|
1 // Copyright (c) 1998-2009 Nokia Corporation and/or its subsidiary(-ies). |
|
2 // All rights reserved. |
|
3 // This component and the accompanying materials are made available |
|
4 // under the terms of the License "Eclipse Public License v1.0" |
|
5 // which accompanies this distribution, and is available |
|
6 // at the URL "http://www.eclipse.org/legal/epl-v10.html". |
|
7 // |
|
8 // Initial Contributors: |
|
9 // Nokia Corporation - initial contribution. |
|
10 // |
|
11 // Contributors: |
|
12 // |
|
13 // Description: |
|
14 // e32\euser\us_encode.cpp |
|
15 // |
|
16 // |
|
17 |
|
18 #include "e32huffman.h" |
|
19 #include <e32base.h> |
|
20 #include <e32base_private.h> |
|
21 #include <e32panic.h> |
|
22 |
|
23 _LIT(KCat,"Huffman"); |
|
24 // local definitions used for Huffman code generation |
|
25 typedef TUint16 THuff; /** @internal */ |
|
26 const THuff KLeaf=0x8000; /** @internal */ |
|
27 struct TNode |
|
28 /** @internal */ |
|
29 { |
|
30 TUint iCount; |
|
31 THuff iLeft; |
|
32 THuff iRight; |
|
33 }; |
|
34 |
|
35 /** recursive function to calculate the code lengths from the node tree |
|
36 |
|
37 @internalComponent |
|
38 */ |
|
39 void HuffmanLengthsL(TUint32* aLengths,const TNode* aNodes,TInt aNode,TInt aLen) |
|
40 { |
|
41 if (++aLen>Huffman::KMaxCodeLength) |
|
42 User::Leave(KErrOverflow); |
|
43 |
|
44 const TNode& node=aNodes[aNode]; |
|
45 TUint x=node.iLeft; |
|
46 if (x&KLeaf) |
|
47 aLengths[x&~KLeaf]=aLen; |
|
48 else |
|
49 HuffmanLengthsL(aLengths,aNodes,x,aLen); |
|
50 x=node.iRight; |
|
51 if (x&KLeaf) |
|
52 aLengths[x&~KLeaf]=aLen; |
|
53 else |
|
54 HuffmanLengthsL(aLengths,aNodes,x,aLen); |
|
55 } |
|
56 |
|
57 /** Insert the {aCount,aValue} pair into the already sorted array of nodes |
|
58 |
|
59 @internalComponent |
|
60 */ |
|
61 void InsertInOrder(TNode* aNodes, TInt aSize, TUint aCount, TInt aVal) |
|
62 { |
|
63 // Uses Insertion sort following a binary search... |
|
64 TInt l=0, r=aSize; |
|
65 while (l < r) |
|
66 { |
|
67 TInt m = (l+r) >> 1; |
|
68 if (aNodes[m].iCount<aCount) |
|
69 r=m; |
|
70 else |
|
71 l=m+1; |
|
72 } |
|
73 Mem::Copy(aNodes+l+1,aNodes+l,sizeof(TNode)*(aSize-l)); |
|
74 aNodes[l].iCount=aCount; |
|
75 aNodes[l].iRight=TUint16(aVal); |
|
76 } |
|
77 |
|
78 /** Generate a Huffman code |
|
79 |
|
80 This generates a Huffman code for a given set of code frequencies. The output |
|
81 is a table of code lengths which can be used to build canonincal encoding tables |
|
82 or decoding trees for use with the TBitInput and TBitOutput classes. |
|
83 |
|
84 Entries in the table with a frequency of zero will have a zero code length |
|
85 and thus no associated huffman encoding. If each such symbol should have a |
|
86 maximum length encoding, they must be given at least a frequency of 1. |
|
87 |
|
88 For an alphabet of n symbols, this algorithm has a transient memory overhead |
|
89 of 8n, and a time complexity of O(n*log(n)). |
|
90 |
|
91 @param aFrequency The table of code frequencies |
|
92 @param aNumCodes The number of codes in the table |
|
