1 /* inftrees.cpp -- generate Huffman trees for efficient decoding |
|
2 * Copyright (C) 1995-2005 Mark Adler |
|
3 * For conditions of distribution and use, see copyright notice in zlib.h |
|
4 */ |
|
5 |
|
6 #include "zutil.h" |
|
7 #include "inftrees.h" |
|
8 |
|
9 #define MAXBITS 15 |
|
10 |
|
11 |
|
12 const char inflate_copyright[] = |
|
13 " inflate 1.2.3 Copyright 1995-2005 Mark Adler "; |
|
14 /* |
|
15 If you use the zlib library in a product, an acknowledgment is welcome |
|
16 in the documentation of your product. If for some reason you cannot |
|
17 include such an acknowledgment, I would appreciate that you keep this |
|
18 copyright string in the executable of your product. |
|
19 */ |
|
20 |
|
21 /* |
|
22 Build a set of tables to decode the provided canonical Huffman code. |
|
23 The code lengths are lens[0..codes-1]. The result starts at *table, |
|
24 whose indices are 0..2^bits-1. work is a writable array of at least |
|
25 lens shorts, which is used as a work area. type is the type of code |
|
26 to be generated, CODES, LENS, or DISTS. On return, zero is success, |
|
27 -1 is an invalid code, and +1 means that ENOUGH isn't enough. table |
|
28 on return points to the next available entry's address. bits is the |
|
29 requested root table index bits, and on return it is the actual root |
|
30 table index bits. It will differ if the request is greater than the |
|
31 longest code or if it is less than the shortest code. |
|
32 */ |
|
33 #ifdef __SYMBIAN32__ |
|
34 int inflate_table(codetype type,unsigned short FAR * lens,unsigned codes, code FAR * FAR * table,unsigned FAR * bits,unsigned short FAR * work) |
|
35 #else |
|
36 int inflate_table(type, lens, codes, table, bits, work) |
|
37 codetype type; |
|
38 unsigned short FAR *lens; |
|
39 unsigned codes; |
|
40 code FAR * FAR *table; |
|
41 unsigned FAR *bits; |
|
42 unsigned short FAR *work; |
|
43 #endif //__SYMBIAN32__ |
|
44 { |
|
45 // Line to stop compiler warning about unused mandatory global variable 'inflate_copyright' |
|
46 char dontCare = inflate_copyright[0]; dontCare = dontCare; |
|
47 |
|
48 unsigned len; /* a code's length in bits */ |
|
49 unsigned sym; /* index of code symbols */ |
|
50 unsigned min, max; /* minimum and maximum code lengths */ |
|
51 unsigned root; /* number of index bits for root table */ |
|
52 unsigned curr; /* number of index bits for current table */ |
|
53 unsigned drop; /* code bits to drop for sub-table */ |
|
54 int left; /* number of prefix codes available */ |
|
55 unsigned used; /* code entries in table used */ |
|
56 unsigned huff; /* Huffman code */ |
|
57 unsigned incr; /* for incrementing code, index */ |
|
58 unsigned fill; /* index for replicating entries */ |
|
59 unsigned low; /* low bits for current root entry */ |
|
60 unsigned mask; /* mask for low root bits */ |
|
61 |
|
62 /* Need to replace "this" variable with "current" as "this" is a reserved |
|
63 * keyword in C++ which is prefectly fine for a c code. As this file |
|
64 * has been changed to C++ "this" needs to be changed. |
|
65 */ |
|
66 # define this current |
|
67 code this; /* table entry for duplication */ |
|
68 code FAR *next; /* next available space in table */ |
|
69 const unsigned short FAR *base; /* base value table to use */ |
|
70 const unsigned short FAR *extra; /* extra bits table to use */ |
|
71 int end; /* use base and extra for symbol > end */ |
|
72 unsigned short count[MAXBITS+1]; /* number of codes of each length */ |
|
73 unsigned short offs[MAXBITS+1]; /* offsets in table for each length */ |
|
74 static const unsigned short lbase[31] = { /* Length codes 257..285 base */ |
