|
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 // e32tools\petran\Szip\encode.cpp |
|
15 // |
|
16 // |
|
17 |
|
18 #include "huffman.h" |
|
19 #include "panic.h" |
|
20 #include <e32base.h> |
|
21 #include <e32base_private.h> |
|
22 #include "h_utl.h" |
|
23 #include <assert.h> |
|
24 #include "farray.h" |
|
25 #include <stdlib.h> |
|
26 |
|
27 void User::Invariant() |
|
28 { |
|
29 fprintf(stderr, "User::Invariant() called\n"); |
|
30 exit(1); |
|
31 } |
|
32 |
|
33 // local definitions used for Huffman code generation |
|
34 typedef TUint16 THuff; /** @internal */ |
|
35 const THuff KLeaf=0x8000; /** @internal */ |
|
36 struct TNode |
|
37 /** @internal */ |
|
38 { |
|
39 TUint iCount; |
|
40 THuff iLeft; |
|
41 THuff iRight; |
|
42 }; |
|
43 |
|
44 void HuffmanLengthsL(TUint32* aLengths,const TNode* aNodes,TInt aNode,TInt aLen) |
|
45 /** recursive function to calculate the code lengths from the node tree |
|
46 |
|
47 @internal |
|
48 */ |
|
49 { |
|
50 if (++aLen>Huffman::KMaxCodeLength) |
|
51 Panic(EHuffmanBufferOverflow); |
|
52 |
|
53 const TNode& node=aNodes[aNode]; |
|
54 TUint x=node.iLeft; |
|
55 if (x&KLeaf) |
|
56 aLengths[x&~KLeaf]=aLen; |
|
57 else |
|
58 HuffmanLengthsL(aLengths,aNodes,x,aLen); |
|
59 x=node.iRight; |
|
60 if (x&KLeaf) |
|
61 aLengths[x&~KLeaf]=aLen; |
|
62 else |
|
63 HuffmanLengthsL(aLengths,aNodes,x,aLen); |
|
64 } |
|
65 |
|
66 void InsertInOrder(TNode* aNodes, TInt aSize, TUint aCount, TInt aVal) |
|
67 /** Insert the {aCount,aValue} pair into the already sorted array of nodes |
|
68 |
|
69 @internal |
|
70 */ |
|
71 { |
|
72 // Uses Insertion sort following a binary search... |
|
73 TInt l=0, r=aSize; |
|
74 while (l < r) |
|
75 { |
|
76 TInt m = (l+r) >> 1; |
|
77 if (aNodes[m].iCount<aCount) |
|
78 r=m; |
|
79 else |
|
80 l=m+1; |
|
81 } |
|
82 HMem::Copy(aNodes+l+1,aNodes+l,sizeof(TNode)*(aSize-l)); |
|
83 aNodes[l].iCount=aCount; |
|
84 aNodes[l].iRight=TUint16(aVal); |
|
85 } |
|
86 |
|
87 void Huffman::HuffmanL(const TUint32 aFrequency[],TInt aNumCodes,TUint32 aHuffman[]) |
|
88 /** Generate a Huffman code |
|
89 |
|
90 This generates a Huffman code for a given set of code frequencies. The output |
|
91 is a table of code lengths which can be used to build canonincal encoding tables |
|
92 or decoding trees for use with the TBitInput and TBitOutput classes. |
|
93 |
|
94 Entries in the table with a frequency of zero will have a zero code length |
|
95 and thus no associated huffman encoding. If each such symbol should have a |
|
96 maximum length encoding, they must be given at least a frequency of 1. |
|
97 |
|
98 For an alphabet of n symbols, this algorithm has a transient memory overhead |
|
99 of 8n, and a time complexity of O(n*log(n)). |
|
100 |
|
101 @param "const TUint32 aFrequency[]" The table of code frequencies |
|
102 @param "TInt aNumCodes" The number of codes in the table |
|
103 @param "TUint32 aHuffman[]" The table for the output code-length table. This must be |
|
104 the same size as the frequency table, and can safely be the same table |
