convert to 4-byte UTF-8 and 32-bit Rune

http://codereview.appspot.com/116075

2 files changed, 65 insertions, 24 deletions

diff --git a/src/lib9/fmt/dofmt.c b/src/lib9/fmt/dofmt.c
index 214b71db..353c76e3 100644
--- a/src/lib9/fmt/dofmt.c
+++ b/src/lib9/fmt/dofmt.c
@@ -605,12 +605,13 @@ __flagfmt(Fmt *f)
 int
 __badfmt(Fmt *f)
 {
-	char x[3];
+	char x[2+UTFmax];
+	int n;
 
 	x[0] = '%';
-	x[1] = f->r;
-	x[2] = '%';
-	f->prec = 3;
-	__fmtcpy(f, (const void*)x, 3, 3);
+	n = 1 + runetochar(x+1, &f->r);
+	x[n++] = '%';
+	f->prec = n;
+	__fmtcpy(f, (const void*)x, n, n);
 	return 0;
 }
diff --git a/src/lib9/utf/rune.c b/src/lib9/utf/rune.c
index 3d6831b0..f5944806 100644
--- a/src/lib9/utf/rune.c
+++ b/src/lib9/utf/rune.c
@@ -23,16 +23,19 @@ enum
 	Bit2	= 5,
 	Bit3	= 4,
 	Bit4	= 3,
+	Bit5	= 2,
 
 	T1	= ((1<<(Bit1+1))-1) ^ 0xFF,	/* 0000 0000 */
 	Tx	= ((1<<(Bitx+1))-1) ^ 0xFF,	/* 1000 0000 */
 	T2	= ((1<<(Bit2+1))-1) ^ 0xFF,	/* 1100 0000 */
 	T3	= ((1<<(Bit3+1))-1) ^ 0xFF,	/* 1110 0000 */
 	T4	= ((1<<(Bit4+1))-1) ^ 0xFF,	/* 1111 0000 */
+	T5	= ((1<<(Bit5+1))-1) ^ 0xFF,	/* 1111 1000 */
 
-	Rune1	= (1<<(Bit1+0*Bitx))-1,		/* 0000 0000 0111 1111 */
-	Rune2	= (1<<(Bit2+1*Bitx))-1,		/* 0000 0111 1111 1111 */
-	Rune3	= (1<<(Bit3+2*Bitx))-1,		/* 1111 1111 1111 1111 */
+	Rune1	= (1<<(Bit1+0*Bitx))-1,		/* 0000 0000 0000 0000 0111 1111 */
+	Rune2	= (1<<(Bit2+1*Bitx))-1,		/* 0000 0000 0000 0111 1111 1111 */
+	Rune3	= (1<<(Bit3+2*Bitx))-1,		/* 0000 0000 1111 1111 1111 1111 */
+	Rune4	= (1<<(Bit4+3*Bitx))-1,		/* 0011 1111 1111 1111 1111 1111 */
 
 	Maskx	= (1<<Bitx)-1,			/* 0011 1111 */
 	Testx	= Maskx ^ 0xFF,			/* 1100 0000 */
@@ -43,7 +46,7 @@ enum
 int
 chartorune(Rune *rune, char *str)
 {
-	int c, c1, c2;
+	int c, c1, c2, c3;
 	long l;
 
 	/*
@@ -89,6 +92,25 @@ chartorune(Rune *rune, char *str)
 	}
 
 	/*
+	 * four character sequence
+	 *	10000-10FFFF => T4 Tx Tx Tx
+	 */
+	if(UTFmax >= 4) {
+		c3 = *(uchar*)(str+3) ^ Tx;
+		if(c3 & Testx)
+			goto bad;
+		if(c < T5) {
+			l = ((((((c << Bitx) | c1) << Bitx) | c2) << Bitx) | c3) & Rune4;
+			if(l <= Rune3)
+				goto bad;
+			if(l > Runemax)
+				goto bad;
+			*rune = l;
+			return 4;
+		}
+	}
+
+	/*
 	 * bad decoding
 	 */
 bad:
@@ -113,7 +135,7 @@ runetochar(char *str, Rune *rune)
 
 	/*
 	 * two character sequence
-	 *	0080-07FF => T2 Tx
+	 *	00080-007FF => T2 Tx
 	 */
 	if(c <= Rune2) {
 		str[0] = T2 | (c >> 1*Bitx);
@@ -123,12 +145,26 @@ runetochar(char *str, Rune *rune)
 
 	/*
 	 * three character sequence
-	 *	0800-FFFF => T3 Tx Tx
+	 *	00800-0FFFF => T3 Tx Tx
 	 */
-	str[0] = T3 |  (c >> 2*Bitx);
-	str[1] = Tx | ((c >> 1*Bitx) & Maskx);
-	str[2] = Tx |  (c & Maskx);
-	return 3;
+	if(c > Runemax)
+		c = Runeerror;
+	if(c <= Rune3) {
+		str[0] = T3 |  (c >> 2*Bitx);
+		str[1] = Tx | ((c >> 1*Bitx) & Maskx);
+		str[2] = Tx |  (c & Maskx);
+		return 3;
+	}
+	
+	/*
+	 * four character sequence
+	 *	010000-1FFFFF => T4 Tx Tx Tx
+	 */
+	str[0] = T4 |  (c >> 3*Bitx);
+	str[1] = Tx | ((c >> 2*Bitx) & Maskx);
+	str[2] = Tx | ((c >> 1*Bitx) & Maskx);
+	str[3] = Tx |  (c & Maskx);
+	return 4;
 }
 
 int
@@ -155,7 +191,10 @@ runenlen(Rune *r, int nrune)
 		if(c <= Rune2)
 			nb += 2;
 		else
+		if(c <= Rune3 || c > Runemax)
 			nb += 3;
+		else
+			nb += 4;
 	}
 	return nb;
 }
@@ -165,13 +204,14 @@ fullrune(char *str, int n)
 {
 	int c;
 
-	if(n > 0) {
-		c = *(uchar*)str;
-		if(c < Tx)
-			return 1;
-		if(n > 1)
-			if(c < T3 || n > 2)
-				return 1;
-	}
-	return 0;
+	if(n <= 0)
+		return 0;
+	c = *(uchar*)str;
+	if(c < Tx)
+		return 1;
+	if(c < T3)
+		return n >= 2;
+	if(UTFmax == 3 || c < T4)
+		return n >= 3;
+	return n >= 4;
 }