syntax.c

/* GNU Emacs routines to deal with syntax tables; also word and list parsing.
   Copyright (C) 1985, 1987, 1993, 1994, 1995, 1997, 1998, 1999, 2002,
                 2003, 2004, 2005, 2006 Free Software Foundation, Inc.

This file is part of GNU Emacs.

GNU Emacs is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2, or (at your option)
any later version.

GNU Emacs is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with GNU Emacs; see the file COPYING.  If not, write to
the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
Boston, MA 02110-1301, USA.  */


#include <config.h>
#include <ctype.h>
#include "lisp.h"
#include "commands.h"
#include "buffer.h"
#include "charset.h"
#include "keymap.h"
#include "regex.h"

/* Make syntax table lookup grant data in gl_state.  */
#define SYNTAX_ENTRY_VIA_PROPERTY

#include "syntax.h"
#include "intervals.h"

/* We use these constants in place for comment-style and
   string-ender-char to distinguish  comments/strings started by
   comment_fence and string_fence codes.  */

#define ST_COMMENT_STYLE (256 + 1)
#define ST_STRING_STYLE (256 + 2)
#include "category.h"

Lisp_Object Qsyntax_table_p, Qsyntax_table, Qscan_error;

int words_include_escapes;
int parse_sexp_lookup_properties;

/* Nonzero means `scan-sexps' treat all multibyte characters as symbol.  */
int multibyte_syntax_as_symbol;

/* Used as a temporary in SYNTAX_ENTRY and other macros in syntax.h,
   if not compiled with GCC.  No need to mark it, since it is used
   only very temporarily.  */
Lisp_Object syntax_temp;

/* Non-zero means an open parenthesis in column 0 is always considered
   to be the start of a defun.  Zero means an open parenthesis in
   column 0 has no special meaning.  */

int open_paren_in_column_0_is_defun_start;

/* This is the internal form of the parse state used in parse-partial-sexp.  */

struct lisp_parse_state
  {
    int depth;		/* Depth at end of parsing.  */
    int instring;	/* -1 if not within string, else desired terminator.  */
    int incomment;	/* -1 if in unnestable comment else comment nesting */
    int comstyle;	/* comment style a=0, or b=1, or ST_COMMENT_STYLE.  */
    int quoted;		/* Nonzero if just after an escape char at end of parsing */
    int thislevelstart;	/* Char number of most recent start-of-expression at current level */
    int prevlevelstart; /* Char number of start of containing expression */
    int location;	/* Char number at which parsing stopped.  */
    int mindepth;	/* Minimum depth seen while scanning.  */
    int comstr_start;	/* Position just after last comment/string starter.  */
    Lisp_Object levelstarts;	/* Char numbers of starts-of-expression
				   of levels (starting from outermost).  */
  };

/* These variables are a cache for finding the start of a defun.
   find_start_pos is the place for which the defun start was found.
   find_start_value is the defun start position found for it.
   find_start_value_byte is the corresponding byte position.
   find_start_buffer is the buffer it was found in.
   find_start_begv is the BEGV value when it was found.
   find_start_modiff is the value of MODIFF when it was found.  */

static int find_start_pos;
static int find_start_value;
static int find_start_value_byte;
static struct buffer *find_start_buffer;
static int find_start_begv;
static int find_start_modiff;


static int find_defun_start P_ ((int, int));
static int back_comment P_ ((EMACS_INT, EMACS_INT, EMACS_INT, int, int,
			     EMACS_INT *, EMACS_INT *));
static int char_quoted P_ ((int, int));
static Lisp_Object skip_chars P_ ((int, int, Lisp_Object, Lisp_Object, int));
static Lisp_Object scan_lists P_ ((EMACS_INT, EMACS_INT, EMACS_INT, int));
static void scan_sexps_forward P_ ((struct lisp_parse_state *,
				    int, int, int, int,
				    int, Lisp_Object, int));
static int in_classes P_ ((int, Lisp_Object));


struct gl_state_s gl_state;		/* Global state of syntax parser.  */

INTERVAL interval_of ();
#define INTERVALS_AT_ONCE 10		/* 1 + max-number of intervals
					   to scan to property-change.  */

/* Update gl_state to an appropriate interval which contains CHARPOS.  The
   sign of COUNT give the relative position of CHARPOS wrt the previously
   valid interval.  If INIT, only [be]_property fields of gl_state are
   valid at start, the rest is filled basing on OBJECT.

