nuclear@2: /* deflate.h -- internal compression state nuclear@2: * Copyright (C) 1995-2004 Jean-loup Gailly nuclear@2: * For conditions of distribution and use, see copyright notice in zlib.h nuclear@2: */ nuclear@2: nuclear@2: /* WARNING: this file should *not* be used by applications. It is nuclear@2: part of the implementation of the compression library and is nuclear@2: subject to change. Applications should only use zlib.h. nuclear@2: */ nuclear@2: nuclear@2: /* @(#) $Id$ */ nuclear@2: nuclear@2: #ifndef DEFLATE_H nuclear@2: #define DEFLATE_H nuclear@2: nuclear@2: #include "zutil.h" nuclear@2: nuclear@2: /* define NO_GZIP when compiling if you want to disable gzip header and nuclear@2: trailer creation by deflate(). NO_GZIP would be used to avoid linking in nuclear@2: the crc code when it is not needed. For shared libraries, gzip encoding nuclear@2: should be left enabled. */ nuclear@2: #ifndef NO_GZIP nuclear@2: # define GZIP nuclear@2: #endif nuclear@2: nuclear@2: /* =========================================================================== nuclear@2: * Internal compression state. nuclear@2: */ nuclear@2: nuclear@2: #define LENGTH_CODES 29 nuclear@2: /* number of length codes, not counting the special END_BLOCK code */ nuclear@2: nuclear@2: #define LITERALS 256 nuclear@2: /* number of literal bytes 0..255 */ nuclear@2: nuclear@2: #define L_CODES (LITERALS+1+LENGTH_CODES) nuclear@2: /* number of Literal or Length codes, including the END_BLOCK code */ nuclear@2: nuclear@2: #define D_CODES 30 nuclear@2: /* number of distance codes */ nuclear@2: nuclear@2: #define BL_CODES 19 nuclear@2: /* number of codes used to transfer the bit lengths */ nuclear@2: nuclear@2: #define HEAP_SIZE (2*L_CODES+1) nuclear@2: /* maximum heap size */ nuclear@2: nuclear@2: #define MAX_BITS 15 nuclear@2: /* All codes must not exceed MAX_BITS bits */ nuclear@2: nuclear@2: #define INIT_STATE 42 nuclear@2: #define EXTRA_STATE 69 nuclear@2: #define NAME_STATE 73 nuclear@2: #define COMMENT_STATE 91 nuclear@2: #define HCRC_STATE 103 nuclear@2: #define BUSY_STATE 113 nuclear@2: #define FINISH_STATE 666 nuclear@2: /* Stream status */ nuclear@2: nuclear@2: nuclear@2: /* Data structure describing a single value and its code string. */ nuclear@2: typedef struct ct_data_s { nuclear@2: union { nuclear@2: ush freq; /* frequency count */ nuclear@2: ush code; /* bit string */ nuclear@2: } fc; nuclear@2: union { nuclear@2: ush dad; /* father node in Huffman tree */ nuclear@2: ush len; /* length of bit string */ nuclear@2: } dl; nuclear@2: } FAR ct_data; nuclear@2: nuclear@2: #define Freq fc.freq nuclear@2: #define Code fc.code nuclear@2: #define Dad dl.dad nuclear@2: #define Len dl.len nuclear@2: nuclear@2: typedef struct static_tree_desc_s static_tree_desc; nuclear@2: nuclear@2: typedef struct tree_desc_s { nuclear@2: ct_data *dyn_tree; /* the dynamic tree */ nuclear@2: int max_code; /* largest code with non zero frequency */ nuclear@2: static_tree_desc *stat_desc; /* the corresponding static tree */ nuclear@2: } FAR tree_desc; nuclear@2: nuclear@2: typedef ush Pos; nuclear@2: typedef Pos FAR Posf; nuclear@2: typedef unsigned IPos; nuclear@2: nuclear@2: /* A Pos is an index in the character window. We use short instead of int to nuclear@2: * save space in the various tables. IPos is used only for parameter passing. nuclear@2: */ nuclear@2: nuclear@2: typedef struct internal_state { nuclear@2: z_streamp strm; /* pointer back to this zlib stream */ nuclear@2: int status; /* as the name implies */ nuclear@2: Bytef *pending_buf; /* output still pending */ nuclear@2: