1 | // tRingBuffer.h  |
2 | //  |
3 | // A ring (or circular) buffer works like a capacitor in that it can temporarily absorb the shock of a large influx  |
4 | // of data, but needs to be emptied eventually. The read rate must on average be >= to the write rate if you want to  |
5 | // avoid a stall. Fortunately, unlike real capacitors, they do not leak information over time.  |
6 | //   |
7 | // Copyright (c) 2016, 2017 Tristan Grimmer.  |
8 | // Permission to use, copy, modify, and/or distribute this software for any purpose with or without fee is hereby  |
9 | // granted, provided that the above copyright notice and this permission notice appear in all copies.  |
10 | //  |
11 | // THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL  |
12 | // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT,  |
13 | // INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN  |
14 | // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR  |
15 | // PERFORMANCE OF THIS SOFTWARE.  |
16 |   |
17 | #pragma once  |
18 | #include <Foundation/tAssert.h>  |
19 | #include <Foundation/tStandard.h>  |
20 |   |
21 |   |
22 | // tRingBuffer allows you to append one or more items (of user-specified type) to the tail of the ring, remove one or  |
23 | // more items from the head of the ring, and randomly access any stored item for read/write.  |
24 | template<typename T> class tRingBuffer  |
25 | {   |
26 | public:  |
27 | tRingBuffer() : OwnsBuffer(false), Capacity(0), Buffer(nullptr), Head(nullptr), Tail(nullptr) { }  |
28 | tRingBuffer(int capacity) /* Buffer is owned by tRingBuffer. */ : OwnsBuffer(false), Capacity(0), Buffer(nullptr), Head(nullptr), Tail(nullptr) { Set(capacity); }  |
29 | tRingBuffer(T* const buffer, int capacity) /* Buffer ownership is external. */ : OwnsBuffer(false), Capacity(0), Buffer(nullptr), Head(nullptr), Tail(nullptr) { Set(buffer, capacity); }  |
30 |   |
31 | void Set(int capacity); // Buffer is owned by tRingBuffer.  |
32 | void Set(T* const buffer, int capacity); // Buffer is owned by tRingBuffer.  |
33 | void Clear();  |
34 | bool IsValid() const { return Buffer ? true : false; }  |
35 |   |
36 | int GetNumItems() const;  |
37 | int GetCapacity() const { return Capacity; }  |
38 | int GetRoom() const /* Returns how many items can be successfully appended. */ { return Capacity - GetNumItems(); }  |
39 | bool IsEmpty() const /* An invalid ring buffer always returns true. */ { return !Buffer || ((Tail != nullptr) && (Head == nullptr)); }  |
40 | bool IsFull() const /* An invalid ring buffer always returns true. */ { return !Buffer || ((Head != nullptr) && (Tail == nullptr)); }  |
41 |   |
42 | // There are two versions each of append and remove. One for single items and one for multiple. The multiple item  |
43 | // versions are faster than multiple single item appends or removes because they perform the operation with either a  |
44 | // single or double tight loop of item copying. If you know the type T can be memory copied (POD types and such),  |
45 | // set useMemCpy to true for increased performance. Returns success for the single item versions, and number items  |
46 | // appended or removed for the multiple item versions.  |
47 | bool Append(const T& item);  |
48 | int Append(const T* items, int numItems, bool useMemCpy = false);  |
49 | bool Remove(T& item);  |
50 | int Remove(T* items, int numItems, bool useMemCpy = false);  |
51 |   |
52 | // These two functions are for testing and verification and need not be called for production use. They return -1  |
53 | // if the head or tail is invalid.  |
54 | int GetHeadIndex() const { if (!Head) return -1; return int(Head - Buffer); }  |
55 | int GetTailIndex() const { if (!Tail) return -1; return int(Tail - Buffer); }  |
56 |   |
57 | private:  |
58 | bool OwnsBuffer; // True if this object owns the buffer.  |
59 | int Capacity; // Maximum number of items that may be stored.  |
60 | T* Buffer; // If null the whole object is invalid.  |
61 |   |
