/*
 * jchuff.c
 *
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
 * This file contains Huffman entropy encoding routines.
 *
 * Much of the complexity here has to do with supporting output suspension.
 * If the data destination module demands suspension, we want to be able to
 * back up to the start of the current MCU.  To do this, we copy state
 * variables into local working storage, and update them back to the
 * permanent JPEG objects only upon successful completion of an MCU.
 */

#define JPEG_INTERNALS
#include "jinclude.h"
#include "jpeglib.h"
#include "jchuff.h"		/* Declarations shared with jcphuff.c */


/* Expanded entropy encoder object for Huffman encoding.
 *
 * The savable_state subrecord contains fields that change within an MCU,
 * but must not be updated permanently until we complete the MCU.
 */

typedef struct {
  INT32 put_buffer;		/* current bit-accumulation buffer */
  int put_bits;			/* # of bits now in it */
  int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
} savable_state;

/* This macro is to work around compilers with missing or broken
 * structure assignment.  You'll need to fix this code if you have
 * such a compiler and you change MAX_COMPS_IN_SCAN.
 */

#ifndef NO_STRUCT_ASSIGN
#define ASSIGN_STATE(dest,src)  ((dest) = (src))
#else
#if MAX_COMPS_IN_SCAN == 4
#define ASSIGN_STATE(dest,src)  \
	((dest).put_buffer = (src).put_buffer, \
	 (dest).put_bits = (src).put_bits, \
	 (dest).last_dc_val[0] = (src).last_dc_val[0], \
	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
	 (dest).last_dc_val[3] = (src).last_dc_val[3])
#endif
#endif


typedef struct {
  struct jpeg_entropy_encoder pub; /* public fields */

  savable_state saved;		/* Bit buffer & DC state at start of MCU */

  /* These fields are NOT loaded into local working state. */
  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
  int next_restart_num;		/* next restart number to write (0-7) */

  /* Pointers to derived tables (these workspaces have image lifespan) */
  c_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS];
  c_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS];

#ifdef ENTROPY_OPT_SUPPORTED	/* Statistics tables for optimization */
  long * dc_count_ptrs[NUM_HUFF_TBLS];
  long * ac_count_ptrs[NUM_HUFF_TBLS];
#endif
} huff_entropy_encoder;

typedef huff_entropy_encoder * huff_entropy_ptr;

/* Working state while writing an MCU.
 * This struct contains all the fields that are needed by subroutines.
 */

typedef struct {
  JOCTET * next_output_byte;	/* => next byte to write in buffer */
  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
  savable_state cur;		/* Current bit buffer & DC state */
  j_compress_ptr cinfo;		/* dump_buffer needs access to this */
} working_state;


/* Forward declarations */
METHODDEF(boolean) encode_mcu_huff JPP((j_compress_ptr cinfo,
					JBLOCKROW *MCU_data));
METHODDEF(void) finish_pass_huff JPP((j_compress_ptr cinfo));
#ifdef ENTROPY_OPT_SUPPORTED
METHODDEF(boolean) encode_mcu_gather JPP((j_compress_ptr cinfo,
					  JBLOCKROW *MCU_data));
METHODDEF(void) finish_pass_gather JPP((j_compress_ptr cinfo));
#endif


/*
 * Initialize for a Huffman-compressed scan.
 * If gather_statistics is TRUE, we do not output anything during the scan,
 * just count the Huffman symbols used and generate Huffman code tables.
 */

METHODDEF(void)
start_pass_huff (j_compress_ptr cinfo, boolean gather_statistics)
{
  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
  int ci, dctbl, actbl;
  jpeg_component_info * compptr;

  if (gather_statistics) {
#ifdef ENTROPY_OPT_SUPPORTED
    entropy->pub.encode_mcu = encode_mcu_gather;
    entropy->pub.finish_pass = finish_pass_gather;
#else
    ERREXIT(cinfo, JERR_NOT_COMPILED);
#endif
  } else {
    entropy->pub.encode_mcu = encode_mcu_huff;
    entropy->pub.finish_pass = finish_pass_huff;
  }

  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
    compptr = cinfo->cur_comp_info[ci];
    dctbl = compptr->dc_tbl_no;
    actbl = compptr->ac_tbl_no;
    if (gather_statistics) {
#ifdef ENTROPY_OPT_SUPPORTED
      /* Check for invalid table indexes */
      /* (make_c_derived_tbl does this in the other path) */
      if (dctbl < 0 || dctbl >= NUM_HUFF_TBLS)
	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, dctbl);
      if (actbl < 0 || actbl >= NUM_HUFF_TBLS)
	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, actbl);
      /* Allocate and zero the statistics tables */
      /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
      if (entropy->dc_count_ptrs[dctbl] == NULL)
	entropy->dc_count_ptrs[dctbl] = (long *)
	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
				      257 * SIZEOF(long));
      MEMZERO(entropy->dc_count_ptrs[dctbl], 257 * SIZEOF(long));
      if (entropy->ac_count_ptrs[actbl] == NULL)
	entropy->ac_count_ptrs[actbl] = (long *)
	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
				      257 * SIZEOF(long));
      MEMZERO(entropy->ac_count_ptrs[actbl], 257 * SIZEOF(long));
#endif
    } else {
      /* Compute derived values for Huffman tables */
      /* We may do this more than once for a table, but it's not expensive */
      jpeg_make_c_derived_tbl(cinfo, TRUE, dctbl,
			      & entropy->dc_derived_tbls[dctbl]);
      jpeg_make_c_derived_tbl(cinfo, FALSE, actbl,
			      & entropy->ac_derived_tbls[actbl]);
    }
    /* Initialize DC predictions to 0 */
    entropy->saved.last_dc_val[ci] = 0;
  }

