tif_lzw.c

Go to the documentation of this file.

/* $Id: tif_lzw.c,v 1.28 2006/03/16 12:38:24 dron Exp $ */

/*
 * Copyright (c) 1988-1997 Sam Leffler
 * Copyright (c) 1991-1997 Silicon Graphics, Inc.
 *
 * Permission to use, copy, modify, distribute, and sell this software and 
 * its documentation for any purpose is hereby granted without fee, provided
 * that (i) the above copyright notices and this permission notice appear in
 * all copies of the software and related documentation, and (ii) the names of
 * Sam Leffler and Silicon Graphics may not be used in any advertising or
 * publicity relating to the software without the specific, prior written
 * permission of Sam Leffler and Silicon Graphics.
 * 
 * THE SOFTWARE IS PROVIDED "AS-IS" AND WITHOUT WARRANTY OF ANY KIND, 
 * EXPRESS, IMPLIED OR OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY 
 * WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.  
 * 
 * IN NO EVENT SHALL SAM LEFFLER OR SILICON GRAPHICS BE LIABLE FOR
 * ANY SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY KIND,
 * OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
 * WHETHER OR NOT ADVISED OF THE POSSIBILITY OF DAMAGE, AND ON ANY THEORY OF 
 * LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE 
 * OF THIS SOFTWARE.
 */

#include "tiffiop.h"
#ifdef LZW_SUPPORT
/*
 * TIFF Library.
 * Rev 5.0 Lempel-Ziv & Welch Compression Support
 *
 * This code is derived from the compress program whose code is
 * derived from software contributed to Berkeley by James A. Woods,
 * derived from original work by Spencer Thomas and Joseph Orost.
 *
 * The original Berkeley copyright notice appears below in its entirety.
 */
#include "tif_predict.h"

#include <stdio.h>

/*
 * NB: The 5.0 spec describes a different algorithm than Aldus
 *     implements.  Specifically, Aldus does code length transitions
 *     one code earlier than should be done (for real LZW).
 *     Earlier versions of this library implemented the correct
 *     LZW algorithm, but emitted codes in a bit order opposite
 *     to the TIFF spec.  Thus, to maintain compatibility w/ Aldus
 *     we interpret MSB-LSB ordered codes to be images written w/
 *     old versions of this library, but otherwise adhere to the
 *     Aldus "off by one" algorithm.
 *
 * Future revisions to the TIFF spec are expected to "clarify this issue".
 */
#define LZW_COMPAT      /* include backwards compatibility code */
/*
 * Each strip of data is supposed to be terminated by a CODE_EOI.
 * If the following #define is included, the decoder will also
 * check for end-of-strip w/o seeing this code.  This makes the
 * library more robust, but also slower.
 */
#define LZW_CHECKEOS        /* include checks for strips w/o EOI code */

#define MAXCODE(n)  ((1L<<(n))-1)
/*
 * The TIFF spec specifies that encoded bit
 * strings range from 9 to 12 bits.
 */
#define BITS_MIN    9       /* start with 9 bits */
#define BITS_MAX    12      /* max of 12 bit strings */
/* predefined codes */
#define CODE_CLEAR  256     /* code to clear string table */
#define CODE_EOI    257     /* end-of-information code */
#define CODE_FIRST  258     /* first free code entry */
#define CODE_MAX    MAXCODE(BITS_MAX)
#define HSIZE       9001L       /* 91% occupancy */
#define HSHIFT      (13-8)
#ifdef LZW_COMPAT
/* NB: +1024 is for compatibility with old files */
#define CSIZE       (MAXCODE(BITS_MAX)+1024L)
#else
#define CSIZE       (MAXCODE(BITS_MAX)+1L)
#endif

/*
 * State block for each open TIFF file using LZW
 * compression/decompression.  Note that the predictor
 * state block must be first in this data structure.
 */
typedef struct {
    TIFFPredictorState predict; /* predictor super class */

    unsigned short  nbits;      /* # of bits/code */
    unsigned short  maxcode;    /* maximum code for lzw_nbits */
    unsigned short  free_ent;   /* next free entry in hash table */
    long        nextdata;   /* next bits of i/o */
    long        nextbits;   /* # of valid bits in lzw_nextdata */

        int             rw_mode;        /* preserve rw_mode from init */
} LZWBaseState;

#define lzw_nbits   base.nbits
#define lzw_maxcode base.maxcode
#define lzw_free_ent    base.free_ent
#define lzw_nextdata    base.nextdata
#define lzw_nextbits    base.nextbits