   `gl_state.*_i' are the intervals, and CHARPOS is further in the search
   direction than the intervals - or in an interval.  We update the
   current syntax-table basing on the property of this interval, and
   update the interval to start further than CHARPOS - or be
   NULL_INTERVAL.  We also update lim_property to be the next value of
   charpos to call this subroutine again - or be before/after the
   start/end of OBJECT.  */

void
update_syntax_table (charpos, count, init, object)
     int charpos, count, init;
     Lisp_Object object;
{
  Lisp_Object tmp_table;
  int cnt = 0, invalidate = 1;
  INTERVAL i;

  if (init)
    {
      gl_state.old_prop = Qnil;
      gl_state.start = gl_state.b_property;
      gl_state.stop = gl_state.e_property;
      i = interval_of (charpos, object);
      gl_state.backward_i = gl_state.forward_i = i;
      invalidate = 0;
      if (NULL_INTERVAL_P (i))
	return;
      /* interval_of updates only ->position of the return value, so
	 update the parents manually to speed up update_interval.  */
      while (!NULL_PARENT (i))
	{
	  if (AM_RIGHT_CHILD (i))
	    INTERVAL_PARENT (i)->position = i->position
	      - LEFT_TOTAL_LENGTH (i) + TOTAL_LENGTH (i) /* right end */
	      - TOTAL_LENGTH (INTERVAL_PARENT (i))
	      + LEFT_TOTAL_LENGTH (INTERVAL_PARENT (i));
	  else
	    INTERVAL_PARENT (i)->position = i->position - LEFT_TOTAL_LENGTH (i)
	      + TOTAL_LENGTH (i);
	  i = INTERVAL_PARENT (i);
	}
      i = gl_state.forward_i;
      gl_state.b_property = i->position - gl_state.offset;
      gl_state.e_property = INTERVAL_LAST_POS (i) - gl_state.offset;
      goto update;
    }
  i = count > 0 ? gl_state.forward_i : gl_state.backward_i;

  /* We are guaranteed to be called with CHARPOS either in i,
     or further off.  */
  if (NULL_INTERVAL_P (i))
    error ("Error in syntax_table logic for to-the-end intervals");
  else if (charpos < i->position)		/* Move left.  */
    {
      if (count > 0)
	error ("Error in syntax_table logic for intervals <-");
      /* Update the interval.  */
      i = update_interval (i, charpos);
      if (INTERVAL_LAST_POS (i) != gl_state.b_property)
	{
	  invalidate = 0;
	  gl_state.forward_i = i;
	  gl_state.e_property = INTERVAL_LAST_POS (i) - gl_state.offset;
	}
    }
  else if (charpos >= INTERVAL_LAST_POS (i)) /* Move right.  */
    {
      if (count < 0)
	error ("Error in syntax_table logic for intervals ->");
      /* Update the interval.  */
      i = update_interval (i, charpos);
      if (i->position != gl_state.e_property)
	{
	  invalidate = 0;
	  gl_state.backward_i = i;
	  gl_state.b_property = i->position - gl_state.offset;
	}
    }

  update:
  tmp_table = textget (i->plist, Qsyntax_table);

  if (invalidate)
    invalidate = !EQ (tmp_table, gl_state.old_prop); /* Need to invalidate? */

  if (invalidate)		/* Did not get to adjacent interval.  */
    {				/* with the same table => */
				/* invalidate the old range.  */
      if (count > 0)
	{
	  gl_state.backward_i = i;
	  gl_state.b_property = i->position - gl_state.offset;
	}
      else
	{
	  gl_state.forward_i = i;
	  gl_state.e_property = INTERVAL_LAST_POS (i) - gl_state.offset;
	}
    }

  if (!EQ (tmp_table, gl_state.old_prop))
    {
      gl_state.current_syntax_table = tmp_table;
      gl_state.old_prop = tmp_table;
      if (EQ (Fsyntax_table_p (tmp_table), Qt))
	{
	  gl_state.use_global = 0;
	}
      else if (CONSP (tmp_table))
	{
	  gl_state.use_global = 1;
	  gl_state.global_code = tmp_table;
	}
      else
	{
	  gl_state.use_global = 0;
	  gl_state.current_syntax_table = current_buffer->syntax_table;
	}
    }

  while (!NULL_INTERVAL_P (i))
    {
      if (cnt && !EQ (tmp_table, textget (i->plist, Qsyntax_table)))
	{
	  if (count > 0)
	    {
	      gl_state.e_property = i->position - gl_state.offset;
	      gl_state.forward_i = i;
	    }
	  else
	    {
	      gl_state.b_property
		= i->position + LENGTH (i) - gl_state.offset;
	      gl_state.backward_i = i;
	    }
	  return;
	}
      else if (cnt == INTERVALS_AT_ONCE)
	{
	  if (count > 0)
	    {
	      gl_state.e_property
		= i->position + LENGTH (i) - gl_state.offset
		/* e_property at EOB is not set to ZV but to ZV+1, so that
		   we can do INC(from);UPDATE_SYNTAX_TABLE_FORWARD without
		   having to check eob between the two.  */
		+ (NULL_INTERVAL_P (next_interval (i)) ? 1 : 0);
	      gl_state.forward_i = i;
	    }
	  else
	    {
	      gl_state.b_property = i->position - gl_state.offset;
	      gl_state.backward_i = i;
	    }
	  return;
	}
      cnt++;
      i = count > 0 ? next_interval (i) : previous_interval (i);
    }
  eassert (NULL_INTERVAL_P (i)); /* This property goes to the end.  */
  if (count > 0)
    gl_state.e_property = gl_state.stop;
  else
    gl_state.b_property = gl_state.start;
}

