diff --git a/‎ChangeLog‎
Lines changed: 12 additions & 0 deletions b/‎ChangeLog‎
Lines changed: 12 additions & 0 deletions
diff --git a/‎src/term.c‎
Lines changed: 4 additions & 2 deletions b/‎src/term.c‎
Lines changed: 4 additions & 2 deletions
diff --git a/‎term/estimate.trm‎
Lines changed: 39 additions & 3 deletions b/‎term/estimate.trm‎
Lines changed: 39 additions & 3 deletions
@@ -1,3 +1,15 @@
+2010-05-20  Ethan A Merritt  <merritt@u.washington.edu>
+
+	* term/estimate.trm (ENHest_writec strlen_utf8)
+	src/term.c (estimate_strlen):
+
+	Teach estimate_strlen() to handle UTF-8 encoded strings.  The estimate
+	is imperfect, but then again the estimate is already imperfect for any
+	proportional font.  To truly do this more accurately would require
+	customized implementations for each terminal type.  Any unicode code
+	point below 0x3000 is treated as requiring one character width; code
+	points above 0x3000 are treated as requiring two character widths.
+
 2010-05-16  Ethan A Merritt  <merritt@u.washington.edu>
 
 	* src/hidden3d.c (draw_edge): Add support for coloring lines in
 
@@ -1,5 +1,5 @@
 #ifndef lint
-static char *RCSid() { return RCSid("$Id: term.c,v 1.202 2010/03/14 22:44:38 sfeam Exp $"); }
+static char *RCSid() { return RCSid("$Id: term.c,v 1.203 2010/05/05 18:03:50 sfeam Exp $"); }
 #endif
 
 /* GNUPLOT - term.c */
@@ -2789,7 +2789,9 @@ int len;
 	FPRINTF((stderr,"Estimating length %d height %g for enhanced text string \"%s\"\n",
 		len, (double)(term->ymax)/10., text));
 	term = tsave;
-    } else
+    } else if (encoding == S_ENC_UTF8)
+	len = strlen_utf8(text);
+    else
 #endif
 	len = strlen(text);
 
 
@@ -1,5 +1,5 @@
 /* Hello, Emacs, this is -*-C-*-
- * $Id: estimate.trm,v 1.7 2008/06/02 19:49:50 sfeam Exp $
+ * $Id: estimate.trm,v 1.8 2009/03/26 00:49:20 sfeam Exp $
  *
  */
 
@@ -44,6 +44,9 @@ static TBOOLEAN ENHest_widthflag = TRUE;
 #define ENHest_font ""
 static double ENHest_base;
 
+/* Internal routines for UTF-8 support */
+static size_t strlen_utf8 __PROTO((const char *s));
+
 TERM_PUBLIC void
 ENHest_OPEN(
     char *fontname,
@@ -123,7 +126,10 @@ ENHest_put_text(unsigned int x, unsigned int y, const char *str)
 
     /* If no enhanced text processing is needed, strlen() is sufficient */
     if (ignore_enhanced_text || !strpbrk(str, "{}^_@&~\n")) {
-	term->xmax = strlen(str);
+	if (encoding == S_ENC_UTF8)
+	    term->xmax = strlen_utf8(str);
+	else
+	    term->xmax = strlen(str);
 	term->ymax = 10;
 	return;
     }
@@ -158,7 +164,37 @@ ENHest_writec(int c)
 	ENHest_x = 0;
     }
 
-    ENHest_fragment_width += ENHest_fontsize;
+    if (encoding == S_ENC_UTF8) {
+	/* Skip all but the first byte of UTF-8 multi-byte characters. */
+	if ((c & 0xc0) != 0x80) {
+	    ENHest_fragment_width += ENHest_fontsize;
+	    /* [most] characters above 0x3000 are square CJK glyphs, */
+	    /* which are wider than western characters.              */
+	    if ((unsigned int)c >= 0xec)
+		ENHest_fragment_width += ENHest_fontsize;
+	}
+    } else
+	ENHest_fragment_width += ENHest_fontsize;
+}
+
+/*
+ * This routine accounts for multi-byte characters in UTF-8.
+ * NB:  It does not return the _number_ of characters in the string, but
+ * rather their approximate _width_ in units of typical character width.
+ * As with the ENHest_writec() routine, it approximates the width of characters
+ * above unicode 0x3000 as being twice that of western alphabetic characters.
+ */
+size_t strlen_utf8(const char *s) {
+    int i = 0, j = 0;
+    while (s[i]) {
+       if ((s[i] & 0xc0) != 0x80) {
+           j++;
+	   if ((unsigned char)(s[i]) >= 0xe3)
+		j++;
+       }
+       i++;
+    }
+    return j;
 }