annotate src/zlib-1.2.7/inftrees.c @ 169:223a55898ab9 tip default

Add null config files
author Chris Cannam <cannam@all-day-breakfast.com>
date Mon, 02 Mar 2020 14:03:47 +0000
parents 8a15ff55d9af
children
rev   line source
cannam@89 1 /* inftrees.c -- generate Huffman trees for efficient decoding
cannam@89 2 * Copyright (C) 1995-2012 Mark Adler
cannam@89 3 * For conditions of distribution and use, see copyright notice in zlib.h
cannam@89 4 */
cannam@89 5
cannam@89 6 #include "zutil.h"
cannam@89 7 #include "inftrees.h"
cannam@89 8
cannam@89 9 #define MAXBITS 15
cannam@89 10
cannam@89 11 const char inflate_copyright[] =
cannam@89 12 " inflate 1.2.7 Copyright 1995-2012 Mark Adler ";
cannam@89 13 /*
cannam@89 14 If you use the zlib library in a product, an acknowledgment is welcome
cannam@89 15 in the documentation of your product. If for some reason you cannot
cannam@89 16 include such an acknowledgment, I would appreciate that you keep this
cannam@89 17 copyright string in the executable of your product.
cannam@89 18 */
cannam@89 19
cannam@89 20 /*
cannam@89 21 Build a set of tables to decode the provided canonical Huffman code.
cannam@89 22 The code lengths are lens[0..codes-1]. The result starts at *table,
cannam@89 23 whose indices are 0..2^bits-1. work is a writable array of at least
cannam@89 24 lens shorts, which is used as a work area. type is the type of code
cannam@89 25 to be generated, CODES, LENS, or DISTS. On return, zero is success,
cannam@89 26 -1 is an invalid code, and +1 means that ENOUGH isn't enough. table
cannam@89 27 on return points to the next available entry's address. bits is the
cannam@89 28 requested root table index bits, and on return it is the actual root
cannam@89 29 table index bits. It will differ if the request is greater than the
cannam@89 30 longest code or if it is less than the shortest code.
cannam@89 31 */
cannam@89 32 int ZLIB_INTERNAL inflate_table(type, lens, codes, table, bits, work)
cannam@89 33 codetype type;
cannam@89 34 unsigned short FAR *lens;
cannam@89 35 unsigned codes;
cannam@89 36 code FAR * FAR *table;
cannam@89 37 unsigned FAR *bits;
cannam@89 38 unsigned short FAR *work;
cannam@89 39 {
cannam@89 40 unsigned len; /* a code's length in bits */
cannam@89 41 unsigned sym; /* index of code symbols */
cannam@89 42 unsigned min, max; /* minimum and maximum code lengths */
cannam@89 43 unsigned root; /* number of index bits for root table */
cannam@89 44 unsigned curr; /* number of index bits for current table */
cannam@89 45 unsigned drop; /* code bits to drop for sub-table */
cannam@89 46 int left; /* number of prefix codes available */
cannam@89 47 unsigned used; /* code entries in table used */
cannam@89 48 unsigned huff; /* Huffman code */
cannam@89 49 unsigned incr; /* for incrementing code, index */
cannam@89 50 unsigned fill; /* index for replicating entries */
cannam@89 51 unsigned low; /* low bits for current root entry */
cannam@89 52 unsigned mask; /* mask for low root bits */
cannam@89 53 code here; /* table entry for duplication */
cannam@89 54 code FAR *next; /* next available space in table */
cannam@89 55 const unsigned short FAR *base; /* base value table to use */
cannam@89 56 const unsigned short FAR *extra; /* extra bits table to use */
cannam@89 57 int end; /* use base and extra for symbol > end */
cannam@89 58 unsigned short count[MAXBITS+1]; /* number of codes of each length */
cannam@89 59 unsigned short offs[MAXBITS+1]; /* offsets in table for each length */
cannam@89 60 static const unsigned short lbase[31] = { /* Length codes 257..285 base */
cannam@89 61 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
cannam@89 62 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
cannam@89 63 static const unsigned short lext[31] = { /* Length codes 257..285 extra */
cannam@89 64 16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18,
cannam@89 65 19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 78, 68};
cannam@89 66 static const unsigned short dbase[32] = { /* Distance codes 0..29 base */
cannam@89 67 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
cannam@89 68 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
cannam@89 69 8193, 12289, 16385, 24577, 0, 0};
cannam@89 70 static const unsigned short dext[32] = { /* Distance codes 0..29 extra */
cannam@89 71 16, 16, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22,
cannam@89 72 23, 23, 24, 24, 25, 25, 26, 26, 27, 27,