  /* Initialize bit buffer to empty */
  entropy->saved.put_buffer = 0;
  entropy->saved.put_bits = 0;

  /* Initialize restart stuff */
  entropy->restarts_to_go = cinfo->restart_interval;
  entropy->next_restart_num = 0;
}


/*
 * Compute the derived values for a Huffman table.
 * This routine also performs some validation checks on the table.
 *
 * Note this is also used by jcphuff.c.
 */

GLOBAL(void)
jpeg_make_c_derived_tbl (j_compress_ptr cinfo, boolean isDC, int tblno,
			 c_derived_tbl ** pdtbl)
{
  JHUFF_TBL *htbl;
  c_derived_tbl *dtbl;
  int p, i, l, lastp, si, maxsymbol;
  char huffsize[257];
  unsigned int huffcode[257];
  unsigned int code;

  /* Note that huffsize[] and huffcode[] are filled in code-length order,
   * paralleling the order of the symbols themselves in htbl->huffval[].
   */

  /* Find the input Huffman table */
  if (tblno < 0 || tblno >= NUM_HUFF_TBLS)
    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
  htbl =
    isDC ? cinfo->dc_huff_tbl_ptrs[tblno] : cinfo->ac_huff_tbl_ptrs[tblno];
  if (htbl == NULL)
    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);

  /* Allocate a workspace if we haven't already done so. */
  if (*pdtbl == NULL)
    *pdtbl = (c_derived_tbl *)
      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
				  SIZEOF(c_derived_tbl));
  dtbl = *pdtbl;
  
  /* Figure C.1: make table of Huffman code length for each symbol */

  p = 0;
  for (l = 1; l <= 16; l++) {
    i = (int) htbl->bits[l];
    if (i < 0 || p + i > 256)	/* protect against table overrun */
      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
    while (i--)
      huffsize[p++] = (char) l;
  }
  huffsize[p] = 0;
  lastp = p;
  
  /* Figure C.2: generate the codes themselves */
  /* We also validate that the counts represent a legal Huffman code tree. */

  code = 0;
  si = huffsize[0];
  p = 0;
  while (huffsize[p]) {
    while (((int) huffsize[p]) == si) {
      huffcode[p++] = code;
      code++;
    }
    /* code is now 1 more than the last code used for codelength si; but
     * it must still fit in si bits, since no code is allowed to be all ones.
     */
    if (((INT32) code) >= (((INT32) 1) << si))
      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
    code <<= 1;
    si++;
  }
  
  /* Figure C.3: generate encoding tables */
  /* These are code and size indexed by symbol value */

  /* Set all codeless symbols to have code length 0;
   * this lets us detect duplicate VAL entries here, and later
   * allows emit_bits to detect any attempt to emit such symbols.
   */
  MEMZERO(dtbl->ehufsi, SIZEOF(dtbl->ehufsi));

  /* This is also a convenient place to check for out-of-range
   * and duplicated VAL entries.  We allow 0..255 for AC symbols
   * but only 0..15 for DC.  (We could constrain them further
   * based on data depth and mode, but this seems enough.)
   */
  maxsymbol = isDC ? 15 : 255;

  for (p = 0; p < lastp; p++) {
    i = htbl->huffval[p];
    if (i < 0 || i > maxsymbol || dtbl->ehufsi[i])
      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
    dtbl->ehufco[i] = huffcode[p];
    dtbl->ehufsi[i] = huffsize[p];
  }
}


/* Outputting bytes to the file */

/* Emit a byte, taking 'action' if must suspend. */
#define emit_byte(state,val,action)  \
	{ *(state)->next_output_byte++ = (JOCTET) (val);  \
	  if (--(state)->free_in_buffer == 0)  \
	    if (! dump_buffer(state))  \
	      { action; } }


LOCAL(boolean)
dump_buffer (working_state * state)
/* Empty the output buffer; return TRUE if successful, FALSE if must suspend */
{
  struct jpeg_destination_mgr * dest = state->cinfo->dest;

  if (! (*dest->empty_output_buffer) (state->cinfo))
    return FALSE;
  /* After a successful buffer dump, must reset buffer pointers */
  state->next_output_byte = dest->next_output_byte;
  state->free_in_buffer = dest->free_in_buffer;
  return TRUE;
}


/* Outputting bits to the file */

/* Only the right 24 bits of put_buffer are used; the valid bits are
 * left-justified in this part.  At most 16 bits can be passed to emit_bits
 * in one call, and we never retain more than 7 bits in put_buffer
 * between calls, so 24 bits are sufficient.
 */

