ncurses 6.2 - patch 20200404
[ncurses.git] / ncurses / tinfo / comp_expand.c
index eb552fadabda0d524ad94d4c949b3ff1ed8c515e..02e38e6e3c7a440a3715b8d0010f0209dd852659 100644 (file)
@@ -1,5 +1,6 @@
 /****************************************************************************
- * Copyright (c) 1998 Free Software Foundation, Inc.                        *
+ * Copyright 2020 Thomas E. Dickey                                          *
+ * Copyright 1998-2016,2017 Free Software Foundation, Inc.                  *
  *                                                                          *
  * Permission is hereby granted, free of charge, to any person obtaining a  *
  * copy of this software and associated documentation files (the            *
@@ -27,7 +28,7 @@
  ****************************************************************************/
 
 /****************************************************************************
- *  Author: Thomas E. Dickey <dickey@clark.net> 1998                        *
+ *  Author: Thomas E. Dickey                    1998                        *
  ****************************************************************************/
 
 #include <curses.priv.h>
 #include <ctype.h>
 #include <tic.h>
 
-MODULE_ID("$Id: comp_expand.c,v 1.11 1999/03/07 00:51:07 tom Exp $")
+MODULE_ID("$Id: comp_expand.c,v 1.32 2020/02/02 23:34:34 tom Exp $")
 
-static int trailing_spaces(const char *src)
+#if 0
+#define DEBUG_THIS(p) DEBUG(9, p)
+#else
+#define DEBUG_THIS(p)          /* nothing */
+#endif
+
+static int
+trailing_spaces(const char *src)
 {
-       while (*src == ' ')
-               src++;
-       return *src == 0;
+    while (*src == ' ')
+       src++;
+    return *src == 0;
 }
 
 /* this deals with differences over whether 0x7f and 0x80..0x9f are controls */
-#define CHAR_OF(s) (*(unsigned const char *)(s))
-#define REALCTL(s) (CHAR_OF(s) < 127 && iscntrl(CHAR_OF(s)))
-#define REALPRINT(s) (CHAR_OF(s) < 127 && isprint(CHAR_OF(s)))
+#define REALPRINT(s) (UChar(*(s)) < 127 && isprint(UChar(*(s))))
+
+#define P_LIMIT(p)   (length - (size_t)(p))
 
-char *_nc_tic_expand(const char *srcp, bool tic_format, int numbers)
+NCURSES_EXPORT(char *)
+_nc_tic_expand(const char *srcp, bool tic_format, int numbers)
 {
-static char *  buffer;
-static size_t  length;
+    static char *buffer;
+    static size_t length;
 
-int            bufp;
-const char     *ptr, *str = VALID_STRING(srcp) ? srcp : "";
-bool           islong = (strlen(str) > 3);
-size_t         need = (2 + strlen(str)) * 4;
-int            ch;
+    int bufp;
+    const char *str = VALID_STRING(srcp) ? srcp : "\0\0";
+    size_t need = (2 + strlen(str)) * 4;
+    int ch;
+    int octals = 0;
+    struct {
+       int ch;
+       int offset;
+    } fixups[MAX_TC_FIXUPS];
 
