cc68a136 |
1 | /* inftrees.c -- 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 | const char inflate_copyright[] = |
12 | " inflate 1.2.3 Copyright 1995-2005 Mark Adler "; |
13 | /* |
14 | If you use the zlib library in a product, an acknowledgment is welcome |
15 | in the documentation of your product. If for some reason you cannot |
16 | include such an acknowledgment, I would appreciate that you keep this |
17 | copyright string in the executable of your product. |
18 | */ |
19 | |
20 | /* |
21 | Build a set of tables to decode the provided canonical Huffman code. |
22 | The code lengths are lens[0..codes-1]. The result starts at *table, |
23 | whose indices are 0..2^bits-1. work is a writable array of at least |
24 | lens shorts, which is used as a work area. type is the type of code |
25 | to be generated, CODES, LENS, or DISTS. On return, zero is success, |
26 | -1 is an invalid code, and +1 means that ENOUGH isn't enough. table |
27 | on return points to the next available entry's address. bits is the |
28 | requested root table index bits, and on return it is the actual root |
29 | table index bits. It will differ if the request is greater than the |
30 | longest code or if it is less than the shortest code. |
31 | */ |
32 | int inflate_table(type, lens, codes, table, bits, work) |
33 | codetype type; |
34 | unsigned short FAR *lens; |
35 | unsigned codes; |
36 | code FAR * FAR *table; |
37 | unsigned FAR *bits; |
38 | unsigned short FAR *work; |
39 | { |
40 | unsigned len; /* a code's length in bits */ |
41 | unsigned sym; /* index of code symbols */ |
42 | unsigned min, max; /* minimum and maximum code lengths */ |
43 | unsigned root; /* number of index bits for root table */ |
44 | unsigned curr; /* number of index bits for current table */ |
45 | unsigned drop; /* code bits to drop for sub-table */ |
46 | int left; /* number of prefix codes available */ |
47 | unsigned used; /* code entries in table used */ |
48 | unsigned huff; /* Huffman code */ |
49 | unsigned incr; /* for incrementing code, index */ |
50 | unsigned fill; /* index for replicating entries */ |
51 | unsigned low; /* low bits for current root entry */ |
52 | unsigned mask; /* mask for low root bits */ |
53 | code this; /* table entry for duplication */ |
54 | code FAR *next; /* next available space in table */ |
55 | const unsigned short FAR *base; /* base value table to use */ |
56 | const unsigned short FAR *extra; /* extra bits table to use */ |
57 | int end; /* use base and extra for symbol > end */ |
58 | unsigned short count[MAXBITS+1]; /* number of codes of each length */ |
59 | unsigned short offs[MAXBITS+1]; /* offsets in table for each length */ |
60 | static const unsigned short lbase[31] = { /* Length codes 257..285 base */ |
61 | 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, |
62 | 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0}; |
63 | static const unsigned short lext[31] = { /* Length codes 257..285 extra */ |
64 | 16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18, |
65 | 19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 201, 196}; |
66 | static const unsigned short dbase[32] = { /* Distance codes 0..29 base */ |
67 | 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, |
68 | 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145, |
69 | 8193, 12289, 16385, 24577, 0, 0}; |
70 | static const unsigned short dext[32] = { /* Distance codes 0..29 extra */ |
71 | 16, 16, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22, |
72 | 23, 23, 24, 24, 25, 25, 26, 26, 27, 27, |
73 | 28, 28, 29, 29, 64, 64}; |
74 | |
75 | /* |
76 | Process a set of code lengths to create a canonical Huffman code. The |
77 | code lengths are lens[0..codes-1]. Each length corresponds to the |
78 | symbols 0..codes-1. The Huffman code is generated by first sorting the |
79 | symbols by length from short to long, and retaining the symbol order |
80 | for codes with equal lengths. Then the code starts with all zero bits |
81 | for the first code of the shortest length, and the codes are integer |
82 | increments for the same length, and zeros are appended as the length |
83 | increases. For the deflate format, these bits are stored backwards |
84 | from their more natural integer increment ordering, and so when the |
85 | decoding tables are built in the large loop below, the integer codes |
86 | are incremented backwards. |
87 | |
88 | This routine assumes, but does not check, that all of the entries in |
