]> code.delx.au - gnu-emacs/blobdiff - src/syntax.c
* xterm.c (x_term_init): Adjust message printed when we can't
[gnu-emacs] / src / syntax.c
index aa40547ba5ac457775c58fc50929982a69902a0d..f00fb3c9dd3dab270d7fa8925b866100e7b662eb 100644 (file)
@@ -1,5 +1,5 @@
 /* GNU Emacs routines to deal with syntax tables; also word and list parsing.
-   Copyright (C) 1985, 1987, 1992 Free Software Foundation, Inc.
+   Copyright (C) 1985, 1987, 1993 Free Software Foundation, Inc.
 
 This file is part of GNU Emacs.
 
@@ -27,6 +27,9 @@ the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.  */
 
 Lisp_Object Qsyntax_table_p;
 
+static void scan_sexps_forward ();
+static int char_quoted ();
+
 int words_include_escapes;
 
 /* This is the internal form of the parse state used in parse-partial-sexp.  */
@@ -45,6 +48,63 @@ struct lisp_parse_state
     int comstart;      /* Position just after last comment starter.  */
   };
 \f
+/* 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_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 struct buffer *find_start_buffer;
+static int find_start_begv;
+static int find_start_modiff;
+
+/* Find a defun-start that is the last one before POS (or nearly the last).
+   We record what we find, so that another call in the same area
+   can return the same value right away.  */
+
+static int
+find_defun_start (pos)
+     int pos;
+{
+  int tem;
+  int shortage;
+
+  /* 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.  */
+  tem = scan_buffer ('\n', pos, -1, &shortage);
+
+  while (tem > BEGV)
+    {
+      /* Open-paren at start of line means we found our defun-start.  */
+      if (SYNTAX (FETCH_CHAR (tem)) == Sopen)
+       break;
+      /* Move to beg of previous line.  */
+      tem = scan_buffer ('\n', tem, -2, &shortage);
+    }
+
+  /* Record what we found, for the next try.  */
+  find_start_value = tem;
+  find_start_buffer = current_buffer;
+  find_start_modiff = MODIFF;
+  find_start_begv = BEGV;
+  find_start_pos = pos;
+
+  return find_start_value;
+}
+\f
 DEFUN ("syntax-table-p", Fsyntax_table_p, Ssyntax_table_p, 1, 1, 0,
   "Return t if ARG is a syntax table.\n\
 Any vector of 256 elements will do.")
@@ -63,7 +123,7 @@ check_syntax_table (obj)
   register Lisp_Object tem;
   while (tem = Fsyntax_table_p (obj),
         NILP (tem))
-    obj = wrong_type_argument (Qsyntax_table_p, obj, 0);
+    obj = wrong_type_argument (Qsyntax_table_p, obj);
   return obj;
 }   
 
@@ -116,7 +176,8 @@ One argument, a syntax table.")
   table = check_syntax_table (table);
   current_buffer->syntax_table = table;
   /* Indicate that this buffer now has a specified syntax table.  */
-  current_buffer->local_var_flags |= buffer_local_flags.syntax_table;
+  current_buffer->local_var_flags
+    |= XFASTINT (buffer_local_flags.syntax_table);
   return table;
 }
 \f
@@ -169,7 +230,7 @@ are listed in the documentation of `modify-syntax-entry'.")
    for make-docfile to see.  We cannot put this in the real DEFUN
    due to limits in the Unix cpp.
 
-DEFUN ("modify-syntax-entry", foo, bar, 0, 0, 0,
+DEFUN ("modify-syntax-entry", foo, bar, 2, 3, 0,
   "Set syntax for character CHAR according to string S.\n\
 The syntax is changed only for table TABLE, which defaults to\n\
  the current buffer's syntax table.\n\
@@ -202,7 +263,7 @@ this flag:\n\
  p means C is a prefix character for `backward-prefix-chars';\n\
    such characters are treated as whitespace when they occur\n\
    between expressions.")
-
+  (char, s, table)
 */
 
 DEFUN ("modify-syntax-entry", Fmodify_syntax_entry, Smodify_syntax_entry, 2, 3, 
@@ -270,6 +331,7 @@ DEFUN ("modify-syntax-entry", Fmodify_syntax_entry, Smodify_syntax_entry, 2, 3,
 \f
 /* Dump syntax table to buffer in human-readable format */
 
+static void
 describe_syntax (value)
     Lisp_Object value;
 {
@@ -365,9 +427,7 @@ describe_syntax (value)
   if (match)
     {
       insert_string (", matches ");
-      
-      str[0] = match, str[1] = 0;
-      insert (str, 1);
+      insert_char (match);
     }
 
   if (start1)
@@ -388,13 +448,13 @@ describe_syntax (value)
   insert_string ("\n");
 }
 