-       if (buffer == 0 || need > length) {
-               if ((buffer = typeRealloc(char, length = need, buffer)) == 0)
-                       return 0;
+    if (srcp == 0) {
+#if NO_LEAKS
+       if (buffer != 0) {
+           FreeAndNull(buffer);
+           length = 0;
        }
-
-       bufp = 0;
-       ptr = str;
-       while ((ch = (*str & 0xff)) != 0) {
-               if (ch == '%' && REALPRINT(str+1)) {
-                       buffer[bufp++] = *str++;
-                       /*
-                        * Though the character literals are more compact, most
-                        * terminal descriptions use numbers and are not easy
-                        * to read in character-literal form.
-                        */
-                       switch (numbers) {
-                       case -1:
-                           if (str[0] == S_QUOTE
-                            && str[1] != '\\'
-                            && REALPRINT(str+1)
-                            && str[2] == S_QUOTE) {
-                               sprintf(buffer+bufp, "{%d}", str[1]);
-                               bufp += strlen(buffer+bufp);
-                               str += 2;
-                           } else {
-                               buffer[bufp++] = *str;
-                           }
-                           break;
-                       /*
-                        * If we have a "%{number}", try to translate it into
-                        * a "%'char'" form, since that will run a little faster
-                        * when we're interpreting it.  Also, having one form
-                        * for the constant makes it simpler to compare terminal
-                        * descriptions.
-                        */
-                       case 1:
-                           if (str[0] == L_BRACE
-                            && isdigit(str[1])) {
-                               char *dst = 0;
-                               long value = strtol(str+1, &dst, 0);
-                               if (dst != 0
-                                && *dst == R_BRACE
-                                && value < 127
-                                && value != '\\'       /* FIXME */
-                                && isprint((int)value)) {
-                                       ch = (int)value;
-                                       buffer[bufp++] = S_QUOTE;
-                                       if (ch == '\\'
-                                        || ch == S_QUOTE)
-                                               buffer[bufp++] = '\\';
-                                       buffer[bufp++] = ch;
-                                       buffer[bufp++] = S_QUOTE;
-                                       str = dst;
-                               } else {
-                                       buffer[bufp++] = *str;
-                               }
-                           } else {
-                                   buffer[bufp++] = *str;
-                           }
-                           break;
-                       default:
-                           buffer[bufp++] = *str;
-                           break;
-                       }
-               }
-               else if (ch == 128) {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = '0';
-               }
-               else if (ch == '\033') {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = 'E';
-               }
-               else if (ch == '\\' && tic_format && (str == srcp || str[-1] != '^')) {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = '\\';
-               }
-               else if (ch == ' ' && tic_format && (str == srcp || trailing_spaces(str))) {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = 's';
-               }
-               else if ((ch == ',' || ch == ':' || ch == '^') && tic_format) {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = ch;
-               }
-               else if (REALPRINT(str) && (ch != ',' && ch != ':' && !(ch == '!' && !tic_format) && ch != '^'))
-                       buffer[bufp++] = ch;
-#if 0          /* FIXME: this would be more readable (in fact the whole 'islong' logic should be removed) */
-               else if (ch == '\b') {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = 'b';
-               }
-               else if (ch == '\f') {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = 'f';
-               }
-               else if (ch == '\t' && islong) {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = 't';
-               }
 #endif
-               else if (ch == '\r' && (islong || (strlen(srcp) > 2 && str[1] == '\0'))) {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = 'r';
+       return 0;
+    }
+    if (buffer == 0 || need > length) {
+       if ((buffer = typeRealloc(char, length = need, buffer)) == 0)
+             return 0;
+    }
+
+    DEBUG_THIS(("_nc_tic_expand %s", _nc_visbuf(srcp)));
+    bufp = 0;
+    while ((ch = UChar(*str)) != 0) {
+       if (ch == '%' && REALPRINT(str + 1)) {
+           buffer[bufp++] = *str++;
+           /*
+            * Though the character literals are more compact, most
+            * terminal descriptions use numbers and are not easy
+            * to read in character-literal form.
+            */
+           switch (numbers) {