|
105 |
|
106 @leave "KErrNoMemory" If memory used for code generation cannot be allocated |
|
107 |
|
108 @panic "USER ???" If the number of codes exceeds Huffman::KMaxCodes |
|
109 */ |
|
110 { |
|
111 if(TUint(aNumCodes)>TUint(KMaxCodes)) |
|
112 Panic(EHuffmanTooManyCodes); |
|
113 |
|
114 // Sort the values into decreasing order of frequency |
|
115 // |
|
116 TNode* nodes = new TNode[aNumCodes]; |
|
117 if(nodes==NULL) |
|
118 Panic(EHuffmanOutOfMemory); |
|
119 |
|
120 TInt lCount=0; |
|
121 |
|
122 for (TInt ii=0;ii<aNumCodes;++ii) |
|
123 { |
|
124 TInt c=aFrequency[ii]; |
|
125 if (c!=0) |
|
126 InsertInOrder(nodes,lCount++,c,ii|KLeaf); |
|
127 } |
|
128 |
|
129 // default code length is zero |
|
130 HMem::FillZ(aHuffman,aNumCodes*sizeof(TUint32)); |
|
131 |
|
132 if (lCount==0) |
|
133 { |
|
134 // no codes with frequency>0. No code has a length |
|
135 } |
|
136 else if (lCount==1) |
|
137 { |
|
138 // special case for a single value (always encode as "0") |
|
139 aHuffman[nodes[0].iRight&~KLeaf]=1; |
|
140 } |
|
141 else |
|
142 { |
|
143 // Huffman algorithm: pair off least frequent nodes and reorder |
|
144 // |
|
145 do |
|
146 { |
|
147 --lCount; |
|
148 TUint c=nodes[lCount].iCount + nodes[lCount-1].iCount; |
|
149 nodes[lCount].iLeft=nodes[lCount-1].iRight; |
|
150 // re-order the leaves now to reflect new combined frequency 'c' |
|
151 InsertInOrder(nodes,lCount-1,c,lCount); |
|
152 } while (lCount>1); |
|
153 // generate code lengths in aHuffman[] |
|
154 HuffmanLengthsL(aHuffman,nodes,1,0); |
|
155 } |
|
156 |
|
157 delete [] nodes; |
|
158 |
|
159 if(!IsValid(aHuffman,aNumCodes)) |
|
160 Panic(EHuffmanInvalidCoding); |
|
161 } |
|
162 |
|
163 TBool Huffman::IsValid(const TUint32 aHuffman[],TInt aNumCodes) |
|
164 /** Validate a Huffman encoding |
|
165 |
|
166 This verifies that a Huffman coding described by the code lengths is valid. |
|
167 In particular, it ensures that no code exceeds the maximum length and |
|
168 that it is possible to generate a canonical coding for the specified lengths. |
|
169 |
|
170 @param "const TUint32 aHuffman[]" The table of code lengths as generated by Huffman::HuffmanL() |
|
171 @param "TInt aNumCodes" The number of codes in the table |
|
172 |
|
173 @return True if the code is valid, otherwise false |
|
174 */ |
|
175 { |
|
176 // The code is valid if one of the following holds: |
|
177 // (a) the code exactly fills the 'code space' |
|
178 // (b) there is only a single symbol with code length 1 |
|
179 // (c) there are no encoded symbols |
|
180 // |
|
181 TUint remain=1<<KMaxCodeLength; |
|
182 TInt totlen=0; |
|
183 for (const TUint32* p=aHuffman+aNumCodes; p>aHuffman;) |
|
184 { |
|
185 TInt len=*--p; |
|
186 if (len>0) |
|
187 { |
|
188 totlen+=len; |
|
189 if (len>KMaxCodeLength) |
|
190 return EFalse; |
|
191 TUint c=1<<(KMaxCodeLength-len); |
|
192 if (c>remain) |
|
193 return EFalse; |
|
194 remain-=c; |
|
195 } |
|
196 } |
|
197 |
|
198 return remain==0 || totlen<=1; |
|
199 } |
|
200 |
|
201 void Huffman::Encoding(const TUint32 aHuffman[],TInt aNumCodes,TUint32 aEncodeTable[]) |