89 | lens[] are in the range 0..MAXBITS. The caller must assure this. |
90 | 1..MAXBITS is interpreted as that code length. zero means that that |
91 | symbol does not occur in this code. |
92 | |
93 | The codes are sorted by computing a count of codes for each length, |
94 | creating from that a table of starting indices for each length in the |
95 | sorted table, and then entering the symbols in order in the sorted |
96 | table. The sorted table is work[], with that space being provided by |
97 | the caller. |
98 | |
99 | The length counts are used for other purposes as well, i.e. finding |
100 | the minimum and maximum length codes, determining if there are any |
101 | codes at all, checking for a valid set of lengths, and looking ahead |
102 | at length counts to determine sub-table sizes when building the |
103 | decoding tables. |
104 | */ |
105 | |
106 | /* accumulate lengths for codes (assumes lens[] all in 0..MAXBITS) */ |
107 | for (len = 0; len <= MAXBITS; len++) |
108 | count[len] = 0; |
109 | for (sym = 0; sym < codes; sym++) |
110 | count[lens[sym]]++; |
111 | |
112 | /* bound code lengths, force root to be within code lengths */ |
113 | root = *bits; |
114 | for (max = MAXBITS; max >= 1; max--) |
115 | if (count[max] != 0) break; |
116 | if (root > max) root = max; |
117 | if (max == 0) { /* no symbols to code at all */ |
118 | this.op = (unsigned char)64; /* invalid code marker */ |
119 | this.bits = (unsigned char)1; |
120 | this.val = (unsigned short)0; |
121 | *(*table)++ = this; /* make a table to force an error */ |
122 | *(*table)++ = this; |
123 | *bits = 1; |
124 | return 0; /* no symbols, but wait for decoding to report error */ |
125 | } |
126 | for (min = 1; min <= MAXBITS; min++) |
127 | if (count[min] != 0) break; |
128 | if (root < min) root = min; |
129 | |
130 | /* check for an over-subscribed or incomplete set of lengths */ |
131 | left = 1; |
132 | for (len = 1; len <= MAXBITS; len++) { |
133 | left <<= 1; |
134 | left -= count[len]; |
135 | if (left < 0) return -1; /* over-subscribed */ |
136 | } |
137 | if (left > 0 && (type == CODES || max != 1)) |
138 | return -1; /* incomplete set */ |
139 | |
140 | /* generate offsets into symbol table for each length for sorting */ |
141 | offs[1] = 0; |
142 | for (len = 1; len < MAXBITS; len++) |
143 | offs[len + 1] = offs[len] + count[len]; |
144 | |
145 | /* sort symbols by length, by symbol order within each length */ |
146 | for (sym = 0; sym < codes; sym++) |
147 | if (lens[sym] != 0) work[offs[lens[sym]]++] = (unsigned short)sym; |
148 | |
149 | /* |
150 | Create and fill in decoding tables. In this loop, the table being |
151 | filled is at next and has curr index bits. The code being used is huff |
152 | with length len. That code is converted to an index by dropping drop |
153 | bits off of the bottom. For codes where len is less than drop + curr, |
154 | those top drop + curr - len bits are incremented through all values to |
155 | fill the table with replicated entries. |
156 | |
157 | root is the number of index bits for the root table. When len exceeds |
158 | root, sub-tables are created pointed to by the root entry with an index |
159 | of the low root bits of huff. This is saved in low to check for when a |
160 | new sub-table should be started. drop is zero when the root table is |
161 | being filled, and drop is root when sub-tables are being filled. |
162 | |
163 | When a new sub-table is needed, it is necessary to look ahead in the |
164 | code lengths to determine what size sub-table is needed. The length |
165 | counts are used for this, and so count[] is decremented as codes are |
166 | entered in the tables. |
167 | |
168 | used keeps track of how many table entries have been allocated from the |
169 | provided *table space. It is checked when a LENS table is being made |
170 | against the space in *table, ENOUGH, minus the maximum space needed by |
171 | the worst case distance code, MAXD. This should never happen, but the |
172 | sufficiency of ENOUGH has not been proven exhaustively, hence the check. |
173 | This assumes that when type == LENS, bits == 9. |
174 | |
175 | sym increments through all symbols, and the loop terminates when |
176 | all codes of length max, i.e. all codes, have been processed. This |
177 | routine permits incomplete codes, so another loop after this one fills |
178 | in the rest of the decoding tables with invalid code markers. |
179 | */ |
180 | |
181 | /* set up for code type */ |
182 | switch (type) { |
183 | case CODES: |
