|
1 /* |
|
2 * Copyright (c) 1998-2009 Nokia Corporation and/or its subsidiary(-ies). |
|
3 * All rights reserved. |
|
4 * This component and the accompanying materials are made available |
|
5 * under the terms of the License "Eclipse Public License v1.0" |
|
6 * which accompanies this distribution, and is available |
|
7 * at the URL "http://www.eclipse.org/legal/epl-v10.html". |
|
8 * |
|
9 * Initial Contributors: |
|
10 * Nokia Corporation - initial contribution. |
|
11 * |
|
12 * Contributors: |
|
13 * |
|
14 * Description: |
|
15 * e32tools\petran\Szip\decode.cpp |
|
16 * |
|
17 */ |
|
18 |
|
19 |
|
20 #include "huffman.h" |
|
21 #include "panic.h" |
|
22 #include <cpudefs.h> |
|
23 #include "h_utl.h" |
|
24 #include "farray.h" |
|
25 |
|
26 |
|
27 const TInt KHuffTerminate=0x0001; |
|
28 const TUint32 KBranch1=sizeof(TUint32)<<16; |
|
29 |
|
30 TUint32* HuffmanSubTree(TUint32* aPtr,const TUint32* aValue,TUint32** aLevel) |
|
31 // |
|
32 // write the subtree below aPtr and return the head |
|
33 // |
|
34 { |
|
35 TUint32* l=*aLevel++; |
|
36 if (l>aValue) |
|
37 { |
|
38 TUint32* sub0=HuffmanSubTree(aPtr,aValue,aLevel); // 0-tree first |
|
39 aPtr=HuffmanSubTree(sub0,aValue-(aPtr-sub0)-1,aLevel); // 1-tree |
|
40 TInt branch0=(TUint8*)sub0-(TUint8*)(aPtr-1); |
|
41 *--aPtr=KBranch1|branch0; |
|
42 } |
|
43 else if (l==aValue) |
|
44 { |
|
45 TUint term0=*aValue--; // 0-term |
|
46 aPtr=HuffmanSubTree(aPtr,aValue,aLevel); // 1-tree |
|
47 *--aPtr=KBranch1|(term0>>16); |
|
48 } |
|
49 else // l<iNext |
|
50 { |
|
51 TUint term0=*aValue--; // 0-term |
|
52 TUint term1=*aValue--; |
|
53 *--aPtr=(term1>>16<<16)|(term0>>16); |
|
54 } |
|
55 return aPtr; |
|
56 } |
|
57 |
|
58 void Huffman::Decoding(const TUint32 aHuffman[],TInt aNumCodes,TUint32 aDecodeTree[],TInt aSymbolBase) |
|
59 /** Create a canonical Huffman decoding tree |
|
60 |
|
61 This generates the huffman decoding tree used by TBitInput::HuffmanL() to read huffman |
|
62 encoded data. The input is table of code lengths, as generated by Huffman::HuffmanL() |
|
63 and must represent a valid huffman code. |
|
64 |
|
65 @param "const TUint32 aHuffman[]" The table of code lengths as generated by Huffman::HuffmanL() |
|
66 @param "TInt aNumCodes" The number of codes in the table |
|
67 @param "TUint32 aDecodeTree[]" The space for the decoding tree. This must be the same |
|
68 size as the code-length table, and can safely be the same memory |
|
69 @param "TInt aSymbolBase" the base value for the output 'symbols' from the decoding tree, by default |
|
70 this is zero. |
|
71 |
|
72 @panic "USER ???" If the provided code is not a valid Huffman coding |
|
73 |
|
74 @see IsValid() |
|
75 @see HuffmanL() |
|
76 */ |
|
77 { |
|
78 if(!IsValid(aHuffman,aNumCodes)) |
|
79 Panic(EHuffmanInvalidCoding); |
|
80 // |
|
81 TFixedArray<TInt,KMaxCodeLength> counts; |
|
82 counts.Reset(); |
|
83 TInt codes=0; |
|
84 TInt ii; |
|
85 for (ii=0;ii<aNumCodes;++ii) |
|
86 { |
|
87 TInt len=aHuffman[ii]; |
|
88 aDecodeTree[ii]=len; |
|
89 if (--len>=0) |
|
90 { |
|
91 ++counts[len]; |
|
92 ++codes; |
|
93 } |
|
94 } |
|
95 // |
|