/*
 * Encoding-specific state.
 */
typedef uint16 hcode_t;         /* codes fit in 16 bits */
typedef struct {
    long    hash;
    hcode_t code;
} hash_t;

/*
 * Decoding-specific state.
 */
typedef struct code_ent {
    struct code_ent *next;
    unsigned short  length;     /* string len, including this token */
    unsigned char   value;      /* data value */
    unsigned char   firstchar;  /* first token of string */
} code_t;

typedef int (*decodeFunc)(TIFF*, tidata_t, tsize_t, tsample_t);

typedef struct {
    LZWBaseState base;

    /* Decoding specific data */
    long    dec_nbitsmask;      /* lzw_nbits 1 bits, right adjusted */
    long    dec_restart;        /* restart count */
#ifdef LZW_CHECKEOS
    long    dec_bitsleft;       /* available bits in raw data */
#endif
    decodeFunc dec_decode;      /* regular or backwards compatible */
    code_t* dec_codep;      /* current recognized code */
    code_t* dec_oldcodep;       /* previously recognized code */
    code_t* dec_free_entp;      /* next free entry */
    code_t* dec_maxcodep;       /* max available entry */
    code_t* dec_codetab;        /* kept separate for small machines */

    /* Encoding specific data */
    int enc_oldcode;        /* last code encountered */
    long    enc_checkpoint;     /* point at which to clear table */
#define CHECK_GAP   10000       /* enc_ratio check interval */
    long    enc_ratio;      /* current compression ratio */
    long    enc_incount;        /* (input) data bytes encoded */
    long    enc_outcount;       /* encoded (output) bytes */
    tidata_t enc_rawlimit;      /* bound on tif_rawdata buffer */
    hash_t* enc_hashtab;        /* kept separate for small machines */
} LZWCodecState;

#define LZWState(tif)       ((LZWBaseState*) (tif)->tif_data)
#define DecoderState(tif)   ((LZWCodecState*) LZWState(tif))
#define EncoderState(tif)   ((LZWCodecState*) LZWState(tif))

static  int LZWDecode(TIFF*, tidata_t, tsize_t, tsample_t);
#ifdef LZW_COMPAT
static  int LZWDecodeCompat(TIFF*, tidata_t, tsize_t, tsample_t);
#endif
static  void cl_hash(LZWCodecState*);

/*
 * LZW Decoder.
 */

#ifdef LZW_CHECKEOS
/*
 * This check shouldn't be necessary because each
 * strip is suppose to be terminated with CODE_EOI.
 */
#define NextCode(_tif, _sp, _bp, _code, _get) {             \
    if ((_sp)->dec_bitsleft < nbits) {              \
        TIFFWarningExt(_tif->tif_clientdata, _tif->tif_name,                \
            "LZWDecode: Strip %d not terminated with EOI code", \
            _tif->tif_curstrip);                \
        _code = CODE_EOI;                   \
    } else {                            \
        _get(_sp,_bp,_code);                    \
        (_sp)->dec_bitsleft -= nbits;               \
    }                               \
}
#else
#define NextCode(tif, sp, bp, code, get) get(sp, bp, code)
#endif

static int
LZWSetupDecode(TIFF* tif)
{
    LZWCodecState* sp = DecoderState(tif);
    static const char module[] = " LZWSetupDecode";
    int code;

        if( sp == NULL )
        {
            /*
             * Allocate state block so tag methods have storage to record 
             * values.
             */
            tif->tif_data = (tidata_t) _TIFFmalloc(sizeof(LZWCodecState));
            if (tif->tif_data == NULL)
            {
                TIFFErrorExt(tif->tif_clientdata, "LZWPreDecode", "No space for LZW state block");
                return (0);
            }

            DecoderState(tif)->dec_codetab = NULL;
            DecoderState(tif)->dec_decode = NULL;
            
            /*
             * Setup predictor setup.
             */
            (void) TIFFPredictorInit(tif);

            sp = DecoderState(tif);
        }
            
    assert(sp != NULL);

    if (sp->dec_codetab == NULL) {
        sp->dec_codetab = (code_t*)_TIFFmalloc(CSIZE*sizeof (code_t));
        if (sp->dec_codetab == NULL) {
            TIFFErrorExt(tif->tif_clientdata, module, "No space for LZW code table");
            return (0);
        }
        /*
         * Pre-load the table.
         */
                code = 255;
                do {
                    sp->dec_codetab[code].value = code;
                    sp->dec_codetab[code].firstchar = code;
                    sp->dec_codetab[code].length = 1;
                    sp->dec_codetab[code].next = NULL;
                } while (code--);
    }
    return (1);
}