184 | base = extra = work; /* dummy value--not used */ |
185 | end = 19; |
186 | break; |
187 | case LENS: |
188 | base = lbase; |
189 | base -= 257; |
190 | extra = lext; |
191 | extra -= 257; |
192 | end = 256; |
193 | break; |
194 | default: /* DISTS */ |
195 | base = dbase; |
196 | extra = dext; |
197 | end = -1; |
198 | } |
199 | |
200 | /* initialize state for loop */ |
201 | huff = 0; /* starting code */ |
202 | sym = 0; /* starting code symbol */ |
203 | len = min; /* starting code length */ |
204 | next = *table; /* current table to fill in */ |
205 | curr = root; /* current table index bits */ |
206 | drop = 0; /* current bits to drop from code for index */ |
207 | low = (unsigned)(-1); /* trigger new sub-table when len > root */ |
208 | used = 1U << root; /* use root table entries */ |
209 | mask = used - 1; /* mask for comparing low */ |
210 | |
211 | /* check available table space */ |
212 | if (type == LENS && used >= ENOUGH - MAXD) |
213 | return 1; |
214 | |
215 | /* process all codes and make table entries */ |
216 | for (;;) { |
217 | /* create table entry */ |
218 | this.bits = (unsigned char)(len - drop); |
219 | if ((int)(work[sym]) < end) { |
220 | this.op = (unsigned char)0; |
221 | this.val = work[sym]; |
222 | } |
223 | else if ((int)(work[sym]) > end) { |
224 | this.op = (unsigned char)(extra[work[sym]]); |
225 | this.val = base[work[sym]]; |
226 | } |
227 | else { |
228 | this.op = (unsigned char)(32 + 64); /* end of block */ |
229 | this.val = 0; |
230 | } |
231 | |
232 | /* replicate for those indices with low len bits equal to huff */ |
233 | incr = 1U << (len - drop); |
234 | fill = 1U << curr; |
235 | min = fill; /* save offset to next table */ |
236 | do { |
237 | fill -= incr; |
238 | next[(huff >> drop) + fill] = this; |
239 | } while (fill != 0); |
240 | |
241 | /* backwards increment the len-bit code huff */ |
242 | incr = 1U << (len - 1); |
243 | while (huff & incr) |
244 | incr >>= 1; |
245 | if (incr != 0) { |
246 | huff &= incr - 1; |
247 | huff += incr; |
248 | } |
249 | else |
250 | huff = 0; |
251 | |
252 | /* go to next symbol, update count, len */ |
253 | sym++; |
254 | if (--(count[len]) == 0) { |
255 | if (len == max) break; |
256 | len = lens[work[sym]]; |
257 | } |
258 | |
259 | /* create new sub-table if needed */ |
260 | if (len > root && (huff & mask) != low) { |
261 | /* if first time, transition to sub-tables */ |
262 | if (drop == 0) |
263 | drop = root; |
264 | |
265 | /* increment past last table */ |
266 | next += min; /* here min is 1 << curr */ |
267 | |
268 | /* determine length of next table */ |
269 | curr = len - drop; |
270 | left = (int)(1 << curr); |
271 | while (curr + drop < max) { |
272 | left -= count[curr + drop]; |
273 | if (left <= 0) break; |
274 | curr++; |
275 | left <<= 1; |
276 | } |
277 | |
278 | /* check for enough space */ |
279 | used += 1U << curr; |
280 | if (type == LENS && used >= ENOUGH - MAXD) |
281 | return 1; |
282 | |
283 | /* point entry in root table to sub-table */ |
284 | low = huff & mask; |
285 | (*table)[low].op = (unsigned char)curr; |
286 | (*table)[low].bits = (unsigned char)root; |
287 | (*table)[low].val = (unsigned short)(next - *table); |
288 | } |
289 | } |
290 | |
291 | /* |
292 | Fill in rest of table for incomplete codes. This loop is similar to the |
293 | loop above in incrementing huff for table indices. It is assumed that |
294 | len is equal to curr + drop, so there is no loop needed to increment |
295 | through high index bits. When the current sub-table is filled, the loop |
296 | drops back to the root table to fill in any remaining entries there. |
297 | */ |
298 | this.op = (unsigned char)64; /* invalid code marker */ |
299 | this.bits = (unsigned char)(len - drop); |
300 | this.val = (unsigned short)0; |
301 | while (huff != 0) { |
302 | /* when done with sub-table, drop back to root table */ |
303 | if (drop != 0 && (huff & mask) != low) { |
304 | drop = 0; |
305 | len = root; |
306 | next = *table; |
307 | this.bits = (unsigned char)len; |
308 | } |
309 | |
310 | /* put invalid code marker in table */ |
311 | next[huff >> drop] = this; |
312 | |
313 | /* backwards increment the len-bit code huff */ |
314 | incr = 1U << (len - 1); |
315 | while (huff & incr) |
316 | incr >>= 1; |
317 | if (incr != 0) { |
318 | huff &= incr - 1; |
319 | huff += incr; |
320 | } |
321 | else |
322 | huff = 0; |
323 | } |
324 | |
325 | /* set return parameters */ |
326 | *table += used; |
327 | *bits = root; |
328 | return 0; |
329 | } |