Home | History | Annotate | Download | only in gzip
      1 /* inftrees.c -- generate Huffman trees for efficient decoding
      2  * Copyright (C) 1995-2002 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 #if !defined(BUILDFIXED) && !defined(STDC)
     10 #  define BUILDFIXED   /* non ANSI compilers may not accept inffixed.h */
     11 #endif
     12 
     13 
     14 #if 0
     15 local const char inflate_copyright[] =
     16    " inflate 1.1.4 Copyright 1995-2002 Mark Adler ";
     17 #endif
     18 /*
     19   If you use the zlib library in a product, an acknowledgment is welcome
     20   in the documentation of your product. If for some reason you cannot
     21   include such an acknowledgment, I would appreciate that you keep this
     22   copyright string in the executable of your product.
     23  */
     24 
     25 /* simplify the use of the inflate_huft type with some defines */
     26 #define exop word.what.Exop
     27 #define bits word.what.Bits
     28 
     29 
     30 local int huft_build OF((
     31     uIntf *,            /* code lengths in bits */
     32     uInt,               /* number of codes */
     33     uInt,               /* number of "simple" codes */
     34     const uIntf *,      /* list of base values for non-simple codes */
     35     const uIntf *,      /* list of extra bits for non-simple codes */
     36     inflate_huft * FAR*,/* result: starting table */
     37     uIntf *,            /* maximum lookup bits (returns actual) */
     38     inflate_huft *,     /* space for trees */
     39     uInt *,             /* hufts used in space */
     40     uIntf * ));         /* space for values */
     41 
     42 /* Tables for deflate from PKZIP's appnote.txt. */
     43 local const uInt cplens[31] = { /* Copy lengths for literal codes 257..285 */
     44         3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
     45         35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
     46         /* see note #13 above about 258 */
     47 local const uInt cplext[31] = { /* Extra bits for literal codes 257..285 */
     48         0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
     49         3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 112, 112}; /* 112==invalid */
     50 local const uInt cpdist[30] = { /* Copy offsets for distance codes 0..29 */
     51         1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
     52         257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
     53         8193, 12289, 16385, 24577};
     54 local const uInt cpdext[30] = { /* Extra bits for distance codes */
     55         0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
     56         7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
     57         12, 12, 13, 13};
     58 
     59 /*
     60    Huffman code decoding is performed using a multi-level table lookup.
     61    The fastest way to decode is to simply build a lookup table whose
     62    size is determined by the longest code.  However, the time it takes
     63    to build this table can also be a factor if the data being decoded
     64    is not very long.  The most common codes are necessarily the
     65    shortest codes, so those codes dominate the decoding time, and hence
     66    the speed.  The idea is you can have a shorter table that decodes the
     67    shorter, more probable codes, and then point to subsidiary tables for
     68    the longer codes.  The time it costs to decode the longer codes is
     69    then traded against the time it takes to make longer tables.
     70 
     71    This results of this trade are in the variables lbits and dbits
     72    below.  lbits is the number of bits the first level table for literal/
     73    length codes can decode in one step, and dbits is the same thing for
     74    the distance codes.  Subsequent tables are also less than or equal to
     75    those sizes.  These values may be adjusted either when all of the
     76    codes are shorter than that, in which case the longest code length in
     77    bits is used, or when the shortest code is *longer* than the requested
     78    table size, in which case the length of the shortest code in bits is
     79    used.
     80 
     81    There are two different values for the two tables, since they code a
     82    different number of possibilities each.  The literal/length table
     83    codes 286 possible values, or in a flat code, a little over eight
     84    bits.  The distance table codes 30 possible values, or a little less
     85    than five bits, flat.  The optimum values for speed end up being
     86    about one bit more than those, so lbits is 8+1 and dbits is 5+1.
     87    The optimum values may differ though from machine to machine, and
     88    possibly even between compilers.  Your mileage may vary.