-Lisp_Object
+static Lisp_Object
 describe_syntax_1 (vector)
      Lisp_Object vector;
 {
   struct buffer *old = current_buffer;
   set_buffer_internal (XBUFFER (Vstandard_output));
-  describe_vector (vector, Qnil, describe_syntax, 0, Qnil, Qnil);
+  describe_vector (vector, Qnil, describe_syntax, 0, Qnil);
   set_buffer_internal (old);
   return Qnil;
 }
@@ -508,6 +568,274 @@ and nil is returned.")
   return Qt;
 }
 \f
+DEFUN ("forward-comment", Fforward_comment, Sforward_comment, 1, 1, 0,
+  "Move forward across up to N comments.  If N is negative, move backward.\n\
+Stop scanning if we find something other than a comment or whitespace.\n\
+Set point to where scanning stops.\n\
+If N comments are found as expected, with nothing except whitespace\n\
+between them, return t; otherwise return nil.")
+  (count)
+     Lisp_Object count;
+{
+  register int from;
+  register int stop;
+  register int c;
+  register enum syntaxcode code;
+  int comstyle = 0;        /* style of comment encountered */
+  int found;
+  int count1;
+
+  CHECK_NUMBER (count, 0);
+  count1 = XINT (count);
+
+  immediate_quit = 1;
+  QUIT;
+
+  from = PT;
+
+  while (count1 > 0)
+    {
+      stop = ZV;
+      while (from < stop)
+       {
+         c = FETCH_CHAR (from);
+         code = SYNTAX (c);
+         from++;
+         comstyle = 0;
+         if (from < stop && SYNTAX_COMSTART_FIRST (c)
+             && SYNTAX_COMSTART_SECOND (FETCH_CHAR (from)))
+           {
+             /* we have encountered a comment start sequence and we 
+                are ignoring all text inside comments. we must record
+                the comment style this sequence begins so that later,
+                only a comment end of the same style actually ends
+                the comment section */
+             code = Scomment;
+             comstyle = SYNTAX_COMMENT_STYLE (FETCH_CHAR (from));
+             from++;
+           }
+
+         if (code == Scomment)
+           {
+             while (1)
+               {
+                 if (from == stop)
+                   {
+                     immediate_quit = 0;
+                     SET_PT (from);
+                     return Qnil;
+                   }
+                 c = FETCH_CHAR (from);
+                 if (SYNTAX (c) == Sendcomment
+                     && SYNTAX_COMMENT_STYLE (c) == comstyle)
+                   /* we have encountered a comment end of the same style
+                      as the comment sequence which began this comment
+                      section */
+                   break;
+                 from++;
+                 if (from < stop && SYNTAX_COMEND_FIRST (c)
+                     && SYNTAX_COMEND_SECOND (FETCH_CHAR (from))
+                     && SYNTAX_COMMENT_STYLE (c) == comstyle)
+                   /* we have encountered a comment end of the same style
+                      as the comment sequence which began this comment
+                      section */
+                   { from++; break; }
+               }
+             /* We have skipped one comment.  */
+             break;
+           }
+         else if (code != Swhitespace && code != Sendcomment)
+           {
+             immediate_quit = 0;
+             SET_PT (from - 1);
+             return Qnil;
+           }
+       }
+
+      /* End of comment reached */
+      count1--;
+    }
+
+  while (count1 < 0)
+    {
+      stop = BEGV;
+      while (from > stop)
+       {
+         int quoted;
+
+         from--;
+         quoted = char_quoted (from);
+         if (quoted)
+           from--;
+         c = FETCH_CHAR (from);
+         code = SYNTAX (c);
+         comstyle = 0;
+         if (from > stop && SYNTAX_COMEND_SECOND (c)
+             && SYNTAX_COMEND_FIRST (FETCH_CHAR (from - 1))
+             && !char_quoted (from - 1))
+           {
+             /* we must record the comment style encountered so that
+                later, we can match only the proper comment begin
+                sequence of the same style */
+             code = Sendcomment;
+             comstyle = SYNTAX_COMMENT_STYLE (FETCH_CHAR (from - 1));
+             from--;
+           }
+
+         if (code == Sendcomment && !quoted)
+           {
+#if 0
+             if (code != SYNTAX (c))
+               /* For a two-char comment ender, we can assume
+                  it does end a comment.  So scan back in a simple way.  */
+               {
+                 if (from != stop) from--;
+                 while (1)
+                   {
+                     if (SYNTAX (c = FETCH_CHAR (from)) == Scomment
+                         && SYNTAX_COMMENT_STYLE (c) == comstyle)
+                       break;
+                     if (from == stop)
+                       {
+                         immediate_quit = 0;
+                         SET_PT (from);
+                         return Qnil;
+                       }
+                     from--;
+                     if (SYNTAX_COMSTART_SECOND (c)
+                         && SYNTAX_COMSTART_FIRST (FETCH_CHAR (from))
+                         && SYNTAX_COMMENT_STYLE (c) == comstyle
+                         && !char_quoted (from))
+                       break;
+                   }
+                 break;
+               }
+#endif /* 0 */
+
+             /* 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 parity = 0;
+               char my_stringend = 0;
+               int string_lossage = 0;
+               int comment_end = from;
+               int comstart_pos = 0;
+               int comstart_parity = 0;
+
+               /* At beginning of range to scan, we're outside of strings;
+                  that determines quote parity to the comment-end.  */
+               while (from != stop)
+                 {
+                   /* Move back and examine a character.  */
+                   from--;
+
+                   c = FETCH_CHAR (from);
+                   code = SYNTAX (c);
+
+                   /* If this char is the second of a 2-char comment sequence,
+                      back up and give the pair the appropriate syntax.  */
+                   if (from > stop && SYNTAX_COMEND_SECOND (c)
+                       && SYNTAX_COMEND_FIRST (FETCH_CHAR (from - 1)))
+                     {
+                       code = Sendcomment;
+                       from--;
+                     }
+                       
+                   else if (from > stop && SYNTAX_COMSTART_SECOND (c)
+                            && SYNTAX_COMSTART_FIRST (FETCH_CHAR (from - 1))
+                            && comstyle == SYNTAX_COMMENT_STYLE (c))
+                     {
+                       code = Scomment;
+                       from--;
+                     }
+
+                   /* Ignore escaped characters.  */
+                   if (char_quoted (from))
+                     continue;
+
+                   /* Track parity of quotes.  */
+                   if (code == Sstring)
+                     {
+                       parity ^= 1;
+                       if (my_stringend == 0)
+                         my_stringend = c;
+                       /* If we have two kinds of string delimiters.
+                          There's no way to grok this scanning backwards.  */
+                       else if (my_stringend != c)
+                         string_lossage = 1;
+                     }
+
+                   /* Record comment-starters according to that
+                      quote-parity to the comment-end.  */
+                   if (code == Scomment)
+                     {
+                       comstart_parity = parity;
+                       comstart_pos = from;
+                     }
+
+                   /* If we find another earlier comment-ender,
+                      any comment-starts earlier than that don't count
+                      (because they go with the earlier comment-ender).  */
+                   if (code == Sendcomment
+                       && SYNTAX_COMMENT_STYLE (FETCH_CHAR (from)) == comstyle)
+                     break;
+
+                   /* Assume a defun-start point is outside of strings.  */
+                   if (code == Sopen
+                       && (from == stop || FETCH_CHAR (from - 1) == '\n'))
+                     break;
+                 }
+
+               if (comstart_pos == 0)
+                 from = comment_end;
+               /* If the earliest comment starter
+                  is followed by uniform paired string quotes or none,
+                  we know it can't be inside a string
+                  since if it were then the comment ender would be inside one.
+                  So it does start a comment.  Skip back to it.  */
+               else if (comstart_parity == 0 && !string_lossage)
+                 from = comstart_pos;
+               else
+                 {
+                   /* We had two kinds of string delimiters mixed up
+                      together.  Decode this going forwards.
+                      Scan fwd from the previous comment ender
+                      to the one in question; this records where we
+                      last passed a comment starter.  */
+                   struct lisp_parse_state state;
+                   scan_sexps_forward (&state, find_defun_start (comment_end),
+                                       comment_end - 1, -10000, 0, Qnil, 0);
+                   if (state.incomment)
+                     from = state.comstart;
+                   else
+                     /* We can't grok this as a comment; scan it normally.  */
+                     from = comment_end;
+                 }
+             }
+           }
+         else if ((code != Swhitespace && code != Scomment) || quoted)
+           {
+             immediate_quit = 0;
+             SET_PT (from + 1);
+             return Qnil;
+           }
+       }
+
+      count1++;
+    }
+
+  SET_PT (from);
+  immediate_quit = 0;
+  return Qt;
+}
+\f
 int parse_sexp_ignore_comments;
 
