/* Copyright (C) 1993, 1996, 1998, 1999 Aladdin Enterprises.  All rights reserved.
  
  This software is provided AS-IS with no warranty, either express or
  implied.
  
  This software is distributed under license and may not be copied,
  modified or distributed except as expressly authorized under the terms
  of the license contained in the file LICENSE in this distribution.
  
  For more information about licensing, please refer to
  http://www.ghostscript.com/licensing/. For information on
  commercial licensing, go to http://www.artifex.com/licensing/ or
  contact Artifex Software, Inc., 101 Lucas Valley Road #110,
  San Rafael, CA  94903, U.S.A., +1(415)492-9861.
*/

/* $Id$ */
/* LZW decoding filter */
#include "stdio_.h"		/* includes std.h */
#include "gdebug.h"
#include "strimpl.h"
#include "slzwx.h"

/* ------ LZWDecode ------ */

/********************************************************/
/* LZW routines are based on:                           */
/* Dr. Dobbs Journal --- Oct. 1989.                     */
/* Article on LZW Data Compression by Mark R. Nelson    */
/********************************************************/

/* Define the special codes in terms of code_escape, which is */
/* 1 << InitialCodeLength. */
#define code_reset (code_escape + 0)
#define code_eod (code_escape + 1)
#define code_0 (code_escape + 2)	/* first assignable code */

struct lzw_decode_s {
    byte datum;
    byte len;			/* length of code */
    ushort prefix;		/* code to be prefixed */
};

gs_private_st_simple(st_lzw_decode, lzw_decode, "lzw_decode");
/* We can use a simple type as the element type, */
/* because there are no pointers to enumerate or relocate. */
#define st_lzw_decode_element st_lzw_decode
#define lzw_decode_max 4096	/* must be 4096 */

/* Initialize LZWDecode filter */
/* We separate out the reset function for some non-stream clients. */
private int
s_LZWD_reset(stream_state * st)
{
    stream_LZW_state *const ss = (stream_LZW_state *) st;
    register lzw_decode *dc = ss->table.decode;
    register int i;
    uint code_escape = 1 << ss->InitialCodeLength;

    ss->bits_left = 0;
    ss->bytes_left = 0;
    ss->next_code = code_0;
    ss->code_size = ss->InitialCodeLength + 1;
    ss->prev_code = -1;
    ss->copy_code = -1;
    dc[code_reset].len = 255;
    dc[code_eod].len = 255;
    for (i = 0; i < code_escape; i++, dc++)
	dc->datum = i, dc->len = 1, dc->prefix = code_eod;
    return 0;
}
private int
s_LZWD_init(stream_state * st)
{
    stream_LZW_state *const ss = (stream_LZW_state *) st;
    lzw_decode *dc =
    gs_alloc_struct_array(st->memory, lzw_decode_max + 1,
			  lzw_decode, &st_lzw_decode_element,
			  "LZWDecode(init)");

    if (dc == 0)
	return ERRC;
/****** WRONG ******/
    ss->table.decode = dc;
    ss->min_left = 1;
    return s_LZWD_reset(st);
}

/* Process a buffer */
private int
s_LZWD_process(stream_state * st, stream_cursor_read * pr,
	       stream_cursor_write * pw, bool last)
{
    stream_LZW_state *const ss = (stream_LZW_state *) st;
    register const byte *p = pr->ptr;
    register byte *q = pw->ptr;

#ifdef DEBUG
    byte *q0 = q;

#endif
    const byte *rlimit = pr->limit;	/* constant pointer */
    byte *wlimit = pw->limit;	/* constant pointer */
    int status = 0;
    int code = ss->copy_code;
    int prev_code = ss->prev_code;
    uint prev_len = ss->prev_len;
    byte bits = ss->bits;
    int bits_left = ss->bits_left;
    int bytes_left = ss->bytes_left;
    int code_size = ss->code_size;
    int code_mask;
    int switch_code;
    int next_code = ss->next_code;
    lzw_decode *table = ss->table.decode;	/* constant pointer */
    lzw_decode *dc_next = table + next_code;	/* invariant */
    lzw_decode *dc;
    int code_escape = 1 << ss->InitialCodeLength;
    int eod = code_eod;
    bool low_order = ss->FirstBitLowOrder;
    uint len;
    int c;
    byte b;
    byte *q1;