|
202 /** Create a canonical Huffman encoding table |
|
203 |
|
204 This generates the huffman codes used by TBitOutput::HuffmanL() to write huffman |
|
205 encoded data. The input is table of code lengths, as generated by Huffman::HuffmanL() |
|
206 and must represent a valid huffman code. |
|
207 |
|
208 @param "const TUint32 aHuffman[]" The table of code lengths as generated by Huffman::HuffmanL() |
|
209 @param "TInt aNumCodes" The number of codes in the table |
|
210 @param "TUint32 aEncodeTable[]" The table for the output huffman codes. This must be |
|
211 the same size as the code-length table, and can safely be the same table |
|
212 |
|
213 @panic "USER ???" If the provided code is not a valid Huffman coding |
|
214 |
|
215 @see IsValid() |
|
216 @see HuffmanL() |
|
217 */ |
|
218 { |
|
219 __ASSERT_ALWAYS(IsValid(aHuffman,aNumCodes),Panic(EHuffmanInvalidCoding)); |
|
220 |
|
221 TFixedArray<TInt,KMaxCodeLength> lenCount; |
|
222 lenCount.Reset(); |
|
223 |
|
224 TInt ii; |
|
225 for (ii=0;ii<aNumCodes;++ii) |
|
226 { |
|
227 TInt len=aHuffman[ii]-1; |
|
228 if (len>=0) |
|
229 ++lenCount[len]; |
|
230 } |
|
231 |
|
232 TFixedArray<TUint,KMaxCodeLength> nextCode; |
|
233 TUint code=0; |
|
234 for (ii=0;ii<KMaxCodeLength;++ii) |
|
235 { |
|
236 code<<=1; |
|
237 nextCode[ii]=code; |
|
238 code+=lenCount[ii]; |
|
239 } |
|
240 |
|
241 for (ii=0;ii<aNumCodes;++ii) |
|
242 { |
|
243 TInt len=aHuffman[ii]; |
|
244 if (len==0) |
|
245 aEncodeTable[ii]=0; |
|
246 else |
|
247 { |
|
248 aEncodeTable[ii] = (nextCode[len-1]<<(KMaxCodeLength-len))|(len<<KMaxCodeLength); |
|
249 ++nextCode[len-1]; |
|
250 } |
|
251 } |
|
252 } |
|
253 |
|
254 /** the encoding table for the externalised code |
|
255 @internal |
|
256 */ |
|
257 const TUint32 HuffmanEncoding[]= |
|
258 { |
|
259 0x10000000, |
|
260 0x1c000000, |
|
261 0x12000000, |
|
262 0x1d000000, |
|
263 0x26000000, |
|
264 0x26800000, |
|
265 0x2f000000, |
|
266 0x37400000, |
|
267 0x37600000, |
|
268 0x37800000, |
|
269 0x3fa00000, |
|
270 0x3fb00000, |
|
271 0x3fc00000, |
|
272 0x3fd00000, |
|
273 0x47e00000, |
|
274 0x47e80000, |
|
275 0x47f00000, |
|
276 0x4ff80000, |
|
277 0x57fc0000, |
|
278 0x5ffe0000, |
|
279 0x67ff0000, |
|
280 0x77ff8000, |
|
281 0x7fffa000, |
|
282 0x7fffb000, |
|
283 0x7fffc000, |
|
284 0x7fffd000, |
|
285 0x7fffe000, |
|
286 0x87fff000, |
|
287 0x87fff800 |
|
288 }; |
|
289 |
|
290 void EncodeRunLengthL(TBitOutput& aOutput, TInt aLength) |
|
291 /** encode 0a as '0' and 0b as '1', return number of symbols created |
|
292 |
|
293 @internal |
|
294 */ |
|
295 { |
|
296 if (aLength>0) |
|
297 { |
|
298 EncodeRunLengthL(aOutput,(aLength-1)>>1); |
|
299 aOutput.HuffmanL(HuffmanEncoding[1-(aLength&1)]); |
|
300 } |
|
301 } |
|
302 |
|
303 void Huffman::ExternalizeL(TBitOutput& aOutput,const TUint32 aHuffman[],TInt aNumCodes) |
|
304 /** Store a canonical huffman encoding in compact form |
|
305 |
|
306 As the encoding is canonical, only the code lengths of each code needs to be saved. |
|
307 |
|
