comparison src/coding.c @ 36087:05ef3feab39b

Doc and message fixes.

36086:60d14f3aae0a

      c = translate_char (translation_table, c, -1, 0, 0);	\
    src += bytes;						\
  } while (0)


/* Produce a multibyte form of characater C to `dst'.  Jump to
   `label_end_of_loop' if there's not enough space at `dst'.

   If we are now in the middle of a composition sequence, the decoded
   character may be ALTCHAR (for the current composition).  In that
   case, the character goes to coding->cmp_data->data instead of

/* Table of pointers to coding systems corresponding to each coding
   categories.  */
struct coding_system *coding_system_table[CODING_CATEGORY_IDX_MAX];

/* Table of coding category masks.  Nth element is a mask for a coding
   cateogry of which priority is Nth.  */
static
int coding_priorities[CODING_CATEGORY_IDX_MAX];

/* Flag to tell if we look up translation table on character code
   conversion.  */

	BYTES is 0x20 plus a byte length of this composition data,

	CHARS is 0x20 plus a number of characters composed by this
	data,

	COMPONENTs are characters of multibye form or composition
	rules encoded by two-byte of ASCII codes.

   In addition, for backward compatibility, the following formats are
   also recognized as composition data on decoding.

#define CODING_ADD_COMPOSITION_COMPONENT(coding, component)	\
  (coding->cmp_data->data[coding->cmp_data->used++] = component)


/* Get one byte from a data pointed by SRC and increment SRC.  If SRC
   is not less than SRC_END, return -1 without inccrementing Src.  */

#define SAFE_ONE_MORE_BYTE() (src >= src_end ? -1 : *src++)


/* Decode a character represented as a component of composition

     int dst_bytes;
{
  unsigned char *dst = *destination;
  int method, data_len, nchars;
  unsigned char *src_base = src++;
  /* Store compoments of composition.  */
  int component[COMPOSITION_DATA_MAX_BUNCH_LENGTH];
  int ncomponent;
  /* Store multibyte form of characters to be composed.  This is for
     Emacs 20 style composition sequence.  */
  unsigned char buf[MAX_COMPOSITION_COMPONENTS * MAX_MULTIBYTE_LENGTH];

   ISO2022 provides many mechanisms to encode several character sets
   in 7-bit and 8-bit environments.  For 7-bit environments, all text
   is encoded using bytes less than 128.  This may make the encoded
   text a little bit longer, but the text passes more easily through
   several types of gateway, some of which strip off the MSB (Most
   Signigant Bit).

   There are two kinds of character sets: control character sets and
   graphic character sets.  The former contain control characters such
   as `newline' and `escape' to provide control functions (control
   functions are also provided by escape sequences).  The latter

   '(' can be omitted.  We refer to this as "short-form" hereafter.

   Now you may notice that there are a lot of ways of encoding the
   same multilingual text in ISO2022.  Actually, there exist many
   coding systems such as Compound Text (used in X11's inter client
   communication, ISO-2022-JP (used in Japanese internet), ISO-2022-KR
   (used in Korean internet), EUC (Extended UNIX Code, used in Asian
   localized platforms), and all of these are variants of ISO2022.

   In addition to the above, Emacs handles two more kinds of escape
   sequences: ISO6429's direction specification and Emacs' private
   sequence for specifying character composition.

  Here's a list of example usages of these composition escape
  sequences (categorized by `enum composition_method').

  COMPOSITION_RELATIVE:
	ESC 0 CHAR [ CHAR ] ESC 1
  COMPOSITOIN_WITH_RULE:
	ESC 2 CHAR [ RULE CHAR ] ESC 1
  COMPOSITION_WITH_ALTCHARS:
	ESC 3 ALTCHAR [ ALTCHAR ] ESC 0 CHAR [ CHAR ] ESC 1
  COMPOSITION_WITH_RULE_ALTCHARS:
	ESC 4 ALTCHAR [ RULE ALTCHAR ] ESC 0 CHAR [ CHAR ] ESC 1 */