/*
 * Setup state for decoding a strip.
 */
static int
LZWPreDecode(TIFF* tif, tsample_t s)
{
    LZWCodecState *sp = DecoderState(tif);

    (void) s;
    assert(sp != NULL);
    /*
     * Check for old bit-reversed codes.
     */
    if (tif->tif_rawdata[0] == 0 && (tif->tif_rawdata[1] & 0x1)) {
#ifdef LZW_COMPAT
        if (!sp->dec_decode) {
            TIFFWarningExt(tif->tif_clientdata, tif->tif_name,
                "Old-style LZW codes, convert file");
            /*
             * Override default decoding methods with
             * ones that deal with the old coding.
             * Otherwise the predictor versions set
             * above will call the compatibility routines
             * through the dec_decode method.
             */
            tif->tif_decoderow = LZWDecodeCompat;
            tif->tif_decodestrip = LZWDecodeCompat;
            tif->tif_decodetile = LZWDecodeCompat;
            /*
             * If doing horizontal differencing, must
             * re-setup the predictor logic since we
             * switched the basic decoder methods...
             */
            (*tif->tif_setupdecode)(tif);
            sp->dec_decode = LZWDecodeCompat;
        }
        sp->lzw_maxcode = MAXCODE(BITS_MIN);
#else /* !LZW_COMPAT */
        if (!sp->dec_decode) {
            TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
                "Old-style LZW codes not supported");
            sp->dec_decode = LZWDecode;
        }
        return (0);
#endif/* !LZW_COMPAT */
    } else {
        sp->lzw_maxcode = MAXCODE(BITS_MIN)-1;
        sp->dec_decode = LZWDecode;
    }
    sp->lzw_nbits = BITS_MIN;
    sp->lzw_nextbits = 0;
    sp->lzw_nextdata = 0;

    sp->dec_restart = 0;
    sp->dec_nbitsmask = MAXCODE(BITS_MIN);
#ifdef LZW_CHECKEOS
    sp->dec_bitsleft = tif->tif_rawcc << 3;
#endif
    sp->dec_free_entp = sp->dec_codetab + CODE_FIRST;
    /*
     * Zero entries that are not yet filled in.  We do
     * this to guard against bogus input data that causes
     * us to index into undefined entries.  If you can
     * come up with a way to safely bounds-check input codes
     * while decoding then you can remove this operation.
     */
    _TIFFmemset(sp->dec_free_entp, 0, (CSIZE-CODE_FIRST)*sizeof (code_t));
    sp->dec_oldcodep = &sp->dec_codetab[-1];
    sp->dec_maxcodep = &sp->dec_codetab[sp->dec_nbitsmask-1];
    return (1);
}

/*
 * Decode a "hunk of data".
 */
#define GetNextCode(sp, bp, code) {             \
    nextdata = (nextdata<<8) | *(bp)++;         \
    nextbits += 8;                      \
    if (nextbits < nbits) {                 \
        nextdata = (nextdata<<8) | *(bp)++;     \
        nextbits += 8;                  \
    }                           \
    code = (hcode_t)((nextdata >> (nextbits-nbits)) & nbitsmask);   \
    nextbits -= nbits;                  \
}

static void
codeLoop(TIFF* tif)
{
    TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
        "LZWDecode: Bogus encoding, loop in the code table; scanline %d",
        tif->tif_row);
}

static int
LZWDecode(TIFF* tif, tidata_t op0, tsize_t occ0, tsample_t s)
{
    LZWCodecState *sp = DecoderState(tif);
    char *op = (char*) op0;
    long occ = (long) occ0;
    char *tp;
    unsigned char *bp;
    hcode_t code;
    int len;
    long nbits, nextbits, nextdata, nbitsmask;
    code_t *codep, *free_entp, *maxcodep, *oldcodep;

