1 | // Licensed to the .NET Foundation under one or more agreements. |
2 | // The .NET Foundation licenses this file to you under the MIT license. |
3 | // See the LICENSE file in the project root for more information. |
4 | |
5 | #ifndef _SIMPLERHASHTABLE_H_ |
6 | #define _SIMPLERHASHTABLE_H_ |
7 | |
8 | #include "iallocator.h" |
9 | |
10 | // SimplerHashTable implements a mapping from a Key type to a Value type, |
11 | // via a hash table. |
12 | |
13 | // Synchronization is the responsibility of the caller: if a |
14 | // SimplerHash is used in a multithreaded context, the table should be |
15 | // associated with a lock. |
16 | |
17 | // SimplerHashTable actually takes four template arguments: Key, |
18 | // KeyFuncs, Value, and Behavior. We don't assume that Key has hash or equality |
19 | // functions specific names; rather, we assume that KeyFuncs has |
20 | // statics methods |
21 | // int GetHashCode(Key) |
22 | // and |
23 | // bool Equals(Key, Key) |
24 | // and use those. An |
25 | // instantiator can thus make a small "adaptor class" to invoke |
26 | // existing instance method hash and/or equality functions. If the |
27 | // implementor of a candidate Key class K understands this convention, |
28 | // these static methods can be implemented by K, so that K can be used |
29 | // as the actual arguments for the both Key and KeyTrait classes. |
30 | // |
31 | // The "Behavior" argument provides the following static members: |
32 | // |
33 | // s_growth_factor_numerator |
34 | // s_growth_factor_denominator Factor to grow allocation (numerator/denominator). |
35 | // Typically inherited from default traits (3/2) |
36 | // |
37 | // s_density_factor_numerator |
38 | // s_density_factor_denominator Maxium occupied density of table before growth |
39 | // occurs (num/denom). Typically inherited (3/4). |
40 | // |
41 | // s_minimum_allocation Minimum table allocation count (size on first growth.) It is |
42 | // probably preferable to call Reallocate on initialization rather |
43 | // than override his from the default traits. (7) |
44 | // |
45 | // NoMemory() Called when the hash table is unable to grow due to potential |
46 | // overflow or the lack of a sufficiently large prime. |
47 | |
48 | void DECLSPEC_NORETURN ThrowOutOfMemory(); |
49 | |
50 | class DefaultSimplerHashBehavior |
51 | { |
52 | public: |
53 | static const unsigned s_growth_factor_numerator = 3; |
54 | static const unsigned s_growth_factor_denominator = 2; |
55 | |
56 | static const unsigned s_density_factor_numerator = 3; |
57 | static const unsigned s_density_factor_denominator = 4; |
58 | |
59 | static const unsigned s_minimum_allocation = 7; |
60 | |
61 | inline static void DECLSPEC_NORETURN NoMemory() |
62 | { |
63 | ::ThrowOutOfMemory(); |
64 | } |
65 | }; |
66 | |
67 | // Stores info about primes, including the magic number and shift amount needed |
68 | // to implement a divide without using the divide instruction |
69 | class PrimeInfo |
70 | { |
71 | public: |
72 | PrimeInfo() : prime(0), magic(0), shift(0) {} |
73 | PrimeInfo(unsigned p, unsigned m, unsigned s) : prime(p), magic(m), shift(s) {} |
74 | unsigned prime; |
75 | unsigned magic; |
76 | unsigned shift; |
77 | }; |
78 | |
79 | |
80 | // Hash table class definition |
81 | |
82 | template <typename Key, typename KeyFuncs, typename Value, typename Behavior> |
83 | class SimplerHashTable |
84 | { |
85 | public: |
86 | // Forward declaration. |
87 | class KeyIterator; |
88 | |
89 | // Constructor/destructor. SHash tables always start out empty, with no |
