File indexing completed on 2024-04-14 03:46:46

0001 /* deflate.h -- internal compression state
0002  * Copyright (C) 1995-2012 Jean-loup Gailly
0003  * For conditions of distribution and use, see copyright notice in zlib.h
0004  */
0005 
0006 /* WARNING: this file should *not* be used by applications. It is
0007    part of the implementation of the compression library and is
0008    subject to change. Applications should only use zlib.h.
0009  */
0010 
0011 /* @(#) $Id$ */
0012 
0013 #ifndef DEFLATE_H
0014 #define DEFLATE_H
0015 
0016 #include "zutil.h"
0017 
0018 /* define NO_GZIP when compiling if you want to disable gzip header and
0019    trailer creation by deflate().  NO_GZIP would be used to avoid linking in
0020    the crc code when it is not needed.  For shared libraries, gzip encoding
0021    should be left enabled. */
0022 #ifndef NO_GZIP
0023 #  define GZIP
0024 #endif
0025 
0026 /* ===========================================================================
0027  * Internal compression state.
0028  */
0029 
0030 #define LENGTH_CODES 29
0031 /* number of length codes, not counting the special END_BLOCK code */
0032 
0033 #define LITERALS  256
0034 /* number of literal bytes 0..255 */
0035 
0036 #define L_CODES (LITERALS+1+LENGTH_CODES)
0037 /* number of Literal or Length codes, including the END_BLOCK code */
0038 
0039 #define D_CODES   30
0040 /* number of distance codes */
0041 
0042 #define BL_CODES  19
0043 /* number of codes used to transfer the bit lengths */
0044 
0045 #define HEAP_SIZE (2*L_CODES+1)
0046 /* maximum heap size */
0047 
0048 #define MAX_BITS 15
0049 /* All codes must not exceed MAX_BITS bits */
0050 
0051 #define Buf_size 16
0052 /* size of bit buffer in bi_buf */
0053 
0054 #define INIT_STATE    42
0055 #define EXTRA_STATE   69
0056 #define NAME_STATE    73
0057 #define COMMENT_STATE 91
0058 #define HCRC_STATE   103
0059 #define BUSY_STATE   113
0060 #define FINISH_STATE 666
0061 /* Stream status */
0062 
0063 
0064 /* Data structure describing a single value and its code string. */
0065 typedef struct ct_data_s {
0066     union {
0067         ush  freq;       /* frequency count */
0068         ush  code;       /* bit string */
0069     } fc;
0070     union {
0071         ush  dad;        /* father node in Huffman tree */
0072         ush  len;        /* length of bit string */
0073     } dl;
0074 } FAR ct_data;
0075 
0076 #define Freq fc.freq
0077 #define Code fc.code
0078 #define Dad  dl.dad
0079 #define Len  dl.len
0080 
0081 typedef struct static_tree_desc_s  static_tree_desc;
0082 
0083 typedef struct tree_desc_s {
0084     ct_data *dyn_tree;           /* the dynamic tree */
0085     int     max_code;            /* largest code with non zero frequency */
0086     static_tree_desc *stat_desc; /* the corresponding static tree */
0087 } FAR tree_desc;
0088 
0089 typedef ush Pos;
0090 typedef Pos FAR Posf;
0091 typedef unsigned IPos;
0092 
0093 /* A Pos is an index in the character window. We use short instead of int to
0094  * save space in the various tables. IPos is used only for parameter passing.
0095  */
0096 
0097 typedef struct internal_state {
0098     z_streamp strm;      /* pointer back to this zlib stream */
0099     int   status;        /* as the name implies */
0100     Bytef *pending_buf;  /* output still pending */
0101     ulg   pending_buf_size; /* size of pending_buf */
0102     Bytef *pending_out;  /* next pending byte to output to the stream */
0103     uInt   pending;      /* nb of bytes in the pending buffer */
0104     int   wrap;          /* bit 0 true for zlib, bit 1 true for gzip */
0105     gz_headerp  gzhead;  /* gzip header information to write */
0106     uInt   gzindex;      /* where in extra, name, or comment */
0107     Byte  method;        /* can only be DEFLATED */
0108     int   last_flush;    /* value of flush param for previous deflate call */
0109 
0110                 /* used by deflate.c: */
0111 
0112     uInt  w_size;        /* LZ77 window size (32K by default) */
0113     uInt  w_bits;        /* log2(w_size)  (8..16) */
0114     uInt  w_mask;        /* w_size - 1 */
0115 
0116     Bytef *window;
0117     /* Sliding window. Input bytes are read into the second half of the window,
0118      * and move to the first half later to keep a dictionary of at least wSize
0119      * bytes. With this organization, matches are limited to a distance of
0120      * wSize-MAX_MATCH bytes, but this ensures that IO is always
0121      * performed with a length multiple of the block size. Also, it limits
0122      * the window size to 64K, which is quite useful on MSDOS.