      {									   \
	/* This is surely the start of a composition.  We must be sure	   \
           that coding->cmp_data has enough space to store the		   \
           information about the composition.  If not, terminate the	   \
           current decoding loop, allocate one more memory block for	   \
           coding->cmp_data in the calller, then start the decoding	   \
           loop again.  We can't allocate memory here directly because	   \
           it may cause buffer/string relocation.  */			   \
	if (!coding->cmp_data						   \
	    || (coding->cmp_data->used + COMPOSITION_DATA_MAX_BUNCH_LENGTH \
		>= COMPOSITION_DATA_SIZE))				   \

	    coding->composition_rule_follows = 0;			   \
	  }								   \
      }									   \
  } while (0)

/* Handle compositoin end sequence ESC 1.  */

#define DECODE_COMPOSITION_END(c1)					\
  do {									\
    if (coding->composing == COMPOSITION_DISABLED)			\
      {									\

	    case '[':		/* specification of direction */
	      if (coding->flags & CODING_FLAG_ISO_NO_DIRECTION)
		goto label_invalid_code;
	      /* For the moment, nested direction is not supported.
		 So, `coding->mode & CODING_MODE_DIRECTION' zero means
		 left-to-right, and nozero means right-to-left.  */
	      ONE_MORE_BYTE (c1);
	      switch (c1)
		{
		case ']':	/* end of the current direction */
		  coding->mode &= ~CODING_MODE_DIRECTION;

/*
   It is not enough to say just "ISO2022" on encoding, we have to
   specify more details.  In Emacs, each ISO2022 coding system
   variant has the following specifications:
	1. Initial designation to G0 thru G3.
	2. Allows short-form designation?
	3. ASCII should be designated to G0 before control characters?
	4. ASCII should be designated to G0 at end of line?
	5. 7-bit environment or 8-bit environment?
	6. Use locking-shift?

	  else if (COMPOSING_P (coding))
	    {
	      /* COMPOSITION_WITH_ALTCHARS or COMPOSITION_WITH_RULE_ALTCHAR  */
	      if (coding->cmp_data_index == coding->cmp_data_start + data[0])
		/* We have consumed components of the composition.
                   What follows in SRC is the compositions's base
                   text.  */
		ENCODE_COMPOSITION_FAKE_START (coding);
	      else
		{
		  int c = cmp_data->data[coding->cmp_data_index++];

  return;
}

/* See "GENERAL NOTES about `encode_coding_XXX ()' functions".  Encode
   format of end-of-line according to `coding->eol_type'.  It also
   convert multibyte form 8-bit characers to unibyte if
   CODING->src_multibyte is nonzero.  If `coding->mode &
   CODING_MODE_SELECTIVE_DISPLAY' is nonzero, code '\r' in source text
   also means end-of-line.  */

static void

  /* Get values of coding system properties:
     `post-read-conversion', `pre-write-conversion',
     `translation-table-for-decode', `translation-table-for-encode'.  */
  plist = XVECTOR (coding_spec)->contents[3];
  /* Pre & post conversion functions should be disabled if
     inhibit_eol_conversion is nozero.  This is the case that a code
     conversion function is called while those functions are running.  */
  if (! inhibit_pre_post_conversion)
    {
      coding->post_read_conversion = Fplist_get (plist, Qpost_read_conversion);
      coding->pre_write_conversion = Fplist_get (plist, Qpre_write_conversion);

	symbol) `japanese-iso-8bit' by default.

   o coding-category-iso-7-else

   	The category for a coding system which has the same code range
	as ISO2022 of 7-bit environemnt but uses locking shift or
	single shift functions.  Assigned the coding-system (Lisp
	symbol) `iso-2022-7bit-lock' by default.

   o coding-category-iso-8-else

   	The category for a coding system which has the same code range
	as ISO2022 of 8-bit environemnt but uses locking shift or
	single shift functions.  Assigned the coding-system (Lisp
	symbol) `iso-2022-8bit-ss2' by default.