308 Due to the nature of code length tables, these can usually be stored very compactly |
|
309 by encoding the encoding itself, hence the use of the bit output stream. |
|
310 |
|
311 @param "TBitOutput& aOutput" The output stream for the encoding |
|
312 @param "const TUint32 aHuffman[]" The table of code lengths as generated by Huffman::HuffmanL() |
|
313 @param "TInt aNumCodes" The number of huffman codes in the table |
|
314 |
|
315 @leave "TBitOutput::HuffmanL()" |
|
316 */ |
|
317 { |
|
318 // We assume that the code length table is generated by the huffman generator, |
|
319 // in which case the maxmimum code length is 27 bits. |
|
320 // |
|
321 // We apply three transformations to the data: |
|
322 // 1. the data goes through a move-to-front coder |
|
323 // 2. apply a rle-0 coder which replace runs of '0' with streams of '0a' and '0b' |
|
324 // 3. encode the result using a predefined (average) huffman coding |
|
325 // |
|
326 // This can be done in a single pass over the data, avoiding the need for additional |
|
327 // memory. |
|
328 // |
|
329 // initialise the list for the MTF coder |
|
330 TFixedArray<TUint8,Huffman::KMetaCodes> list; |
|
331 TInt i; |
|
332 for (i=0;i<list.Count();++i) |
|
333 list[i]=TUint8(i); |
|
334 TInt last=0; |
|
335 |
|
336 TInt rl=0; |
|
337 const TUint32* p32=aHuffman; |
|
338 const TUint32* e32=p32+aNumCodes; |
|
339 while (p32<e32) |
|
340 { |
|
341 TInt c=*p32++; |
|
342 if (c==last) |
|
343 ++rl; // repeat of last symbol |
|
344 else |
|
345 { |
|
346 // encode run-length |
|
347 EncodeRunLengthL(aOutput,rl); |
|
348 rl=0; |
|
349 // find code in MTF list |
|
350 TInt j; |
|
351 for (j=1;list[j]!=c;++j) |
|
352 ; |
|
353 // store this code |
|
354 aOutput.HuffmanL(HuffmanEncoding[j+1]); |
|
355 // adjust list for MTF algorithm |
|
356 while (--j>0) |
|
357 list[j+1]=list[j]; |
|
358 list[1]=TUint8(last); |
|
359 last=c; |
|
360 } |
|
361 } |
|
362 // encod any remaining run-length |
|
363 EncodeRunLengthL(aOutput,rl); |
|
364 } |
|
365 |
|
366 |
|
367 TBitOutput::TBitOutput() |
|
368 /** Construct a bit stream output object |
|
369 |
|
370 Following construction the bit stream is ready for writing bits, but will first call |
|
371 OverflowL() as the output buffer is 'full'. A derived class can detect this state as |
|
372 Ptr() will return null. |
|
373 */ |
|
374 :iCode(0),iBits(-8),iPtr(0),iEnd(0) |
|
375 {} |
|
376 |
|
377 TBitOutput::TBitOutput(TUint8* aBuf,TInt aSize) |
|
378 /** Construct a bit stream output object over a buffer |
|
379 |
|
380 Data will be written to the buffer until it is full, at which point OverflowL() will |
|
381 be called. This should handle the data and then can Set() again to reset the buffer |
|
382 for further output. |
|
383 |
|
384 @param "TUint8* aBuf" The buffer for output |
|
385 @param "TInt aSize" The size of the buffer in bytes |
|
386 */ |
|
387 :iCode(0),iBits(-8),iPtr(aBuf),iEnd(aBuf+aSize) |
|
388 {} |
|
389 |
|
390 void TBitOutput::HuffmanL(TUint aHuffCode) |
|
391 /** Write a huffman code |
|
392 |
|