|
75 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, |
|
76 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0}; |
|
77 static const unsigned short lext[31] = { /* Length codes 257..285 extra */ |
|
78 16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18, |
|
79 19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 201, 196}; |
|
80 static const unsigned short dbase[32] = { /* Distance codes 0..29 base */ |
|
81 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, |
|
82 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145, |
|
83 8193, 12289, 16385, 24577, 0, 0}; |
|
84 static const unsigned short dext[32] = { /* Distance codes 0..29 extra */ |
|
85 16, 16, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22, |
|
86 23, 23, 24, 24, 25, 25, 26, 26, 27, 27, |
|
87 28, 28, 29, 29, 64, 64}; |
|
88 |
|
89 /* |
|
90 Process a set of code lengths to create a canonical Huffman code. The |
|
91 code lengths are lens[0..codes-1]. Each length corresponds to the |
|
92 symbols 0..codes-1. The Huffman code is generated by first sorting the |
|
93 symbols by length from short to long, and retaining the symbol order |
|
94 for codes with equal lengths. Then the code starts with all zero bits |
|
95 for the first code of the shortest length, and the codes are integer |
|
96 increments for the same length, and zeros are appended as the length |
|
97 increases. For the deflate format, these bits are stored backwards |
|
98 from their more natural integer increment ordering, and so when the |
|
99 decoding tables are built in the large loop below, the integer codes |
|
100 are incremented backwards. |
|
101 |
|
102 This routine assumes, but does not check, that all of the entries in |
|
103 lens[] are in the range 0..MAXBITS. The caller must assure this. |
|
104 1..MAXBITS is interpreted as that code length. zero means that that |
|
105 symbol does not occur in this code. |
|
106 |
|
107 The codes are sorted by computing a count of codes for each length, |
|
108 creating from that a table of starting indices for each length in the |
|
109 sorted table, and then entering the symbols in order in the sorted |
|
110 table. The sorted table is work[], with that space being provided by |
|
111 the caller. |
|
112 |
|
113 The length counts are used for other purposes as well, i.e. finding |
|
114 the minimum and maximum length codes, determining if there are any |
|
115 codes at all, checking for a valid set of lengths, and looking ahead |
|
116 at length counts to determine sub-table sizes when building the |
|
117 decoding tables. |
|
118 */ |
|
119 |
|
120 /* accumulate lengths for codes (assumes lens[] all in 0..MAXBITS) */ |
|
121 for (len = 0; len <= MAXBITS; len++) |
|
122 count[len] = 0; |
|
123 for (sym = 0; sym < codes; sym++) |
|
124 count[lens[sym]]++; |
|
125 |
|
126 /* bound code lengths, force root to be within code lengths */ |
|
127 root = *bits; |
|
128 for (max = MAXBITS; max >= 1; max--) |
|
129 if (count[max] != 0) break; |
|
130 if (root > max) root = max; |
|
131 if (max == 0) { /* no symbols to code at all */ |
|
132 this.op = (unsigned char)64; /* invalid code marker */ |
|
133 this.bits = (unsigned char)1; |
|
134 this.val = (unsigned short)0; |
|
135 *(*table)++ = this; /* make a table to force an error */ |
|
136 *(*table)++ = this; |
|
137 *bits = 1; |
|
138 return 0; /* no symbols, but wait for decoding to report error */ |
|
139 } |
|
140 for (min = 1; min <= MAXBITS; min++) |
|
141 if (count[min] != 0) break; |
|
142 if (root < min) root = min; |
|
143 |
|
144 /* check for an over-subscribed or incomplete set of lengths */ |