90 | // allocation overhead. Call Reallocate to prime with an initial size if |
91 | // desired. Pass NULL as the IAllocator* if you want to use DefaultAllocator |
92 | // (basically, operator new/delete). |
93 | |
94 | SimplerHashTable(IAllocator* alloc); |
95 | ~SimplerHashTable(); |
96 | |
97 | // operators new/delete when an IAllocator is to be used. |
98 | void * operator new(size_t sz, IAllocator * alloc); |
99 | void * operator new[](size_t sz, IAllocator * alloc); |
100 | void operator delete(void * p, IAllocator * alloc); |
101 | void operator delete[](void * p, IAllocator * alloc); |
102 | |
103 | // If the table contains a mapping for "key", returns "true" and |
104 | // sets "*pVal" to the value to which "key" maps. Otherwise, |
105 | // returns false, and does not modify "*pVal". |
106 | bool Lookup(Key k, Value* pVal = NULL) const; |
107 | |
108 | Value *LookupPointer(Key k) const; |
109 | |
110 | // Causes the table to map "key" to "val". Returns "true" if |
111 | // "key" had already been mapped by the table, "false" otherwise. |
112 | bool Set(Key k, Value val); |
113 | |
114 | // Ensures that "key" is not mapped to a value by the "table." |
115 | // Returns "true" iff it had been mapped. |
116 | bool Remove(Key k); |
117 | |
118 | // Remove all mappings in the table. |
119 | void RemoveAll(); |
120 | |
121 | // Begin and End pointers for iteration over entire table. |
122 | KeyIterator Begin() const; |
123 | KeyIterator End() const; |
124 | |
125 | // Return the number of elements currently stored in the table |
126 | unsigned GetCount() const; |
127 | |
128 | private: |
129 | // Forward declaration of the linked-list node class. |
130 | struct Node; |
131 | |
132 | unsigned GetIndexForKey(Key k) const; |
133 | |
134 | // If the table has a mapping for "k", return the node containing |
135 | // that mapping, else "NULL". |
136 | Node* FindNode(Key k) const; |
137 | |
138 | // Resizes a hash table for growth. The new size is computed based |
139 | // on the current population, growth factor, and maximum density factor. |
140 | void Grow(); |
141 | |
142 | // See if it is OK to grow the hash table by one element. If not, reallocate |
143 | // the hash table. |
144 | void CheckGrowth(); |
145 | |
146 | public: |
147 | // Reallocates a hash table to a specific size. The size must be big enough |
148 | // to hold all elements in the table appropriately. |
149 | // |
150 | // Note that the actual table size must always be a prime number; the number |
151 | // passed in will be upward adjusted if necessary. |
152 | void Reallocate(unsigned newTableSize); |
153 | |
154 | // For iteration, we use a pattern similar to the STL "forward |
155 | // iterator" pattern. It basically consists of wrapping an |
156 | // "iteration variable" in an object, and providing pointer-like |
157 | // operators on the iterator. Example usage: |
158 | // |
159 | // for (SimplerHashTable::KeyIterator iter = foo->Begin(), end = foo->End(); !iter.Equal(end); iter++) |
160 | // { |
161 | // // use foo, iter. |
162 | // } |
163 | // iter.Get() will yield (a reference to) the |
164 | // current key. It will assert the equivalent of "iter != end." |
165 | class KeyIterator |
166 | { |
167 | private: |
168 | friend class SimplerHashTable; |
169 | |
170 | // The method implementations have to be here for portability. |
171 | // Some compilers won't compile the separate implementation in shash.inl |
172 | |
173 | Node** m_table; |
174 | Node* m_node; |