/* Returns TRUE if char at CHARPOS is quoted.
   Global syntax-table data should be set up already to be good at CHARPOS
   or after.  On return global syntax data is good for lookup at CHARPOS. */

static int
char_quoted (charpos, bytepos)
     register int charpos, bytepos;
{
  register enum syntaxcode code;
  register int beg = BEGV;
  register int quoted = 0;
  int orig = charpos;

  DEC_BOTH (charpos, bytepos);

  while (charpos >= beg)
    {
      int c;

      UPDATE_SYNTAX_TABLE_BACKWARD (charpos);
      c = FETCH_CHAR (bytepos);
      code = SYNTAX (c);
      if (! (code == Scharquote || code == Sescape))
	break;

      DEC_BOTH (charpos, bytepos);
      quoted = !quoted;
    }

  UPDATE_SYNTAX_TABLE (orig);
  return quoted;
}

/* Return the bytepos one character after BYTEPOS.
   We assume that BYTEPOS is not at the end of the buffer.  */

INLINE int
inc_bytepos (bytepos)
     int bytepos;
{
  if (NILP (current_buffer->enable_multibyte_characters))
    return bytepos + 1;

  INC_POS (bytepos);
  return bytepos;
}

/* Return the bytepos one character before BYTEPOS.
   We assume that BYTEPOS is not at the start of the buffer.  */

INLINE int
dec_bytepos (bytepos)
     int bytepos;
{
  if (NILP (current_buffer->enable_multibyte_characters))
    return bytepos - 1;

  DEC_POS (bytepos);
  return bytepos;
}

/* Return a defun-start position before before POS and not too far before.
   It should be the last one before POS, or nearly the last.

   When open_paren_in_column_0_is_defun_start is nonzero,
   only the beginning of the buffer is treated as a defun-start.

   We record the information about where the scan started
   and what its result was, so that another call in the same area
   can return the same value very quickly.

   There is no promise at which position the global syntax data is
   valid on return from the subroutine, so the caller should explicitly
   update the global data.  */

static int
find_defun_start (pos, pos_byte)
     int pos, pos_byte;
{
  int opoint = PT, opoint_byte = PT_BYTE;

  if (!open_paren_in_column_0_is_defun_start)
    {
      find_start_value_byte = BEGV_BYTE;
      return BEGV;
    }

  /* Use previous finding, if it's valid and applies to this inquiry.  */
  if (current_buffer == find_start_buffer
      /* Reuse the defun-start even if POS is a little farther on.
	 POS might be in the next defun, but that's ok.
	 Our value may not be the best possible, but will still be usable.  */
      && pos <= find_start_pos + 1000
      && pos >= find_start_value
      && BEGV == find_start_begv
      && MODIFF == find_start_modiff)
    return find_start_value;

  /* Back up to start of line.  */
  scan_newline (pos, pos_byte, BEGV, BEGV_BYTE, -1, 1);

  /* We optimize syntax-table lookup for rare updates.  Thus we accept
     only those `^\s(' which are good in global _and_ text-property
     syntax-tables.  */
  gl_state.current_syntax_table = current_buffer->syntax_table;
  gl_state.use_global = 0;
  while (PT > BEGV)
    {
      int c;

      /* Open-paren at start of line means we may have found our
	 defun-start.  */
      c = FETCH_CHAR (PT_BYTE);
      if (SYNTAX (c) == Sopen)
	{
	  SETUP_SYNTAX_TABLE (PT + 1, -1);	/* Try again... */
	  c = FETCH_CHAR (PT_BYTE);
	  if (SYNTAX (c) == Sopen)
	    break;
	  /* Now fallback to the default value.  */
	  gl_state.current_syntax_table = current_buffer->syntax_table;
	  gl_state.use_global = 0;
	}
      /* Move to beg of previous line.  */
      scan_newline (PT, PT_BYTE, BEGV, BEGV_BYTE, -2, 1);
    }

  /* Record what we found, for the next try.  */
  find_start_value = PT;
  find_start_value_byte = PT_BYTE;
  find_start_buffer = current_buffer;
  find_start_modiff = MODIFF;
  find_start_begv = BEGV;
  find_start_pos = pos;

