/* deflate.h -- internal compression state | |

* Copyright (C) 1995 Jean-loup Gailly | |

* For conditions of distribution and use, see copyright notice in zlib.h | |

*/ | |

/* WARNING: this file should *not* be used by applications. It is | |

part of the implementation of the compression library and is | |

subject to change. Applications should only use zlib.h. | |

*/ | |

/* $Id: deflate.h,v 1.5 1995/05/03 17:27:09 jloup Exp $ */ | |

#include "zutil.h" | |

/* =========================================================================== | |

* Internal compression state. | |

*/ | |

/* Data type */ | |

#define BINARY 0 | |

#define ASCII 1 | |

#define UNKNOWN 2 | |

#define LENGTH_CODES 29 | |

/* number of length codes, not counting the special END_BLOCK code */ | |

#define LITERALS 256 | |

/* number of literal bytes 0..255 */ | |

#define L_CODES (LITERALS+1+LENGTH_CODES) | |

/* number of Literal or Length codes, including the END_BLOCK code */ | |

#define D_CODES 30 | |

/* number of distance codes */ | |

#define BL_CODES 19 | |

/* number of codes used to transfer the bit lengths */ | |

#define HEAP_SIZE (2*L_CODES+1) | |

/* maximum heap size */ | |

#define MAX_BITS 15 | |

/* All codes must not exceed MAX_BITS bits */ | |

#define INIT_STATE 42 | |

#define BUSY_STATE 113 | |

#define FINISH_STATE 666 | |

/* Stream status */ | |

/* Data structure describing a single value and its code string. */ | |

typedef struct ct_data_s { | |

union { | |

ush freq; /* frequency count */ | |

ush code; /* bit string */ | |

} fc; | |

union { | |

ush dad; /* father node in Huffman tree */ | |

ush len; /* length of bit string */ | |

} dl; | |

} ct_data; | |

#define Freq fc.freq | |

#define Code fc.code | |

#define Dad dl.dad | |

#define Len dl.len | |

typedef struct static_tree_desc_s static_tree_desc; | |

typedef struct tree_desc_s { | |

ct_data *dyn_tree; /* the dynamic tree */ | |

int max_code; /* largest code with non zero frequency */ | |

static_tree_desc *stat_desc; /* the corresponding static tree */ | |

} tree_desc; | |

typedef ush Pos; | |

typedef unsigned IPos; | |

/* A Pos is an index in the character window. We use short instead of int to | |

* save space in the various tables. IPos is used only for parameter passing. | |

*/ | |

typedef struct internal_state { | |

z_stream *strm; /* pointer back to this zlib stream */ | |

int status; /* as the name implies */ | |

Byte *pending_buf; /* output still pending */ | |

Byte *pending_out; /* next pending byte to output to the stream */ | |

int pending; /* nb of bytes in the pending buffer */ | |

uLong adler; /* adler32 of uncompressed data */ | |

int noheader; /* suppress zlib header and adler32 */ | |

Byte data_type; /* UNKNOWN, BINARY or ASCII */ | |

Byte method; /* STORED (for zip only) or DEFLATED */ | |

/* used by deflate.c: */ | |

uInt w_size; /* LZ77 window size (32K by default) */ | |

uInt w_bits; /* log2(w_size) (8..16) */ | |

uInt w_mask; /* w_size - 1 */ | |

Byte *window; | |

/* Sliding window. Input bytes are read into the second half of the window, | |

* and move to the first half later to keep a dictionary of at least wSize | |

* bytes. With this organization, matches are limited to a distance of | |

* wSize-MAX_MATCH bytes, but this ensures that IO is always | |

* performed with a length multiple of the block size. Also, it limits | |

* the window size to 64K, which is quite useful on MSDOS. | |

* To do: use the user input buffer as sliding window. | |

*/ | |

ulg window_size; | |

/* Actual size of window: 2*wSize, except when the user input buffer | |

* is directly used as sliding window. | |

*/ | |

Pos *prev; | |

/* Link to older string with same hash index. To limit the size of this | |

* array to 64K, this link is maintained only for the last 32K strings. | |

* An index in this array is thus a window index modulo 32K. | |

*/ | |

Pos *head; /* Heads of the hash chains or NIL. */ | |

uInt ins_h; /* hash index of string to be inserted */ | |

uInt hash_size; /* number of elements in hash table */ | |

uInt hash_bits; /* log2(hash_size) */ | |

uInt hash_mask; /* hash_size-1 */ | |

uInt hash_shift; | |

/* Number of bits by which ins_h must be shifted at each input | |

* step. It must be such that after MIN_MATCH steps, the oldest | |

* byte no longer takes part in the hash key, that is: | |

* hash_shift * MIN_MATCH >= hash_bits | |

*/ | |

long block_start; | |

/* Window position at the beginning of the current output block. Gets | |

* negative when the window is moved backwards. | |

*/ | |

uInt match_length; /* length of best match */ | |

IPos prev_match; /* previous match */ | |

int match_available; /* set if previous match exists */ | |

uInt strstart; /* start of string to insert */ | |

uInt match_start; /* start of matching string */ | |

uInt lookahead; /* number of valid bytes ahead in window */ | |

uInt prev_length; | |