ulg pending_buf_size; /* size of pending_buf */ nuclear@2: Bytef *pending_out; /* next pending byte to output to the stream */ nuclear@2: uInt pending; /* nb of bytes in the pending buffer */ nuclear@2: int wrap; /* bit 0 true for zlib, bit 1 true for gzip */ nuclear@2: gz_headerp gzhead; /* gzip header information to write */ nuclear@2: uInt gzindex; /* where in extra, name, or comment */ nuclear@2: Byte method; /* STORED (for zip only) or DEFLATED */ nuclear@2: int last_flush; /* value of flush param for previous deflate call */ nuclear@2: nuclear@2: /* used by deflate.c: */ nuclear@2: nuclear@2: uInt w_size; /* LZ77 window size (32K by default) */ nuclear@2: uInt w_bits; /* log2(w_size) (8..16) */ nuclear@2: uInt w_mask; /* w_size - 1 */ nuclear@2: nuclear@2: Bytef *window; nuclear@2: /* Sliding window. Input bytes are read into the second half of the window, nuclear@2: * and move to the first half later to keep a dictionary of at least wSize nuclear@2: * bytes. With this organization, matches are limited to a distance of nuclear@2: * wSize-MAX_MATCH bytes, but this ensures that IO is always nuclear@2: * performed with a length multiple of the block size. Also, it limits nuclear@2: * the window size to 64K, which is quite useful on MSDOS. nuclear@2: * To do: use the user input buffer as sliding window. nuclear@2: */ nuclear@2: nuclear@2: ulg window_size; nuclear@2: /* Actual size of window: 2*wSize, except when the user input buffer nuclear@2: * is directly used as sliding window. nuclear@2: */ nuclear@2: nuclear@2: Posf *prev; nuclear@2: /* Link to older string with same hash index. To limit the size of this nuclear@2: * array to 64K, this link is maintained only for the last 32K strings. nuclear@2: * An index in this array is thus a window index modulo 32K. nuclear@2: */ nuclear@2: nuclear@2: Posf *head; /* Heads of the hash chains or NIL. */ nuclear@2: nuclear@2: uInt ins_h; /* hash index of string to be inserted */ nuclear@2: uInt hash_size; /* number of elements in hash table */ nuclear@2: uInt hash_bits; /* log2(hash_size) */ nuclear@2: uInt hash_mask; /* hash_size-1 */ nuclear@2: nuclear@2: uInt hash_shift; nuclear@2: /* Number of bits by which ins_h must be shifted at each input nuclear@2: * step. It must be such that after MIN_MATCH steps, the oldest nuclear@2: * byte no longer takes part in the hash key, that is: nuclear@2: * hash_shift * MIN_MATCH >= hash_bits nuclear@2: */ nuclear@2: nuclear@2: long block_start; nuclear@2: /* Window position at the beginning of the current output block. Gets nuclear@2: * negative when the window is moved backwards. nuclear@2: */ nuclear@2: nuclear@2: uInt match_length; /* length of best match */ nuclear@2: IPos prev_match; /* previous match */ nuclear@2: int match_available; /* set if previous match exists */ nuclear@2: uInt strstart; /* start of string to insert */ nuclear@2: uInt match_start; /* start of matching string */ nuclear@2: uInt lookahead; /* number of valid bytes ahead in window */ nuclear@2: nuclear@2: uInt prev_length; nuclear@2: /* Length of the best match at previous step. Matches not greater than this nuclear@2: * are discarded. This is used in the lazy match evaluation. nuclear@2: */ nuclear@2: nuclear@2: uInt max_chain_length; nuclear@2: /* To speed up deflation, hash chains are never searched beyond this nuclear@2: * length. A higher limit improves compression ratio but degrades the nuclear@2: * speed. nuclear@2: */ nuclear@2: nuclear@2: uInt max_lazy_match; nuclear@2: /* Attempt to find a better match only when the current match is strictly nuclear@2: * smaller