96 TFixedArray<TUint32*,KMaxCodeLength> level; |
|
97 TUint32* lit=aDecodeTree+codes; |
|
98 for (ii=0;ii<KMaxCodeLength;++ii) |
|
99 { |
|
100 level[ii]=lit; |
|
101 lit-=counts[ii]; |
|
102 } |
|
103 aSymbolBase=(aSymbolBase<<17)+(KHuffTerminate<<16); |
|
104 for (ii=0;ii<aNumCodes;++ii) |
|
105 { |
|
106 TUint len=TUint8(aDecodeTree[ii]); |
|
107 if (len) |
|
108 *--level[len-1]|=(ii<<17)+aSymbolBase; |
|
109 } |
|
110 if (codes==1) // codes==1 special case: the tree is not complete |
|
111 { |
|
112 TUint term=aDecodeTree[0]>>16; |
|
113 aDecodeTree[0]=term|(term<<16); // 0- and 1-terminate at root |
|
114 } |
|
115 else if (codes>1) |
|
116 HuffmanSubTree(aDecodeTree+codes-1,aDecodeTree+codes-1,&level[0]); |
|
117 } |
|
118 |
|
119 // The decoding tree for the externalised code |
|
120 const TUint32 HuffmanDecoding[]= |
|
121 { |
|
122 0x0004006c, |
|
123 0x00040064, |
|
124 0x0004005c, |
|
125 0x00040050, |
|
126 0x00040044, |
|
127 0x0004003c, |
|
128 0x00040034, |
|
129 0x00040021, |
|
130 0x00040023, |
|
131 0x00040025, |
|
132 0x00040027, |
|
133 0x00040029, |
|
134 0x00040014, |
|
135 0x0004000c, |
|
136 0x00040035, |
|
137 0x00390037, |
|
138 0x00330031, |
|
139 0x0004002b, |
|
140 0x002f002d, |
|
141 0x001f001d, |
|
142 0x001b0019, |
|
143 0x00040013, |
|
144 0x00170015, |
|
145 0x0004000d, |
|
146 0x0011000f, |
|
147 0x000b0009, |
|
148 0x00070003, |
|
149 0x00050001 |
|
150 }; |
|
151 |
|
152 void Huffman::InternalizeL(TBitInput& aInput,TUint32 aHuffman[],TInt aNumCodes) |
|
153 /** Restore a canonical huffman encoding from a bit stream |
|
154 |
|
155 The encoding must have been stored using Huffman::ExternalizeL(). The resulting |
|
156 code-length table can be used to create an encoding table using Huffman::Encoding() |
|
157 or a decoding tree using Huffman::Decoding(). |
|
158 |
|
159 @param "TBitInput& aInput" The input stream with the encoding |
|
160 @param "TUint32 aHuffman[]" The internalized code-length table is placed here |
|
161 @param "TInt aNumCodes" The number of huffman codes in the table |
|
162 |
|
163 @leave "TBitInput::HuffmanL()" |
|
164 |
|
165 @see ExternalizeL() |
|
166 */ |
|
167 // See ExternalizeL for a description of the format |
|
168 { |
|
169 // initialise move-to-front list |
|
170 TFixedArray<TUint8,Huffman::KMetaCodes> list; |
|
171 for (TInt i=0;i<list.Count();++i) |
|
172 list[i]=TUint8(i); |
|
173 TInt last=0; |
|
174 // extract codes, reverse rle-0 and mtf encoding in one pass |
|
175 TUint32* p=aHuffman; |
|
176 const TUint32* end=aHuffman+aNumCodes; |
|
177 TInt rl=0; |
|
178 while (p+rl<end) |
|
179 { |
|
180 TInt c=aInput.HuffmanL(HuffmanDecoding); |
|
181 if (c<2) |
|
182 { |
|
183 // one of the zero codes used by RLE-0 |
|
184 // update he run-length |
|
185 rl+=rl+c+1; |
|
186 } |
|
187 else |
|
188 { |
|
189 while (rl>0) |
|
190 { |
|
191 if (p>end) |
|
192 { |
|
193 Panic(EHuffmanCorruptFile); |
|
194 } |
|
195 *p++=last; |
|
196 --rl; |
|
197 } |
|
198 --c; |
|
199 list[0]=TUint8(last); |
|
200 last=list[c]; |
|
201 HMem::Copy(&list[1],&list[0],c); |
|
202 if (p>end) |
|
203 { |
|
204 Panic(EHuffmanCorruptFile); |
|
205 } |
|
206 *p++=last; |
|