   o coding-category-big5

  if (coding->eol_type == CODING_EOL_UNDECIDED
      && coding->type != coding_type_ccl)
    {
      detect_eol (coding, source, src_bytes);
      /* We had better recover the original eol format if we
	 encounter an inconsitent eol format while decoding.  */
      coding->mode |= CODING_MODE_INHIBIT_INCONSISTENT_EOL;
    }

  coding->produced = coding->produced_char = 0;
  coding->consumed = coding->consumed_char = 0;

	  break;

	case CODING_CATEGORY_IDX_ISO_7:
	case CODING_CATEGORY_IDX_ISO_7_TIGHT:
	  {
	    /* We can skip all charactes at the tail except for 8-bit
	       codes and ESC and the following 2-byte at the tail.  */
	    unsigned char *eight_bit = NULL;

	    if (eol_conversion)
	      while (begp < endp

      int i;
      for (i = 0; i < 128; i++)
	if (!NILP (CHAR_TABLE_REF (translation_table, i)))
	  break;
      if (i < 128)
	/* Some ASCII character should be tranlsated.  We give up
	   shrinking.  */
	return;
    }

  if (str)

    coding->cmp_data = coding->cmp_data->prev;
  coding->cmp_data_start = 0;
}

/* Reflect the saved information about compositions to OBJ.
   CODING->cmp_data points to a memory block for the informaiton.  OBJ
   is a buffer or a string, defaults to the current buffer.  */

void
coding_restore_composition (coding, obj)
     struct coding_system *coding;

   If REPLACE is nonzero, we do various things as if the original text
   is deleted and a new text is inserted.  See the comments in
   replace_range (insdel.c) to know what we are doing.

   If REPLACE is zero, it is assumed that the source text is unibyte.
   Otherwize, it is assumed that the source text is multibyte.  */

int
code_convert_region (from, from_byte, to, to_byte, coding, encodep, replace)
     int from, from_byte, to, to_byte, encodep, replace;
     struct coding_system *coding;

	{
	  detect_eol (coding, BYTE_POS_ADDR (from_byte), len_byte);
	  if (coding->eol_type == CODING_EOL_UNDECIDED)
	    coding->eol_type = CODING_EOL_LF;
	  /* We had better recover the original eol format if we
	     encounter an inconsitent eol format while decoding.  */
	  coding->mode |= CODING_MODE_INHIBIT_INCONSISTENT_EOL;
	}
    }

  /* Now we convert the text.  */

      from += head_skip;
      to -= tail_skip;
      len -= total_skip; len_byte -= total_skip;
    }

  /* For converion, we must put the gap before the text in addition to
     making the gap larger for efficient decoding.  The required gap
     size starts from 2000 which is the magic number used in make_gap.
     But, after one batch of conversion, it will be incremented if we
     find that it is not enough .  */
  require = 2000;

	  break;
	}
      if (first)
	{
	  /* We have just done the first batch of conversion which was
	     stoped because of insufficient gap.  Let's reconsider the
	     required gap size (i.e. SRT - DST) now.

	     We have converted ORIG bytes (== coding->consumed) into
	     NEW bytes (coding->produced).  To convert the remaining
	     LEN bytes, we may need REQUIRE bytes of gap, where:

	  GPT -= inserted_byte; GPT_BYTE -= inserted_byte;
	}
      inserted_byte = str_to_multibyte (GPT_ADDR, GAP_SIZE, inserted_byte);
    }

  /* If we have shrinked the conversion area, adjust it now.  */ 
  if (total_skip > 0)
    {
      if (tail_skip > 0)
	safe_bcopy (GAP_END_ADDR, GPT_ADDR + inserted_byte, tail_skip);
      inserted += total_skip; inserted_byte += total_skip;

	  saved_coding_symbol = coding->symbol;
	  detect_eol (coding, XSTRING (str)->data, to_byte);
	  if (coding->eol_type == CODING_EOL_UNDECIDED)
	    coding->eol_type = CODING_EOL_LF;
	  /* We had better recover the original eol format if we
	     encounter an inconsitent eol format while decoding.  */
	  coding->mode |= CODING_MODE_INHIBIT_INCONSISTENT_EOL;
	}
    }

  if (coding->type == coding_type_no_conversion

}

DEFUN ("decode-coding-string", Fdecode_coding_string, Sdecode_coding_string,
       2, 3, 0,
  "Decode STRING which is encoded in CODING-SYSTEM, and return the result.\n\
Optional arg NOCOPY non-nil means it is ok to return STRING itself\n\
if the decoding operation is trivial.\n\
This function sets `last-coding-system-used' to the precise coding system\n\
used (which may be different from CODING-SYSTEM if CODING-SYSTEM is\n\
not fully specified.)")
  (string, coding_system, nocopy)