93 @param aHuffman The table for the output code-length table. This must be |
|
94 the same size as the frequency table, and can safely be the same table |
|
95 |
|
96 @leave KErrNoMemory If memory used for code generation cannot be allocated |
|
97 |
|
98 @panic "USER ???" If the number of codes exceeds Huffman::KMaxCodes |
|
99 */ |
|
100 EXPORT_C void Huffman::HuffmanL(const TUint32 aFrequency[],TInt aNumCodes,TUint32 aHuffman[]) |
|
101 { |
|
102 __ASSERT_ALWAYS(TUint(aNumCodes)<=TUint(KMaxCodes),User::Panic(KCat,EHuffmanTooManyCodes)); |
|
103 |
|
104 // Sort the values into decreasing order of frequency |
|
105 // |
|
106 TNode* nodes = new(ELeave) TNode[aNumCodes]; |
|
107 CleanupArrayDeletePushL(nodes); |
|
108 TInt lCount=0; |
|
109 |
|
110 for (TInt ii=0;ii<aNumCodes;++ii) |
|
111 { |
|
112 TInt c=aFrequency[ii]; |
|
113 if (c!=0) |
|
114 InsertInOrder(nodes,lCount++,c,ii|KLeaf); |
|
115 } |
|
116 |
|
117 // default code length is zero |
|
118 Mem::FillZ(aHuffman,aNumCodes*sizeof(TUint32)); |
|
119 |
|
120 if (lCount==0) |
|
121 { |
|
122 // no codes with frequency>0. No code has a length |
|
123 } |
|
124 else if (lCount==1) |
|
125 { |
|
126 // special case for a single value (always encode as "0") |
|
127 aHuffman[nodes[0].iRight&~KLeaf]=1; |
|
128 } |
|
129 else |
|
130 { |
|
131 // Huffman algorithm: pair off least frequent nodes and reorder |
|
132 // |
|
133 do |
|
134 { |
|
135 --lCount; |
|
136 TUint c=nodes[lCount].iCount + nodes[lCount-1].iCount; |
|
137 nodes[lCount].iLeft=nodes[lCount-1].iRight; |
|
138 // re-order the leaves now to reflect new combined frequency 'c' |
|
139 InsertInOrder(nodes,lCount-1,c,lCount); |
|
140 } while (lCount>1); |
|
141 // generate code lengths in aHuffman[] |
|
142 HuffmanLengthsL(aHuffman,nodes,1,0); |
|
143 } |
|
144 CleanupStack::PopAndDestroy(nodes); |
|
145 |
|
146 __ASSERT_DEBUG(IsValid(aHuffman,aNumCodes),User::Panic(KCat,EHuffmanInvalidCoding)); |
|
147 } |
|
148 |
|
149 /** Validate a Huffman encoding |
|
150 |
|
151 This verifies that a Huffman coding described by the code lengths is valid. |
|
152 In particular, it ensures that no code exceeds the maximum length and |
|
153 that it is possible to generate a canonical coding for the specified lengths. |
|
154 |
|
155 @param aHuffman The table of code lengths as generated by Huffman::HuffmanL() |
|
156 @param aNumCodes The number of codes in the table |
|
157 |
|
158 @return True if the code is valid, otherwise false |
|
159 */ |
|
160 EXPORT_C TBool Huffman::IsValid(const TUint32 aHuffman[],TInt aNumCodes) |
|
161 { |
|
162 // The code is valid if one of the following holds: |
|
163 // (a) the code exactly fills the 'code space' |
|
164 // (b) there is only a single symbol with code length 1 |
|
165 // (c) there are no encoded symbols |
|
166 // |
|
167 TUint remain=1<<KMaxCodeLength; |
|
168 TInt totlen=0; |
|
169 for (const TUint32* p=aHuffman+aNumCodes; p>aHuffman;) |
|
170 { |
|
171 TInt len=*--p; |
|
172 if (len>0) |
|
173 { |
|
174 totlen+=len; |
|
175 if (len>KMaxCodeLength) |
|
176 return EFalse; |
|
177 TUint c=1<<(KMaxCodeLength-len); |
|
178 if (c>remain) |
|