|
145 left = 1; |
|
146 for (len = 1; len <= MAXBITS; len++) { |
|
147 left <<= 1; |
|
148 left -= count[len]; |
|
149 if (left < 0) return -1; /* over-subscribed */ |
|
150 } |
|
151 if (left > 0 && (type == CODES || max != 1)) |
|
152 return -1; /* incomplete set */ |
|
153 |
|
154 /* generate offsets into symbol table for each length for sorting */ |
|
155 offs[1] = 0; |
|
156 for (len = 1; len < MAXBITS; len++) |
|
157 offs[len + 1] = offs[len] + count[len]; |
|
158 |
|
159 /* sort symbols by length, by symbol order within each length */ |
|
160 for (sym = 0; sym < codes; sym++) |
|
161 if (lens[sym] != 0) work[offs[lens[sym]]++] = (unsigned short)sym; |
|
162 |
|
163 /* |
|
164 Create and fill in decoding tables. In this loop, the table being |
|
165 filled is at next and has curr index bits. The code being used is huff |
|
166 with length len. That code is converted to an index by dropping drop |
|
167 bits off of the bottom. For codes where len is less than drop + curr, |
|
168 those top drop + curr - len bits are incremented through all values to |
|
169 fill the table with replicated entries. |
|
170 |
|
171 root is the number of index bits for the root table. When len exceeds |
|
172 root, sub-tables are created pointed to by the root entry with an index |
|
173 of the low root bits of huff. This is saved in low to check for when a |
|
174 new sub-table should be started. drop is zero when the root table is |
|
175 being filled, and drop is root when sub-tables are being filled. |
|
176 |
|
177 When a new sub-table is needed, it is necessary to look ahead in the |
|
178 code lengths to determine what size sub-table is needed. The length |
|
179 counts are used for this, and so count[] is decremented as codes are |
|
180 entered in the tables. |
|
181 |
|
182 used keeps track of how many table entries have been allocated from the |
|
183 provided *table space. It is checked when a LENS table is being made |
|
184 against the space in *table, ENOUGH, minus the maximum space needed by |
|
185 the worst case distance code, MAXD. This should never happen, but the |
|
186 sufficiency of ENOUGH has not been proven exhaustively, hence the check. |
|
187 This assumes that when type == LENS, bits == 9. |
|
188 |
|
189 sym increments through all symbols, and the loop terminates when |
|
190 all codes of length max, i.e. all codes, have been processed. This |
|
191 routine permits incomplete codes, so another loop after this one fills |
|
192 in the rest of the decoding tables with invalid code markers. |
|
193 */ |
|
194 |
|
195 /* set up for code type */ |
|
196 switch (type) { |
|
197 case CODES: |
|
198 base = extra = work; /* dummy value--not used */ |
|
199 end = 19; |
|
200 break; |
|
201 case LENS: |
|
202 base = lbase; |
|
203 base -= 257; |
|
204 extra = lext; |
|
205 extra -= 257; |
|
206 end = 256; |
|
207 break; |
|
208 default: /* DISTS */ |
|
209 base = dbase; |
|
210 extra = dext; |
|
211 end = -1; |
|
212 } |
|
213 |
|
214 /* initialize state for loop */ |
|
215 huff = 0; /* starting code */ |
|
216 sym = 0; /* starting code symbol */ |
|
217 len = min; /* starting code length */ |
|
218 next = *table; /* current table to fill in */ |
|
219 curr = root; /* current table index bits */ |
|
220 drop = 0; /* current bits to drop from code for index */ |
|
221 low = (unsigned)(-1); /* trigger new sub-table when len > root */ |
|
222 used = 1U << root; /* use root table entries */ |
|
223 mask = used - 1; /* mask for comparing low */ |
|
224 |
|
225 /* check available table space */ |
|
226 if (type == LENS && used >= ENOUGH - MAXD) |