     89  */
     90 
     91 
     92 /* If BMAX needs to be larger than 16, then h and x[] should be uLong. */
     93 #define BMAX 15         /* maximum bit length of any code */
     94 
     95 local int huft_build( /* b, n, s, d, e, t, m, hp, hn, v) */
     96 uIntf *b,               /* code lengths in bits (all assumed <= BMAX) */
     97 uInt n,                 /* number of codes (assumed <= 288) */
     98 uInt s,                 /* number of simple-valued codes (0..s-1) */
     99 const uIntf *d,         /* list of base values for non-simple codes */
    100 const uIntf *e,         /* list of extra bits for non-simple codes */
    101 inflate_huft * FAR *t,  /* result: starting table */
    102 uIntf *m,               /* maximum lookup bits, returns actual */
    103 inflate_huft *hp,       /* space for trees */
    104 uInt *hn,               /* hufts used in space */
    105 uIntf *v                /* working area: values in order of bit length */
    106 /* Given a list of code lengths and a maximum table size, make a set of
    107    tables to decode that set of codes.  Return Z_OK on success, Z_BUF_ERROR
    108    if the given code set is incomplete (the tables are still built in this
    109    case), or Z_DATA_ERROR if the input is invalid. */
    110 )
    111 {
    112 
    113   uInt a;                       /* counter for codes of length k */
    114   uInt c[BMAX+1];               /* bit length count table */
    115   uInt f;                       /* i repeats in table every f entries */
    116   int g;                        /* maximum code length */
    117   int h;                        /* table level */
    118   uInt i;                       /* counter, current code */
    119   uInt j;                       /* counter */
    120   int k;                        /* number of bits in current code */
    121   int l;                        /* bits per table (returned in m) */
    122   uInt mask;                    /* (1 << w) - 1, to avoid cc -O bug on HP */
    123   uIntf *p;                     /* pointer into c[], b[], or v[] */
    124   inflate_huft *q;              /* points to current table */
    125   struct inflate_huft_s r;      /* table entry for structure assignment */
    126   inflate_huft *u[BMAX];        /* table stack */
    127   int w;                        /* bits before this table == (l * h) */
    128   uInt x[BMAX+1];               /* bit offsets, then code stack */
    129   uIntf *xp;                    /* pointer into x */
    130   int y;                        /* number of dummy codes added */
    131   uInt z;                       /* number of entries in current table */
    132 
    133 
    134   /* Make compiler happy */
    135   r.base = 0;
    136 
    137   /* Generate counts for each bit length */
    138   p = c;
    139 #define C0 *p++ = 0;
    140 #define C2 C0 C0 C0 C0
    141 #define C4 C2 C2 C2 C2
    142   C4                            /* clear c[]--assume BMAX+1 is 16 */
    143   p = b;  i = n;
    144   do {
    145     c[*p++]++;                  /* assume all entries <= BMAX */
    146   } while (--i);
    147   if (c[0] == n)                /* null input--all zero length codes */
    148   {
    149     *t = (inflate_huft *)Z_NULL;
    150     *m = 0;
    151     return Z_OK;
    152   }
    153 
    154 
    155   /* Find minimum and maximum length, bound *m by those */
    156   l = *m;
    157   for (j = 1; j <= BMAX; j++)
    158     if (c[j])
    159       break;
    160   k = j;                        /* minimum code length */
    161   if ((uInt)l < j)
    162     l = j;
    163   for (i = BMAX; i; i--)
    164     if (c[i])
    165       break;
    166   g = i;                        /* maximum code length */