than this value. This mechanism is used only for compression nuclear@2: * levels >= 4. nuclear@2: */ nuclear@2: # define max_insert_length max_lazy_match nuclear@2: /* Insert new strings in the hash table only if the match length is not nuclear@2: * greater than this length. This saves time but degrades compression. nuclear@2: * max_insert_length is used only for compression levels <= 3. nuclear@2: */ nuclear@2: nuclear@2: int level; /* compression level (1..9) */ nuclear@2: int strategy; /* favor or force Huffman coding*/ nuclear@2: nuclear@2: uInt good_match; nuclear@2: /* Use a faster search when the previous match is longer than this */ nuclear@2: nuclear@2: int nice_match; /* Stop searching when current match exceeds this */ nuclear@2: nuclear@2: /* used by trees.c: */ nuclear@2: /* Didn't use ct_data typedef below to supress compiler warning */ nuclear@2: struct ct_data_s dyn_ltree[HEAP_SIZE]; /* literal and length tree */ nuclear@2: struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */ nuclear@2: struct ct_data_s bl_tree[2*BL_CODES+1]; /* Huffman tree for bit lengths */ nuclear@2: nuclear@2: struct tree_desc_s l_desc; /* desc. for literal tree */ nuclear@2: struct tree_desc_s d_desc; /* desc. for distance tree */ nuclear@2: struct tree_desc_s bl_desc; /* desc. for bit length tree */ nuclear@2: nuclear@2: ush bl_count[MAX_BITS+1]; nuclear@2: /* number of codes at each bit length for an optimal tree */ nuclear@2: nuclear@2: int heap[2*L_CODES+1]; /* heap used to build the Huffman trees */ nuclear@2: int heap_len; /* number of elements in the heap */ nuclear@2: int heap_max; /* element of largest frequency */ nuclear@2: /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used. nuclear@2: * The same heap array is used to build all trees. nuclear@2: */ nuclear@2: nuclear@2: uch depth[2*L_CODES+1]; nuclear@2: /* Depth of each subtree used as tie breaker for trees of equal frequency nuclear@2: */ nuclear@2: nuclear@2: uchf *l_buf; /* buffer for literals or lengths */ nuclear@2: nuclear@2: uInt lit_bufsize; nuclear@2: /* Size of match buffer for literals/lengths. There are 4 reasons for nuclear@2: * limiting lit_bufsize to 64K: nuclear@2: * - frequencies can be kept in 16 bit counters nuclear@2: * - if compression is not successful for the first block, all input nuclear@2: * data is still in the window so we can still emit a stored block even nuclear@2: * when input comes from standard input. (This can also be done for nuclear@2: * all blocks if lit_bufsize is not greater than 32K.) nuclear@2: * - if compression is not successful for a file smaller than 64K, we can nuclear@2: * even emit a stored file instead of a stored block (saving 5 bytes). nuclear@2: * This is applicable only for zip (not gzip or zlib). nuclear@2: * - creating new Huffman trees less frequently may not provide fast nuclear@2: * adaptation to changes in the input data statistics. (Take for nuclear@2: * example a binary file with poorly compressible code followed by nuclear@2: * a highly compressible string table.) Smaller buffer sizes give nuclear@2: * fast adaptation but have of course the overhead of transmitting nuclear@2: * trees more frequently. nuclear@2: * - I can't count above 4 nuclear@2: */ nuclear@2: nuclear@2: uInt last_lit; /* running index in l_buf */ nuclear@2: nuclear@2: ushf *d_buf; nuclear@2: /* Buffer for distances. To simplify the code, d_buf and l_buf have nuclear@2: * the same number of elements. To use different lengths, an extra flag nuclear@2: * array would be necessary. nuclear@2: */ nuclear@2: nuclear@2: ulg opt_len; /* bit