 Lisp_Object
@@ -750,6 +1078,29 @@ scan_lists (from, count, depth, sexpflag)
            case Sendcomment:
              if (!parse_sexp_ignore_comments)
                break;
+#if 0
+             if (code != SYNTAX (c))
+               /* For a two-char comment ender, we can assume
+                  it does end a comment.  So scan back in a simple way.  */
+               {
+                 if (from != stop) from--;
+                 while (1)
+                   {
+                     if (SYNTAX (c = FETCH_CHAR (from)) == Scomment
+                         && SYNTAX_COMMENT_STYLE (c) == comstyle)
+                       break;
+                     if (from == stop) goto done;
+                     from--;
+                     if (SYNTAX_COMSTART_SECOND (c)
+                         && SYNTAX_COMSTART_FIRST (FETCH_CHAR (from))
+                         && SYNTAX_COMMENT_STYLE (c) == comstyle
+                         && !char_quoted (from))
+                       break;
+                   }
+                 break;
+               }
+#endif /* 0 */
+
              /* 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;
@@ -760,13 +1111,12 @@ scan_lists (from, count, depth, sexpflag)
                 which is I+2X quotes from the comment-end.
                 PARITY is current parity of quotes from the comment end.  */
              {
-               int ofrom[2];
                int parity = 0;
                char my_stringend = 0;
                int string_lossage = 0;
                int comment_end = from;
-
-               ofrom[0] = ofrom[1] = from;
+               int comstart_pos = 0;
+               int comstart_parity = 0;
 
