trees.c

GZip Compress Souce Code

        } else if (curlen == nextlen) {
            max_count = 6, min_count = 3;
        } else {
            max_count = 7, min_count = 4;
        }
    }

  }
  DBGPrintfo(("scan_tree(out)\r\n"));
}

/* ===========================================================================
 * Send a literal or distance tree in compressed form, using the codes in
 * bl_tree.
 */
local void send_tree (tree, max_code)
    ct_data near *tree; /* the tree to be scanned */
    int max_code;       /* and its largest code of non zero frequency */
{
  DBGPrintfi(("send_tree(In)\r\n"));
  {
    int n;                     /* iterates over all tree elements */
    int prevlen = -1;          /* last emitted length */
    int curlen;                /* length of current code */
    int nextlen = tree[0].Len; /* length of next code */
    int count = 0;             /* repeat count of the current code */
    int max_count = 7;         /* max repeat count */
    int min_count = 4;         /* min repeat count */

    /* tree[max_code+1].Len = -1; */  /* guard already set */
    if (nextlen == 0) max_count = 138, min_count = 3;

    for (n = 0; n <= max_code; n++) {
        curlen = nextlen; nextlen = tree[n+1].Len;
        if (++count < max_count && curlen == nextlen) {
            continue;
        } else if (count < min_count) {
            do { send_code(curlen, bl_tree); } while (--count != 0);

        } else if (curlen != 0) {
            if (curlen != prevlen) {
                send_code(curlen, bl_tree); count--;
            }
            Assert(count >= 3 && count <= 6, " 3_6?");
            send_code(REP_3_6, bl_tree); send_bits(count-3, 2);

        } else if (count <= 10) {
            send_code(REPZ_3_10, bl_tree); send_bits(count-3, 3);

        } else {
            send_code(REPZ_11_138, bl_tree); send_bits(count-11, 7);
        }
        count = 0; prevlen = curlen;
        if (nextlen == 0) {
            max_count = 138, min_count = 3;
        } else if (curlen == nextlen) {
            max_count = 6, min_count = 3;
        } else {
            max_count = 7, min_count = 4;
        }
    }

  }
  DBGPrintfo(("send_tree(out)\r\n"));
}

/* ===========================================================================
 * Construct the Huffman tree for the bit lengths and return the index in
 * bl_order of the last bit length code to send.
 */
local int build_bl_tree()
{
  DBGPrintfi(("build_bl_tree(In)\r\n"));
  {
    int max_blindex;  /* index of last bit length code of non zero freq */

    /* Determine the bit length frequencies for literal and distance trees */
    scan_tree((ct_data near *)dyn_ltree, l_desc.max_code);
    scan_tree((ct_data near *)dyn_dtree, d_desc.max_code);

    /* Build the bit length tree: */
    build_tree((tree_desc near *)(&bl_desc));
    /* opt_len now includes the length of the tree representations, except
     * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
     */

    /* Determine the number of bit length codes to send. The pkzip format
     * requires that at least 4 bit length codes be sent. (appnote.txt says
     * 3 but the actual value used is 4.)
     */
    for (max_blindex = BL_CODES-1; max_blindex >= 3; max_blindex--) {
        if (bl_tree[bl_order[max_blindex]].Len != 0) break;
    }
    /* Update opt_len to include the bit length tree and counts */
    opt_len += 3*(max_blindex+1) + 5+5+4;
    Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld", opt_len, static_len));

    DBGPrintfo(("build_bl_tree(out1)\r\n"));
    return  max_blindex;

  }
}

/* ===========================================================================
 * Send the header for a block using dynamic Huffman trees: the counts, the
 * lengths of the bit length codes, the literal tree and the distance tree.
 * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
 */
local void send_all_trees(lcodes, dcodes, blcodes)
    int lcodes, dcodes, blcodes; /* number of codes for each tree */
{
  DBGPrintfi(("send_all_trees(In)\r\n"));
  {
    int rank;                    /* index in bl_order */

    Assert (lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
    Assert (lcodes <= L_CODES && dcodes <= D_CODES && blcodes <= BL_CODES,
            "too many codes");
    Tracev((stderr, "\nbl counts: "));
    send_bits(lcodes-257, 5); /* not +255 as stated in appnote.txt */
    send_bits(dcodes-1,   5);
    send_bits(blcodes-4,  4); /* not -3 as stated in appnote.txt */
    for (rank = 0; rank < blcodes; rank++) {
        Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
        send_bits(bl_tree[bl_order[rank]].Len, 3);
    }
    Tracev((stderr, "\nbl tree: sent %ld", bits_sent));

    send_tree((ct_data near *)dyn_ltree, lcodes-1); /* send the literal tree */
    Tracev((stderr, "\nlit tree: sent %ld", bits_sent));

    send_tree((ct_data near *)dyn_dtree, dcodes-1); /* send the distance tree */
    Tracev((stderr, "\ndist tree: sent %ld", bits_sent));