179 return EFalse; |
|
180 remain-=c; |
|
181 } |
|
182 } |
|
183 |
|
184 return remain==0 || totlen<=1; |
|
185 } |
|
186 |
|
187 /** Create a canonical Huffman encoding table |
|
188 |
|
189 This generates the huffman codes used by TBitOutput::HuffmanL() to write huffman |
|
190 encoded data. The input is table of code lengths, as generated by Huffman::HuffmanL() |
|
191 and must represent a valid huffman code. |
|
192 |
|
193 @param aHuffman The table of code lengths as generated by Huffman::HuffmanL() |
|
194 @param aNumCodes The number of codes in the table |
|
195 @param aEncodeTable The table for the output huffman codes. This must be |
|
196 the same size as the code-length table, and can safely be the same table |
|
197 |
|
198 @panic "USER ???" If the provided code is not a valid Huffman coding |
|
199 |
|
200 @see IsValid() |
|
201 @see HuffmanL() |
|
202 */ |
|
203 EXPORT_C void Huffman::Encoding(const TUint32 aHuffman[],TInt aNumCodes,TUint32 aEncodeTable[]) |
|
204 { |
|
205 __ASSERT_ALWAYS(IsValid(aHuffman,aNumCodes),User::Panic(KCat,EHuffmanInvalidCoding)); |
|
206 |
|
207 TFixedArray<TInt,KMaxCodeLength> lenCount; |
|
208 lenCount.Reset(); |
|
209 |
|
210 TInt ii; |
|
211 for (ii=0;ii<aNumCodes;++ii) |
|
212 { |
|
213 TInt len=aHuffman[ii]-1; |
|
214 if (len>=0) |
|
215 ++lenCount[len]; |
|
216 } |
|
217 |
|
218 TFixedArray<TUint,KMaxCodeLength> nextCode; |
|
219 TUint code=0; |
|
220 for (ii=0;ii<KMaxCodeLength;++ii) |
|
221 { |
|
222 code<<=1; |
|
223 nextCode[ii]=code; |
|
224 code+=lenCount[ii]; |
|
225 } |
|
226 |
|
227 for (ii=0;ii<aNumCodes;++ii) |
|
228 { |
|
229 TInt len=aHuffman[ii]; |
|
230 if (len==0) |
|
231 aEncodeTable[ii]=0; |
|
232 else |
|
233 { |
|
234 aEncodeTable[ii] = (nextCode[len-1]<<(KMaxCodeLength-len))|(len<<KMaxCodeLength); |
|
235 ++nextCode[len-1]; |
|
236 } |
|
237 } |
|
238 } |
|
239 |
|
240 /** the encoding table for the externalised code |
|
241 @internalComponent |
|
242 */ |
|
243 const TUint32 HuffmanEncoding[]= |
|
244 { |
|
245 0x10000000, |
|
246 0x1c000000, |
|
247 0x12000000, |
|
248 0x1d000000, |
|
249 0x26000000, |
|
250 0x26800000, |
|
251 0x2f000000, |
|
252 0x37400000, |
|
253 0x37600000, |
|
254 0x37800000, |
|
255 0x3fa00000, |
|
256 0x3fb00000, |
|
257 0x3fc00000, |
|
258 0x3fd00000, |
|
259 0x47e00000, |
|
260 0x47e80000, |
|
261 0x47f00000, |
|
262 0x4ff80000, |
|
263 0x57fc0000, |
|
264 0x5ffe0000, |
|
265 0x67ff0000, |
|
266 0x77ff8000, |
|
267 0x7fffa000, |
|
268 0x7fffb000, |
|
269 0x7fffc000, |
|
270 0x7fffd000, |
|
271 0x7fffe000, |
|
272 0x87fff000, |
|
273 0x87fff800 |
|
274 }; |
|
275 |
|
276 /** encode 0a as '0' and 0b as '1', return number of symbols created |
|
277 |
|
278 @internalComponent |
|
279 */ |
|
280 void EncodeRunLengthL(TBitOutput& aOutput, TInt aLength) |
|
281 { |
|
282 if (aLength>0) |
|
283 { |
|
284 EncodeRunLengthL(aOutput,(aLength-1)>>1); |
|
285 aOutput.HuffmanL(HuffmanEncoding[1-(aLength&1)]); |
|
286 } |
|
287 } |
|
288 |
|
289 /** Store a canonical huffman encoding in compact form |
|
290 |
|
291 As the encoding is canonical, only the code lengths of each code needs to be saved. |