}

DEFUN ("encode-coding-string", Fencode_coding_string, Sencode_coding_string,
       2, 3, 0,
  "Encode STRING to CODING-SYSTEM, and return the result.\n\
Optional arg NOCOPY non-nil means it is ok to return STRING itself\n\
if the encoding operation is trivial.\n\
This function sets `last-coding-system-used' to the precise coding system\n\
used (which may be different from CODING-SYSTEM if CODING-SYSTEM is\n\
not fully specified.)")
  (string, coding_system, nocopy)

{
  CHECK_SYMBOL (coding_system, 0);
  setup_coding_system (Fcheck_coding_system (coding_system), &terminal_coding);
  /* We had better not send unsafe characters to terminal.  */
  terminal_coding.flags |= CODING_FLAG_ISO_SAFE;
  /* Characer composition should be disabled.  */
  terminal_coding.composing = COMPOSITION_DISABLED;
  /* Error notification should be suppressed.  */
  terminal_coding.suppress_error = 1;
  terminal_coding.src_multibyte = 1;
  terminal_coding.dst_multibyte = 0;

     Lisp_Object coding_system;
{
  CHECK_SYMBOL (coding_system, 0);
  setup_coding_system (Fcheck_coding_system (coding_system),
		       &safe_terminal_coding);
  /* Characer composition should be disabled.  */
  safe_terminal_coding.composing = COMPOSITION_DISABLED;
  /* Error notification should be suppressed.  */
  terminal_coding.suppress_error = 1;
  safe_terminal_coding.src_multibyte = 1;
  safe_terminal_coding.dst_multibyte = 0;

  (coding_system)
     Lisp_Object coding_system;
{
  CHECK_SYMBOL (coding_system, 0);
  setup_coding_system (Fcheck_coding_system (coding_system), &keyboard_coding);
  /* Characer composition should be disabled.  */
  keyboard_coding.composing = COMPOSITION_DISABLED;
  return Qnil;
}

DEFUN ("keyboard-coding-system",

  if (nargs < 2)
    error ("Too few arguments");
  operation = args[0];
  if (!SYMBOLP (operation)
      || !INTEGERP (target_idx = Fget (operation, Qtarget_idx)))
    error ("Invalid first arguement");
  if (nargs < 1 + XINT (target_idx))
    error ("Too few arguments for operation: %s",
	   XSYMBOL (operation)->name->data);
  target = args[XINT (target_idx) + 1];
  if (!(STRINGP (target)
	|| (EQ (operation, Qopen_network_stream) && INTEGERP (target))))
    error ("Invalid %dth argument", XINT (target_idx) + 1);

  chain = ((EQ (operation, Qinsert_file_contents)
	    || EQ (operation, Qwrite_region))
	   ? Vfile_coding_system_alist
	   : (EQ (operation, Qopen_network_stream)

    "Table for translating characters while decoding.");
  Vstandard_translation_table_for_decode = Qnil;

  DEFVAR_LISP ("standard-translation-table-for-encode",
    &Vstandard_translation_table_for_encode,
    "Table for translationg characters while encoding.");
  Vstandard_translation_table_for_encode = Qnil;

  DEFVAR_LISP ("charset-revision-table", &Vcharset_revision_alist,
    "Alist of charsets vs revision numbers.\n\
While encoding, if a charset (car part of an element) is found,\n\
designate it with the escape sequence identifing revision (cdr part of the element).");
  Vcharset_revision_alist = Qnil;

  DEFVAR_LISP ("default-process-coding-system",
	       &Vdefault_process_coding_system,
    "Cons of coding systems used for process I/O by default.\n\

36087:05ef3feab39b

      c = translate_char (translation_table, c, -1, 0, 0);	\
    src += bytes;						\
  } while (0)


/* Produce a multibyte form of character C to `dst'.  Jump to
   `label_end_of_loop' if there's not enough space at `dst'.