  TEMP_SET_PT_BOTH (opoint, opoint_byte);

  return find_start_value;
}

/* Return the SYNTAX_COMEND_FIRST of the character before POS, POS_BYTE.  */

static int
prev_char_comend_first (pos, pos_byte)
     int pos, pos_byte;
{
  int c, val;

  DEC_BOTH (pos, pos_byte);
  UPDATE_SYNTAX_TABLE_BACKWARD (pos);
  c = FETCH_CHAR (pos_byte);
  val = SYNTAX_COMEND_FIRST (c);
  UPDATE_SYNTAX_TABLE_FORWARD (pos + 1);
  return val;
}

/* Return the SYNTAX_COMSTART_FIRST of the character before POS, POS_BYTE.  */

/* static int
 * prev_char_comstart_first (pos, pos_byte)
 *      int pos, pos_byte;
 * {
 *   int c, val;
 *
 *   DEC_BOTH (pos, pos_byte);
 *   UPDATE_SYNTAX_TABLE_BACKWARD (pos);
 *   c = FETCH_CHAR (pos_byte);
 *   val = SYNTAX_COMSTART_FIRST (c);
 *   UPDATE_SYNTAX_TABLE_FORWARD (pos + 1);
 *   return val;
 * } */

/* Checks whether charpos FROM is at the end of a comment.
   FROM_BYTE is the bytepos corresponding to FROM.
   Do not move back before STOP.

   Return a positive value if we find a comment ending at FROM/FROM_BYTE;
   return -1 otherwise.

   If successful, store the charpos of the comment's beginning
   into *CHARPOS_PTR, and the bytepos into *BYTEPOS_PTR.

   Global syntax data remains valid for backward search starting at
   the returned value (or at FROM, if the search was not successful).  */

static int
back_comment (from, from_byte, stop, comnested, comstyle, charpos_ptr, bytepos_ptr)
     EMACS_INT from, from_byte, stop;
     int comnested, comstyle;
     EMACS_INT *charpos_ptr, *bytepos_ptr;
{
  /* Look back, counting the parity of string-quotes,
     and recording the comment-starters seen.
     When we reach a safe place, assume that's not in a string;
     then step the main scan to the earliest comment-starter seen
     an even number of string quotes away from the safe place.

     OFROM[I] is position of the earliest comment-starter seen
     which is I+2X quotes from the comment-end.
     PARITY is current parity of quotes from the comment end.  */
  int string_style = -1;	/* Presumed outside of any string. */
  int string_lossage = 0;
  /* Not a real lossage: indicates that we have passed a matching comment
     starter plus a non-matching comment-ender, meaning that any matching
     comment-starter we might see later could be a false positive (hidden
     inside another comment).
     Test case:  { a (* b } c (* d *) */
  int comment_lossage = 0;
  int comment_end = from;
  int comment_end_byte = from_byte;
  int comstart_pos = 0;
  int comstart_byte;
  /* Place where the containing defun starts,
     or 0 if we didn't come across it yet.  */
  int defun_start = 0;
  int defun_start_byte = 0;
  register enum syntaxcode code;
  int nesting = 1;		/* current comment nesting */
  int c;
  int syntax = 0;

  /* FIXME: A }} comment-ender style leads to incorrect behavior
     in the case of {{ c }}} because we ignore the last two chars which are
     assumed to be comment-enders although they aren't.  */

  /* At beginning of range to scan, we're outside of strings;
     that determines quote parity to the comment-end.  */
  while (from != stop)
    {
      int temp_byte, prev_syntax;
      int com2start, com2end;

      /* Move back and examine a character.  */
      DEC_BOTH (from, from_byte);
      UPDATE_SYNTAX_TABLE_BACKWARD (from);

      prev_syntax = syntax;
      c = FETCH_CHAR (from_byte);
      syntax = SYNTAX_WITH_FLAGS (c);
      code = SYNTAX (c);

      /* Check for 2-char comment markers.  */
      com2start = (SYNTAX_FLAGS_COMSTART_FIRST (syntax)
		   && SYNTAX_FLAGS_COMSTART_SECOND (prev_syntax)
		   && comstyle == SYNTAX_FLAGS_COMMENT_STYLE (prev_syntax)
		   && (SYNTAX_FLAGS_COMMENT_NESTED (prev_syntax)
		       || SYNTAX_FLAGS_COMMENT_NESTED (syntax)) == comnested);
      com2end = (SYNTAX_FLAGS_COMEND_FIRST (syntax)
		 && SYNTAX_FLAGS_COMEND_SECOND (prev_syntax));

      /* Nasty cases with overlapping 2-char comment markers:
	 - snmp-mode: -- c -- foo -- c --
	              --- c --
		      ------ c --
	 - c-mode:    *||*
		      |* *|* *|
		      |*| |* |*|
		      ///   */