                /* At beginning of range to scan, we're outside of strings;
                   that determines quote parity to the comment-end.  */
@@ -799,13 +1149,13 @@ scan_lists (from, count, depth, sexpflag)
                    if (char_quoted (from))
                      continue;
 
-                   /* Track parity of quotes between here and comment-end.  */
+                   /* Track parity of quotes.  */
                    if (code == Sstring)
                      {
                        parity ^= 1;
                        if (my_stringend == 0)
                          my_stringend = c;
-                       /* We have two kinds of string delimiters.
+                       /* If we have two kinds of string delimiters.
                           There's no way to grok this scanning backwards.  */
                        else if (my_stringend != c)
                          string_lossage = 1;
@@ -814,18 +1164,34 @@ scan_lists (from, count, depth, sexpflag)
                    /* Record comment-starters according to that
                       quote-parity to the comment-end.  */
                    if (code == Scomment)
-                     ofrom[parity] = from;
+                     {
+                       comstart_parity = parity;
+                       comstart_pos = from;
+                     }
 
-                   /* If we come to another comment-end,
-                      assume it's not inside a string.
-                      That determines the quote parity to the comment-end.
-                      Note that the comment style this character ends must
-                      match the style that we have begun */
+                   /* If we find another earlier comment-ender,
+                      any comment-starts earlier than that don't count
+                      (because they go with the earlier comment-ender).  */
                    if (code == Sendcomment
                        && SYNTAX_COMMENT_STYLE (FETCH_CHAR (from)) == comstyle)
                      break;
+
+                   /* Assume a defun-start point is outside of strings.  */
+                   if (code == Sopen
+                       && (from == stop || FETCH_CHAR (from - 1) == '\n'))
+                     break;
                  }
-               if (string_lossage)
+
+               if (comstart_pos == 0)
+                 from = comment_end;
+               /* If the earliest comment starter
+                  is followed by uniform paired string quotes or none,
+                  we know it can't be inside a string
+                  since if it were then the comment ender would be inside one.
+                  So it does start a comment.  Skip back to it.  */
+               else if (comstart_parity == 0 && !string_lossage)
+                 from = comstart_pos;
+               else
                  {
                    /* We had two kinds of string delimiters mixed up
                       together.  Decode this going forwards.
@@ -833,16 +1199,14 @@ scan_lists (from, count, depth, sexpflag)
                       to the one in question; this records where we
                       last passed a comment starter.  */
                    struct lisp_parse_state state;
-                   scan_sexps_forward (&state, from + 1, comment_end - 1,
-                                       -10000, 0, Qnil);
+                   scan_sexps_forward (&state, find_defun_start (comment_end),
+                                       comment_end - 1, -10000, 0, Qnil, 0);
                    if (state.incomment)
                      from = state.comstart;
                    else
                      /* We can't grok this as a comment; scan it normally.  */
                      from = comment_end;
                  }
-               else
-                 from = ofrom[parity];
              }
              break;
 
@@ -882,6 +1246,7 @@ scan_lists (from, count, depth, sexpflag)
   /* NOTREACHED */
 }
 