207 } |
|
208 } |
|
209 while (rl>0) |
|
210 { |
|
211 if (p>end) |
|
212 { |
|
213 Panic(EHuffmanCorruptFile); |
|
214 } |
|
215 *p++=last; |
|
216 --rl; |
|
217 } |
|
218 } |
|
219 |
|
220 // bit-stream input class |
|
221 |
|
222 inline TUint reverse(TUint aVal) |
|
223 // |
|
224 // Reverse the byte-order of a 32 bit value |
|
225 // This generates optimal ARM code (4 instructions) |
|
226 // |
|
227 { |
|
228 TUint v=(aVal<<16)|(aVal>>16); |
|
229 v^=aVal; |
|
230 v&=0xff00ffff; |
|
231 aVal=(aVal>>8)|(aVal<<24); |
|
232 return aVal^(v>>8); |
|
233 } |
|
234 |
|
235 TBitInput::TBitInput() |
|
236 /** Construct a bit stream input object |
|
237 |
|
238 Following construction the bit stream is ready for reading bits, but will |
|
239 immediately call UnderflowL() as the input buffer is empty. |
|
240 */ |
|
241 :iCount(0),iRemain(0) |
|
242 {} |
|
243 |
|
244 TBitInput::TBitInput(const TUint8* aPtr, TInt aLength, TInt aOffset) |
|
245 /** Construct a bit stream input object over a buffer |
|
246 |
|
247 Following construction the bit stream is ready for reading bits from |
|
248 the specified buffer. |
|
249 |
|
250 @param "const TUint8* aPtr" The address of the buffer containing the bit stream |
|
251 @param "TInt aLength" The length of the bitstream in bits |
|
252 @param "TInt aOffset" The bit offset from the start of the buffer to the bit stream (defaults to zero) |
|
253 */ |
|
254 { |
|
255 Set(aPtr,aLength,aOffset); |
|
256 } |
|
257 |
|
258 void TBitInput::Set(const TUint8* aPtr, TInt aLength, TInt aOffset) |
|
259 /** Set the memory buffer to use for input |
|
260 |
|
261 Bits will be read from this buffer until it is empty, at which point |
|
262 UnderflowL() will be called. |
|
263 |
|
264 @param "const TUint8* aPtr" The address of the buffer containing the bit stream |
|
265 @param "TInt aLength" The length of the bitstream in bits |
|
266 @param "TInt aOffset" The bit offset from the start of the buffer to the bit stream (defaults to zero) |
|
267 */ |
|
268 { |
|
269 TUint p=(TUint)aPtr; |
|
270 p+=aOffset>>3; // nearest byte to the specified bit offset |
|
271 aOffset&=7; // bit offset within the byte |
|
272 const TUint32* ptr=(const TUint32*)(p&~3); // word containing this byte |
|
273 aOffset+=(p&3)<<3; // bit offset within the word |
|
274 if (aLength==0) |
|
275 iCount=0; |
|
276 else |
|
277 { |
|
278 // read the first few bits of the stream |
|
279 iBits=reverse(*ptr++)<<aOffset; |
|
280 aOffset=32-aOffset; |
|
281 aLength-=aOffset; |
|
282 if (aLength<0) |
|
283 aOffset+=aLength; |
|
284 iCount=aOffset; |
|
285 } |
|
286 iRemain=aLength; |
|
287 iPtr=ptr; |
|
288 } |
|
289 |
|
290 #ifndef __HUFFMAN_MACHINE_CODED__ |
|
291 |
|
292 TUint TBitInput::ReadL() |
|
293 /** Read a single bit from the input |
|
294 |
|
295 Return the next bit in the input stream. This will call UnderflowL() if |
|
296 there are no more bits available. |
|
297 |
|
298 @return The next bit in the stream |
|
299 |
|
300 @leave "UnderflowL()" It the bit stream is exhausted more UnderflowL is called |
|
301 to get more data |
|
302 */ |
|
303 { |
|
304 TInt c=iCount; |
|
305 TUint bits=iBits; |