length of current block with optimal trees */ nuclear@2: ulg static_len; /* bit length of current block with static trees */ nuclear@2: uInt matches; /* number of string matches in current block */ nuclear@2: int last_eob_len; /* bit length of EOB code for last block */ nuclear@2: nuclear@2: #ifdef DEBUG nuclear@2: ulg compressed_len; /* total bit length of compressed file mod 2^32 */ nuclear@2: ulg bits_sent; /* bit length of compressed data sent mod 2^32 */ nuclear@2: #endif nuclear@2: nuclear@2: ush bi_buf; nuclear@2: /* Output buffer. bits are inserted starting at the bottom (least nuclear@2: * significant bits). nuclear@2: */ nuclear@2: int bi_valid; nuclear@2: /* Number of valid bits in bi_buf. All bits above the last valid bit nuclear@2: * are always zero. nuclear@2: */ nuclear@2: nuclear@2: } FAR deflate_state; nuclear@2: nuclear@2: /* Output a byte on the stream. nuclear@2: * IN assertion: there is enough room in pending_buf. nuclear@2: */ nuclear@2: #define put_byte(s, c) {s->pending_buf[s->pending++] = (c);} nuclear@2: nuclear@2: nuclear@2: #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1) nuclear@2: /* Minimum amount of lookahead, except at the end of the input file. nuclear@2: * See deflate.c for comments about the MIN_MATCH+1. nuclear@2: */ nuclear@2: nuclear@2: #define MAX_DIST(s) ((s)->w_size-MIN_LOOKAHEAD) nuclear@2: /* In order to simplify the code, particularly on 16 bit machines, match nuclear@2: * distances are limited to MAX_DIST instead of WSIZE. nuclear@2: */ nuclear@2: nuclear@2: /* in trees.c */ nuclear@2: void _tr_init OF((deflate_state *s)); nuclear@2: int _tr_tally OF((deflate_state *s, unsigned dist, unsigned lc)); nuclear@2: void _tr_flush_block OF((deflate_state *s, charf *buf, ulg stored_len, nuclear@2: int eof)); nuclear@2: void _tr_align OF((deflate_state *s)); nuclear@2: void _tr_stored_block OF((deflate_state *s, charf *buf, ulg stored_len, nuclear@2: int eof)); nuclear@2: nuclear@2: #define d_code(dist) \ nuclear@2: ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)]) nuclear@2: /* Mapping from a distance to a distance code. dist is the distance - 1 and nuclear@2: * must not have side effects. _dist_code[256] and _dist_code[257] are never nuclear@2: * used. nuclear@2: */ nuclear@2: nuclear@2: #ifndef DEBUG nuclear@2: /* Inline versions of _tr_tally for speed: */ nuclear@2: nuclear@2: #if defined(GEN_TREES_H) || !defined(STDC) nuclear@2: extern uch _length_code[]; nuclear@2: extern uch _dist_code[]; nuclear@2: #else nuclear@2: extern const uch _length_code[]; nuclear@2: extern const uch _dist_code[]; nuclear@2: #endif nuclear@2: nuclear@2: # define _tr_tally_lit(s, c, flush) \ nuclear@2: { uch cc = (c); \ nuclear@2: s->d_buf[s->last_lit] = 0; \ nuclear@2: s->l_buf[s->last_lit++] = cc; \ nuclear@2: s->dyn_ltree[cc].Freq++; \ nuclear@2: flush = (s->last_lit == s->lit_bufsize-1); \ nuclear@2: } nuclear@2: # define _tr_tally_dist(s, distance, length, flush) \ nuclear@2: { uch len = (length); \ nuclear@2: ush dist = (distance); \ nuclear@2: s->d_buf[s->last_lit] = dist; \ nuclear@2: s->l_buf[s->last_lit++] = len; \ nuclear@2: dist--; \ nuclear@2: s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \ nuclear@2: s->dyn_dtree[d_code(dist)].Freq++; \ nuclear@2: flush = (s->last_lit == s->lit_bufsize-1); \ nuclear@2: } nuclear@2: #else nuclear@2: # define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c) nuclear@2: # define _tr_tally_dist(s, distance, length, flush) \ nuclear@2: flush = _tr_tally(s, distance, length) nuclear@2: #endif nuclear@2: nuclear@2: #endif /* DEFLATE_H */