trees.c

给出了 zip 压缩算法的完整实现过程。

/* Number of valid bits in bi_buf.  All bits above the last valid bit
 * are always zero.
 */

#if (!defined(ASMV) || !defined(RISCOS))
local char *out_buf;
#else
char *out_buf;
#endif
/* Current output buffer. */

#if (!defined(ASMV) || !defined(RISCOS))
local unsigned out_offset;
#else
unsigned out_offset;
#endif
/* Current offset in output buffer.
 * On 16 bit machines, the buffer is limited to 64K.
 */

#if !defined(ASMV) || !defined(RISCOS)
local unsigned out_size;
#else
unsigned out_size;
#endif
/* Size of current output buffer */

/* Output a 16 bit value to the bit stream, lower (oldest) byte first */
#define PUTSHORT(w) \
{ if (out_offset >= out_size-1) \
    flush_outbuf(out_buf, &out_offset); \
  out_buf[out_offset++] = (char) ((w) & 0xff); \
  out_buf[out_offset++] = (char) ((ush)(w) >> 8); \
}

#define PUTBYTE(b) \
{ if (out_offset >= out_size) \
    flush_outbuf(out_buf, &out_offset); \
  out_buf[out_offset++] = (char) (b); \
}

#ifdef DEBUG
local ulg bits_sent;   /* bit length of the compressed data */
extern ulg isize;      /* byte length of input file */
#endif

extern long block_start;       /* window offset of current block */
extern unsigned near strstart; /* window offset of current string */


/* ===========================================================================
 * Local (static) routines in this file.
 */

local void init_block     OF((void));
local void pqdownheap     OF((ct_data near *tree, int k));
local void gen_bitlen     OF((tree_desc near *desc));
local void gen_codes      OF((ct_data near *tree, int max_code));
local void build_tree     OF((tree_desc near *desc));
local void scan_tree      OF((ct_data near *tree, int max_code));
local void send_tree      OF((ct_data near *tree, int max_code));
local int  build_bl_tree  OF((void));
local void send_all_trees OF((int lcodes, int dcodes, int blcodes));
local void compress_block OF((ct_data near *ltree, ct_data near *dtree));
local void set_file_type  OF((void));
#if (!defined(ASMV) || !defined(RISCOS))
local void send_bits      OF((int value, int length));
local unsigned bi_reverse OF((unsigned code, int len));
#endif
local void bi_windup      OF((void));
local void copy_block     OF((char *buf, unsigned len, int header));


#ifndef DEBUG
#  define send_code(c, tree) send_bits(tree[c].Code, tree[c].Len)
   /* Send a code of the given tree. c and tree must not have side effects */

#else /* DEBUG */
#  define send_code(c, tree) \
     { if (verbose>1) fprintf(stderr,"\ncd %3d ",(c)); \
       send_bits(tree[c].Code, tree[c].Len); }
#endif

#define d_code(dist) \
   ((dist) < 256 ? dist_code[dist] : dist_code[256+((dist)>>7)])
/* Mapping from a distance to a distance code. dist is the distance - 1 and
 * must not have side effects. dist_code[256] and dist_code[257] are never
 * used.
 */

#define Max(a,b) (a >= b ? a : b)
/* the arguments must not have side effects */

/* ===========================================================================
 * Allocate the match buffer, initialize the various tables and save the
 * location of the internal file attribute (ascii/binary) and method
 * (DEFLATE/STORE).
 */
void ct_init(attr, method)
    ush  *attr;   /* pointer to internal file attribute */
    int  *method; /* pointer to compression method */
{
    int n;        /* iterates over tree elements */
    int bits;     /* bit counter */
    int length;   /* length value */
    int code;     /* code value */
    int dist;     /* distance index */

    file_type = attr;
    file_method = method;
    cmpr_bytelen = cmpr_len_bits = 0L;
#ifdef DEBUG
    input_len = 0L;
#endif

    if (static_dtree[0].Len != 0) return; /* ct_init already called */

#ifdef DYN_ALLOC
    d_buf = (ush far *) zcalloc(DIST_BUFSIZE, sizeof(ush));
    l_buf = (uch far *) zcalloc(LIT_BUFSIZE/2, 2);
    /* Avoid using the value 64K on 16 bit machines */
    if (l_buf == NULL || d_buf == NULL)
        ziperr(ZE_MEM, "ct_init: out of memory");
#endif