      /* If a 2-char comment sequence partly overlaps with another,
	 we don't try to be clever.  */
      if (from > stop && (com2end || com2start))
	{
	  int next = from, next_byte = from_byte, next_c, next_syntax;
	  DEC_BOTH (next, next_byte);
	  UPDATE_SYNTAX_TABLE_BACKWARD (next);
	  next_c = FETCH_CHAR (next_byte);
	  next_syntax = SYNTAX_WITH_FLAGS (next_c);
	  if (((com2start || comnested)
	       && SYNTAX_FLAGS_COMEND_SECOND (syntax)
	       && SYNTAX_FLAGS_COMEND_FIRST (next_syntax))
	      || ((com2end || comnested)
		  && SYNTAX_FLAGS_COMSTART_SECOND (syntax)
		  && comstyle == SYNTAX_FLAGS_COMMENT_STYLE (syntax)
		  && SYNTAX_FLAGS_COMSTART_FIRST (next_syntax)))
	    goto lossage;
	  /* UPDATE_SYNTAX_TABLE_FORWARD (next + 1); */
	}

      if (com2start && comstart_pos == 0)
	/* We're looking at a comment starter.  But it might be a comment
	   ender as well (see snmp-mode).  The first time we see one, we
	   need to consider it as a comment starter,
	   and the subsequent times as a comment ender.  */
	com2end = 0;

      /* Turn a 2-char comment sequences into the appropriate syntax.  */
      if (com2end)
	code = Sendcomment;
      else if (com2start)
	code = Scomment;
      /* Ignore comment starters of a different style.  */
      else if (code == Scomment
	       && (comstyle != SYNTAX_FLAGS_COMMENT_STYLE (syntax)
		   || SYNTAX_FLAGS_COMMENT_NESTED (syntax) != comnested))
	continue;

      /* Ignore escaped characters, except comment-enders.  */
      if (code != Sendcomment && char_quoted (from, from_byte))
	continue;

      switch (code)
	{
	case Sstring_fence:
	case Scomment_fence:
	  c = (code == Sstring_fence ? ST_STRING_STYLE : ST_COMMENT_STYLE);
	case Sstring:
	  /* Track parity of quotes.  */
	  if (string_style == -1)
	    /* Entering a string.  */
	    string_style = c;
	  else if (string_style == c)
	    /* Leaving the string.  */
	    string_style = -1;
	  else
	    /* If we have two kinds of string delimiters.
	       There's no way to grok this scanning backwards.  */
	    string_lossage = 1;
	  break;

	case Scomment:
	  /* We've already checked that it is the relevant comstyle.  */
	  if (string_style != -1 || comment_lossage || string_lossage)
	    /* There are odd string quotes involved, so let's be careful.
	       Test case in Pascal: " { " a { " } */
	    goto lossage;

	  if (!comnested)
	    {
	      /* Record best comment-starter so far.  */
	      comstart_pos = from;
	      comstart_byte = from_byte;
	    }
	  else if (--nesting <= 0)
	    /* nested comments have to be balanced, so we don't need to
	       keep looking for earlier ones.  We use here the same (slightly
	       incorrect) reasoning as below:  since it is followed by uniform
	       paired string quotes, this comment-start has to be outside of
	       strings, else the comment-end itself would be inside a string. */
	    goto done;
	  break;

	case Sendcomment:
	  if (SYNTAX_FLAGS_COMMENT_STYLE (syntax) == comstyle
	      && ((com2end && SYNTAX_FLAGS_COMMENT_NESTED (prev_syntax))
		  || SYNTAX_FLAGS_COMMENT_NESTED (syntax)) == comnested)
	    /* This is the same style of comment ender as ours. */
	    {
	      if (comnested)
		nesting++;
	      else
		/* Anything before that can't count because it would match
		   this comment-ender rather than ours.  */
		from = stop;	/* Break out of the loop.  */
	    }
	  else if (comstart_pos != 0 || c != '\n')
	    /* We're mixing comment styles here, so we'd better be careful.
	       The (comstart_pos != 0 || c != '\n') check is not quite correct
	       (we should just always set comment_lossage), but removing it
	       would imply that any multiline comment in C would go through
	       lossage, which seems overkill.
	       The failure should only happen in the rare cases such as
	         { (* } *)   */
	    comment_lossage = 1;
	  break;

	case Sopen:
	  /* Assume a defun-start point is outside of strings.  */
	  if (open_paren_in_column_0_is_defun_start
	      && (from == stop
		  || (temp_byte = dec_bytepos (from_byte),
		      FETCH_CHAR (temp_byte) == '\n')))
	    {
	      defun_start = from;
	      defun_start_byte = from_byte;
	      from = stop;	/* Break out of the loop.  */
	    }
	  break;