    (void) s;
    assert(sp != NULL);
    /*
     * Restart interrupted output operation.
     */
    if (sp->dec_restart) {
        long residue;

        codep = sp->dec_codep;
        residue = codep->length - sp->dec_restart;
        if (residue > occ) {
            /*
             * Residue from previous decode is sufficient
             * to satisfy decode request.  Skip to the
             * start of the decoded string, place decoded
             * values in the output buffer, and return.
             */
            sp->dec_restart += occ;
            do {
                codep = codep->next;
            } while (--residue > occ && codep);
            if (codep) {
                tp = op + occ;
                do {
                    *--tp = codep->value;
                    codep = codep->next;
                } while (--occ && codep);
            }
            return (1);
        }
        /*
         * Residue satisfies only part of the decode request.
         */
        op += residue, occ -= residue;
        tp = op;
        do {
            int t;
            --tp;
            t = codep->value;
            codep = codep->next;
            *tp = t;
        } while (--residue && codep);
        sp->dec_restart = 0;
    }

    bp = (unsigned char *)tif->tif_rawcp;
    nbits = sp->lzw_nbits;
    nextdata = sp->lzw_nextdata;
    nextbits = sp->lzw_nextbits;
    nbitsmask = sp->dec_nbitsmask;
    oldcodep = sp->dec_oldcodep;
    free_entp = sp->dec_free_entp;
    maxcodep = sp->dec_maxcodep;

    while (occ > 0) {
        NextCode(tif, sp, bp, code, GetNextCode);
        if (code == CODE_EOI)
            break;
        if (code == CODE_CLEAR) {
            free_entp = sp->dec_codetab + CODE_FIRST;
            nbits = BITS_MIN;
            nbitsmask = MAXCODE(BITS_MIN);
            maxcodep = sp->dec_codetab + nbitsmask-1;
            NextCode(tif, sp, bp, code, GetNextCode);
            if (code == CODE_EOI)
                break;
            *op++ = (char)code, occ--;
            oldcodep = sp->dec_codetab + code;
            continue;
        }
        codep = sp->dec_codetab + code;

        /*
         * Add the new entry to the code table.
         */
        if (free_entp < &sp->dec_codetab[0] ||
            free_entp >= &sp->dec_codetab[CSIZE]) {
            TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
            "LZWDecode: Corrupted LZW table at scanline %d",
            tif->tif_row);
            return (0);
        }

        free_entp->next = oldcodep;
        if (free_entp->next < &sp->dec_codetab[0] ||
            free_entp->next >= &sp->dec_codetab[CSIZE]) {
            TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
            "LZWDecode: Corrupted LZW table at scanline %d",
            tif->tif_row);
            return (0);
        }
        free_entp->firstchar = free_entp->next->firstchar;
        free_entp->length = free_entp->next->length+1;
        free_entp->value = (codep < free_entp) ?
            codep->firstchar : free_entp->firstchar;
        if (++free_entp > maxcodep) {
            if (++nbits > BITS_MAX)     /* should not happen */
                nbits = BITS_MAX;
            nbitsmask = MAXCODE(nbits);
            maxcodep = sp->dec_codetab + nbitsmask-1;
        }
        oldcodep = codep;
        if (code >= 256) {
            /*
             * Code maps to a string, copy string
             * value to output (written in reverse).
             */
            if(codep->length == 0) {
                TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
                    "LZWDecode: Wrong length of decoded string: "
                "data probably corrupted at scanline %d",
                tif->tif_row);  
                return (0);
            }
            if (codep->length > occ) {
                /*
                 * String is too long for decode buffer,
                 * locate portion that will fit, copy to
                 * the decode buffer, and setup restart
                 * logic for the next decoding call.
                 */
                sp->dec_codep = codep;
                do {
                    codep = codep->next;
                } while (codep && codep->length > occ);
                if (codep) {
                    sp->dec_restart = occ;
                    tp = op + occ;
                    do  {
                        *--tp = codep->value;
                        codep = codep->next;
                    }  while (--occ && codep);
                    if (codep)
                        codeLoop(tif);
                }
                break;
            }
            len = codep->length;
            tp = op + len;
            do {
                int t;
                --tp;
                t = codep->value;
                codep = codep->next;
                *tp = t;
            } while (codep && tp > op);
            if (codep) {
                codeLoop(tif);
                break;
            }
            op += len, occ -= len;
        } else
            *op++ = (char)code, occ--;
    }

    tif->tif_rawcp = (tidata_t) bp;
    sp->lzw_nbits = (unsigned short) nbits;
    sp->lzw_nextdata = nextdata;
    sp->lzw_nextbits = nextbits;
    sp->dec_nbitsmask = nbitsmask;
    sp->dec_oldcodep = oldcodep;
    sp->dec_free_entp = free_entp;
    sp->dec_maxcodep = maxcodep;

    if (occ > 0) {
        TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
        "LZWDecode: Not enough data at scanline %d (short %d bytes)",
            tif->tif_row, occ);
        return (0);
    }
    return (1);
}