   If we are now in the middle of a composition sequence, the decoded
   character may be ALTCHAR (for the current composition).  In that
   case, the character goes to coding->cmp_data->data instead of

/* Table of pointers to coding systems corresponding to each coding
   categories.  */
struct coding_system *coding_system_table[CODING_CATEGORY_IDX_MAX];

/* Table of coding category masks.  Nth element is a mask for a coding
   category of which priority is Nth.  */
static
int coding_priorities[CODING_CATEGORY_IDX_MAX];

/* Flag to tell if we look up translation table on character code
   conversion.  */

	BYTES is 0x20 plus a byte length of this composition data,

	CHARS is 0x20 plus a number of characters composed by this
	data,

	COMPONENTs are characters of multibyte form or composition
	rules encoded by two-byte of ASCII codes.

   In addition, for backward compatibility, the following formats are
   also recognized as composition data on decoding.

#define CODING_ADD_COMPOSITION_COMPONENT(coding, component)	\
  (coding->cmp_data->data[coding->cmp_data->used++] = component)


/* Get one byte from a data pointed by SRC and increment SRC.  If SRC
   is not less than SRC_END, return -1 without incrementing Src.  */

#define SAFE_ONE_MORE_BYTE() (src >= src_end ? -1 : *src++)


/* Decode a character represented as a component of composition

     int dst_bytes;
{
  unsigned char *dst = *destination;
  int method, data_len, nchars;
  unsigned char *src_base = src++;
  /* Store components of composition.  */
  int component[COMPOSITION_DATA_MAX_BUNCH_LENGTH];
  int ncomponent;
  /* Store multibyte form of characters to be composed.  This is for
     Emacs 20 style composition sequence.  */
  unsigned char buf[MAX_COMPOSITION_COMPONENTS * MAX_MULTIBYTE_LENGTH];

   ISO2022 provides many mechanisms to encode several character sets
   in 7-bit and 8-bit environments.  For 7-bit environments, all text
   is encoded using bytes less than 128.  This may make the encoded
   text a little bit longer, but the text passes more easily through
   several types of gateway, some of which strip off the MSB (Most
   Significant Bit).

   There are two kinds of character sets: control character sets and
   graphic character sets.  The former contain control characters such
   as `newline' and `escape' to provide control functions (control
   functions are also provided by escape sequences).  The latter

   '(' can be omitted.  We refer to this as "short-form" hereafter.

   Now you may notice that there are a lot of ways of encoding the
   same multilingual text in ISO2022.  Actually, there exist many
   coding systems such as Compound Text (used in X11's inter client
   communication, ISO-2022-JP (used in Japanese Internet), ISO-2022-KR
   (used in Korean Internet), EUC (Extended UNIX Code, used in Asian
   localized platforms), and all of these are variants of ISO2022.

   In addition to the above, Emacs handles two more kinds of escape
   sequences: ISO6429's direction specification and Emacs' private
   sequence for specifying character composition.

  Here's a list of example usages of these composition escape
  sequences (categorized by `enum composition_method').

  COMPOSITION_RELATIVE:
	ESC 0 CHAR [ CHAR ] ESC 1
  COMPOSITION_WITH_RULE:
	ESC 2 CHAR [ RULE CHAR ] ESC 1
  COMPOSITION_WITH_ALTCHARS:
	ESC 3 ALTCHAR [ ALTCHAR ] ESC 0 CHAR [ CHAR ] ESC 1
  COMPOSITION_WITH_RULE_ALTCHARS:
	ESC 4 ALTCHAR [ RULE ALTCHAR ] ESC 0 CHAR [ CHAR ] ESC 1 */