	default:
	  break;
	}
    }

  if (comstart_pos == 0)
    {
      from = comment_end;
      from_byte = comment_end_byte;
      UPDATE_SYNTAX_TABLE_FORWARD (comment_end - 1);
    }
  /* If comstart_pos is set and we get here (ie. didn't jump to `lossage'
     or `done'), then we've found the beginning of the non-nested comment.  */
  else if (1)	/* !comnested */
    {
      from = comstart_pos;
      from_byte = comstart_byte;
      UPDATE_SYNTAX_TABLE_FORWARD (from - 1);
    }
  else
    {
      struct lisp_parse_state state;
    lossage:
      /* We had two kinds of string delimiters mixed up
	 together.  Decode this going forwards.
	 Scan fwd from a known safe place (beginning-of-defun)
	 to the one in question; this records where we
	 last passed a comment starter.  */
      /* If we did not already find the defun start, find it now.  */
      if (defun_start == 0)
	{
	  defun_start = find_defun_start (comment_end, comment_end_byte);
	  defun_start_byte = find_start_value_byte;
	}
      do
	{
	  scan_sexps_forward (&state,
			      defun_start, defun_start_byte,
			      comment_end, -10000, 0, Qnil, 0);
	  defun_start = comment_end;
	  if (state.incomment == (comnested ? 1 : -1)
	      && state.comstyle == comstyle)
	    from = state.comstr_start;
	  else
	    {
	      from = comment_end;
	      if (state.incomment)
		/* If comment_end is inside some other comment, maybe ours
		   is nested, so we need to try again from within the
		   surrounding comment.  Example: { a (* " *)  */
		{
		  /* FIXME: We should advance by one or two chars. */
		  defun_start = state.comstr_start + 2;
		  defun_start_byte = CHAR_TO_BYTE (defun_start);
		}
	    }
	} while (defun_start < comment_end);

      from_byte = CHAR_TO_BYTE (from);
      UPDATE_SYNTAX_TABLE_FORWARD (from - 1);
    }

 done:
  *charpos_ptr = from;
  *bytepos_ptr = from_byte;

  return (from == comment_end) ? -1 : from;
}

DEFUN ("syntax-table-p", Fsyntax_table_p, Ssyntax_table_p, 1, 1, 0,
       doc: /* Return t if OBJECT is a syntax table.
Currently, any char-table counts as a syntax table.  */)
     (object)
     Lisp_Object object;
{
  if (CHAR_TABLE_P (object)
      && EQ (XCHAR_TABLE (object)->purpose, Qsyntax_table))
    return Qt;
  return Qnil;
}

static void
check_syntax_table (obj)
     Lisp_Object obj;
{
  CHECK_TYPE (CHAR_TABLE_P (obj) && EQ (XCHAR_TABLE (obj)->purpose, Qsyntax_table),
	      Qsyntax_table_p, obj);
}

DEFUN ("syntax-table", Fsyntax_table, Ssyntax_table, 0, 0, 0,
       doc: /* Return the current syntax table.
This is the one specified by the current buffer.  */)
     ()
{
  return current_buffer->syntax_table;
}

DEFUN ("standard-syntax-table", Fstandard_syntax_table,
   Sstandard_syntax_table, 0, 0, 0,
       doc: /* Return the standard syntax table.
This is the one used for new buffers.  */)
     ()
{
  return Vstandard_syntax_table;
}

DEFUN ("copy-syntax-table", Fcopy_syntax_table, Scopy_syntax_table, 0, 1, 0,
       doc: /* Construct a new syntax table and return it.
It is a copy of the TABLE, which defaults to the standard syntax table.  */)
     (table)
     Lisp_Object table;
{
  Lisp_Object copy;

  if (!NILP (table))
    check_syntax_table (table);
  else
    table = Vstandard_syntax_table;

  copy = Fcopy_sequence (table);

  /* Only the standard syntax table should have a default element.
     Other syntax tables should inherit from parents instead.  */
  XCHAR_TABLE (copy)->defalt = Qnil;

  /* Copied syntax tables should all have parents.
     If we copied one with no parent, such as the standard syntax table,
     use the standard syntax table as the copy's parent.  */
  if (NILP (XCHAR_TABLE (copy)->parent))
    Fset_char_table_parent (copy, Vstandard_syntax_table);
  return copy;
}

DEFUN ("set-syntax-table", Fset_syntax_table, Sset_syntax_table, 1, 1, 0,
       doc: /* Select a new syntax table for the current buffer.
One argument, a syntax table.  */)
     (table)
     Lisp_Object table;
{
  int idx;
  check_syntax_table (table);
  current_buffer->syntax_table = table;
  /* Indicate that this buffer now has a specified syntax table.  */
  idx = PER_BUFFER_VAR_IDX (syntax_table);
  SET_PER_BUFFER_VALUE_P (current_buffer, idx, 1);
  return table;
}