    if_debug2('w', "[w]process decode: code_size=%d next_code=%d\n",
	      code_size, next_code);
#define set_code_size()\
  code_mask = (1 << code_size) - 1,\
  switch_code = code_mask + 1 - ss->EarlyChange
    set_code_size();
    if (!ss->BlockData)
	bytes_left = rlimit - p + 2;	/* never stop for bytes_left */
    /* If we are in the middle of copying a string, */
    /* do some more now. */
    if (code >= 0) {
	int rlen = ss->copy_left;
	int wlen = wlimit - q;
	int n = len = min(rlen, wlen);

	c = code;
	ss->copy_left = rlen -= len;
	if_debug3('W', "[W]copying 0x%x, %d byte(s) out of %d left\n",
		  code, len, rlen + len);
	while (rlen)
	    c = table[c].prefix,
		rlen--;
	q1 = q += len;
	n = len;
	while (--n >= 0) {
	    *q1-- = (dc = &table[c])->datum;
	    c = dc->prefix;
	}
	if (ss->copy_left) {	/* more to do */
	    pw->ptr = q;
	    return 1;
	}
	ss->copy_code = -1;
	len = ss->copy_len;
	/* Retrieve the first byte of the code just copied. */
	if (c == eod) {		/* We just copied the entire code, */
	    /* so the byte we want is immediately available. */
	    b = q1[1];
	} else {		/* We have to scan to the beginning of the code. */
	    for (; c != eod; c = table[c].prefix)
		b = (byte) c;
	}
	goto add;
    }
  top:if (code_size > bits_left) {
	if (bytes_left == 0) {
	    if (p == rlimit)
		goto out;
	    bytes_left = *++p;
	    if_debug1('W', "[W]block count %d\n", bytes_left);
	    if (bytes_left == 0) {
		status = EOFC;
		goto out;
	    }
	    goto top;
	}
	if (low_order)
	    code = bits >> (8 - bits_left);
	else
	    code = (uint) bits << (code_size - bits_left);
	if (bits_left + 8 < code_size) {	/* Need 2 more data bytes */
	    if (bytes_left == 1) {
		if (rlimit - p < 3)
		    goto out;
		bytes_left = p[2];
		if_debug1('W', "[W]block count %d\n",
			  bytes_left);
		if (bytes_left == 0) {
		    status = EOFC;
		    goto out;
		}
		bytes_left++;
		bits = p[1];
		p++;
	    } else {
		if (rlimit - p < 2)
		    goto out;
		bits = p[1];
	    }
	    if (low_order)
		code += (uint) bits << bits_left;
	    else
		code += (uint) bits << (code_size - 8 - bits_left);
	    bits_left += 8;
	    bits = p[2];
	    p += 2;
	    bytes_left -= 2;
	} else {
	    if (p == rlimit)
		goto out;
	    bits = *++p;
	    bytes_left--;
	}
	if (low_order)
	    code += (uint) bits << bits_left,
		bits_left += 8 - code_size;
	else
	    bits_left += 8 - code_size,
		code += bits >> bits_left;
    } else {
	if (low_order)
	    code = bits >> (8 - bits_left),
		bits_left -= code_size;
	else
	    bits_left -= code_size,
		code = bits >> bits_left;
    }
    code &= code_mask;
    if_debug2('W', "[W]reading 0x%x,%d\n", code, code_size);
    /*
     * There is an anomalous case where a code S is followed
     * immediately by another occurrence of the S string.
     * In this case, the next available code will be defined as
     * S followed by the first character of S, and will be
     * emitted immediately after the code S.  We have to
     * recognize this case specially, by noting that the code is
     * equal to next_code.
     */
    if (code >= next_code) {
	if (code > next_code) {
#ifdef DEBUG
	    lprintf2("[W]code = %d > next_code = %d\n",
		     code, next_code);
#endif
	    status = ERRC;
	    goto out;
	}
	/* Fabricate the entry for the code.  It will be */
	/* overwritten immediately, of course. */
	for (c = prev_code; c != eod; c = table[c].prefix)
	    dc_next->datum = c;
	len = prev_len + 1;
	dc_next->len = min(len, 255);
	dc_next->prefix = prev_code;
	if_debug3('w', "[w]decoding anomalous 0x%x=0x%x+%c\n",
		  next_code, prev_code, dc_next->datum);
    }
    /* See if there is enough room for the code. */
reset:
    len = table[code].len;
    if (len == 255) {		/* Check for special code (reset or end). */
	/* We set their lengths to 255 to avoid doing */
	/* an extra check in the normal case. */
	if (code == code_reset) {
	    if_debug1('w', "[w]reset: next_code was %d\n",
		      next_code);
	    next_code = code_0;
	    dc_next = table + code_0;
	    code_size = ss->InitialCodeLength + 1;
	    set_code_size();
	    prev_code = -1;
	    goto top;
	} else if (code == eod) {
	    status = EOFC;
	    goto out;
	}
	/* The code length won't fit in a byte, */
	/* compute it the hard way. */
	for (c = code, len = 0; c != eod; len++)
	    c = table[c].prefix;
	if_debug2('w', "[w]long code %d, length=%d\n", code, len);
    }
    if (wlimit - q < len) {
	ss->copy_code = code;
	ss->copy_left = ss->copy_len = len;
	status = 1;
	goto out;
    }
    /* Copy the string to the buffer (back to front). */
    /* Optimize for short codes, which are the most frequent. */
    dc = &table[code];
    switch (len) {
	default:
	    {
		byte *q1 = q += len;