175 | unsigned m_tableSize; |
176 | unsigned m_index; |
177 | |
178 | public: |
179 | KeyIterator(const SimplerHashTable *hash, BOOL begin) |
180 | : m_table(hash->m_table), |
181 | m_node(NULL), |
182 | m_tableSize(hash->m_tableSizeInfo.prime), |
183 | m_index(begin ? 0 : m_tableSize) |
184 | { |
185 | if (begin && hash->m_tableCount > 0) |
186 | { |
187 | assert(m_table != NULL); |
188 | while (m_index < m_tableSize && m_table[m_index] == NULL) |
189 | m_index++; |
190 | |
191 | if (m_index >= m_tableSize) |
192 | { |
193 | return; |
194 | } |
195 | else |
196 | { |
197 | m_node = m_table[m_index]; |
198 | } |
199 | assert(m_node != NULL); |
200 | } |
201 | } |
202 | |
203 | const Key& Get() const |
204 | { |
205 | assert(m_node != NULL); |
206 | |
207 | return m_node->m_key; |
208 | } |
209 | |
210 | const Value& GetValue() const |
211 | { |
212 | assert(m_node != NULL); |
213 | |
214 | return m_node->m_val; |
215 | } |
216 | |
217 | void SetValue(const Value & value) const |
218 | { |
219 | assert(m_node != NULL); |
220 | |
221 | m_node->m_val = value; |
222 | } |
223 | |
224 | void Next() |
225 | { |
226 | if (m_node != NULL) |
227 | { |
228 | m_node = m_node->m_next; |
229 | if (m_node != NULL) |
230 | { |
231 | return; |
232 | } |
233 | |
234 | // Otherwise... |
235 | m_index++; |
236 | } |
237 | while (m_index < m_tableSize && m_table[m_index] == NULL) |
238 | m_index++; |
239 | |
240 | if (m_index >= m_tableSize) |
241 | { |
242 | m_node = NULL; |
243 | return; |
244 | } |
245 | else |
246 | { |
247 | m_node = m_table[m_index]; |
248 | } |
249 | assert(m_node != NULL); |
250 | } |
251 | |
252 | bool Equal(const KeyIterator &i) const |
253 | { |
254 | return i.m_node == m_node; |
255 | } |
256 | |
257 | void operator++() { |
258 | Next(); |
259 | } |
260 | |
261 | void operator++(int) { |
262 | Next(); |
263 | } |
264 | }; |
265 | |
266 | // HashTableRef only exists to support operator[] |
267 | // operator[] returns a HashTableRef which enables operator[] to support reading and writing |
268 | // in a normal array, it just returns a ref an actual element, which is not possible here. |
269 | class HashTableRef |
270 | { |
271 | public: |
272 | // this is really the getter for the array. |
273 | operator Value() |
274 | { |
275 | |
276 | Value result; |
277 | table->Lookup(key, &result); |
278 | return result; |
279 | } |
280 | |
281 | void operator =(const Value v) |
282 | { |
283 | table->Set(key, v); |
284 | } |
285 | |
286 | friend class SimplerHashTable; |
287 | |
288 | protected: |
289 | HashTableRef(SimplerHashTable *t, Key k) |
290 | { |
291 | table = t; |
292 | key = k; |
293 | } |
294 | |
295 | SimplerHashTable *table; |
296 | Key key; |
297 | }; |
298 | |
299 | Value &operator[](Key k) const |
300 | { |
301 | Value* p = LookupPointer(k); |
302 | assert(p); |
303 | return *p; |
304 | } |
305 | |
306 | private: |
307 | // Find the next prime number >= the given value. |
308 | static PrimeInfo NextPrime(unsigned number); |
309 | |
310 | // Instance members |
311 | IAllocator* m_alloc; // IAllocator to use in this |
312 | // table. |
313 | // The node type. |
314 | struct Node { |
315 | Node* m_next; // Assume that the alignment requirement of Key and Value are no greater than Node*, so put m_next to avoid unnecessary padding. |
316 | Key m_key; |
317 | Value m_val; |
318 | |
319 | Node(Key k, Value v, Node* next) : m_next(next), m_key(k), m_val(v) {} |
320 | |
321 | void* operator new(size_t sz, IAllocator* alloc) |
322 | { |
323 | return alloc->Alloc(sz); |