|
292 |
|
293 Due to the nature of code length tables, these can usually be stored very compactly |
|
294 by encoding the encoding itself, hence the use of the bit output stream. |
|
295 |
|
296 @param aOutput The output stream for the encoding |
|
297 @param aHuffman The table of code lengths as generated by Huffman::HuffmanL() |
|
298 @param aNumCodes The number of huffman codes in the table |
|
299 |
|
300 @leave TBitOutput::HuffmanL() |
|
301 */ |
|
302 EXPORT_C void Huffman::ExternalizeL(TBitOutput& aOutput,const TUint32 aHuffman[],TInt aNumCodes) |
|
303 { |
|
304 // We assume that the code length table is generated by the huffman generator, |
|
305 // in which case the maxmimum code length is 27 bits. |
|
306 // |
|
307 // We apply three transformations to the data: |
|
308 // 1. the data goes through a move-to-front coder |
|
309 // 2. apply a rle-0 coder which replace runs of '0' with streams of '0a' and '0b' |
|
310 // 3. encode the result using a predefined (average) huffman coding |
|
311 // |
|
312 // This can be done in a single pass over the data, avoiding the need for additional |
|
313 // memory. |
|
314 // |
|
315 // initialise the list for the MTF coder |
|
316 TFixedArray<TUint8,Huffman::KMetaCodes> list; |
|
317 TInt i; |
|
318 for (i=0;i<list.Count();++i) |
|
319 list[i]=TUint8(i); |
|
320 TInt last=0; |
|
321 |
|
322 TInt rl=0; |
|
323 const TUint32* p32=aHuffman; |
|
324 const TUint32* e32=p32+aNumCodes; |
|
325 while (p32<e32) |
|
326 { |
|
327 TInt c=*p32++; |
|
328 if (c==last) |
|
329 ++rl; // repeat of last symbol |
|
330 else |
|
331 { |
|
332 // encode run-length |
|
333 EncodeRunLengthL(aOutput,rl); |
|
334 rl=0; |
|
335 // find code in MTF list |
|
336 TInt j; |
|
337 for (j=1;list[j]!=c;++j) |
|
338 ; |
|
339 // store this code |
|
340 aOutput.HuffmanL(HuffmanEncoding[j+1]); |
|
341 // adjust list for MTF algorithm |
|
342 while (--j>0) |
|
343 list[j+1]=list[j]; |
|
344 list[1]=TUint8(last); |
|
345 last=c; |
|
346 } |
|
347 } |
|
348 // encod any remaining run-length |
|
349 EncodeRunLengthL(aOutput,rl); |
|
350 } |
|
351 |
|
352 |
|
353 /** Construct a bit stream output object |
|
354 |
|
355 Following construction the bit stream is ready for writing bits, but will first call |
|
356 OverflowL() as the output buffer is 'full'. A derived class can detect this state as |
|
357 Ptr() will return null. |
|
358 */ |
|
359 EXPORT_C TBitOutput::TBitOutput() |
|
360 :iCode(0),iBits(-8),iPtr(0),iEnd(0) |
|
361 {} |
|
362 |
|
363 /** Construct a bit stream output object over a buffer |
|
364 |
|
365 Data will be written to the buffer until it is full, at which point OverflowL() will |
|
366 be called. This should handle the data and then can Set() again to reset the buffer |
|
367 for further output. |
|
368 |
|
369 @param aBuf The buffer for output |
|
370 @param aSize The size of the buffer in bytes |
|
371 */ |
|
372 EXPORT_C TBitOutput::TBitOutput(TUint8* aBuf,TInt aSize) |
|
373 :iCode(0),iBits(-8),iPtr(aBuf),iEnd(aBuf+aSize) |
|
374 {} |
|
375 |
|
376 /** Write a huffman code |
|
377 |
|
