Replace utf8codepoint with Chris Wellons' utf8_decode

Code under a different license should be kept in a separate file. This
implemention is a single header file with ~65 lines, so it better fits this
requirement.

4 files changed, 75 insertions, 34 deletions

diff --git a/sxiv.h b/sxiv.h
index 29bdb2b..4e384c8 100644
--- a/sxiv.h
+++ b/sxiv.h
@@ -361,7 +361,6 @@ int r_opendir(r_dir_t*, const char*, bool);
 int r_closedir(r_dir_t*);
 char* r_readdir(r_dir_t*);
 int r_mkdir(char*);
-void* utf8codepoint(const void * __restrict__ str, long * __restrict__ out_codepoint);
 
 
 /* window.c */
diff --git a/utf8.h b/utf8.h
new file mode 100644
index 0000000..8c6a7a0
--- /dev/null
+++ b/utf8.h
@@ -0,0 +1,68 @@
+/* Branchless UTF-8 decoder
+ *
+ * This is free and unencumbered software released into the public domain.
+ */
+#ifndef UTF8_H
+#define UTF8_H
+
+#include <stdint.h>
+
+/* Decode the next character, C, from BUF, reporting errors in E.
+ *
+ * Since this is a branchless decoder, four bytes will be read from the
+ * buffer regardless of the actual length of the next character. This
+ * means the buffer _must_ have at least three bytes of zero padding
+ * following the end of the data stream.
+ *
+ * Errors are reported in E, which will be non-zero if the parsed
+ * character was somehow invalid: invalid byte sequence, non-canonical
+ * encoding, or a surrogate half.
+ *
+ * The function returns a pointer to the next character. When an error
+ * occurs, this pointer will be a guess that depends on the particular
+ * error, but it will always advance at least one byte.
+ */
+static void *
+utf8_decode(void *buf, uint32_t *c, int *e)
+{
+    static const char lengths[] = {
+        1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+        0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 3, 3, 4, 0
+    };
+    static const int masks[]  = {0x00, 0x7f, 0x1f, 0x0f, 0x07};
+    static const uint32_t mins[] = {4194304, 0, 128, 2048, 65536};
+    static const int shiftc[] = {0, 18, 12, 6, 0};
+    static const int shifte[] = {0, 6, 4, 2, 0};
+
+    unsigned char *s = buf;
+    int len = lengths[s[0] >> 3];
+
+    /* Compute the pointer to the next character early so that the next
+     * iteration can start working on the next character. Neither Clang
+     * nor GCC figure out this reordering on their own.
+     */
+    unsigned char *next = s + len + !len;
+
+    /* Assume a four-byte character and load four bytes. Unused bits are
+     * shifted out.
+     */
+    *c  = (uint32_t)(s[0] & masks[len]) << 18;
+    *c |= (uint32_t)(s[1] & 0x3f) << 12;
+    *c |= (uint32_t)(s[2] & 0x3f) <<  6;
+    *c |= (uint32_t)(s[3] & 0x3f) <<  0;
+    *c >>= shiftc[len];
+
+    /* Accumulate the various error conditions. */
+    *e  = (*c < mins[len]) << 6; // non-canonical encoding
+    *e |= ((*c >> 11) == 0x1b) << 7;  // surrogate half?
+    *e |= (*c > 0x10FFFF) << 8;  // out of range?
+    *e |= (s[1] & 0xc0) >> 2;
+    *e |= (s[2] & 0xc0) >> 4;
+    *e |= (s[3]       ) >> 6;
+    *e ^= 0x2a; // top two bits of each tail byte correct?
+    *e >>= shifte[len];
+
+    return next;
+}
+
+#endif
diff --git a/util.c b/util.c
index 973bc80..8f41d78 100644
--- a/util.c
+++ b/util.c
@@ -205,31 +205,3 @@ int r_mkdir(char *path)
 	return 0;
 }
 
-/* copied from sheredom's utf8.h (public domain) https://github.com/sheredom/utf8.h */
-
-void* utf8codepoint(const void* __restrict__ str, long* __restrict__ out_codepoint)
-{
-	 const char *s = (const char *)str;
-
-	 if (0xf0 == (0xf8 & s[0])) {
-		// 4 byte utf8 codepoint
-		*out_codepoint = ((0x07 & s[0]) << 18) | ((0x3f & s[1]) << 12) |
-				 ((0x3f & s[2]) << 6) | (0x3f & s[3]);
-		s += 4;
-	} else if (0xe0 == (0xf0 & s[0])) {
-		// 3 byte utf8 codepoint
-		*out_codepoint = ((0x0f & s[0]) << 12) | ((0x3f & s[1]) << 6) | (0x3f & s[2]);
-		s += 3;
-	} else if (0xc0 == (0xe0 & s[0])) {
-		// 2 byte utf8 codepoint
-		*out_codepoint = ((0x1f & s[0]) << 6) | (0x3f & s[1]);
-		s += 2;
-	} else {
-		// 1 byte utf8 codepoint otherwise
-		*out_codepoint = s[0];
-		s += 1;
-	}
-
-	return (void *)s;
-}
-
diff --git a/window.c b/window.c
index d1a96fb..b2d72d1 100644
--- a/window.c
+++ b/window.c
@@ -20,6 +20,7 @@
 #define _WINDOW_CONFIG
 #include "config.h"
 #include "icon/data.h"
+#include "utf8.h"
 
 #include <stdlib.h>
 #include <string.h>
@@ -131,8 +132,9 @@ void win_init(win_t *win)
 
 	win->bar.l.size = BAR_L_LEN;
 	win->bar.r.size = BAR_R_LEN;
-	win->bar.l.buf = emalloc(win->bar.l.size);
-	win->bar.r.buf = emalloc(win->bar.r.size);
+	/* 3 padding bytes needed by utf8_decode */
+	win->bar.l.buf = emalloc(win->bar.l.size + 3);
+	win->bar.r.buf = emalloc(win->bar.r.size + 3);
 	win->bar.h = options->hide_bar ? 0 : barheight;
 
 	INIT_ATOM_(WM_DELETE_WINDOW);
@@ -371,14 +373,14 @@ int win_textwidth(const win_env_t *e, const char *text, unsigned int len, bool w
 void win_draw_bar_text(win_t *win, XftDraw *d, XftColor *color, XftFont *font, int x, int y, char *text, int maxlen, int maximum_x)
 {
 	size_t len = 0;
-	int xshift = 0, newshift;
-	long codep;
+	int err, xshift = 0, newshift;
+	uint32_t codep;
 	char *p, *nextp;
 	FcCharSet* fccharset;
 	XftFont* fallback = NULL;
 
 	for (p = text; *p && (len < maxlen); p = nextp, len++) {
-		nextp = utf8codepoint(p, &codep);
+		nextp = utf8_decode(p, &codep, &err);
 		if (!XftCharExists(win->env.dpy, font, codep)) {
 			fccharset = FcCharSetCreate();
 			FcCharSetAddChar(fccharset, codep);