#ifdef LZW_COMPAT
/*
 * Decode a "hunk of data" for old images.
 */
#define GetNextCodeCompat(sp, bp, code) {           \
    nextdata |= (unsigned long) *(bp)++ << nextbits;    \
    nextbits += 8;                      \
    if (nextbits < nbits) {                 \
        nextdata |= (unsigned long) *(bp)++ << nextbits;\
        nextbits += 8;                  \
    }                           \
    code = (hcode_t)(nextdata & nbitsmask);         \
    nextdata >>= nbits;                 \
    nextbits -= nbits;                  \
}

static int
LZWDecodeCompat(TIFF* tif, tidata_t op0, tsize_t occ0, tsample_t s)
{
    LZWCodecState *sp = DecoderState(tif);
    char *op = (char*) op0;
    long occ = (long) occ0;
    char *tp;
    unsigned char *bp;
    int code, nbits;
    long nextbits, nextdata, nbitsmask;
    code_t *codep, *free_entp, *maxcodep, *oldcodep;

    (void) s;
    assert(sp != NULL);
    /*
     * Restart interrupted output operation.
     */
    if (sp->dec_restart) {
        long residue;

        codep = sp->dec_codep;
        residue = codep->length - sp->dec_restart;
        if (residue > occ) {
            /*
             * Residue from previous decode is sufficient
             * to satisfy decode request.  Skip to the
             * start of the decoded string, place decoded
             * values in the output buffer, and return.
             */
            sp->dec_restart += occ;
            do {
                codep = codep->next;
            } while (--residue > occ);
            tp = op + occ;
            do {
                *--tp = codep->value;
                codep = codep->next;
            } while (--occ);
            return (1);
        }
        /*
         * Residue satisfies only part of the decode request.
         */
        op += residue, occ -= residue;
        tp = op;
        do {
            *--tp = codep->value;
            codep = codep->next;
        } while (--residue);
        sp->dec_restart = 0;
    }

    bp = (unsigned char *)tif->tif_rawcp;
    nbits = sp->lzw_nbits;
    nextdata = sp->lzw_nextdata;
    nextbits = sp->lzw_nextbits;
    nbitsmask = sp->dec_nbitsmask;
    oldcodep = sp->dec_oldcodep;
    free_entp = sp->dec_free_entp;
    maxcodep = sp->dec_maxcodep;

    while (occ > 0) {
        NextCode(tif, sp, bp, code, GetNextCodeCompat);
        if (code == CODE_EOI)
            break;
        if (code == CODE_CLEAR) {
            free_entp = sp->dec_codetab + CODE_FIRST;
            nbits = BITS_MIN;
            nbitsmask = MAXCODE(BITS_MIN);
            maxcodep = sp->dec_codetab + nbitsmask;
            NextCode(tif, sp, bp, code, GetNextCodeCompat);
            if (code == CODE_EOI)
                break;
            *op++ = code, occ--;
            oldcodep = sp->dec_codetab + code;
            continue;
        }
        codep = sp->dec_codetab + code;

        /*
         * Add the new entry to the code table.
         */
        if (free_entp < &sp->dec_codetab[0] ||
            free_entp >= &sp->dec_codetab[CSIZE]) {
            TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
            "LZWDecodeCompat: Corrupted LZW table at scanline %d",
            tif->tif_row);
            return (0);
        }

        free_entp->next = oldcodep;
        if (free_entp->next < &sp->dec_codetab[0] ||
            free_entp->next >= &sp->dec_codetab[CSIZE]) {
            TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
            "LZWDecodeCompat: Corrupted LZW table at scanline %d",
            tif->tif_row);
            return (0);
        }
        free_entp->firstchar = free_entp->next->firstchar;
        free_entp->length = free_entp->next->length+1;
        free_entp->value = (codep < free_entp) ?
            codep->firstchar : free_entp->firstchar;
        if (++free_entp > maxcodep) {
            if (++nbits > BITS_MAX)     /* should not happen */
                nbits = BITS_MAX;
            nbitsmask = MAXCODE(nbits);
            maxcodep = sp->dec_codetab + nbitsmask;
        }
        oldcodep = codep;
        if (code >= 256) {
            /*
             * Code maps to a string, copy string
             * value to output (written in reverse).
             */
            if(codep->length == 0) {
                TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
                    "LZWDecodeCompat: Wrong length of decoded "
                "string: data probably corrupted at scanline %d",
                tif->tif_row);  
                return (0);
            }
            if (codep->length > occ) {
                /*
                 * String is too long for decode buffer,
                 * locate portion that will fit, copy to
                 * the decode buffer, and setup restart
                 * logic for the next decoding call.
                 */
                sp->dec_codep = codep;
                do {
                    codep = codep->next;
                } while (codep->length > occ);
                sp->dec_restart = occ;
                tp = op + occ;
                do  {
                    *--tp = codep->value;
                    codep = codep->next;
                }  while (--occ);
                break;
            }
            op += codep->length, occ -= codep->length;
            tp = op;
            do {
                *--tp = codep->value;
            } while( (codep = codep->next) != NULL);
        } else
            *op++ = code, occ--;
    }