    167   if ((uInt)l > i)
    168     l = i;
    169   *m = l;
    170 
    171 
    172   /* Adjust last length count to fill out codes, if needed */
    173   for (y = 1 << j; j < i; j++, y <<= 1)
    174     if ((y -= c[j]) < 0)
    175       return Z_DATA_ERROR;
    176   if ((y -= c[i]) < 0)
    177     return Z_DATA_ERROR;
    178   c[i] += y;
    179 
    180 
    181   /* Generate starting offsets into the value table for each length */
    182   x[1] = j = 0;
    183   p = c + 1;  xp = x + 2;
    184   while (--i) {                 /* note that i == g from above */
    185     *xp++ = (j += *p++);
    186   }
    187 
    188 
    189   /* Make a table of values in order of bit lengths */
    190   p = b;  i = 0;
    191   do {
    192     if ((j = *p++) != 0)
    193       v[x[j]++] = i;
    194   } while (++i < n);
    195   n = x[g];                     /* set n to length of v */
    196 
    197 
    198   /* Generate the Huffman codes and for each, make the table entries */
    199   x[0] = i = 0;                 /* first Huffman code is zero */
    200   p = v;                        /* grab values in bit order */
    201   h = -1;                       /* no tables yet--level -1 */
    202   w = -l;                       /* bits decoded == (l * h) */
    203   u[0] = (inflate_huft *)Z_NULL;        /* just to keep compilers happy */
    204   q = (inflate_huft *)Z_NULL;   /* ditto */
    205   z = 0;                        /* ditto */
    206 
    207   /* go through the bit lengths (k already is bits in shortest code) */
    208   for (; k <= g; k++)
    209   {
    210     a = c[k];
    211     while (a--)
    212     {
    213       /* here i is the Huffman code of length k bits for value *p */
    214       /* make tables up to required level */
    215       while (k > w + l)
    216       {
    217         h++;
    218         w += l;                 /* previous table always l bits */
    219 
    220         /* compute minimum size table less than or equal to l bits */
    221         z = g - w;
    222         z = z > (uInt)l ? (uInt)l : z;        /* table size upper limit */
    223         if ((f = 1 << (j = k - w)) > a + 1)     /* try a k-w bit table */
    224         {                       /* too few codes for k-w bit table */
    225           f -= a + 1;           /* deduct codes from patterns left */
    226           xp = c + k;
    227           if (j < z)
    228             while (++j < z)     /* try smaller tables up to z bits */
    229             {
    230               if ((f <<= 1) <= *++xp)
    231                 break;          /* enough codes to use up j bits */
    232               f -= *xp;         /* else deduct codes from patterns */
    233             }
    234         }
    235         z = 1 << j;             /* table entries for j-bit table */
    236 
    237         /* allocate new table */
    238         if (*hn + z > MANY)     /* (note: doesn't matter for fixed) */
    239           return Z_DATA_ERROR;  /* overflow of MANY */
    240         u[h] = q = hp + *hn;
    241         *hn += z;
    242 
    243         /* connect to last table, if there is one */
    244         if (h)
    245         {
    246           x[h] = i;             /* save pattern for backing up */
    247           r.bits = (Byte)l;     /* bits to dump before this table */
    248           r.exop = (Byte)j;     /* bits in this table */
    249           j = i >> (w - l);
    250           r.base = (uInt)(q - u[h-1] - j);   /* offset to this table */
    251           u[h-1][j] = r;        /* connect to last table */
    252         }
    253         else
    254           *t = q;               /* first table is returned result */
