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