INLINE
LOCAL(boolean)
emit_bits (working_state * state, unsigned int code, int size)
/* Emit some bits; return TRUE if successful, FALSE if must suspend */
{
  /* This routine is heavily used, so it's worth coding tightly. */
  register INT32 put_buffer = (INT32) code;
  register int put_bits = state->cur.put_bits;

  /* if size is 0, caller used an invalid Huffman table entry */
  if (size == 0)
    ERREXIT(state->cinfo, JERR_HUFF_MISSING_CODE);

  put_buffer &= (((INT32) 1)<<size) - 1; /* mask off any extra bits in code */
  
  put_bits += size;		/* new number of bits in buffer */
  
  put_buffer <<= 24 - put_bits; /* align incoming bits */

  put_buffer |= state->cur.put_buffer; /* and merge with old buffer contents */
  
  while (put_bits >= 8) {
    int c = (int) ((put_buffer >> 16) & 0xFF);
    
    emit_byte(state, c, return FALSE);
    if (c == 0xFF) {		/* need to stuff a zero byte? */
      emit_byte(state, 0, return FALSE);
    }
    put_buffer <<= 8;
    put_bits -= 8;
  }

  state->cur.put_buffer = put_buffer; /* update state variables */
  state->cur.put_bits = put_bits;

  return TRUE;
}


LOCAL(boolean)
flush_bits (working_state * state)
{
  if (! emit_bits(state, 0x7F, 7)) /* fill any partial byte with ones */
    return FALSE;
  state->cur.put_buffer = 0;	/* and reset bit-buffer to empty */
  state->cur.put_bits = 0;
  return TRUE;
}


/* Encode a single block's worth of coefficients */

LOCAL(boolean)
encode_one_block (working_state * state, JCOEFPTR block, int last_dc_val,
		  c_derived_tbl *dctbl, c_derived_tbl *actbl)
{
  register int temp, temp2;
  register int nbits;
  register int k, r, i;
  
  /* Encode the DC coefficient difference per section F.1.2.1 */
  
  temp = temp2 = block[0] - last_dc_val;

  if (temp < 0) {
    temp = -temp;		/* temp is abs value of input */
    /* For a negative input, want temp2 = bitwise complement of abs(input) */
    /* This code assumes we are on a two's complement machine */
    temp2--;
  }
  
  /* Find the number of bits needed for the magnitude of the coefficient */
  nbits = 0;
  while (temp) {
    nbits++;
    temp >>= 1;
  }
  /* Check for out-of-range coefficient values.
   * Since we're encoding a difference, the range limit is twice as much.
   */
  if (nbits > MAX_COEF_BITS+1)
    ERREXIT(state->cinfo, JERR_BAD_DCT_COEF);
  
  /* Emit the Huffman-coded symbol for the number of bits */
  if (! emit_bits(state, dctbl->ehufco[nbits], dctbl->ehufsi[nbits]))
    return FALSE;

  /* Emit that number of bits of the value, if positive, */
  /* or the complement of its magnitude, if negative. */
  if (nbits)			/* emit_bits rejects calls with size 0 */
    if (! emit_bits(state, (unsigned int) temp2, nbits))
      return FALSE;

  /* Encode the AC coefficients per section F.1.2.2 */
  
  r = 0;			/* r = run length of zeros */
  
  for (k = 1; k < DCTSIZE2; k++) {
    if ((temp = block[jpeg_natural_order[k]]) == 0) {
      r++;
    } else {
      /* if run length > 15, must emit special run-length-16 codes (0xF0) */
      while (r > 15) {
	if (! emit_bits(state, actbl->ehufco[0xF0], actbl->ehufsi[0xF0]))
	  return FALSE;
	r -= 16;
      }

      temp2 = temp;
      if (temp < 0) {
	temp = -temp;		/* temp is abs value of input */
	/* This code assumes we are on a two's complement machine */
	temp2--;
      }
      
      /* Find the number of bits needed for the magnitude of the coefficient */
      nbits = 1;		/* there must be at least one 1 bit */
      while ((temp >>= 1))
	nbits++;
      /* Check for out-of-range coefficient values */
      if (nbits > MAX_COEF_BITS)
	ERREXIT(state->cinfo, JERR_BAD_DCT_COEF);
      
      /* Emit Huffman symbol for run length / number of bits */
      i = (r << 4) + nbits;
      if (! emit_bits(state, actbl->ehufco[i], actbl->ehufsi[i]))
	return FALSE;

      /* Emit that number of bits of the value, if positive, */
      /* or the complement of its magnitude, if negative. */
      if (! emit_bits(state, (unsigned int) temp2, nbits))
	return FALSE;
      
      r = 0;
    }
  }

  /* If the last coef(s) were zero, emit an end-of-block code */
  if (r > 0)
    if (! emit_bits(state, actbl->ehufco[0], actbl->ehufsi[0]))
      return FALSE;

  return TRUE;
}


/*
 * Emit a restart marker & resynchronize predictions.
 */

LOCAL(boolean)
emit_restart (working_state * state, int restart_num)
{
  int ci;

  if (! flush_bits(state))