+static int
 char_quoted (pos)
      register int pos;
 {
@@ -962,13 +1327,17 @@ This includes chars with \"quote\" or \"prefix\" syntax (' or p).")
 /* Parse forward from FROM to END,
    assuming that FROM has state OLDSTATE (nil means FROM is start of function),
    and return a description of the state of the parse at END.
-   If STOPBEFORE is nonzero, stop at the start of an atom.  */
+   If STOPBEFORE is nonzero, stop at the start of an atom.
+   If COMMENTSTOP is nonzero, stop at the start of a comment.  */
 
-scan_sexps_forward (stateptr, from, end, targetdepth, stopbefore, oldstate)
+static void
+scan_sexps_forward (stateptr, from, end, targetdepth,
+                   stopbefore, oldstate, commentstop)
      struct lisp_parse_state *stateptr;
      register int from;
      int end, targetdepth, stopbefore;
      Lisp_Object oldstate;
+     int commentstop;
 {
   struct lisp_parse_state state;
 
@@ -1045,14 +1414,18 @@ scan_sexps_forward (stateptr, from, end, targetdepth, stopbefore, oldstate)
     {
       code = SYNTAX (FETCH_CHAR (from));
       from++;
-      if (from < end && SYNTAX_COMSTART_FIRST (FETCH_CHAR (from - 1))
-         && SYNTAX_COMSTART_SECOND (FETCH_CHAR (from)))
+      if (code == Scomment)
+       state.comstart = from-1;
+      
+      else if (from < end && SYNTAX_COMSTART_FIRST (FETCH_CHAR (from - 1))
+              && SYNTAX_COMSTART_SECOND (FETCH_CHAR (from)))
        {
          /* Record the comment style we have entered so that only
             the comment-end sequence of the same style actually
             terminates the comment section.  */
          code = Scomment;
          state.comstyle = SYNTAX_COMMENT_STYLE (FETCH_CHAR (from));
+         state.comstart = from-1;
          from++;
        }
 
@@ -1106,8 +1479,9 @@ scan_sexps_forward (stateptr, from, end, targetdepth, stopbefore, oldstate)
 
        case Scomment:
          state.incomment = 1;
-         state.comstart = from;
        startincomment:
+         if (commentstop)
+           goto done;
          while (1)
            {
              if (from == end) goto done;
@@ -1209,7 +1583,7 @@ scan_sexps_forward (stateptr, from, end, targetdepth, stopbefore, oldstate)
    for make-docfile to see.  We cannot put this in the real DEFUN
    due to limits in the Unix cpp.
 
-DEFUN ("parse-partial-sexp", Ffoo, Sfoo, 2, 5, 0,
+DEFUN ("parse-partial-sexp", Ffoo, Sfoo, 2, 6, 0,
   "Parse Lisp syntax starting at FROM until TO; return status of parse at TO.\n\
 Parsing stops at TO or when certain criteria are met;\n\
  point is set to where parsing stops.\n\
@@ -1231,14 +1605,15 @@ Fourth arg STOPBEFORE non-nil means stop when come to\n\
  any character that starts a sexp.\n\
 Fifth arg STATE is a seven-list like what this function returns.\n\
 It is used to initialize the state of the parse.  Its second and third
-elements are ignored.")
-  (from, to, targetdepth, stopbefore, state)
+elements are ignored.
+Sixth args COMMENTSTOP non-nil means stop at the start of a comment.")
+  (from, to, targetdepth, stopbefore, state, commentstop)
 */
 
-DEFUN ("parse-partial-sexp", Fparse_partial_sexp, Sparse_partial_sexp, 2, 5, 0,
+DEFUN ("parse-partial-sexp", Fparse_partial_sexp, Sparse_partial_sexp, 2, 6, 0,
   0 /* See immediately above */)
-  (from, to, targetdepth, stopbefore, oldstate)
-     Lisp_Object from, to, targetdepth, stopbefore, oldstate;
+  (from, to, targetdepth, stopbefore, oldstate, commentstop)
+     Lisp_Object from, to, targetdepth, stopbefore, oldstate, commentstop;
 {
   struct lisp_parse_state state;
   int target;
@@ -1253,7 +1628,8 @@ DEFUN ("parse-partial-sexp", Fparse_partial_sexp, Sparse_partial_sexp, 2, 5, 0,
 
   validate_region (&from, &to);
   scan_sexps_forward (&state, XINT (from), XINT (to),
-                     target, !NILP (stopbefore), oldstate);
+                     target, !NILP (stopbefore), oldstate,
+                     !NILP (commentstop));
 
   SET_PT (state.location);
   
@@ -1329,6 +1705,7 @@ syms_of_syntax ()
 
   defsubr (&Sforward_word);
 
+  defsubr (&Sforward_comment);
   defsubr (&Sscan_lists);
   defsubr (&Sscan_sexps);
   defsubr (&Sbackward_prefix_chars);