|
227 return 1; |
|
228 |
|
229 /* process all codes and make table entries */ |
|
230 for (;;) { |
|
231 /* create table entry */ |
|
232 this.bits = (unsigned char)(len - drop); |
|
233 if ((int)(work[sym]) < end) { |
|
234 this.op = (unsigned char)0; |
|
235 this.val = work[sym]; |
|
236 } |
|
237 else if ((int)(work[sym]) > end) { |
|
238 this.op = (unsigned char)(extra[work[sym]]); |
|
239 this.val = base[work[sym]]; |
|
240 } |
|
241 else { |
|
242 this.op = (unsigned char)(32 + 64); /* end of block */ |
|
243 this.val = 0; |
|
244 } |
|
245 |
|
246 /* replicate for those indices with low len bits equal to huff */ |
|
247 incr = 1U << (len - drop); |
|
248 fill = 1U << curr; |
|
249 min = fill; /* save offset to next table */ |
|
250 do { |
|
251 fill -= incr; |
|
252 next[(huff >> drop) + fill] = this; |
|
253 } while (fill != 0); |
|
254 |
|
255 /* backwards increment the len-bit code huff */ |
|
256 incr = 1U << (len - 1); |
|
257 while (huff & incr) |
|
258 incr >>= 1; |
|
259 if (incr != 0) { |
|
260 huff &= incr - 1; |
|
261 huff += incr; |
|
262 } |
|
263 else |
|
264 huff = 0; |
|
265 |
|
266 /* go to next symbol, update count, len */ |
|
267 sym++; |
|
268 if (--(count[len]) == 0) { |
|
269 if (len == max) break; |
|
270 len = lens[work[sym]]; |
|
271 } |
|
272 |
|
273 /* create new sub-table if needed */ |
|
274 if (len > root && (huff & mask) != low) { |
|
275 /* if first time, transition to sub-tables */ |
|
276 if (drop == 0) |
|
277 drop = root; |
|
278 |
|
279 /* increment past last table */ |
|
280 next += min; /* here min is 1 << curr */ |
|
281 |
|
282 /* determine length of next table */ |
|
283 curr = len - drop; |
|
284 left = (int)(1 << curr); |
|
285 while (curr + drop < max) { |
|
286 left -= count[curr + drop]; |
|
287 if (left <= 0) break; |
|
288 curr++; |
|
289 left <<= 1; |
|
290 } |
|
291 |
|
292 /* check for enough space */ |
|
293 used += 1U << curr; |
|
294 if (type == LENS && used >= ENOUGH - MAXD) |
|
295 return 1; |
|
296 |
|
297 /* point entry in root table to sub-table */ |
|
298 low = huff & mask; |
|
299 (*table)[low].op = (unsigned char)curr; |
|
300 (*table)[low].bits = (unsigned char)root; |
|
301 (*table)[low].val = (unsigned short)(next - *table); |
|
302 } |
|
303 } |
|
304 |
|
305 /* |
|
306 Fill in rest of table for incomplete codes. This loop is similar to the |
|
307 loop above in incrementing huff for table indices. It is assumed that |
|
308 len is equal to curr + drop, so there is no loop needed to increment |
|
309 through high index bits. When the current sub-table is filled, the loop |
|
310 drops back to the root table to fill in any remaining entries there. |
|
311 */ |
|
312 this.op = (unsigned char)64; /* invalid code marker */ |
|
313 this.bits = (unsigned char)(len - drop); |
|
314 this.val = (unsigned short)0; |
|
315 while (huff != 0) { |
|
316 /* when done with sub-table, drop back to root table */ |
|
317 if (drop != 0 && (huff & mask) != low) { |
|
318 drop = 0; |
|
319 len = root; |
|
320 next = *table; |
|
321 this.bits = (unsigned char)len; |
|
322 } |
|
323 |
|
324 /* put invalid code marker in table */ |
|
325 next[huff >> drop] = this; |
|
326 |
|
327 /* backwards increment the len-bit code huff */ |
|
328 incr = 1U << (len - 1); |
|
329 while (huff & incr) |
|
330 incr >>= 1; |
|
331 if (incr != 0) { |
|
332 huff &= incr - 1; |
|
333 huff += incr; |
|
334 } |
|
335 else |
|
336 huff = 0; |
|
337 } |
|
338 |
|
339 /* set return parameters */ |
|
340 *table += used; |
|
341 *bits = root; |
|
342 return 0; |
|
343 } |
|
344 |
|
345 |
|
346 |
|
347 |
|
348 |
|
349 |
|