		c = code;
		do {
		    *q1-- = (dc = &table[c])->datum;
		}
		while ((c = dc->prefix) != eod);
		b = q1[1];
	    }
	    break;
	case 3:
	    q[3] = dc->datum;
	    dc = &table[dc->prefix];
	case 2:
	    q[2] = dc->datum;
	    dc = &table[dc->prefix];
	case 1:
	    q[1] = b = dc->datum;
	    q += len;
    }
  add:				/* Add a new entry to the table */
    if (prev_code >= 0) {
	/*
	 * Unfortunately, we have to check for next_code ==
	 * lzw_decode_max every time: just checking at power
	 * of 2 boundaries stops us one code too soon.
	 */
	if (next_code == lzw_decode_max) {
	    /*
	     * A few anomalous files have one data item too many before the
	     * reset code.  We think this is a bug in the application that
	     * produced the files, but Acrobat accepts the files, so we do
	     * too.
	     */
	    if (!ss->BlockData) { /* don't do this for GIF */
		if (bits_left < 8 && p >= rlimit && last) {
		    /* We're at EOD. */
		    goto out;
		}
		if (bits_left + ((rlimit - p) << 3) < code_size) {
		    /*
		     * We need more data to decide whether a reset is next.
		     * Return an error if we cannot get more.
                     */
		    if (last)
                        status = ERRC;
		    goto out;
		}
		if (low_order) {
		    code = bits >> (8 - bits_left);
		    code += (bits = *++p) << bits_left;
		    if (bits_left + 8 < code_size)
			code += (bits = *++p) << (bits_left + 8);
		} else {
		    code = bits & ((1 << bits_left) - 1);
		    code = (code << 8) + (bits = *++p);
		    if (bits_left + 8 < code_size)
			code = (code << 8) + (bits = *++p);
		    code >>= (bits_left - code_size) & 7;
		}
		bits_left = (bits_left - code_size) & 7;
		if (code == code_reset)
		    goto reset;
	    }
	    status = ERRC;
	    goto out;
	}
	dc_next->datum = b;	/* added char of string */
	dc_next->len = min(prev_len, 254) + 1;
	dc_next->prefix = prev_code;
	dc_next++;
	if_debug4('W', "[W]adding 0x%x=0x%x+%c(%d)\n",
		  next_code, prev_code, b, min(len, 255));
	if (++next_code == switch_code) {	/* Crossed a power of 2. */
	    /* We have to make a strange special check for */
	    /* reaching the end of the code space. */
	    if (next_code < lzw_decode_max - 1) {
		code_size++;
		set_code_size();
		if_debug2('w', "[w]crossed power of 2: new code_size=%d, next_code=%d\n",
			  code_size, next_code);
	    }
	}
    }
    prev_code = code;
    prev_len = len;
    goto top;
  out:pr->ptr = p;
    pw->ptr = q;
    ss->code_size = code_size;
    ss->prev_code = prev_code;
    ss->prev_len = prev_len;
    ss->bits = bits;
    ss->bits_left = bits_left;
    ss->bytes_left = bytes_left;
    ss->next_code = next_code;
    if_debug3('w', "[w]decoded %d bytes, prev_code=%d, next_code=%d\n",
	      (int)(q - q0), prev_code, next_code);
    return status;
}

/* Stream template */
const stream_template s_LZWD_template =
{&st_LZW_state, s_LZWD_init, s_LZWD_process, 3, 1, s_LZW_release,
 s_LZW_set_defaults, s_LZWD_reset
};