    tif->tif_rawcp = (tidata_t) bp;
    sp->lzw_nbits = nbits;
    sp->lzw_nextdata = nextdata;
    sp->lzw_nextbits = nextbits;
    sp->dec_nbitsmask = nbitsmask;
    sp->dec_oldcodep = oldcodep;
    sp->dec_free_entp = free_entp;
    sp->dec_maxcodep = maxcodep;

    if (occ > 0) {
        TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
        "LZWDecodeCompat: Not enough data at scanline %d (short %d bytes)",
            tif->tif_row, occ);
        return (0);
    }
    return (1);
}
#endif /* LZW_COMPAT */

/*
 * LZW Encoding.
 */

static int
LZWSetupEncode(TIFF* tif)
{
    LZWCodecState* sp = EncoderState(tif);
    static const char module[] = "LZWSetupEncode";

    assert(sp != NULL);
    sp->enc_hashtab = (hash_t*) _TIFFmalloc(HSIZE*sizeof (hash_t));
    if (sp->enc_hashtab == NULL) {
        TIFFErrorExt(tif->tif_clientdata, module, "No space for LZW hash table");
        return (0);
    }
    return (1);
}

/*
 * Reset encoding state at the start of a strip.
 */
static int
LZWPreEncode(TIFF* tif, tsample_t s)
{
    LZWCodecState *sp = EncoderState(tif);

    (void) s;
    assert(sp != NULL);
    sp->lzw_nbits = BITS_MIN;
    sp->lzw_maxcode = MAXCODE(BITS_MIN);
    sp->lzw_free_ent = CODE_FIRST;
    sp->lzw_nextbits = 0;
    sp->lzw_nextdata = 0;
    sp->enc_checkpoint = CHECK_GAP;
    sp->enc_ratio = 0;
    sp->enc_incount = 0;
    sp->enc_outcount = 0;
    /*
     * The 4 here insures there is space for 2 max-sized
     * codes in LZWEncode and LZWPostDecode.
     */
    sp->enc_rawlimit = tif->tif_rawdata + tif->tif_rawdatasize-1 - 4;
    cl_hash(sp);        /* clear hash table */
    sp->enc_oldcode = (hcode_t) -1; /* generates CODE_CLEAR in LZWEncode */
    return (1);
}

#define CALCRATIO(sp, rat) {                    \
    if (incount > 0x007fffff) { /* NB: shift will overflow */\
        rat = outcount >> 8;                \
        rat = (rat == 0 ? 0x7fffffff : incount/rat);    \
    } else                          \
        rat = (incount<<8) / outcount;          \
}
#define PutNextCode(op, c) {                    \
    nextdata = (nextdata << nbits) | c;         \
    nextbits += nbits;                  \
    *op++ = (unsigned char)(nextdata >> (nextbits-8));      \
    nextbits -= 8;                      \
    if (nextbits >= 8) {                    \
        *op++ = (unsigned char)(nextdata >> (nextbits-8));  \
        nextbits -= 8;                  \
    }                           \
    outcount += nbits;                  \
}

/*
 * Encode a chunk of pixels.
 *
 * Uses an open addressing double hashing (no chaining) on the 
 * prefix code/next character combination.  We do a variant of
 * Knuth's algorithm D (vol. 3, sec. 6.4) along with G. Knott's
 * relatively-prime secondary probe.  Here, the modular division
 * first probe is gives way to a faster exclusive-or manipulation. 
 * Also do block compression with an adaptive reset, whereby the
 * code table is cleared when the compression ratio decreases,
 * but after the table fills.  The variable-length output codes
 * are re-sized at this point, and a CODE_CLEAR is generated
 * for the decoder. 
 */
static int
LZWEncode(TIFF* tif, tidata_t bp, tsize_t cc, tsample_t s)
{
    register LZWCodecState *sp = EncoderState(tif);
    register long fcode;
    register hash_t *hp;
    register int h, c;
    hcode_t ent;
    long disp;
    long incount, outcount, checkpoint;
    long nextdata, nextbits;
    int free_ent, maxcode, nbits;
    tidata_t op, limit;