324 | } |
325 | |
326 | void operator delete(void* p, IAllocator* alloc) |
327 | { |
328 | alloc->Free(p); |
329 | } |
330 | }; |
331 | |
332 | Node** m_table; // pointer to table |
333 | PrimeInfo m_tableSizeInfo; // size of table (a prime) and information about it |
334 | unsigned m_tableCount; // number of elements in table |
335 | unsigned m_tableMax; // maximum occupied count |
336 | }; |
337 | |
338 | #include "simplerhash.inl" |
339 | |
340 | // A few simple KeyFuncs types... |
341 | |
342 | // Base class for types whose equality function is the same as their "==". |
343 | template<typename T> |
344 | struct KeyFuncsDefEquals |
345 | { |
346 | static bool Equals(const T& x, const T& y) |
347 | { |
348 | return x == y; |
349 | } |
350 | }; |
351 | |
352 | template<typename T> |
353 | struct PtrKeyFuncs: public KeyFuncsDefEquals<const T*> |
354 | { |
355 | public: |
356 | static unsigned GetHashCode(const T* ptr) |
357 | { |
358 | // Hmm. Maybe (unsigned) ought to be "ssize_t" -- or this ought to be ifdef'd by size. |
359 | return static_cast<unsigned>(reinterpret_cast<uintptr_t>(ptr)); |
360 | } |
361 | }; |
362 | |
363 | template<typename T> // Must be coercable to "unsigned" with no loss of information. |
364 | struct SmallPrimitiveKeyFuncs: public KeyFuncsDefEquals<T> |
365 | { |
366 | static unsigned GetHashCode(const T& val) |
367 | { |
368 | return static_cast<unsigned>(val); |
369 | } |
370 | }; |
371 | |
372 | template<typename T> // Assumed to be of size sizeof(UINT64). |
373 | struct LargePrimitiveKeyFuncs: public KeyFuncsDefEquals<T> |
374 | { |
375 | static unsigned GetHashCode(const T val) |
376 | { |
377 | // A static cast when T is a float or a double converts the value (i.e. 0.25 converts to 0) |
378 | // |
379 | // Instead we want to use all of the bits of a float to create the hash value |
380 | // So we cast the address of val to a pointer to an equivalent sized unsigned int |
381 | // This allows us to read the actual bit representation of a float type |
382 | // |
383 | // We can't read beyond the end of val, so we use sizeof(T) to determine |
384 | // exactly how many bytes to read |
385 | // |
386 | if (sizeof(T) == 8) |
387 | { |
388 | // cast &val to (UINT64 *) then deref to get the bits |
389 | UINT64 asUINT64 = *(reinterpret_cast<const UINT64 *>(&val)); |
390 | |
391 | // Get the upper and lower 32-bit values from the 64-bit value |
392 | UINT32 upper32 = static_cast<UINT32> (asUINT64 >> 32); |
393 | UINT32 lower32 = static_cast<UINT32> (asUINT64 & 0xFFFFFFFF); |
394 | |
395 | // Exclusive-Or the upper32 and the lower32 values |
396 | return static_cast<unsigned>(upper32 ^ lower32); |
397 | |
398 | } |
399 | else if (sizeof(T) == 4) |
400 | { |
401 | // cast &val to (UINT32 *) then deref to get the bits |
402 | UINT32 asUINT32 = *(reinterpret_cast<const UINT32 *>(&val)); |
403 | |
404 | // Just return the 32-bit value |
405 | return static_cast<unsigned>(asUINT32); |
406 | } |
407 | else if ((sizeof(T) == 2) || (sizeof(T) == 1)) |
408 | { |
409 | // For small sizes we must have an integer type |
410 | // so we can just use the static_cast. |
411 | // |
412 | return static_cast<unsigned>(val); |
413 | } |
414 | else |
415 | { |
416 | // Only support Hashing for types that are 8,4,2 or 1 bytes in size |
417 | assert(!"Unsupported size" ); |
418 | return static_cast<unsigned>(val); // compile-time error here when we have a illegal size |
419 | } |
420 | } |
421 | }; |
422 | |
423 | #endif // _SIMPLERHASHTABLE_H_ |
424 | |