62 | // There are few choices regarding head and tail here. I wanted to use the full capacity rather than making  |
63 | // head == tail represent 'empty' or something. In the following implementation the head always represents the read  |
64 | // location, and is null if nothing can be read. The tail always represents the write location, and is null if  |
65 | // nothing can be written. As such, if buffer is null (invalid ring buffer), both head and tail are null. For valid  |
66 | // ring buffers the following is true. An empty buffer has a non-null tail and a null head. That is, you can write but  |
67 | // not read. A full buffer has a non-null head and a null tail. That is, you can read but not write. eg.  |
68 | // - - - - - - 0 items.  |
69 | // T H=null  |
70 | //  |
71 | // - - X - - - 1 item.  |
72 | // H T  |
73 | //  |
74 | // - - X X - - 2 items.  |
75 | // H T  |
76 | //  |
77 | // X - X X X X 5 items.  |
78 | // T H  |
79 | //  |
80 | // X X X X X X 6 items (full).  |
81 | // H T=null.  |
82 | T* Head; // Points to where you can read from.  |
83 | T* Tail; // Points to where you can write to.  |
84 | };  |
85 | template<typename T> using tRing = tRingBuffer<T>;  |
86 |   |
87 |   |
88 | // Implementation below this line.  |
89 |   |
90 |   |
91 | template<typename T> inline void tRingBuffer<T>::Set(int capacity)  |
92 | {  |
93 | Clear();  |
94 | tAssert(capacity > 0);  |
95 | Capacity = capacity;  |
96 | Buffer = new T[capacity];  |
97 |   |
98 | // We're empty so read head is null, and write tail is valid.  |
99 | Tail = Buffer;  |
100 | OwnsBuffer = true;  |
101 | }  |
102 |   |
103 |   |
104 | template<typename T> inline void tRingBuffer<T>::Set(T* const buffer, int capacity)  |
105 | {  |
106 | Clear();  |
107 | tAssert((capacity > 0) && buffer);  |
108 | Capacity = capacity;  |
109 | Buffer = buffer;  |
110 | Tail = Buffer;  |
111 | OwnsBuffer = false;  |
112 | }  |
113 |   |
114 |   |
115 | template<typename T> inline void tRingBuffer<T>::Clear()  |
116 | {  |
117 | if (OwnsBuffer)  |
118 | delete Buffer;  |
119 | Buffer = nullptr;  |
120 | Head = nullptr;  |
121 | Tail = nullptr;  |
122 | OwnsBuffer = false;  |
123 | Capacity = 0;  |
124 | }  |
125 |   |
126 |   |
127 | template<typename T> inline int tRingBuffer<T>::GetNumItems() const  |
128 | {  |
129 | if (!Head)  |
130 | return 0;  |
131 |   |
132 | if (!Tail)  |
133 | return Capacity;  |
134 |   |
135 | if (Head > Tail)  |
136 | return Capacity - int(Head - Tail);  |
137 |   |
138 | return int(Tail - Head);  |
139 | }  |
140 |   |
141 |   |
142 | template<typename T> inline bool tRingBuffer<T>::Append(const T& item)  |
143 | {  |
144 | if (!Tail)  |
145 | return false;  |
146 |   |
147 | *Tail = item;  |
148 | if (!Head)  |
149 | Head = Tail;  |
150 |   |
151 | Tail++;  |
152 | if (Tail == Buffer+Capacity)  |
153 | Tail = Buffer;  |
154 |   |
155 | // Are we full?  |
156 | if (Tail == Head)  |
157 | Tail = nullptr;  |
158 |   |
159 | return true;  |
160 | }  |
161 |   |
162 |   |
163 | template<typename T> inline int tRingBuffer<T>::Append(const T* items, int numItems, bool useMemCpy)  |
164 | {  |
165 | // This could have been written naively by just sequentially calling the Append(T) above. The reason  |
166 | // it is so complicated is because it organized the writes into two separate tight copy loops.  |
167 | int room = GetRoom();  |
168 | int numAppended = 0;  |
169 | if (!items || (numItems <= 0) || (room <= 0))  |
170 | return numAppended;  |
171 |   |
172 | int numToAppend = (numItems < room) ? numItems : room;  |
173 |   |
174 | // We may need two memcpy/write-loops if we go past the end.  |
175 | int loopAvail = 0;  |
176 | if (Tail > Head) // This works if pHead is nullptr as well.  |
177 | loopAvail = int((Buffer+Capacity) - Tail);  |
178 | int itemCount = (numToAppend < loopAvail) ? numToAppend : loopAvail;  |
179 |   |
180 | // Copy loop 1.  |
181 | if (useMemCpy)  |
182 | tStd::tMemcpy(Tail, items, itemCount*sizeof(T));  |
183 | else  |
