Added fast utf8 string detection routine fl_is_valid_utf8(). Solves the drawing problem in #2080 related fl_expand_text() use. The particularity of this routine is that the complete scan is cached so that it executes 0 to 1 time in the fl_expand_text() loop. Now chinese example in str 2080 seems to works perfectly here. Still, incorrect windows title in Win32 need to be adressed.

git-svn-id: file:///fltk/svn/fltk/branches/branch-1.3@6497 ea41ed52-d2ee-0310-a9c1-e6b18d33e121

1 files changed, 54 insertions, 2 deletions

diff --git a/src/fl_draw.cxx b/src/fl_draw.cxx
index e4a617b85..3815ab657 100644
--- a/src/fl_draw.cxx
+++ b/src/fl_draw.cxx
@@ -55,6 +55,56 @@ static char* underline_at;
 // Sets n to the number of characters put into the buffer.
 // Sets width to the width of the string in the current font.
 
+#define C_IN(c,a,b) ((c)>=(a) && (c)<=(b)) 
+#define C_UTF8(c)   C_IN(c,0x80,0xBF)
+
+/** fast utf8 string detection routine. \reval 0 if not utf8, 1 otherwise */
+int fl_is_valid_utf8(int& init_scan, int& scan_ret, const char* s) {
+    if (init_scan) return scan_ret; // scan only once the string
+    init_scan=1;
+    if ( !s || !(*s) ) return 0;
+
+    register const unsigned char* p=(const unsigned char*)s;
+    while (*p) {
+	if ( p[0]==0x09 || p[0]==0x0d || p[0]==0x0a || (p[0]>0x1f && p[0]<0x80) ) {
+	    p++;
+	    continue; // Ascii 
+	}
+	if ( C_IN(p[0], 0xc2, 0xdf) && C_UTF8(p[1]) ) {
+	    p+=2;
+	    continue; // non-overlong 2-byte
+	}
+	if ( p[0]==0xe0 && C_IN(p[1], 0xa0, 0xbf) && C_UTF8(p[2]) ) {
+	    p+=3;
+	    continue; //  excluding overlongs
+	}
+	if (p[0]==0xed && C_IN(p[1], 0x80, 0x9f) && C_UTF8(p[2]) ) {
+	    p+=3;
+	    continue; //  excluding surrogates
+	}
+	if (p[0]!=0xed && C_IN(p[0], 0xe1, 0xef) && C_UTF8(p[1]) && C_UTF8(p[2]) ) {
+	    p+=3;
+	    continue; // straight 3-byte
+	}
+	if (p[0]==0xf0 && C_IN(p[1], 0x90, 0xbf)   && C_UTF8(p[2]) && C_UTF8(p[3]) ) {
+	    p+=4;
+	    continue; // planes 1-3
+	}
+	if (C_IN(p[0], 0xf1, 0xf3) && C_UTF8(p[1]) && C_UTF8(p[2]) && C_UTF8(p[3]) ) {
+	    p+=4;
+	    continue; // planes 4-15
+	}
+	if (p[0]==0xf4 && C_IN(p[1], 0x80, 0x8f)   && C_UTF8(p[2]) && C_UTF8(p[3]) ) {
+	    p+=4;
+	    continue; // planes 16
+	}
+	scan_ret=0;
+	return scan_ret; // not utf8
+    }
+    scan_ret=1;
+    return scan_ret;
+}
+
 const char*
 fl_expand_text(const char* from, char* buf, int maxbuf, double maxw, int& n, 
 	double &width, int wrap, int draw_symbols) {
@@ -65,6 +115,8 @@ fl_expand_text(const char* from, char* buf, int maxbuf, double maxw, int& n,
   const char* word_start = from;
   double w = 0;
 
+  int init_scan=0, scan_ret;
+  
   const char* p = from;
   for (;; p++) {
 
@@ -99,9 +151,9 @@ fl_expand_text(const char* from, char* buf, int maxbuf, double maxw, int& n,
       *o++ = '^';
       *o++ = c ^ 0x40;
 #ifdef __APPLE__
-    } else if (c == 0xCA) { // non-breaking space in MacRoman
+    } else if (c == 0xCA && !fl_is_valid_utf8(init_scan, scan_ret,from) ) { // non-breaking space in MacRoman
 #else
-    } else if (c == 0xA0) { // non-breaking space in ISO 8859
+    } else if (c == 0xA0 && !fl_is_valid_utf8(init_scan, scan_ret,from) ) { // non-breaking space in ISO 8859
 #endif
       *o++ = ' ';
     } else if (c == '@' && draw_symbols) { // Symbol???