    (void) s;
    if (sp == NULL)
        return (0);
    /*
     * Load local state.
     */
    incount = sp->enc_incount;
    outcount = sp->enc_outcount;
    checkpoint = sp->enc_checkpoint;
    nextdata = sp->lzw_nextdata;
    nextbits = sp->lzw_nextbits;
    free_ent = sp->lzw_free_ent;
    maxcode = sp->lzw_maxcode;
    nbits = sp->lzw_nbits;
    op = tif->tif_rawcp;
    limit = sp->enc_rawlimit;
    ent = sp->enc_oldcode;

    if (ent == (hcode_t) -1 && cc > 0) {
        /*
         * NB: This is safe because it can only happen
         *     at the start of a strip where we know there
         *     is space in the data buffer.
         */
        PutNextCode(op, CODE_CLEAR);
        ent = *bp++; cc--; incount++;
    }
    while (cc > 0) {
        c = *bp++; cc--; incount++;
        fcode = ((long)c << BITS_MAX) + ent;
        h = (c << HSHIFT) ^ ent;    /* xor hashing */
#ifdef _WINDOWS
        /*
         * Check hash index for an overflow.
         */
        if (h >= HSIZE)
            h -= HSIZE;
#endif
        hp = &sp->enc_hashtab[h];
        if (hp->hash == fcode) {
            ent = hp->code;
            continue;
        }
        if (hp->hash >= 0) {
            /*
             * Primary hash failed, check secondary hash.
             */
            disp = HSIZE - h;
            if (h == 0)
                disp = 1;
            do {
                /*
                 * Avoid pointer arithmetic 'cuz of
                 * wraparound problems with segments.
                 */
                if ((h -= disp) < 0)
                    h += HSIZE;
                hp = &sp->enc_hashtab[h];
                if (hp->hash == fcode) {
                    ent = hp->code;
                    goto hit;
                }
            } while (hp->hash >= 0);
        }
        /*
         * New entry, emit code and add to table.
         */
        /*
         * Verify there is space in the buffer for the code
         * and any potential Clear code that might be emitted
         * below.  The value of limit is setup so that there
         * are at least 4 bytes free--room for 2 codes.
         */
        if (op > limit) {
            tif->tif_rawcc = (tsize_t)(op - tif->tif_rawdata);
            TIFFFlushData1(tif);
            op = tif->tif_rawdata;
        }
        PutNextCode(op, ent);
        ent = c;
        hp->code = free_ent++;
        hp->hash = fcode;
        if (free_ent == CODE_MAX-1) {
            /* table is full, emit clear code and reset */
            cl_hash(sp);
            sp->enc_ratio = 0;
            incount = 0;
            outcount = 0;
            free_ent = CODE_FIRST;
            PutNextCode(op, CODE_CLEAR);
            nbits = BITS_MIN;
            maxcode = MAXCODE(BITS_MIN);
        } else {
            /*
             * If the next entry is going to be too big for
             * the code size, then increase it, if possible.
             */
            if (free_ent > maxcode) {
                nbits++;
                assert(nbits <= BITS_MAX);
                maxcode = (int) MAXCODE(nbits);
            } else if (incount >= checkpoint) {
                long rat;
                /*
                 * Check compression ratio and, if things seem
                 * to be slipping, clear the hash table and
                 * reset state.  The compression ratio is a
                 * 24+8-bit fractional number.
                 */
                checkpoint = incount+CHECK_GAP;
                CALCRATIO(sp, rat);
                if (rat <= sp->enc_ratio) {
                    cl_hash(sp);
                    sp->enc_ratio = 0;
                    incount = 0;
                    outcount = 0;
                    free_ent = CODE_FIRST;
                    PutNextCode(op, CODE_CLEAR);
                    nbits = BITS_MIN;
                    maxcode = MAXCODE(BITS_MIN);
                } else
                    sp->enc_ratio = rat;
            }
        }
    hit:
        ;
    }

    /*
     * Restore global state.
     */
    sp->enc_incount = incount;
    sp->enc_outcount = outcount;
    sp->enc_checkpoint = checkpoint;
    sp->enc_oldcode = ent;
    sp->lzw_nextdata = nextdata;
    sp->lzw_nextbits = nextbits;
    sp->lzw_free_ent = free_ent;
    sp->lzw_maxcode = maxcode;
    sp->lzw_nbits = nbits;
    tif->tif_rawcp = op;
    return (1);
}