cannam@89 73 28, 28, 29, 29, 64, 64};
cannam@89 74
cannam@89 75 /*
cannam@89 76 Process a set of code lengths to create a canonical Huffman code. The
cannam@89 77 code lengths are lens[0..codes-1]. Each length corresponds to the
cannam@89 78 symbols 0..codes-1. The Huffman code is generated by first sorting the
cannam@89 79 symbols by length from short to long, and retaining the symbol order
cannam@89 80 for codes with equal lengths. Then the code starts with all zero bits
cannam@89 81 for the first code of the shortest length, and the codes are integer
cannam@89 82 increments for the same length, and zeros are appended as the length
cannam@89 83 increases. For the deflate format, these bits are stored backwards
cannam@89 84 from their more natural integer increment ordering, and so when the
cannam@89 85 decoding tables are built in the large loop below, the integer codes
cannam@89 86 are incremented backwards.
cannam@89 87
cannam@89 88 This routine assumes, but does not check, that all of the entries in
cannam@89 89 lens[] are in the range 0..MAXBITS. The caller must assure this.
cannam@89 90 1..MAXBITS is interpreted as that code length. zero means that that
cannam@89 91 symbol does not occur in this code.
cannam@89 92
cannam@89 93 The codes are sorted by computing a count of codes for each length,
cannam@89 94 creating from that a table of starting indices for each length in the
cannam@89 95 sorted table, and then entering the symbols in order in the sorted
cannam@89 96 table. The sorted table is work[], with that space being provided by
cannam@89 97 the caller.
cannam@89 98
cannam@89 99 The length counts are used for other purposes as well, i.e. finding
cannam@89 100 the minimum and maximum length codes, determining if there are any
cannam@89 101 codes at all, checking for a valid set of lengths, and looking ahead
cannam@89 102 at length counts to determine sub-table sizes when building the
cannam@89 103 decoding tables.
cannam@89 104 */
cannam@89 105
cannam@89 106 /* accumulate lengths for codes (assumes lens[] all in 0..MAXBITS) */
cannam@89 107 for (len = 0; len <= MAXBITS; len++)
cannam@89 108 count[len] = 0;
cannam@89 109 for (sym = 0; sym < codes; sym++)
cannam@89 110 count[lens[sym]]++;
cannam@89 111
cannam@89 112 /* bound code lengths, force root to be within code lengths */
cannam@89 113 root = *bits;
cannam@89 114 for (max = MAXBITS; max >= 1; max--)
cannam@89 115 if (count[max] != 0) break;
cannam@89 116 if (root > max) root = max;
cannam@89 117 if (max == 0) { /* no symbols to code at all */
cannam@89 118 here.op = (unsigned char)64; /* invalid code marker */
cannam@89 119 here.bits = (unsigned char)1;
cannam@89 120 here.val = (unsigned short)0;
cannam@89 121 *(*table)++ = here; /* make a table to force an error */
cannam@89 122 *(*table)++ = here;
cannam@89 123 *bits = 1;
cannam@89 124 return 0; /* no symbols, but wait for decoding to report error */
cannam@89 125 }
cannam@89 126 for (min = 1; min < max; min++)
cannam@89 127 if (count[min] != 0) break;
cannam@89 128 if (root < min) root = min;
cannam@89 129
cannam@89 130 /* check for an over-subscribed or incomplete set of lengths */
cannam@89 131 left = 1;
cannam@89 132 for (len = 1; len <= MAXBITS; len++) {
cannam@89 133 left <<= 1;
cannam@89 134 left -= count[len];
cannam@89 135 if (left < 0) return -1; /* over-subscribed */
cannam@89 136 }
cannam@89 137 if (left > 0 && (type == CODES || max != 1))
cannam@89 138 return -1; /* incomplete set */
cannam@89 139
cannam@89 140 /* generate offsets into symbol table for each length for sorting */
cannam@89 141 offs[1] = 0;
cannam@89 142 for (len = 1; len < MAXBITS; len++)
cannam@89 143 offs[len + 1] = offs[len] + count[len];
cannam@89 144
cannam@89 145 /* sort symbols by length, by symbol order within each length */
cannam@89 146 for (sym = 0; sym < codes; sym++)
cannam@89 147 if (lens[sym] != 0) work[offs[lens[sym]]++] = (unsigned short)sym;
cannam@89 148
cannam@89 149 /*
cannam@89 150 Create and fill in decoding tables. In this loop, the table being
cannam@89 151 filled is at next and has curr index bits. The code being used is huff
cannam@89 152 with length len. That code is converted to an index by dropping drop
cannam@89 153 bits off of the bottom. For codes where len is less than drop + curr,
cannam@89 154 those top drop + curr - len bits are incremented through all values to
cannam@89 155 fill the table with replicated entries.