      {									   \
	/* This is surely the start of a composition.  We must be sure	   \
           that coding->cmp_data has enough space to store the		   \
           information about the composition.  If not, terminate the	   \
           current decoding loop, allocate one more memory block for	   \
           coding->cmp_data in the caller, then start the decoding	   \
           loop again.  We can't allocate memory here directly because	   \
           it may cause buffer/string relocation.  */			   \
	if (!coding->cmp_data						   \
	    || (coding->cmp_data->used + COMPOSITION_DATA_MAX_BUNCH_LENGTH \
		>= COMPOSITION_DATA_SIZE))				   \

	    coding->composition_rule_follows = 0;			   \
	  }								   \
      }									   \
  } while (0)

/* Handle composition end sequence ESC 1.  */

#define DECODE_COMPOSITION_END(c1)					\
  do {									\
    if (coding->composing == COMPOSITION_DISABLED)			\
      {									\

	    case '[':		/* specification of direction */
	      if (coding->flags & CODING_FLAG_ISO_NO_DIRECTION)
		goto label_invalid_code;
	      /* For the moment, nested direction is not supported.
		 So, `coding->mode & CODING_MODE_DIRECTION' zero means
		 left-to-right, and nonzero means right-to-left.  */
	      ONE_MORE_BYTE (c1);
	      switch (c1)
		{
		case ']':	/* end of the current direction */
		  coding->mode &= ~CODING_MODE_DIRECTION;

/*
   It is not enough to say just "ISO2022" on encoding, we have to
   specify more details.  In Emacs, each ISO2022 coding system
   variant has the following specifications:
	1. Initial designation to G0 through G3.
	2. Allows short-form designation?
	3. ASCII should be designated to G0 before control characters?
	4. ASCII should be designated to G0 at end of line?
	5. 7-bit environment or 8-bit environment?
	6. Use locking-shift?

	  else if (COMPOSING_P (coding))
	    {
	      /* COMPOSITION_WITH_ALTCHARS or COMPOSITION_WITH_RULE_ALTCHAR  */
	      if (coding->cmp_data_index == coding->cmp_data_start + data[0])
		/* We have consumed components of the composition.
                   What follows in SRC is the composition's base
                   text.  */
		ENCODE_COMPOSITION_FAKE_START (coding);
	      else
		{
		  int c = cmp_data->data[coding->cmp_data_index++];

  return;
}

/* See "GENERAL NOTES about `encode_coding_XXX ()' functions".  Encode
   format of end-of-line according to `coding->eol_type'.  It also
   convert multibyte form 8-bit characters to unibyte if
   CODING->src_multibyte is nonzero.  If `coding->mode &
   CODING_MODE_SELECTIVE_DISPLAY' is nonzero, code '\r' in source text
   also means end-of-line.  */

static void

  /* Get values of coding system properties:
     `post-read-conversion', `pre-write-conversion',
     `translation-table-for-decode', `translation-table-for-encode'.  */
  plist = XVECTOR (coding_spec)->contents[3];
  /* Pre & post conversion functions should be disabled if
     inhibit_eol_conversion is nonzero.  This is the case that a code
     conversion function is called while those functions are running.  */
  if (! inhibit_pre_post_conversion)
    {
      coding->post_read_conversion = Fplist_get (plist, Qpost_read_conversion);
      coding->pre_write_conversion = Fplist_get (plist, Qpre_write_conversion);

	symbol) `japanese-iso-8bit' by default.

   o coding-category-iso-7-else

   	The category for a coding system which has the same code range
	as ISO2022 of 7-bit environment but uses locking shift or
	single shift functions.  Assigned the coding-system (Lisp
	symbol) `iso-2022-7bit-lock' by default.

   o coding-category-iso-8-else

   	The category for a coding system which has the same code range
	as ISO2022 of 8-bit environment but uses locking shift or
	single shift functions.  Assigned the coding-system (Lisp
	symbol) `iso-2022-8bit-ss2' by default.