    255       }
    256 
    257       /* set up table entry in r */
    258       r.bits = (Byte)(k - w);
    259       if (p >= v + n)
    260         r.exop = 128 + 64;      /* out of values--invalid code */
    261       else if (*p < s)
    262       {
    263         r.exop = (Byte)(*p < 256 ? 0 : 32 + 64);     /* 256 is end-of-block */
    264         r.base = *p++;          /* simple code is just the value */
    265       }
    266       else
    267       {
    268         r.exop = (Byte)(e[*p - s] + 16 + 64);/* non-simple--look up in lists */
    269         r.base = d[*p++ - s];
    270       }
    271 
    272       /* fill code-like entries with r */
    273       f = 1 << (k - w);
    274       for (j = i >> w; j < z; j += f)
    275         q[j] = r;
    276 
    277       /* backwards increment the k-bit code i */
    278       for (j = 1 << (k - 1); i & j; j >>= 1)
    279         i ^= j;
    280       i ^= j;
    281 
    282       /* backup over finished tables */
    283       mask = (1 << w) - 1;      /* needed on HP, cc -O bug */
    284       while ((i & mask) != x[h])
    285       {
    286         h--;                    /* don't need to update q */
    287         w -= l;
    288         mask = (1 << w) - 1;
    289       }
    290     }
    291   }
    292 
    293 
    294   /* Return Z_BUF_ERROR if we were given an incomplete table */
    295   return y != 0 && g != 1 ? Z_BUF_ERROR : Z_OK;
    296 }
    297 
    298 
    299 local int inflate_trees_bits( /* c, bb, tb, hp, z) */
    300 uIntf *c,               /* 19 code lengths */
    301 uIntf *bb,              /* bits tree desired/actual depth */
    302 inflate_huft * FAR *tb, /* bits tree result */
    303 inflate_huft *hp,       /* space for trees */
    304 z_streamp z             /* for messages */
    305 )
    306 {
    307   int r;
    308   uInt hn = 0;          /* hufts used in space */
    309   uIntf *v;             /* work area for huft_build */
    310 
    311   if ((v = (uIntf*)ZALLOC(z, 19, sizeof(uInt))) == Z_NULL)
    312     return Z_MEM_ERROR;
    313   r = huft_build(c, 19, 19, (uIntf*)Z_NULL, (uIntf*)Z_NULL,
    314                  tb, bb, hp, &hn, v);
    315   if (r == Z_DATA_ERROR)
    316     z->msg = (char*)"oversubscribed dynamic bit lengths tree";
    317   else if (r == Z_BUF_ERROR || *bb == 0)
    318   {
    319     z->msg = (char*)"incomplete dynamic bit lengths tree";
    320     r = Z_DATA_ERROR;
    321   }
    322   ZFREE(z, v);
    323   return r;
    324 }
    325 
    326 
    327 local int inflate_trees_dynamic( /* nl, nd, c, bl, bd, tl, td, hp, z) */
    328 uInt nl,                /* number of literal/length codes */
    329 uInt nd,                /* number of distance codes */
    330 uIntf *c,               /* that many (total) code lengths */
    331 uIntf *bl,              /* literal desired/actual bit depth */
    332 uIntf *bd,              /* distance desired/actual bit depth */
    333 inflate_huft * FAR *tl, /* literal/length tree result */
    334 inflate_huft * FAR *td, /* distance tree result */
    335 inflate_huft *hp,       /* space for trees */
    336 z_streamp z             /* for messages */
    337 )
    338 {
    339   int r;
    340   uInt hn = 0;          /* hufts used in space */
    341   uIntf *v;             /* work area for huft_build */
    342 
    343   /* allocate work area */
    344   if ((v = (uIntf*)ZALLOC(z, 288, sizeof(uInt))) == Z_NULL)
    345     return Z_MEM_ERROR;
    346 
    347   /* build literal/length tree */
    348   r = huft_build(c, nl, 257, cplens, cplext, tl, bl, hp, &hn, v);
    349   if (r != Z_OK || *bl == 0)
    350   {
    351     if (r == Z_DATA_ERROR)