+           case -1:
+               if (str[0] == S_QUOTE
+                   && str[1] != '\\'
+                   && REALPRINT(str + 1)
+                   && str[2] == S_QUOTE) {
+                   _nc_SPRINTF(buffer + bufp, _nc_SLIMIT(P_LIMIT(bufp))
+                               "{%d}", str[1]);
+                   bufp += (int) strlen(buffer + bufp);
+                   str += 2;
+               } else {
+                   buffer[bufp++] = *str;
                }
-               else if (ch == '\n' && islong) {
-                       buffer[bufp++] = '\\';
-                       buffer[bufp++] = 'n';
+               break;
+               /*
+                * If we have a "%{number}", try to translate it into
+                * a "%'char'" form, since that will run a little faster
+                * when we're interpreting it.  Also, having one form
+                * for the constant makes it simpler to compare terminal
+                * descriptions.
+                */
+           case 1:
+               if (str[0] == L_BRACE
+                   && isdigit(UChar(str[1]))) {
+                   char *dst = 0;
+                   long value = strtol(str + 1, &dst, 0);
+                   if (dst != 0
+                       && *dst == R_BRACE
+                       && value < 127
+                       && value != '\\'        /* FIXME */
+                       && isprint((int) value)) {
+                       ch = (int) value;
+                       buffer[bufp++] = S_QUOTE;
+                       if (ch == '\\'
+                           || ch == S_QUOTE)
+                           buffer[bufp++] = '\\';
+                       buffer[bufp++] = (char) ch;
+                       buffer[bufp++] = S_QUOTE;
+                       str = dst;
+                   } else {
+                       buffer[bufp++] = *str;
+                   }
+               } else {
+                   buffer[bufp++] = *str;
                }
+               break;
+           default:
+               if (*str == ',')        /* minitel1 uses this */
+                   buffer[bufp++] = '\\';
+               buffer[bufp++] = *str;
+               break;
+           }
+       } else if (ch == 128) {
+           buffer[bufp++] = '\\';
+           buffer[bufp++] = '0';
+       } else if (ch == '\033') {
+           buffer[bufp++] = '\\';
+           buffer[bufp++] = 'E';
+       } else if (ch == '\\' && tic_format && (str == srcp || str[-1] != '^')) {
+           buffer[bufp++] = '\\';
+           buffer[bufp++] = '\\';
+       } else if (ch == ' ' && tic_format && (str == srcp ||
+                                              trailing_spaces(str))) {
+           buffer[bufp++] = '\\';
+           buffer[bufp++] = 's';
+       } else if ((ch == ',' || ch == ':' || ch == '^') && tic_format) {
+           buffer[bufp++] = '\\';
+           buffer[bufp++] = (char) ch;
+       } else if (REALPRINT(str)
+                  && (ch != ','
+                      && ch != ':'
+                      && !(ch == '!' && !tic_format)
+                      && ch != '^'))
+           buffer[bufp++] = (char) ch;
+       else if (ch == '\r') {
+           buffer[bufp++] = '\\';
+           buffer[bufp++] = 'r';
+       } else if (ch == '\n') {
+           buffer[bufp++] = '\\';
+           buffer[bufp++] = 'n';
+       }
 #define UnCtl(c) ((c) + '@')
-               else if (REALCTL(str) && ch != '\\' && (!islong || isdigit(str[1])))
-               {
-                       (void) sprintf(&buffer[bufp], "^%c", UnCtl(ch));
-                       bufp += 2;
-               }
-               else
-               {
-                       (void) sprintf(&buffer[bufp], "\\%03o", ch);
-                       bufp += 4;
-               }
-
-               str++;
+       else if (UChar(ch) < 32
+                && isdigit(UChar(str[1]))) {
+           _nc_SPRINTF(&buffer[bufp], _nc_SLIMIT(P_LIMIT(bufp))
+                       "^%c", UnCtl(ch));
+           bufp += 2;
+       } else {
+           _nc_SPRINTF(&buffer[bufp], _nc_SLIMIT(P_LIMIT(bufp))
+                       "\\%03o", ch);
+           if ((octals < MAX_TC_FIXUPS) &&
+               ((tic_format && (ch == 127)) || ch < 32)) {
+               fixups[octals].ch = UChar(ch);
+               fixups[octals].offset = bufp;
+               ++octals;
+           }
+           bufp += 4;
        }
 
-       buffer[bufp] = '\0';
-       return(buffer);
+       str++;
+    }
+
+    buffer[bufp] = '\0';
+
+    /*
+     * If most of a short string is ASCII control characters, reformat the
+     * string to show those in up-arrow format.  For longer strings, it's
+     * more likely that the characters are just binary coding.
+     *
+     * If we're formatting termcap, just use the shorter format (up-arrows).
+     */
+    if (octals != 0 && (!tic_format || (bufp - (4 * octals)) < MIN_TC_FIXUPS)) {
+       while (--octals >= 0) {
+           char *p = buffer + fixups[octals].offset;
+           *p++ = '^';
+           *p++ = (char) ((fixups[octals].ch == 127)
+                          ? '?'
+                          : (fixups[octals].ch + (int) '@'));
+           while ((p[0] = p[2]) != 0) {
+               ++p;
+           }
+       }
+    }
+    DEBUG_THIS(("... %s", _nc_visbuf(buffer)));
+    return (buffer);
 }