/* Length of the best match at previous step. Matches not greater than this | |

* are discarded. This is used in the lazy match evaluation. | |

*/ | |

uInt max_chain_length; | |

/* To speed up deflation, hash chains are never searched beyond this | |

* length. A higher limit improves compression ratio but degrades the | |

* speed. | |

*/ | |

uInt max_lazy_match; | |

/* Attempt to find a better match only when the current match is strictly | |

* smaller than this value. This mechanism is used only for compression | |

* levels >= 4. | |

*/ | |

# define max_insert_length max_lazy_match | |

/* Insert new strings in the hash table only if the match length is not | |

* greater than this length. This saves time but degrades compression. | |

* max_insert_length is used only for compression levels <= 3. | |

*/ | |

int level; /* compression level (1..9) */ | |

int strategy; /* favor or force Huffman coding*/ | |

uInt good_match; | |

/* Use a faster search when the previous match is longer than this */ | |

int nice_match; /* Stop searching when current match exceeds this */ | |

/* used by trees.c: */ | |

ct_data dyn_ltree[HEAP_SIZE]; /* literal and length tree */ | |

ct_data dyn_dtree[2*D_CODES+1]; /* distance tree */ | |

ct_data bl_tree[2*BL_CODES+1]; /* Huffman tree for the bit lengths */ | |

tree_desc l_desc; /* descriptor for literal tree */ | |

tree_desc d_desc; /* descriptor for distance tree */ | |

tree_desc bl_desc; /* descriptor for bit length tree */ | |

ush bl_count[MAX_BITS+1]; | |

/* number of codes at each bit length for an optimal tree */ | |

int heap[2*L_CODES+1]; /* heap used to build the Huffman trees */ | |

int heap_len; /* number of elements in the heap */ | |

int heap_max; /* element of largest frequency */ | |

/* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used. | |

* The same heap array is used to build all trees. | |

*/ | |

uch depth[2*L_CODES+1]; | |

/* Depth of each subtree used as tie breaker for trees of equal frequency | |

*/ | |

uch *l_buf; /* buffer for literals or lengths */ | |

uInt lit_bufsize; | |

/* Size of match buffer for literals/lengths. There are 4 reasons for | |

* limiting lit_bufsize to 64K: | |

* - frequencies can be kept in 16 bit counters | |

* - if compression is not successful for the first block, all input | |

* data is still in the window so we can still emit a stored block even | |

* when input comes from standard input. (This can also be done for | |

* all blocks if lit_bufsize is not greater than 32K.) | |

* - if compression is not successful for a file smaller than 64K, we can | |

* even emit a stored file instead of a stored block (saving 5 bytes). | |

* This is applicable only for zip (not gzip or zlib). | |

* - creating new Huffman trees less frequently may not provide fast | |

* adaptation to changes in the input data statistics. (Take for | |

* example a binary file with poorly compressible code followed by | |

* a highly compressible string table.) Smaller buffer sizes give | |

* fast adaptation but have of course the overhead of transmitting | |

* trees more frequently. | |

* - I can't count above 4 | |

*/ | |

uInt last_lit; /* running index in l_buf */ | |

ush *d_buf; | |

/* Buffer for distances. To simplify the code, d_buf and l_buf have | |

* the same number of elements. To use different lengths, an extra flag | |

* array would be necessary. | |

*/ | |

ulg opt_len; /* bit length of current block with optimal trees */ | |

ulg static_len; /* bit length of current block with static trees */ | |

ulg compressed_len; /* total bit length of compressed file */ | |

uInt matches; /* number of string matches in current block */ | |

#ifdef DEBUG | |

ulg bits_sent; /* bit length of the compressed data */ | |

#endif | |

ush bi_buf; | |

/* Output buffer. bits are inserted starting at the bottom (least | |

* significant bits). | |

*/ | |

int bi_valid; | |

/* Number of valid bits in bi_buf. All bits above the last valid bit | |

* are always zero. | |

*/ | |

} deflate_state; | |

/* Output a byte on the stream. | |

* IN assertion: there is enough room in pending_buf. | |

*/ | |

#define put_byte(s, c) {s->pending_buf[s->pending++] = (c);} | |

#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1) | |

/* Minimum amount of lookahead, except at the end of the input file. | |

* See deflate.c for comments about the MIN_MATCH+1. | |

*/ | |

#define MAX_DIST(s) ((s)->w_size-MIN_LOOKAHEAD) | |

/* In order to simplify the code, particularly on 16 bit machines, match | |

* distances are limited to MAX_DIST instead of WSIZE. | |

*/ | |

/* in trees.c */ | |

void ct_init __P((deflate_state *s)); | |

int ct_tally __P((deflate_state *s, int dist, int lc)); | |

ulg ct_flush_block __P((deflate_state *s, char *buf, ulg stored_len, int eof)); | |

void ct_align __P((deflate_state *s)); | |

void ct_stored_block __P((deflate_state *s, char *buf, ulg stored_len, | |

int eof)); |