cannam@89 156
cannam@89 157 root is the number of index bits for the root table. When len exceeds
cannam@89 158 root, sub-tables are created pointed to by the root entry with an index
cannam@89 159 of the low root bits of huff. This is saved in low to check for when a
cannam@89 160 new sub-table should be started. drop is zero when the root table is
cannam@89 161 being filled, and drop is root when sub-tables are being filled.
cannam@89 162
cannam@89 163 When a new sub-table is needed, it is necessary to look ahead in the
cannam@89 164 code lengths to determine what size sub-table is needed. The length
cannam@89 165 counts are used for this, and so count[] is decremented as codes are
cannam@89 166 entered in the tables.
cannam@89 167
cannam@89 168 used keeps track of how many table entries have been allocated from the
cannam@89 169 provided *table space. It is checked for LENS and DIST tables against
cannam@89 170 the constants ENOUGH_LENS and ENOUGH_DISTS to guard against changes in
cannam@89 171 the initial root table size constants. See the comments in inftrees.h
cannam@89 172 for more information.
cannam@89 173
cannam@89 174 sym increments through all symbols, and the loop terminates when
cannam@89 175 all codes of length max, i.e. all codes, have been processed. This
cannam@89 176 routine permits incomplete codes, so another loop after this one fills
cannam@89 177 in the rest of the decoding tables with invalid code markers.
cannam@89 178 */
cannam@89 179
cannam@89 180 /* set up for code type */
cannam@89 181 switch (type) {
cannam@89 182 case CODES:
cannam@89 183 base = extra = work; /* dummy value--not used */
cannam@89 184 end = 19;
cannam@89 185 break;
cannam@89 186 case LENS:
cannam@89 187 base = lbase;
cannam@89 188 base -= 257;
cannam@89 189 extra = lext;
cannam@89 190 extra -= 257;
cannam@89 191 end = 256;
cannam@89 192 break;
cannam@89 193 default: /* DISTS */
cannam@89 194 base = dbase;
cannam@89 195 extra = dext;
cannam@89 196 end = -1;
cannam@89 197 }
cannam@89 198
cannam@89 199 /* initialize state for loop */
cannam@89 200 huff = 0; /* starting code */
cannam@89 201 sym = 0; /* starting code symbol */
cannam@89 202 len = min; /* starting code length */
cannam@89 203 next = *table; /* current table to fill in */
cannam@89 204 curr = root; /* current table index bits */
cannam@89 205 drop = 0; /* current bits to drop from code for index */
cannam@89 206 low = (unsigned)(-1); /* trigger new sub-table when len > root */
cannam@89 207 used = 1U << root; /* use root table entries */
cannam@89 208 mask = used - 1; /* mask for comparing low */
cannam@89 209
cannam@89 210 /* check available table space */
cannam@89 211 if ((type == LENS && used >= ENOUGH_LENS) ||
cannam@89 212 (type == DISTS && used >= ENOUGH_DISTS))
cannam@89 213 return 1;
cannam@89 214
cannam@89 215 /* process all codes and make table entries */
cannam@89 216 for (;;) {
cannam@89 217 /* create table entry */
cannam@89 218 here.bits = (unsigned char)(len - drop);
cannam@89 219 if ((int)(work[sym]) < end) {
cannam@89 220 here.op = (unsigned char)0;
cannam@89 221 here.val = work[sym];