/* Convert a letter which signifies a syntax code
 into the code it signifies.
 This is used by modify-syntax-entry, and other things.  */

unsigned char syntax_spec_code[0400] =
  { 0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    (char) Swhitespace, (char) Scomment_fence, (char) Sstring, 0377,
        (char) Smath, 0377, 0377, (char) Squote,
    (char) Sopen, (char) Sclose, 0377, 0377,
	0377, (char) Swhitespace, (char) Spunct, (char) Scharquote,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377,
	(char) Scomment, 0377, (char) Sendcomment, 0377,
    (char) Sinherit, 0377, 0377, 0377, 0377, 0377, 0377, 0377,   /* @, A ... */
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, (char) Sword,
    0377, 0377, 0377, 0377, (char) Sescape, 0377, 0377, (char) Ssymbol,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,   /* `, a, ... */
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, (char) Sword,
    0377, 0377, 0377, 0377, (char) Sstring_fence, 0377, 0377, 0377
  };

/* Indexed by syntax code, give the letter that describes it.  */

char syntax_code_spec[16] =
  {
    ' ', '.', 'w', '_', '(', ')', '\'', '\"', '$', '\\', '/', '<', '>', '@',
    '!', '|'
  };

/* Indexed by syntax code, give the object (cons of syntax code and
   nil) to be stored in syntax table.  Since these objects can be
   shared among syntax tables, we generate them in advance.  By
   sharing objects, the function `describe-syntax' can give a more
   compact listing.  */
static Lisp_Object Vsyntax_code_object;


/* Look up the value for CHARACTER in syntax table TABLE's parent
   and its parents.  SYNTAX_ENTRY calls this, when TABLE itself has nil
   for CHARACTER.  It's actually used only when not compiled with GCC.  */

Lisp_Object
syntax_parent_lookup (table, character)
     Lisp_Object table;
     int character;
{
  Lisp_Object value;

  while (1)
    {
      table = XCHAR_TABLE (table)->parent;
      if (NILP (table))
	return Qnil;

      value = XCHAR_TABLE (table)->contents[character];
      if (!NILP (value))
	return value;
    }
}

DEFUN ("char-syntax", Fchar_syntax, Schar_syntax, 1, 1, 0,
       doc: /* Return the syntax code of CHARACTER, described by a character.
For example, if CHARACTER is a word constituent,
the character `w' is returned.
The characters that correspond to various syntax codes
are listed in the documentation of `modify-syntax-entry'.  */)
     (character)
     Lisp_Object character;
{
  int char_int;
  gl_state.current_syntax_table = current_buffer->syntax_table;

  gl_state.use_global = 0;
  CHECK_NUMBER (character);
  char_int = XINT (character);
  return make_number (syntax_code_spec[(int) SYNTAX (char_int)]);
}

DEFUN ("matching-paren", Fmatching_paren, Smatching_paren, 1, 1, 0,
       doc: /* Return the matching parenthesis of CHARACTER, or nil if none.  */)
     (character)
     Lisp_Object character;
{
  int char_int, code;
  gl_state.current_syntax_table = current_buffer->syntax_table;
  gl_state.use_global = 0;
  CHECK_NUMBER (character);
  char_int = XINT (character);
  code = SYNTAX (char_int);
  if (code == Sopen || code == Sclose)
    return SYNTAX_MATCH (char_int);
  return Qnil;
}

DEFUN ("string-to-syntax", Fstring_to_syntax, Sstring_to_syntax, 1, 1, 0,
       doc: /* Convert a syntax specification STRING into syntax cell form.
STRING should be a string as it is allowed as argument of
`modify-syntax-entry'.  Value is the equivalent cons cell
(CODE . MATCHING-CHAR) that can be used as value of a `syntax-table'
text property.  */)
     (string)
     Lisp_Object string;
{
  register const unsigned char *p;
  register enum syntaxcode code;
  int val;
  Lisp_Object match;

