 403 void erase(iterator I) { 404 MapEntryTy &V = *I; 405 remove(&V); 406 V.Destroy(Allocator); 407 } 408 409 bool erase(StringRef Key) { 410 iterator I = find(Key); 411 if (I == end()) return false; 412 erase(I); 413 return true; 414 } 415 416 ~StringMap() { 417 clear(); 418 free(TheTable); 419 } 420}; 421 422 423template<typename ValueTy> 424class StringMapConstIterator { 425protected: 426 StringMapImpl::ItemBucket *Ptr; 427public: 428 typedef StringMapEntry<ValueTy> value_type; 429 430 explicit StringMapConstIterator(StringMapImpl::ItemBucket *Bucket, 431 bool NoAdvance = false) 432 : Ptr(Bucket) { 433 if (!NoAdvance) AdvancePastEmptyBuckets(); 434 } 435 436 const value_type &operator*() const { 437 return *static_cast<StringMapEntry<ValueTy>*>(Ptr->Item); 438 } 439 const value_type *operator->() const { 440 return static_cast<StringMapEntry<ValueTy>*>(Ptr->Item); 441 } 442 443 bool operator==(const StringMapConstIterator &RHS) const { 444 return Ptr == RHS.Ptr; 445 } 446 bool operator!=(const StringMapConstIterator &RHS) const { 447 return Ptr != RHS.Ptr; 448 } 449 450 inline StringMapConstIterator& operator++() { // Preincrement 451 ++Ptr; 452 AdvancePastEmptyBuckets(); 453 return *this; 454 } 455 StringMapConstIterator operator++(int) { // Postincrement 456 StringMapConstIterator tmp = *this; ++*this; return tmp; 457 } 458 459private: 460 void AdvancePastEmptyBuckets() { 461 while (Ptr->Item == 0 || Ptr->Item == StringMapImpl::getTombstoneVal()) 462 ++Ptr; 463 } 464}; 465 466template<typename ValueTy> 467class StringMapIterator : public StringMapConstIterator<ValueTy> { 468public: 469 explicit StringMapIterator(StringMapImpl::ItemBucket *Bucket, 470 bool NoAdvance = false) 471 : StringMapConstIterator<ValueTy>(Bucket, NoAdvance) { 472 } 473 StringMapEntry<ValueTy> &operator*() const { 474 return *static_cast<StringMapEntry<ValueTy>*>(this->Ptr->Item); 475 } 476 StringMapEntry<ValueTy> *operator->() const { 477 return static_cast<StringMapEntry<ValueTy>*>(this->Ptr->Item); 478 } 479}; 480 481} 482 483#endif 484