|
306 if (--c<0) |
|
307 return ReadL(1); |
|
308 iCount=c; |
|
309 iBits=bits<<1; |
|
310 return bits>>31; |
|
311 } |
|
312 |
|
313 TUint TBitInput::ReadL(TInt aSize) |
|
314 /** Read a multi-bit value from the input |
|
315 |
|
316 Return the next few bits as an unsigned integer. The last bit read is |
|
317 the least significant bit of the returned value, and the value is |
|
318 zero extended to return a 32-bit result. |
|
319 |
|
320 A read of zero bits will always reaturn zero. |
|
321 |
|
322 This will call UnderflowL() if there are not enough bits available. |
|
323 |
|
324 @param "TInt aSize" The number of bits to read |
|
325 |
|
326 @return The bits read from the stream |
|
327 |
|
328 @leave "UnderflowL()" It the bit stream is exhausted more UnderflowL is called |
|
329 to get more data |
|
330 */ |
|
331 { |
|
332 if (!aSize) |
|
333 return 0; |
|
334 TUint val=0; |
|
335 TUint bits=iBits; |
|
336 iCount-=aSize; |
|
337 while (iCount<0) |
|
338 { |
|
339 // need more bits |
|
340 #ifdef __CPU_X86 |
|
341 // X86 does not allow shift-by-32 |
|
342 if (iCount+aSize!=0) |
|
343 val|=bits>>(32-(iCount+aSize))<<(-iCount); // scrub low order bits |
|
344 #else |
|
345 val|=bits>>(32-(iCount+aSize))<<(-iCount); // scrub low order bits |
|
346 #endif |
|
347 aSize=-iCount; // bits still required |
|
348 if (iRemain>0) |
|
349 { |
|
350 bits=reverse(*iPtr++); |
|
351 iCount+=32; |
|
352 iRemain-=32; |
|
353 if (iRemain<0) |
|
354 iCount+=iRemain; |
|
355 } |
|
356 else |
|
357 { |
|
358 UnderflowL(); |
|
359 bits=iBits; |
|
360 iCount-=aSize; |
|
361 } |
|
362 } |
|
363 #ifdef __CPU_X86 |
|
364 // X86 does not allow shift-by-32 |
|
365 iBits=aSize==32?0:bits<<aSize; |
|
366 #else |
|
367 iBits=bits<<aSize; |
|
368 #endif |
|
369 return val|(bits>>(32-aSize)); |
|
370 } |
|
371 |
|
372 TUint TBitInput::HuffmanL(const TUint32* aTree) |
|
373 /** Read and decode a Huffman Code |
|
374 |
|
375 Interpret the next bits in the input as a Huffman code in the specified |
|
376 decoding. The decoding tree should be the output from Huffman::Decoding(). |
|
377 |
|
378 @param "const TUint32* aTree" The huffman decoding tree |
|
379 |
|
380 @return The symbol that was decoded |
|
381 |
|
382 @leave "UnderflowL()" It the bit stream is exhausted more UnderflowL is called |
|
383 to get more data |
|
384 */ |
|
385 { |
|
386 TUint huff=0; |
|
387 do |
|
388 { |
|
389 aTree=PtrAdd(aTree,huff>>16); |
|
390 huff=*aTree; |
|
391 if (ReadL()==0) |
|
392 huff<<=16; |
|
393 } while ((huff&0x10000u)==0); |
|
394 return huff>>17; |
|
395 } |
|
396 |
|
397 #endif |
|
398 |
|
399 void TBitInput::UnderflowL() |
|
400 /** Handle an empty input buffer |
|
401 |
|
402 This virtual function is called when the input buffer is empty and |
|
403 more bits are required. It should reset the input buffer with more |
|
404 data using Set(). |
|
405 |
|
406 A derived class can replace this to read the data from a file |
|
407 (for example) before reseting the input buffer. |
|
408 |
|
409 @leave "KErrUnderflow" The default implementation leaves |
|
410 */ |
|
411 { |
|
412 Panic(EHuffmanBufferOverflow); |
|
413 } |
|
414 |