184 | for (int i = 0; i < itemCount; i++)  |
185 | *(Tail+i) = *(items+i);  |
186 |   |
187 | if (!Head && itemCount)  |
188 | Head = Tail;  |
189 |   |
190 | numAppended += itemCount;  |
191 | Tail += itemCount;  |
192 | if (Tail == Buffer+Capacity)  |
193 | Tail = Buffer;  |
194 |   |
195 | // Are we full?  |
196 | if (Tail == Head)  |
197 | {  |
198 | Tail = nullptr;  |
199 | return numAppended;  |
200 | }  |
201 |   |
202 | numToAppend -= itemCount;  |
203 | if (numToAppend == 0)  |
204 | return numAppended;  |
205 |   |
206 | // At this point we still have more to append and we have looped around into 'tail < head' territory.  |
207 | items += numAppended;  |
208 | loopAvail = int(Head - Tail);  |
209 | itemCount = (numToAppend < loopAvail) ? numToAppend : loopAvail;  |
210 |   |
211 | // Copy loop 2.  |
212 | if (useMemCpy)  |
213 | tStd::tMemcpy(Buffer, items, itemCount*sizeof(T));  |
214 | else  |
215 | for (int i = 0; i < itemCount; i++)  |
216 | *(Buffer+i) = *(items+i);  |
217 |   |
218 | if (!Head && itemCount)  |
219 | Head = Tail;  |
220 |   |
221 | numAppended += itemCount;  |
222 | Tail += itemCount;  |
223 |   |
224 | // Are we full?  |
225 | if (Tail == Head)  |
226 | Tail = nullptr;  |
227 |   |
228 | return numAppended;  |
229 | }  |
230 |   |
231 |   |
232 | template<typename T> inline bool tRingBuffer<T>::Remove(T& item)  |
233 | {  |
234 | if (!Head)  |
235 | return false;  |
236 |   |
237 | item = *Head;  |
238 | if (!Tail)  |
239 | Tail = Head;  |
240 |   |
241 | Head++;  |
242 | if (Head == Buffer+Capacity)  |
243 | Head = Buffer;  |
244 |   |
245 | // Are we empty?  |
246 | if (Head == Tail)  |
247 | Head = nullptr;  |
248 |   |
249 | return true;  |
250 | }  |
251 |   |
252 |   |
253 | template<typename T> inline int tRingBuffer<T>::Remove(T* items, int numItems, bool useMemCpy)  |
254 | {  |
255 | // This could have been written naively by just sequentially calling the Remove(T) above. The reason  |
256 | // it is so complicated is because it organized the writes into two separate tight copy loops.  |
257 | int avail = GetNumItems();  |
258 | int numRemoved = 0;  |
259 | if (!items || (numItems <= 0) || (avail <= 0))  |
260 | return numRemoved;  |
261 |   |
262 | int numToRemove = (numItems < avail) ? numItems : avail;  |
263 |   |
264 | // We may need two memcpy/write-loops if we go past the end.  |
265 | int loopAvail = 0;  |
266 | if (Head > Tail) // This works if pTail is nullptr as well.  |
267 | loopAvail = int((Buffer+Capacity) - Head);  |
268 | int itemCount = (numToRemove < loopAvail) ? numToRemove : loopAvail;  |
269 |   |
270 | // Copy loop 1.  |
271 | if (useMemCpy)  |
272 | tStd::tMemcpy(items, Head, itemCount*sizeof(T));  |
273 | else  |
274 | for (int i = 0; i < itemCount; i++)  |
275 | *(items+i) = *(Head+i);  |
276 |   |
277 | if (!Tail && itemCount)  |
278 | Tail = Head;  |
279 |   |
280 | numRemoved += itemCount;  |
281 | Head += itemCount;  |
282 | if (Head == Buffer+Capacity)  |
283 | Head = Buffer;  |
284 |   |
285 | // Are we empty?  |
286 | if (Head == Tail)  |
287 | {  |
288 | Head = nullptr;  |
289 | return numRemoved;  |
290 | }  |
291 |   |
292 | numToRemove -= itemCount;  |
293 | if (numToRemove == 0)  |
294 | return numRemoved;  |
295 |   |
296 | // At this point we still have more to remove and we have looped around into 'head < tail' territory.  |
297 | items += numRemoved;  |
298 | loopAvail = int(Tail - Head);  |
299 | itemCount = (numToRemove < loopAvail) ? numToRemove : loopAvail;  |
300 |   |
301 | // Copy loop 2.  |
302 | if (useMemCpy)  |
303 | tStd::tMemcpy(items, Buffer, itemCount*sizeof(T));  |
304 | else  |
305 | for (int i = 0; i < itemCount; i++)  |
306 | *(items+i) = *(Buffer+i);  |
307 |   |
308 | if (!Tail && itemCount)  |
309 | Tail = Head;  |
310 |   |
311 | numRemoved += itemCount;  |
312 | Head += itemCount;  |
313 |   |
314 | // Are we empty?  |
315 | if (Head == Tail)  |
316 | Head = nullptr;  |
317 |   |
318 | return numRemoved;  |
319 | }  |
320 | |