378 This expects a huffman code value as generated by Huffman::Encoding() |
|
379 |
|
380 @param aHuffCode The huffman code write to the stream |
|
381 |
|
382 @leave OverflowL() If the output buffer is full, OverflowL() is called |
|
383 */ |
|
384 EXPORT_C void TBitOutput::HuffmanL(TUint aHuffCode) |
|
385 { |
|
386 DoWriteL(aHuffCode<<(32-Huffman::KMaxCodeLength),aHuffCode>>Huffman::KMaxCodeLength); |
|
387 } |
|
388 |
|
389 /** Write an arbitrary integer value |
|
390 |
|
391 Write an unsigned integer using the number of bits specified. Only |
|
392 the low order bits of the value are written to the output, most |
|
393 significant bit first. |
|
394 |
|
395 @param aValue The value to write to the stream |
|
396 @param aLength The number of bits to output |
|
397 |
|
398 @leave OverflowL() If the output buffer is full, OverflowL() is called |
|
399 */ |
|
400 EXPORT_C void TBitOutput::WriteL(TUint aValue,TInt aLength) |
|
401 { |
|
402 if (aLength) |
|
403 DoWriteL(aValue<<=32-aLength,aLength); |
|
404 } |
|
405 |
|
406 /** Pad the bitstream to the next byte boundary |
|
407 |
|
408 Terminate the bitstream by padding the last byte with the requested value. |
|
409 Following this operation the bitstream can continue to be used, the data will |
|
410 start at the next byte. |
|
411 |
|
412 @param aPadding The bit value to pad the final byte with |
|
413 |
|
414 @leave OverflowL() If the output buffer is full, OverflowL() is called |
|
415 */ |
|
416 EXPORT_C void TBitOutput::PadL(TUint aPadding) |
|
417 { |
|
418 if (iBits>-8) |
|
419 WriteL(aPadding?0xffffffffu:0,-iBits); |
|
420 } |
|
421 |
|
422 /** Write the higher order bits to the stream |
|
423 |
|
424 @internalComponent |
|
425 */ |
|
426 void TBitOutput::DoWriteL(TUint aBits,TInt aSize) |
|
427 { |
|
428 if (aSize>25) |
|
429 { |
|
430 // cannot process >25 bits in a single pass |
|
431 // so do the top 8 bits first |
|
432 ASSERT(aSize<=32); |
|
433 DoWriteL(aBits&0xff000000u,8); |
|
434 aBits<<=8; |
|
435 aSize-=8; |
|
436 } |
|
437 |
|
438 TInt bits=iBits; |
|
439 TUint code=iCode|(aBits>>(bits+8)); |
|
440 bits+=aSize; |
|
441 if (bits>=0) |
|
442 { |
|
443 TUint8* p=iPtr; |
|
444 do |
|
445 { |
|
446 if (p==iEnd) |
|
447 { |
|
448 // run out of buffer space so invoke the overflow handler |
|
449 iPtr=p; |
|
450 OverflowL(); |
|
451 p=iPtr; |
|
452 ASSERT(p!=iEnd); |
|
453 } |
|
454 *p++=TUint8(code>>24); |
|
455 code<<=8; |
|
456 bits-=8; |
|
457 } while (bits>=0); |
|
458 iPtr=p; |
|
459 } |
|
460 iCode=code; |
|
461 iBits=bits; |
|
462 } |
|
463 |
|
464 /** Handle a full output buffer |
|
465 |
|
466 This virtual function is called when the output buffer is full. It should deal |
|
467 with the data in the buffer before reseting the buffer using Set(), allowing |
|
468 further data to be written. |
|
469 |
|
470 A derived class can replace this to write the data to a file (for example) |
|
471 before marking the buffer as empty. |
|
472 |
|
473 @leave KErrOverflow The default implementation leaves |
|
474 */ |
|
475 void TBitOutput::OverflowL() |
|
476 { |
|
477 User::Leave(KErrOverflow); |
|
478 } |