0123      * To do: use the user input buffer as sliding window.
0124      */
0125 
0126     ulg window_size;
0127     /* Actual size of window: 2*wSize, except when the user input buffer
0128      * is directly used as sliding window.
0129      */
0130 
0131     Posf *prev;
0132     /* Link to older string with same hash index. To limit the size of this
0133      * array to 64K, this link is maintained only for the last 32K strings.
0134      * An index in this array is thus a window index modulo 32K.
0135      */
0136 
0137     Posf *head; /* Heads of the hash chains or NIL. */
0138 
0139     uInt  ins_h;          /* hash index of string to be inserted */
0140     uInt  hash_size;      /* number of elements in hash table */
0141     uInt  hash_bits;      /* log2(hash_size) */
0142     uInt  hash_mask;      /* hash_size-1 */
0143 
0144     uInt  hash_shift;
0145     /* Number of bits by which ins_h must be shifted at each input
0146      * step. It must be such that after MIN_MATCH steps, the oldest
0147      * byte no longer takes part in the hash key, that is:
0148      *   hash_shift * MIN_MATCH >= hash_bits
0149      */
0150 
0151     long block_start;
0152     /* Window position at the beginning of the current output block. Gets
0153      * negative when the window is moved backwards.
0154      */
0155 
0156     uInt match_length;           /* length of best match */
0157     IPos prev_match;             /* previous match */
0158     int match_available;         /* set if previous match exists */
0159     uInt strstart;               /* start of string to insert */
0160     uInt match_start;            /* start of matching string */
0161     uInt lookahead;              /* number of valid bytes ahead in window */
0162 
0163     uInt prev_length;
0164     /* Length of the best match at previous step. Matches not greater than this
0165      * are discarded. This is used in the lazy match evaluation.
0166      */
0167 
0168     uInt max_chain_length;
0169     /* To speed up deflation, hash chains are never searched beyond this
0170      * length.  A higher limit improves compression ratio but degrades the
0171      * speed.
0172      */
0173 
0174     uInt max_lazy_match;
0175     /* Attempt to find a better match only when the current match is strictly
0176      * smaller than this value. This mechanism is used only for compression
0177      * levels >= 4.
0178      */
0179 #   define max_insert_length  max_lazy_match
0180     /* Insert new strings in the hash table only if the match length is not
0181      * greater than this length. This saves time but degrades compression.
0182      * max_insert_length is used only for compression levels <= 3.
0183      */
0184 
0185     int level;    /* compression level (1..9) */
0186     int strategy; /* favor or force Huffman coding*/
0187 
0188     uInt good_match;
0189     /* Use a faster search when the previous match is longer than this */
0190 
0191     int nice_match; /* Stop searching when current match exceeds this */
0192 
0193                 /* used by trees.c: */
0194     /* Didn't use ct_data typedef below to suppress compiler warning */
0195     struct ct_data_s dyn_ltree[HEAP_SIZE];   /* literal and length tree */
0196     struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */
0197     struct ct_data_s bl_tree[2*BL_CODES+1];  /* Huffman tree for bit lengths */
0198 
0199     struct tree_desc_s l_desc;               /* desc. for literal tree */
0200     struct tree_desc_s d_desc;               /* desc. for distance tree */
0201     struct tree_desc_s bl_desc;              /* desc. for bit length tree */
0202 
0203     ush bl_count[MAX_BITS+1];
0204     /* number of codes at each bit length for an optimal tree */
0205 
0206     int heap[2*L_CODES+1];      /* heap used to build the Huffman trees */
0207     int heap_len;               /* number of elements in the heap */
0208     int heap_max;               /* element of largest frequency */
0209     /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
0210      * The same heap array is used to build all trees.
0211      */
0212 
0213     uch depth[2*L_CODES+1];
0214     /* Depth of each subtree used as tie breaker for trees of equal frequency
0215      */
0216 
0217     uchf *l_buf;          /* buffer for literals or lengths */
0218 
0219     uInt  lit_bufsize;
0220     /* Size of match buffer for literals/lengths.  There are 4 reasons for
0221      * limiting lit_bufsize to 64K:
0222      *   - frequencies can be kept in 16 bit counters
0223      *   - if compression is not successful for the first block, all input
0224      *     data is still in the window so we can still emit a stored block even
0225      *     when input comes from standard input.  (This can also be done for
0226      *     all blocks if lit_bufsize is not greater than 32K.)
0227      *   - if compression is not successful for a file smaller than 64K, we can
0228      *     even emit a stored file instead of a stored block (saving 5 bytes).
0229      *     This is applicable only for zip (not gzip or zlib).
0230      *   - creating new Huffman trees less frequently may not provide fast
0231      *     adaptation to changes in the input data statistics. (Take for
0232      *     example a binary file with poorly compressible code followed by
0233      *     a highly compressible string table.) Smaller buffer sizes give
0234      *     fast adaptation but have of course the overhead of transmitting
0235      *     trees more frequently.