cannam@89 222 }
cannam@89 223 else if ((int)(work[sym]) > end) {
cannam@89 224 here.op = (unsigned char)(extra[work[sym]]);
cannam@89 225 here.val = base[work[sym]];
cannam@89 226 }
cannam@89 227 else {
cannam@89 228 here.op = (unsigned char)(32 + 64); /* end of block */
cannam@89 229 here.val = 0;
cannam@89 230 }
cannam@89 231
cannam@89 232 /* replicate for those indices with low len bits equal to huff */
cannam@89 233 incr = 1U << (len - drop);
cannam@89 234 fill = 1U << curr;
cannam@89 235 min = fill; /* save offset to next table */
cannam@89 236 do {
cannam@89 237 fill -= incr;
cannam@89 238 next[(huff >> drop) + fill] = here;
cannam@89 239 } while (fill != 0);
cannam@89 240
cannam@89 241 /* backwards increment the len-bit code huff */
cannam@89 242 incr = 1U << (len - 1);
cannam@89 243 while (huff & incr)
cannam@89 244 incr >>= 1;
cannam@89 245 if (incr != 0) {
cannam@89 246 huff &= incr - 1;
cannam@89 247 huff += incr;
cannam@89 248 }
cannam@89 249 else
cannam@89 250 huff = 0;
cannam@89 251
cannam@89 252 /* go to next symbol, update count, len */
cannam@89 253 sym++;
cannam@89 254 if (--(count[len]) == 0) {
cannam@89 255 if (len == max) break;
cannam@89 256 len = lens[work[sym]];
cannam@89 257 }
cannam@89 258
cannam@89 259 /* create new sub-table if needed */
cannam@89 260 if (len > root && (huff & mask) != low) {
cannam@89 261 /* if first time, transition to sub-tables */
cannam@89 262 if (drop == 0)
cannam@89 263 drop = root;
cannam@89 264
cannam@89 265 /* increment past last table */
cannam@89 266 next += min; /* here min is 1 << curr */
cannam@89 267
cannam@89 268 /* determine length of next table */
cannam@89 269 curr = len - drop;
cannam@89 270 left = (int)(1 << curr);
cannam@89 271 while (curr + drop < max) {
cannam@89 272 left -= count[curr + drop];
cannam@89 273 if (left <= 0) break;
cannam@89 274 curr++;
cannam@89 275 left <<= 1;
cannam@89 276 }
cannam@89 277
cannam@89 278 /* check for enough space */
cannam@89 279 used += 1U << curr;
cannam@89 280 if ((type == LENS && used >= ENOUGH_LENS) ||
cannam@89 281 (type == DISTS && used >= ENOUGH_DISTS))
cannam@89 282 return 1;
cannam@89 283
cannam@89 284 /* point entry in root table to sub-table */
cannam@89 285 low = huff & mask;
cannam@89 286 (*table)[low].op = (unsigned char)curr;
cannam@89 287 (*table)[low].bits = (unsigned char)root;
cannam@89 288 (*table)[low].val = (unsigned short)(next - *table);
cannam@89 289 }
cannam@89 290 }
cannam@89 291
cannam@89 292 /* fill in remaining table entry if code is incomplete (guaranteed to have
cannam@89 293 at most one remaining entry, since if the code is incomplete, the
cannam@89 294 maximum code length that was allowed to get this far is one bit) */
cannam@89 295 if (huff != 0) {
cannam@89 296 here.op = (unsigned char)64; /* invalid code marker */
cannam@89 297 here.bits = (unsigned char)(len - drop);
cannam@89 298 here.val = (unsigned short)0;
cannam@89 299 next[huff] = here;
cannam@89 300 }
cannam@89 301
cannam@89 302 /* set return parameters */
cannam@89 303 *table += used;
cannam@89 304 *bits = root;
cannam@89 305 return 0;
cannam@89 306 }