  }
  DBGPrintfo(("send_all_trees(out)\r\n"));
}

/* ===========================================================================
 * Determine the best encoding for the current block: dynamic trees, static
 * trees or store, and output the encoded block to the zip file. This function
 * returns the total compressed length for the file so far.
 */
ulg flush_block(buf, stored_len, eof)
    char *buf;        /* input block, or NULL if too old */
    ulg stored_len;   /* length of input block */
    int eof;          /* true if this is the last block for a file */
{
  DBGPrintfi(("flush_block(In)\r\n"));
  {
    ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
    int max_blindex;  /* index of last bit length code of non zero freq */

    flag_buf[last_flags] = flags; /* Save the flags for the last 8 items */

     /* Check if the file is ascii or binary */
    if (*file_type == (ush)UNKNOWN) set_file_type();

    /* Construct the literal and distance trees */
    build_tree((tree_desc near *)(&l_desc));
    Tracev((stderr, "\nlit data: dyn %ld, stat %ld", opt_len, static_len));

    build_tree((tree_desc near *)(&d_desc));
    Tracev((stderr, "\ndist data: dyn %ld, stat %ld", opt_len, static_len));
    /* At this point, opt_len and static_len are the total bit lengths of
     * the compressed block data, excluding the tree representations.
     */

    /* Build the bit length tree for the above two trees, and get the index
     * in bl_order of the last bit length code to send.
     */
    max_blindex = build_bl_tree();

    /* Determine the best encoding. Compute first the block length in bytes */
    opt_lenb = (opt_len+3+7)>>3;
    static_lenb = (static_len+3+7)>>3;
    input_len += stored_len; /* for debugging only */

    Trace((stderr, "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u dist %u ",
            opt_lenb, opt_len, static_lenb, static_len, stored_len,
            last_lit, last_dist));

    if (static_lenb <= opt_lenb) opt_lenb = static_lenb;

    /* If compression failed and this is the first and last block,
     * and if the zip file can be seeked (to rewrite the local header),
     * the whole file is transformed into a stored file:
     */
#ifdef FORCE_METHOD
    if (level == 1 && eof && compressed_len == 0L) { /* force stored file */
#else
    if (stored_len <= opt_lenb && eof && compressed_len == 0L && seekable()) {
#endif
        /* Since LIT_BUFSIZE <= 2*WSIZE, the input data must be there: */
        if (buf == (char*)0) 
		{
			error ("block vanished");
			 DBGPrintfo(("flush_block(out10)\r\n"));
			return -1;
		}

        copy_block(buf, (unsigned)stored_len, 0); /* without header */
        compressed_len = stored_len << 3;
        *file_method = STORED;

#ifdef FORCE_METHOD
    } else if (level == 2 && buf != (char*)0) { /* force stored block */
#else
    } else if (stored_len+4 <= opt_lenb && buf != (char*)0) {
                       /* 4: two words for the lengths */
#endif
        /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
         * Otherwise we can't have processed more than WSIZE input bytes since
         * the last block flush, because compression would have been
         * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
         * transform a block into a stored block.
         */
        send_bits((STORED_BLOCK<<1)+eof, 3);  /* send block type */
        compressed_len = (compressed_len + 3 + 7) & ~7L;
        compressed_len += (stored_len + 4) << 3;

        copy_block(buf, (unsigned)stored_len, 1); /* with header */

#ifdef FORCE_METHOD
    } else if (level == 3) { /* force static trees */
#else
    } else if (static_lenb == opt_lenb) {
#endif
        send_bits((STATIC_TREES<<1)+eof, 3);
        compress_block((ct_data near *)static_ltree, (ct_data near *)static_dtree);
        compressed_len += 3 + static_len;
    } else {
        send_bits((DYN_TREES<<1)+eof, 3);
        send_all_trees(l_desc.max_code+1, d_desc.max_code+1, max_blindex+1);
        compress_block((ct_data near *)dyn_ltree, (ct_data near *)dyn_dtree);
        compressed_len += 3 + opt_len;
    }
    Assert (compressed_len == bits_sent, "bad compressed size");
    init_block();

    if (eof) {
        Assert (input_len == isize, "bad input size");
        bi_windup();
        compressed_len += 7;  /* align on byte boundary */
    }
    Tracev((stderr,"\ncomprlen %lu(%lu) ", compressed_len>>3,
           compressed_len-7*eof));

    DBGPrintfo(("flush_block(out1)\r\n"));
    return  compressed_len >> 3;