   o coding-category-big5

  if (coding->eol_type == CODING_EOL_UNDECIDED
      && coding->type != coding_type_ccl)
    {
      detect_eol (coding, source, src_bytes);
      /* We had better recover the original eol format if we
	 encounter an inconsistent eol format while decoding.  */
      coding->mode |= CODING_MODE_INHIBIT_INCONSISTENT_EOL;
    }

  coding->produced = coding->produced_char = 0;
  coding->consumed = coding->consumed_char = 0;

	  break;

	case CODING_CATEGORY_IDX_ISO_7:
	case CODING_CATEGORY_IDX_ISO_7_TIGHT:
	  {
	    /* We can skip all characters at the tail except for 8-bit
	       codes and ESC and the following 2-byte at the tail.  */
	    unsigned char *eight_bit = NULL;

	    if (eol_conversion)
	      while (begp < endp

      int i;
      for (i = 0; i < 128; i++)
	if (!NILP (CHAR_TABLE_REF (translation_table, i)))
	  break;
      if (i < 128)
	/* Some ASCII character should be translated.  We give up
	   shrinking.  */
	return;
    }

  if (str)

    coding->cmp_data = coding->cmp_data->prev;
  coding->cmp_data_start = 0;
}

/* Reflect the saved information about compositions to OBJ.
   CODING->cmp_data points to a memory block for the information.  OBJ
   is a buffer or a string, defaults to the current buffer.  */

void
coding_restore_composition (coding, obj)
     struct coding_system *coding;

   If REPLACE is nonzero, we do various things as if the original text
   is deleted and a new text is inserted.  See the comments in
   replace_range (insdel.c) to know what we are doing.

   If REPLACE is zero, it is assumed that the source text is unibyte.
   Otherwise, it is assumed that the source text is multibyte.  */

int
code_convert_region (from, from_byte, to, to_byte, coding, encodep, replace)
     int from, from_byte, to, to_byte, encodep, replace;
     struct coding_system *coding;

	{
	  detect_eol (coding, BYTE_POS_ADDR (from_byte), len_byte);
	  if (coding->eol_type == CODING_EOL_UNDECIDED)
	    coding->eol_type = CODING_EOL_LF;
	  /* We had better recover the original eol format if we
	     encounter an inconsistent eol format while decoding.  */
	  coding->mode |= CODING_MODE_INHIBIT_INCONSISTENT_EOL;
	}
    }

  /* Now we convert the text.  */

      from += head_skip;
      to -= tail_skip;
      len -= total_skip; len_byte -= total_skip;
    }

  /* For conversion, we must put the gap before the text in addition to
     making the gap larger for efficient decoding.  The required gap
     size starts from 2000 which is the magic number used in make_gap.
     But, after one batch of conversion, it will be incremented if we
     find that it is not enough .  */
  require = 2000;

	  break;
	}
      if (first)
	{
	  /* We have just done the first batch of conversion which was
	     stopped because of insufficient gap.  Let's reconsider the
	     required gap size (i.e. SRT - DST) now.

	     We have converted ORIG bytes (== coding->consumed) into
	     NEW bytes (coding->produced).  To convert the remaining
	     LEN bytes, we may need REQUIRE bytes of gap, where:

	  GPT -= inserted_byte; GPT_BYTE -= inserted_byte;
	}
      inserted_byte = str_to_multibyte (GPT_ADDR, GAP_SIZE, inserted_byte);
    }

  /* If we shrank the conversion area, adjust it now.  */ 
  if (total_skip > 0)
    {
      if (tail_skip > 0)
	safe_bcopy (GAP_END_ADDR, GPT_ADDR + inserted_byte, tail_skip);
      inserted += total_skip; inserted_byte += total_skip;

	  saved_coding_symbol = coding->symbol;
	  detect_eol (coding, XSTRING (str)->data, to_byte);
	  if (coding->eol_type == CODING_EOL_UNDECIDED)
	    coding->eol_type = CODING_EOL_LF;
	  /* We had better recover the original eol format if we
	     encounter an inconsistent eol format while decoding.  */
	  coding->mode |= CODING_MODE_INHIBIT_INCONSISTENT_EOL;
	}
    }

  if (coding->type == coding_type_no_conversion

}

DEFUN ("decode-coding-string", Fdecode_coding_string, Sdecode_coding_string,
       2, 3, 0,
  "Decode STRING which is encoded in CODING-SYSTEM, and return the result.\n\
Optional arg NOCOPY non-nil means it is OK to return STRING itself\n\
if the decoding operation is trivial.\n\
This function sets `last-coding-system-used' to the precise coding system\n\
used (which may be different from CODING-SYSTEM if CODING-SYSTEM is\n\
not fully specified.)")
  (string, coding_system, nocopy)