  CHECK_STRING (string);

  p = SDATA (string);
  code = (enum syntaxcode) syntax_spec_code[*p++];
  if (((int) code & 0377) == 0377)
    error ("Invalid syntax description letter: %c", p[-1]);

  if (code == Sinherit)
    return Qnil;

  if (*p)
    {
      int len;
      int character = (STRING_CHAR_AND_LENGTH
		       (p, SBYTES (string) - 1, len));
      XSETINT (match, character);
      if (XFASTINT (match) == ' ')
	match = Qnil;
      p += len;
    }
  else
    match = Qnil;

  val = (int) code;
  while (*p)
    switch (*p++)
      {
      case '1':
	val |= 1 << 16;
	break;

      case '2':
	val |= 1 << 17;
	break;

      case '3':
	val |= 1 << 18;
	break;

      case '4':
	val |= 1 << 19;
	break;

      case 'p':
	val |= 1 << 20;
	break;

      case 'b':
	val |= 1 << 21;
	break;

      case 'n':
	val |= 1 << 22;
	break;
      }

  if (val < XVECTOR (Vsyntax_code_object)->size && NILP (match))
    return XVECTOR (Vsyntax_code_object)->contents[val];
  else
    /* Since we can't use a shared object, let's make a new one.  */
    return Fcons (make_number (val), match);
}

/* I really don't know why this is interactive
   help-form should at least be made useful whilst reading the second arg.  */
DEFUN ("modify-syntax-entry", Fmodify_syntax_entry, Smodify_syntax_entry, 2, 3,
  "cSet syntax for character: \nsSet syntax for %s to: ",
       doc: /* Set syntax for character CHAR according to string NEWENTRY.
The syntax is changed only for table SYNTAX-TABLE, which defaults to
 the current buffer's syntax table.
The first character of NEWENTRY should be one of the following:
  Space or -  whitespace syntax.    w   word constituent.
  _           symbol constituent.   .   punctuation.
  (           open-parenthesis.     )   close-parenthesis.
  "           string quote.         \\   escape.
  $           paired delimiter.     '   expression quote or prefix operator.
  <           comment starter.      >   comment ender.
  /           character-quote.      @   inherit from `standard-syntax-table'.
  |           generic string fence. !   generic comment fence.

Only single-character comment start and end sequences are represented thus.
Two-character sequences are represented as described below.
The second character of NEWENTRY is the matching parenthesis,
 used only if the first character is `(' or `)'.
Any additional characters are flags.
Defined flags are the characters 1, 2, 3, 4, b, p, and n.
 1 means CHAR is the start of a two-char comment start sequence.
 2 means CHAR is the second character of such a sequence.
 3 means CHAR is the start of a two-char comment end sequence.
 4 means CHAR is the second character of such a sequence.

There can be up to two orthogonal comment sequences.  This is to support
language modes such as C++.  By default, all comment sequences are of style
a, but you can set the comment sequence style to b (on the second character
of a comment-start, or the first character of a comment-end sequence) using
this flag:
 b means CHAR is part of comment sequence b.
 n means CHAR is part of a nestable comment sequence.

 p means CHAR is a prefix character for `backward-prefix-chars';
   such characters are treated as whitespace when they occur
   between expressions.
usage: (modify-syntax-entry CHAR NEWENTRY &optional SYNTAX-TABLE) */)
     (c, newentry, syntax_table)
     Lisp_Object c, newentry, syntax_table;
{
  CHECK_NUMBER (c);

  if (NILP (syntax_table))
    syntax_table = current_buffer->syntax_table;
  else
    check_syntax_table (syntax_table);

  SET_RAW_SYNTAX_ENTRY (syntax_table, XINT (c), Fstring_to_syntax (newentry));

  /* We clear the regexp cache, since character classes can now have
     different values from those in the compiled regexps.*/
  clear_regexp_cache ();

  return Qnil;
}

/* Dump syntax table to buffer in human-readable format */

DEFUN ("internal-describe-syntax-value", Finternal_describe_syntax_value,
       Sinternal_describe_syntax_value, 1, 1, 0,
       doc: /* Insert a description of the internal syntax description SYNTAX at point.  */)
     (syntax)
     Lisp_Object syntax;
{
  register enum syntaxcode code;
  char desc, start1, start2, end1, end2, prefix, comstyle, comnested;
  char str[2];
  Lisp_Object first, match_lisp, value = syntax;

  if (NILP (value))
    {
      insert_string ("default");
      return syntax;
    }

  if (CHAR_TABLE_P (value))
    {
      insert_string ("deeper char-table ...");
      return syntax;
    }

  if (!CONSP (value))
    {
      insert_string ("invalid");
      return syntax;
    }

  first = XCAR (value);
  match_lisp = XCDR (value);

  if (!INTEGERP (first) || !(NILP (match_lisp) || INTEGERP (match_lisp)))
    {
      insert_string ("invalid");
      return syntax;
    }

  code = (enum syntaxcode) (XINT (first) & 0377);
  start1 = (XINT (first) >> 16) & 1;
  start2 = (XINT (first) >> 17) & 1;
  end1 = (XINT (first) >> 18) & 1;
  end2 = (XINT (first) >> 19) & 1;
  prefix = (XINT (first) >> 20) & 1;
  comstyle = (XINT (first) >> 21) & 1;
  comnested = (XINT (first) >> 22) & 1;