/*
 * Finish off an encoded strip by flushing the last
 * string and tacking on an End Of Information code.
 */
static int
LZWPostEncode(TIFF* tif)
{
    register LZWCodecState *sp = EncoderState(tif);
    tidata_t op = tif->tif_rawcp;
    long nextbits = sp->lzw_nextbits;
    long nextdata = sp->lzw_nextdata;
    long outcount = sp->enc_outcount;
    int nbits = sp->lzw_nbits;

    if (op > sp->enc_rawlimit) {
        tif->tif_rawcc = (tsize_t)(op - tif->tif_rawdata);
        TIFFFlushData1(tif);
        op = tif->tif_rawdata;
    }
    if (sp->enc_oldcode != (hcode_t) -1) {
        PutNextCode(op, sp->enc_oldcode);
        sp->enc_oldcode = (hcode_t) -1;
    }
    PutNextCode(op, CODE_EOI);
    if (nextbits > 0) 
        *op++ = (unsigned char)(nextdata << (8-nextbits));
    tif->tif_rawcc = (tsize_t)(op - tif->tif_rawdata);
    return (1);
}

/*
 * Reset encoding hash table.
 */
static void
cl_hash(LZWCodecState* sp)
{
    register hash_t *hp = &sp->enc_hashtab[HSIZE-1];
    register long i = HSIZE-8;

    do {
        i -= 8;
        hp[-7].hash = -1;
        hp[-6].hash = -1;
        hp[-5].hash = -1;
        hp[-4].hash = -1;
        hp[-3].hash = -1;
        hp[-2].hash = -1;
        hp[-1].hash = -1;
        hp[ 0].hash = -1;
        hp -= 8;
    } while (i >= 0);
        for (i += 8; i > 0; i--, hp--)
        hp->hash = -1;
}

static void
LZWCleanup(TIFF* tif)
{
    (void)TIFFPredictorCleanup(tif);

    assert(tif->tif_data != 0);

    if (DecoderState(tif)->dec_codetab)
        _TIFFfree(DecoderState(tif)->dec_codetab);

    if (EncoderState(tif)->enc_hashtab)
        _TIFFfree(EncoderState(tif)->enc_hashtab);

    _TIFFfree(tif->tif_data);
    tif->tif_data = NULL;

    _TIFFSetDefaultCompressionState(tif);
}

int
TIFFInitLZW(TIFF* tif, int scheme)
{
    assert(scheme == COMPRESSION_LZW);
    /*
     * Allocate state block so tag methods have storage to record values.
     */
    tif->tif_data = (tidata_t) _TIFFmalloc(sizeof (LZWCodecState));
    if (tif->tif_data == NULL)
        goto bad;
    DecoderState(tif)->dec_codetab = NULL;
    DecoderState(tif)->dec_decode = NULL;
    EncoderState(tif)->enc_hashtab = NULL;
        LZWState(tif)->rw_mode = tif->tif_mode;

    /*
     * Install codec methods.
     */
    tif->tif_setupdecode = LZWSetupDecode;
    tif->tif_predecode = LZWPreDecode;
    tif->tif_decoderow = LZWDecode;
    tif->tif_decodestrip = LZWDecode;
    tif->tif_decodetile = LZWDecode;
    tif->tif_setupencode = LZWSetupEncode;
    tif->tif_preencode = LZWPreEncode;
    tif->tif_postencode = LZWPostEncode;
    tif->tif_encoderow = LZWEncode;
    tif->tif_encodestrip = LZWEncode;
    tif->tif_encodetile = LZWEncode;
    tif->tif_cleanup = LZWCleanup;
    /*
     * Setup predictor setup.
     */
    (void) TIFFPredictorInit(tif);
    return (1);
bad:
    TIFFErrorExt(tif->tif_clientdata, "TIFFInitLZW", 
             "No space for LZW state block");
    return (0);
}

/*
 * Copyright (c) 1985, 1986 The Regents of the University of California.
 * All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * James A. Woods, derived from original work by Spencer Thomas
 * and Joseph Orost.
 *
 * Redistribution and use in source and binary forms are permitted
 * provided that the above copyright notice and this paragraph are
 * duplicated in all such forms and that any documentation,
 * advertising materials, and other materials related to such
 * distribution and use acknowledge that the software was developed
 * by the University of California, Berkeley.  The name of the
 * University may not be used to endorse or promote products derived
 * from this software without specific prior written permission.
 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
 * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
 */
#endif /* LZW_SUPPORT */

/* vim: set ts=8 sts=8 sw=8 noet: */