  }
}

/* ===========================================================================
 * Save the match info and tally the frequency counts. Return true if
 * the current block must be flushed.
 */
int ct_tally (dist, lc)
    int dist;  /* distance of matched string */
    int lc;    /* match length-MIN_MATCH or unmatched char (if dist==0) */
{
  DBGPrintfi(("ct_tally(In)\r\n"));
  {
    l_buf[last_lit++] = (uch)lc;
    if (dist == 0) {
        /* lc is the unmatched char */
        dyn_ltree[lc].Freq++;
    } else {
        /* Here, lc is the match length - MIN_MATCH */
        dist--;             /* dist = match distance - 1 */
        Assert((ush)dist < (ush)MAX_DIST &&
               (ush)lc <= (ush)(MAX_MATCH-MIN_MATCH) &&
               (ush)d_code(dist) < (ush)D_CODES,  "ct_tally: bad match");

        dyn_ltree[length_code[lc]+LITERALS+1].Freq++;
        dyn_dtree[d_code(dist)].Freq++;

        d_buf[last_dist++] = (ush)dist;
        flags |= flag_bit;
    }
    flag_bit <<= 1;

    /* Output the flags if they fill a byte: */
    if ((last_lit & 7) == 0) {
        flag_buf[last_flags++] = flags;
        flags = 0, flag_bit = 1;
    }
    /* Try to guess if it is profitable to stop the current block here */
    if (level > 2 && (last_lit & 0xfff) == 0) {
        /* Compute an upper bound for the compressed length */
        ulg out_length = (ulg)last_lit*8L;
        ulg in_length = (ulg)strstart-block_start;
        int dcode;
        for (dcode = 0; dcode < D_CODES; dcode++) {
            out_length += (ulg)dyn_dtree[dcode].Freq*(5L+extra_dbits[dcode]);
        }
        out_length >>= 3;
        Trace((stderr,"\nlast_lit %u, last_dist %u, in %ld, out ~%ld(%ld%%) ",
               last_lit, last_dist, in_length, out_length,
               100L - out_length*100L/in_length));
        if (last_dist < last_lit/2 && out_length < in_length/2) return 1;
    }
    DBGPrintfo(("ct_tally(out1)\r\n"));
    return  (last_lit == LIT_BUFSIZE-1 || last_dist == DIST_BUFSIZE);
    /* We avoid equality with LIT_BUFSIZE because of wraparound at 64K
     * on 16 bit machines and because stored blocks are restricted to
     * 64K-1 bytes.
     */

  }
}

/* ===========================================================================
 * Send the block data compressed using the given Huffman trees
 */
local void compress_block(ltree, dtree)
    ct_data near *ltree; /* literal tree */
    ct_data near *dtree; /* distance tree */
{
  DBGPrintfi(("compress_block(In)\r\n"));
  {
    unsigned dist;      /* distance of matched string */
    int lc;             /* match length or unmatched char (if dist == 0) */
    unsigned lx = 0;    /* running index in l_buf */
    unsigned dx = 0;    /* running index in d_buf */
    unsigned fx = 0;    /* running index in flag_buf */
    uch flag = 0;       /* current flags */
    unsigned code;      /* the code to send */
    int extra;          /* number of extra bits to send */

    if (last_lit != 0) do {
        if ((lx & 7) == 0) flag = flag_buf[fx++];
        lc = l_buf[lx++];
        if ((flag & 1) == 0) {
            send_code(lc, ltree); /* send a literal byte */
            Tracecv(isgraph(lc), (stderr," '%c' ", lc));
        } else {
            /* Here, lc is the match length - MIN_MATCH */
            code = length_code[lc];
            send_code(code+LITERALS+1, ltree); /* send the length code */
            extra = extra_lbits[code];
            if (extra != 0) {
                lc -= base_length[code];
                send_bits(lc, extra);        /* send the extra length bits */
            }
            dist = d_buf[dx++];
            /* Here, dist is the match distance - 1 */
            code = d_code(dist);
            Assert (code < D_CODES, "bad d_code");

            send_code(code, dtree);       /* send the distance code */
            extra = extra_dbits[code];
            if (extra != 0) {
                dist -= base_dist[code];
                send_bits(dist, extra);   /* send the extra distance bits */
            }
        } /* literal or match pair ? */
        flag >>= 1;
    } while (lx < last_lit);

    send_code(END_BLOCK, ltree);

  }
  DBGPrintfo(("compress_block(out)\r\n"));
}

/* ===========================================================================
 * Set the file type to ASCII or BINARY, using a crude approximation:
 * binary if more than 20% of the bytes are <= 6 or >= 128, ascii otherwise.
 * IN assertion: the fields freq of dyn_ltree are set and the total of all
 * frequencies does not exceed 64K (to fit in an int on 16 bit machines).
 */
local void set_file_type()
{
  DBGPrintfi(("set_file_type(In)\r\n"));
  {
    int n = 0;
    unsigned ascii_freq = 0;
    unsigned bin_freq = 0;
    while (n < 7)        bin_freq += dyn_ltree[n++].Freq;
    while (n < 128)    ascii_freq += dyn_ltree[n++].Freq;
    while (n < LITERALS) bin_freq += dyn_ltree[n++].Freq;
    *file_type = bin_freq > (ascii_freq >> 2) ? BINARY : ASCII;
    if (*file_type == BINARY && translate_eol) {
        warn("-l used on binary file", "");
    }

  }
  DBGPrintfo(("set_file_type(out)\r\n"));
}