    352       z->msg = (char*)"oversubscribed literal/length tree";
    353     else if (r != Z_MEM_ERROR)
    354     {
    355       z->msg = (char*)"incomplete literal/length tree";
    356       r = Z_DATA_ERROR;
    357     }
    358     ZFREE(z, v);
    359     return r;
    360   }
    361 
    362   /* build distance tree */
    363   r = huft_build(c + nl, nd, 0, cpdist, cpdext, td, bd, hp, &hn, v);
    364   if (r != Z_OK || (*bd == 0 && nl > 257))
    365   {
    366     if (r == Z_DATA_ERROR)
    367       z->msg = (char*)"oversubscribed distance tree";
    368     else if (r == Z_BUF_ERROR) {
    369 #if 0
    370     {
    371 #endif
    372 #ifdef PKZIP_BUG_WORKAROUND
    373       r = Z_OK;
    374     }
    375 #else
    376       z->msg = (char*)"incomplete distance tree";
    377       r = Z_DATA_ERROR;
    378     }
    379     else if (r != Z_MEM_ERROR)
    380     {
    381       z->msg = (char*)"empty distance tree with lengths";
    382       r = Z_DATA_ERROR;
    383     }
    384     ZFREE(z, v);
    385     return r;
    386 #endif
    387   }
    388 
    389   /* done */
    390   ZFREE(z, v);
    391   return Z_OK;
    392 }
    393 
    394 
    395 /* build fixed tables only once--keep them here */
    396 #ifdef BUILDFIXED
    397 local int fixed_built = 0;
    398 #define FIXEDH 544      /* number of hufts used by fixed tables */
    399 local inflate_huft fixed_mem[FIXEDH];
    400 local uInt fixed_bl;
    401 local uInt fixed_bd;
    402 local inflate_huft *fixed_tl;
    403 local inflate_huft *fixed_td;
    404 #else
    405 #include "inffixed.h"
    406 #endif
    407 
    408 
    409 local int inflate_trees_fixed( /* bl, bd, tl, td, z) */
    410 uIntf *bl,                      /* literal desired/actual bit depth */
    411 uIntf *bd,                      /* distance desired/actual bit depth */
    412 const inflate_huft * FAR *tl,   /* literal/length tree result */
    413 const inflate_huft * FAR *td,   /* distance tree result */
    414 z_streamp z                     /* for memory allocation */
    415 )
    416 {
    417 #ifdef BUILDFIXED
    418   /* build fixed tables if not already */
    419   if (!fixed_built)
    420   {
    421     int k;              /* temporary variable */
    422     uInt f = 0;         /* number of hufts used in fixed_mem */
    423     uIntf *c;           /* length list for huft_build */
    424     uIntf *v;           /* work area for huft_build */
    425 
    426     /* allocate memory */
    427     if ((c = (uIntf*)ZALLOC(z, 288, sizeof(uInt))) == Z_NULL)
    428       return Z_MEM_ERROR;
    429     if ((v = (uIntf*)ZALLOC(z, 288, sizeof(uInt))) == Z_NULL)
    430     {
    431       ZFREE(z, c);
    432       return Z_MEM_ERROR;
    433     }
    434 
    435     /* literal table */
    436     for (k = 0; k < 144; k++)
    437       c[k] = 8;
    438     for (; k < 256; k++)
    439       c[k] = 9;
    440     for (; k < 280; k++)
    441       c[k] = 7;
    442     for (; k < 288; k++)
    443       c[k] = 8;
    444     fixed_bl = 9;
    445     huft_build(c, 288, 257, cplens, cplext, &fixed_tl, &fixed_bl,
    446                fixed_mem, &f, v);
    447 
    448     /* distance table */
    449     for (k = 0; k < 30; k++)
    450       c[k] = 5;
    451     fixed_bd = 5;
    452     huft_build(c, 30, 0, cpdist, cpdext, &fixed_td, &fixed_bd,
    453                fixed_mem, &f, v);
    454 
    455     /* done */
    456     ZFREE(z, v);
    457     ZFREE(z, c);
    458     fixed_built = 1;
    459   }
    460 #else
    461   FT_UNUSED(z);
    462 #endif
    463   *bl = fixed_bl;
    464   *bd = fixed_bd;
    465   *tl = fixed_tl;
    466   *td = fixed_td;
    467   return Z_OK;
    468 }
    469