}

DEFUN ("encode-coding-string", Fencode_coding_string, Sencode_coding_string,
       2, 3, 0,
  "Encode STRING to CODING-SYSTEM, and return the result.\n\
Optional arg NOCOPY non-nil means it is OK to return STRING itself\n\
if the encoding operation is trivial.\n\
This function sets `last-coding-system-used' to the precise coding system\n\
used (which may be different from CODING-SYSTEM if CODING-SYSTEM is\n\
not fully specified.)")
  (string, coding_system, nocopy)

{
  CHECK_SYMBOL (coding_system, 0);
  setup_coding_system (Fcheck_coding_system (coding_system), &terminal_coding);
  /* We had better not send unsafe characters to terminal.  */
  terminal_coding.flags |= CODING_FLAG_ISO_SAFE;
  /* Character composition should be disabled.  */
  terminal_coding.composing = COMPOSITION_DISABLED;
  /* Error notification should be suppressed.  */
  terminal_coding.suppress_error = 1;
  terminal_coding.src_multibyte = 1;
  terminal_coding.dst_multibyte = 0;

     Lisp_Object coding_system;
{
  CHECK_SYMBOL (coding_system, 0);
  setup_coding_system (Fcheck_coding_system (coding_system),
		       &safe_terminal_coding);
  /* Character composition should be disabled.  */
  safe_terminal_coding.composing = COMPOSITION_DISABLED;
  /* Error notification should be suppressed.  */
  terminal_coding.suppress_error = 1;
  safe_terminal_coding.src_multibyte = 1;
  safe_terminal_coding.dst_multibyte = 0;

  (coding_system)
     Lisp_Object coding_system;
{
  CHECK_SYMBOL (coding_system, 0);
  setup_coding_system (Fcheck_coding_system (coding_system), &keyboard_coding);
  /* Character composition should be disabled.  */
  keyboard_coding.composing = COMPOSITION_DISABLED;
  return Qnil;
}

DEFUN ("keyboard-coding-system",

  if (nargs < 2)
    error ("Too few arguments");
  operation = args[0];
  if (!SYMBOLP (operation)
      || !INTEGERP (target_idx = Fget (operation, Qtarget_idx)))
    error ("Invalid first argument");
  if (nargs < 1 + XINT (target_idx))
    error ("Too few arguments for operation: %s",
	   XSYMBOL (operation)->name->data);
  target = args[XINT (target_idx) + 1];
  if (!(STRINGP (target)
	|| (EQ (operation, Qopen_network_stream) && INTEGERP (target))))
    error ("Invalid argument %d", XINT (target_idx) + 1);

  chain = ((EQ (operation, Qinsert_file_contents)
	    || EQ (operation, Qwrite_region))
	   ? Vfile_coding_system_alist
	   : (EQ (operation, Qopen_network_stream)

    "Table for translating characters while decoding.");
  Vstandard_translation_table_for_decode = Qnil;

  DEFVAR_LISP ("standard-translation-table-for-encode",
    &Vstandard_translation_table_for_encode,
    "Table for translating characters while encoding.");
  Vstandard_translation_table_for_encode = Qnil;

  DEFVAR_LISP ("charset-revision-table", &Vcharset_revision_alist,
    "Alist of charsets vs revision numbers.\n\
While encoding, if a charset (car part of an element) is found,\n\
designate it with the escape sequence identifying revision (cdr part of the element).");
  Vcharset_revision_alist = Qnil;

  DEFVAR_LISP ("default-process-coding-system",
	       &Vdefault_process_coding_system,
    "Cons of coding systems used for process I/O by default.\n\