393 This expects a huffman code value as generated by Huffman::Encoding() |
|
394 |
|
395 @param "TUint aHuffCode" The huffman code write to the stream |
|
396 |
|
397 @leave "OverflowL()" If the output buffer is full, OverflowL() is called |
|
398 */ |
|
399 { |
|
400 DoWriteL(aHuffCode<<(32-Huffman::KMaxCodeLength),aHuffCode>>Huffman::KMaxCodeLength); |
|
401 } |
|
402 |
|
403 void TBitOutput::WriteL(TUint aValue,TInt aLength) |
|
404 /** Write an arbitrary integer value |
|
405 |
|
406 Write an unsigned integer using the number of bits specified. Only |
|
407 the low order bits of the value are written to the output, most |
|
408 significant bit first. |
|
409 |
|
410 @param "TUint aValue" The value to write to the stream |
|
411 @param "TUint aLength" The number of bits to output |
|
412 |
|
413 @leave "OverflowL()" If the output buffer is full, OverflowL() is called |
|
414 */ |
|
415 { |
|
416 if (aLength) |
|
417 DoWriteL(aValue<<=32-aLength,aLength); |
|
418 } |
|
419 |
|
420 void TBitOutput::PadL(TUint aPadding) |
|
421 /** Pad the bitstream to the next byte boundary |
|
422 |
|
423 Terminate the bitstream by padding the last byte with the requested value. |
|
424 Following this operation the bitstream can continue to be used, the data will |
|
425 start at the next byte. |
|
426 |
|
427 @param "TUint aPadding" The bit value to pad the final byte with |
|
428 |
|
429 @leave "OverflowL()" If the output buffer is full, OverflowL() is called |
|
430 */ |
|
431 { |
|
432 if (iBits>-8) |
|
433 WriteL(aPadding?0xffffffffu:0,-iBits); |
|
434 } |
|
435 |
|
436 void TBitOutput::DoWriteL(TUint aBits,TInt aSize) |
|
437 /** Write the higher order bits to the stream |
|
438 |
|
439 @internal |
|
440 */ |
|
441 { |
|
442 if (aSize>25) |
|
443 { |
|
444 // cannot process >25 bits in a single pass |
|
445 // so do the top 8 bits first |
|
446 ASSERT(aSize<=32); |
|
447 DoWriteL(aBits&0xff000000u,8); |
|
448 aBits<<=8; |
|
449 aSize-=8; |
|
450 } |
|
451 |
|
452 TInt bits=iBits; |
|
453 TUint code=iCode|(aBits>>(bits+8)); |
|
454 bits+=aSize; |
|
455 if (bits>=0) |
|
456 { |
|
457 TUint8* p=iPtr; |
|
458 do |
|
459 { |
|
460 if (p==iEnd) |
|
461 { |
|
462 // run out of buffer space so invoke the overflow handler |
|
463 iPtr=p; |
|
464 OverflowL(); |
|
465 p=iPtr; |
|
466 ASSERT(p!=iEnd); |
|
467 } |
|
468 *p++=TUint8(code>>24); |
|
469 code<<=8; |
|
470 bits-=8; |
|
471 } while (bits>=0); |
|
472 iPtr=p; |
|
473 } |
|
474 iCode=code; |
|
475 iBits=bits; |
|
476 } |
|
477 |
|
478 void TBitOutput::OverflowL() |
|
479 /** Handle a full output buffer |
|
480 |
|
481 This virtual function is called when the output buffer is full. It should deal |
|
482 with the data in the buffer before reseting the buffer using Set(), allowing |
|
483 further data to be written. |
|
484 |
|
485 A derived class can replace this to write the data to a file (for example) |
|
486 before marking the buffer as empty. |
|
487 |
|
488 @leave "KErrOverflow" The default implementation leaves |
|
489 */ |
|
490 { |
|
491 Panic(EHuffmanBufferOverflow); |
|
492 } |