    /* Initialize the mapping length (0..255) -> length code (0..28) */
    length = 0;
    for (code = 0; code < LENGTH_CODES-1; code++) {
        base_length[code] = length;
        for (n = 0; n < (1<<extra_lbits[code]); n++) {
            length_code[length++] = (uch)code;
        }
    }
    Assert(length == 256, "ct_init: length != 256");
    /* Note that the length 255 (match length 258) can be represented
     * in two different ways: code 284 + 5 bits or code 285, so we
     * overwrite length_code[255] to use the best encoding:
     */
    length_code[length-1] = (uch)code;

    /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
    dist = 0;
    for (code = 0 ; code < 16; code++) {
        base_dist[code] = dist;
        for (n = 0; n < (1<<extra_dbits[code]); n++) {
            dist_code[dist++] = (uch)code;
        }
    }
    Assert(dist == 256, "ct_init: dist != 256");
    dist >>= 7; /* from now on, all distances are divided by 128 */
    for ( ; code < D_CODES; code++) {
        base_dist[code] = dist << 7;
        for (n = 0; n < (1<<(extra_dbits[code]-7)); n++) {
            dist_code[256 + dist++] = (uch)code;
        }
    }
    Assert(dist == 256, "ct_init: 256+dist != 512");

    /* Construct the codes of the static literal tree */
    for (bits = 0; bits <= MAX_BITS; bits++) bl_count[bits] = 0;
    n = 0;
    while (n <= 143) static_ltree[n++].Len = 8, bl_count[8]++;
    while (n <= 255) static_ltree[n++].Len = 9, bl_count[9]++;
    while (n <= 279) static_ltree[n++].Len = 7, bl_count[7]++;
    while (n <= 287) static_ltree[n++].Len = 8, bl_count[8]++;
    /* Codes 286 and 287 do not exist, but we must include them in the
     * tree construction to get a canonical Huffman tree (longest code
     * all ones)
     */
    gen_codes((ct_data near *)static_ltree, L_CODES+1);

    /* The static distance tree is trivial: */
    for (n = 0; n < D_CODES; n++) {
        static_dtree[n].Len = 5;
        static_dtree[n].Code = (ush)bi_reverse(n, 5);
    }

    /* Initialize the first block of the first file: */
    init_block();
}

/* ===========================================================================
 * Initialize a new block.
 */
local void init_block()
{
    int n; /* iterates over tree elements */

    /* Initialize the trees. */
    for (n = 0; n < L_CODES;  n++) dyn_ltree[n].Freq = 0;
    for (n = 0; n < D_CODES;  n++) dyn_dtree[n].Freq = 0;
    for (n = 0; n < BL_CODES; n++) bl_tree[n].Freq = 0;

    dyn_ltree[END_BLOCK].Freq = 1;
    opt_len = static_len = 0L;
    last_lit = last_dist = last_flags = 0;
    flags = 0; flag_bit = 1;
}

#define SMALLEST 1
/* Index within the heap array of least frequent node in the Huffman tree */


/* ===========================================================================
 * Remove the smallest element from the heap and recreate the heap with
 * one less element. Updates heap and heap_len.
 */
#define pqremove(tree, top) \
{\
    top = heap[SMALLEST]; \
    heap[SMALLEST] = heap[heap_len--]; \
    pqdownheap(tree, SMALLEST); \
}

/* ===========================================================================
 * Compares to subtrees, using the tree depth as tie breaker when
 * the subtrees have equal frequency. This minimizes the worst case length.
 */
#define smaller(tree, n, m) \
   (tree[n].Freq < tree[m].Freq || \
   (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m]))

/* ===========================================================================
 * Restore the heap property by moving down the tree starting at node k,
 * exchanging a node with the smallest of its two sons if necessary, stopping
 * when the heap property is re-established (each father smaller than its
 * two sons).
 */
local void pqdownheap(tree, k)
    ct_data near *tree;  /* the tree to restore */
    int k;               /* node to move down */
{
    int v = heap[k];
    int j = k << 1;  /* left son of k */
    int htemp;       /* required because of bug in SASC compiler */

    while (j <= heap_len) {
        /* Set j to the smallest of the two sons: */
        if (j < heap_len && smaller(tree, heap[j+1], heap[j])) j++;

        /* Exit if v is smaller than both sons */
        htemp = heap[j];
        if (smaller(tree, v, htemp)) break;