0236      *   - I can't count above 4
0237      */
0238 
0239     uInt last_lit;      /* running index in l_buf */
0240 
0241     ushf *d_buf;
0242     /* Buffer for distances. To simplify the code, d_buf and l_buf have
0243      * the same number of elements. To use different lengths, an extra flag
0244      * array would be necessary.
0245      */
0246 
0247     ulg opt_len;        /* bit length of current block with optimal trees */
0248     ulg static_len;     /* bit length of current block with static trees */
0249     uInt matches;       /* number of string matches in current block */
0250     uInt insert;        /* bytes at end of window left to insert */
0251 
0252 #ifdef DEBUG
0253     ulg compressed_len; /* total bit length of compressed file mod 2^32 */
0254     ulg bits_sent;      /* bit length of compressed data sent mod 2^32 */
0255 #endif
0256 
0257     ush bi_buf;
0258     /* Output buffer. bits are inserted starting at the bottom (least
0259      * significant bits).
0260      */
0261     int bi_valid;
0262     /* Number of valid bits in bi_buf.  All bits above the last valid bit
0263      * are always zero.
0264      */
0265 
0266     ulg high_water;
0267     /* High water mark offset in window for initialized bytes -- bytes above
0268      * this are set to zero in order to avoid memory check warnings when
0269      * longest match routines access bytes past the input.  This is then
0270      * updated to the new high water mark.
0271      */
0272 
0273 } FAR deflate_state;
0274 
0275 /* Output a byte on the stream.
0276  * IN assertion: there is enough room in pending_buf.
0277  */
0278 #define put_byte(s, c) {s->pending_buf[s->pending++] = (c);}
0279 
0280 
0281 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
0282 /* Minimum amount of lookahead, except at the end of the input file.
0283  * See deflate.c for comments about the MIN_MATCH+1.
0284  */
0285 
0286 #define MAX_DIST(s)  ((s)->w_size-MIN_LOOKAHEAD)
0287 /* In order to simplify the code, particularly on 16 bit machines, match
0288  * distances are limited to MAX_DIST instead of WSIZE.
0289  */
0290 
0291 #define WIN_INIT MAX_MATCH
0292 /* Number of bytes after end of data in window to initialize in order to avoid
0293    memory checker errors from longest match routines */
0294 
0295         /* in trees.c */
0296 void ZLIB_INTERNAL _tr_init OF((deflate_state *s));
0297 int ZLIB_INTERNAL _tr_tally OF((deflate_state *s, unsigned dist, unsigned lc));
0298 void ZLIB_INTERNAL _tr_flush_block OF((deflate_state *s, charf *buf,
0299                         ulg stored_len, int last));
0300 void ZLIB_INTERNAL _tr_flush_bits OF((deflate_state *s));
0301 void ZLIB_INTERNAL _tr_align OF((deflate_state *s));
0302 void ZLIB_INTERNAL _tr_stored_block OF((deflate_state *s, charf *buf,
0303                         ulg stored_len, int last));
0304 
0305 #define d_code(dist) \
0306    ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)])
0307 /* Mapping from a distance to a distance code. dist is the distance - 1 and
0308  * must not have side effects. _dist_code[256] and _dist_code[257] are never
0309  * used.
0310  */
0311 
0312 #ifndef DEBUG
0313 /* Inline versions of _tr_tally for speed: */
0314 
0315 #if defined(GEN_TREES_H) || !defined(STDC)
0316   extern uch ZLIB_INTERNAL _length_code[];
0317   extern uch ZLIB_INTERNAL _dist_code[];
0318 #else
0319   extern const uch ZLIB_INTERNAL _length_code[];
0320   extern const uch ZLIB_INTERNAL _dist_code[];
0321 #endif
0322 
0323 # define _tr_tally_lit(s, c, flush) \
0324   { uch cc = (c); \
0325     s->d_buf[s->last_lit] = 0; \
0326     s->l_buf[s->last_lit++] = cc; \
0327     s->dyn_ltree[cc].Freq++; \
0328     flush = (s->last_lit == s->lit_bufsize-1); \
0329    }
0330 # define _tr_tally_dist(s, distance, length, flush) \
0331   { uch len = (length); \
0332     ush dist = (distance); \
0333     s->d_buf[s->last_lit] = dist; \
0334     s->l_buf[s->last_lit++] = len; \
0335     dist--; \
0336     s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
0337     s->dyn_dtree[d_code(dist)].Freq++; \
0338     flush = (s->last_lit == s->lit_bufsize-1); \
0339   }
0340 #else
0341 # define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c)
0342 # define _tr_tally_dist(s, distance, length, flush) \
0343               flush = _tr_tally(s, distance, length)
0344 #endif
0345 
0346 #endif /* DEFLATE_H */