        /* Exchange v with the smallest son */
        heap[k] = htemp;
        k = j;

        /* And continue down the tree, setting j to the left son of k */
        j <<= 1;
    }
    heap[k] = v;
}

/* ===========================================================================
 * Compute the optimal bit lengths for a tree and update the total bit length
 * for the current block.
 * IN assertion: the fields freq and dad are set, heap[heap_max] and
 *    above are the tree nodes sorted by increasing frequency.
 * OUT assertions: the field len is set to the optimal bit length, the
 *     array bl_count contains the frequencies for each bit length.
 *     The length opt_len is updated; static_len is also updated if stree is
 *     not null.
 */
local void gen_bitlen(desc)
    tree_desc near *desc; /* the tree descriptor */
{
    ct_data near *tree  = desc->dyn_tree;
    int near *extra     = desc->extra_bits;
    int base            = desc->extra_base;
    int max_code        = desc->max_code;
    int max_length      = desc->max_length;
    ct_data near *stree = desc->static_tree;
    int h;              /* heap index */
    int n, m;           /* iterate over the tree elements */
    int bits;           /* bit length */
    int xbits;          /* extra bits */
    ush f;              /* frequency */
    int overflow = 0;   /* number of elements with bit length too large */

    for (bits = 0; bits <= MAX_BITS; bits++) bl_count[bits] = 0;

    /* In a first pass, compute the optimal bit lengths (which may
     * overflow in the case of the bit length tree).
     */
    tree[heap[heap_max]].Len = 0; /* root of the heap */

    for (h = heap_max+1; h < HEAP_SIZE; h++) {
        n = heap[h];
        bits = tree[tree[n].Dad].Len + 1;
        if (bits > max_length) bits = max_length, overflow++;
        tree[n].Len = (ush)bits;
        /* We overwrite tree[n].Dad which is no longer needed */

        if (n > max_code) continue; /* not a leaf node */

        bl_count[bits]++;
        xbits = 0;
        if (n >= base) xbits = extra[n-base];
        f = tree[n].Freq;
        opt_len += (ulg)f * (bits + xbits);
        if (stree) static_len += (ulg)f * (stree[n].Len + xbits);
    }
    if (overflow == 0) return;

    Trace((stderr,"\nbit length overflow\n"));
    /* This happens for example on obj2 and pic of the Calgary corpus */

    /* Find the first bit length which could increase: */
    do {
        bits = max_length-1;
        while (bl_count[bits] == 0) bits--;
        bl_count[bits]--;           /* move one leaf down the tree */
        bl_count[bits+1] += (ush)2; /* move one overflow item as its brother */
        bl_count[max_length]--;
        /* The brother of the overflow item also moves one step up,
         * but this does not affect bl_count[max_length]
         */
        overflow -= 2;
    } while (overflow > 0);

    /* Now recompute all bit lengths, scanning in increasing frequency.
     * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
     * lengths instead of fixing only the wrong ones. This idea is taken
     * from 'ar' written by Haruhiko Okumura.)
     */
    for (bits = max_length; bits != 0; bits--) {
        n = bl_count[bits];
        while (n != 0) {
            m = heap[--h];
            if (m > max_code) continue;
            if (tree[m].Len != (ush)bits) {
                Trace((stderr,"code %d bits %d->%d\n", m, tree[m].Len, bits));
                opt_len += ((long)bits-(long)tree[m].Len)*(long)tree[m].Freq;
                tree[m].Len = (ush)bits;
            }
            n--;
        }
    }
}

/* ===========================================================================
 * Generate the codes for a given tree and bit counts (which need not be
 * optimal).
 * IN assertion: the array bl_count contains the bit length statistics for
 * the given tree and the field len is set for all tree elements.
 * OUT assertion: the field code is set for all tree elements of non
 *     zero code length.
 */
local void gen_codes (tree, max_code)
    ct_data near *tree;        /* the tree to decorate */
    int max_code;              /* largest code with non zero frequency */
{
    ush next_code[MAX_BITS+1]; /* next code value for each bit length */
    ush code = 0;              /* running code value */
    int bits;                  /* bit index */
    int n;                     /* code index */

    /* The distribution counts are first used to generate the code values
     * without bit reversal.
     */
    for (bits = 1; bits <= MAX_BITS; bits++) {
        next_code[bits] = code = (ush)((code + bl_count[bits-1]) << 1);
    }
    /* Check that the bit counts in bl_count are consistent. The last code