Commit 3c7d6f3

nsxiv3c7d6f3

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.

Author: Bert Münnich
Date: 2017-12-07 14:57 +0100
Committer: Bert Münnich · 2017-12-07 21:19 +0100
Parent: 69b2d3c
Commit: 3c7d6f3528432a433cd368605e4dc6f0fecaa898

+75 −34

4 files changed

Browse files Patch

−Context 3+

M sxiv.h +0 −1
A utf8.h +68 −0
M util.c +0 −28
M window.c +7 −5

M sxiv.h View

A utf8.h View

M util.c View

M window.c View

		@@ -361,7 +361,6 @@ int r_opendir(r_dir_t, const char, bool);
361	361		int r_closedir(r_dir_t*);
362	362		char* r_readdir(r_dir_t*);
363	363		int r_mkdir(char*);
364		−	void* utf8codepoint(const void * __restrict__ str, long * __restrict__ out_codepoint);
365	364
366	365
367	366		/* window.c */

	@@ -0,0 +1,68 @@
1	+	/* Branchless UTF-8 decoder
2	+	*
3	+	* This is free and unencumbered software released into the public domain.
4	+	*/
5	+	#ifndef UTF8_H
6	+	#define UTF8_H
7	+
8	+	#include <stdint.h>
9	+
10	+	/* Decode the next character, C, from BUF, reporting errors in E.
11	+	*
12	+	* Since this is a branchless decoder, four bytes will be read from the
13	+	* buffer regardless of the actual length of the next character. This
14	+	* means the buffer _must_ have at least three bytes of zero padding
15	+	* following the end of the data stream.
16	+	*
17	+	* Errors are reported in E, which will be non-zero if the parsed
18	+	* character was somehow invalid: invalid byte sequence, non-canonical
19	+	* encoding, or a surrogate half.
20	+	*
21	+	* The function returns a pointer to the next character. When an error
22	+	* occurs, this pointer will be a guess that depends on the particular
23	+	* error, but it will always advance at least one byte.
24	+	*/
25	+	static void *
26	+	utf8_decode(void buf, uint32_t c, int *e)
27	+	{
28	+	static const char lengths[] = {
29	+	1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
30	+	0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 3, 3, 4, 0
31	+	};
32	+	static const int masks[] = {0x00, 0x7f, 0x1f, 0x0f, 0x07};
33	+	static const uint32_t mins[] = {4194304, 0, 128, 2048, 65536};
34	+	static const int shiftc[] = {0, 18, 12, 6, 0};
35	+	static const int shifte[] = {0, 6, 4, 2, 0};
36	+
37	+	unsigned char *s = buf;
38	+	int len = lengths[s[0] >> 3];
39	+
40	+	/* Compute the pointer to the next character early so that the next
41	+	* iteration can start working on the next character. Neither Clang
42	+	* nor GCC figure out this reordering on their own.
43	+	*/
44	+	unsigned char *next = s + len + !len;
45	+
46	+	/* Assume a four-byte character and load four bytes. Unused bits are
47	+	* shifted out.
48	+	*/
49	+	*c = (uint32_t)(s[0] & masks[len]) << 18;
50	+	*c \|= (uint32_t)(s[1] & 0x3f) << 12;
51	+	*c \|= (uint32_t)(s[2] & 0x3f) << 6;
52	+	*c \|= (uint32_t)(s[3] & 0x3f) << 0;
53	+	*c >>= shiftc[len];
54	+
55	+	/* Accumulate the various error conditions. */
56	+	e = (c < mins[len]) << 6; // non-canonical encoding
57	+	e \|= ((c >> 11) == 0x1b) << 7; // surrogate half?
58	+	e \|= (c > 0x10FFFF) << 8; // out of range?
59	+	*e \|= (s[1] & 0xc0) >> 2;
60	+	*e \|= (s[2] & 0xc0) >> 4;
61	+	*e \|= (s[3] ) >> 6;
62	+	*e ^= 0x2a; // top two bits of each tail byte correct?
63	+	*e >>= shifte[len];
64	+
65	+	return next;
66	+	}
67	+
68	+	#endif

		@@ -205,31 +205,3 @@ int r_mkdir(char *path)
205	205		return 0;
206	206		}
207	207
208		−	/* copied from sheredom's utf8.h (public domain) https://github.com/sheredom/utf8.h */
209		−
210		−	void* utf8codepoint(const void* __restrict__ str, long* __restrict__ out_codepoint)
211		−	{
212		−	const char s = (const char )str;
213		−
214		−	if (0xf0 == (0xf8 & s[0])) {
215		−	// 4 byte utf8 codepoint
216		−	*out_codepoint = ((0x07 & s[0]) << 18) \| ((0x3f & s[1]) << 12) \|
217		−	((0x3f & s[2]) << 6) \| (0x3f & s[3]);
218		−	s += 4;
219		−	} else if (0xe0 == (0xf0 & s[0])) {
220		−	// 3 byte utf8 codepoint
221		−	*out_codepoint = ((0x0f & s[0]) << 12) \| ((0x3f & s[1]) << 6) \| (0x3f & s[2]);
222		−	s += 3;
223		−	} else if (0xc0 == (0xe0 & s[0])) {
224		−	// 2 byte utf8 codepoint
225		−	*out_codepoint = ((0x1f & s[0]) << 6) \| (0x3f & s[1]);
226		−	s += 2;
227		−	} else {
228		−	// 1 byte utf8 codepoint otherwise
229		−	*out_codepoint = s[0];
230		−	s += 1;
231		−	}
232		−
233		−	return (void *)s;
234		−	}
235		−

		@@ -20,6 +20,7 @@
20	20		#define _WINDOW_CONFIG
21	21		#include "config.h"
22	22		#include "icon/data.h"
	23	+	#include "utf8.h"
23	24
24	25		#include <stdlib.h>
25	26		#include <string.h>
		@@ -131,8 +132,9 @@ void win_init(win_t *win)
131	132
132	133		win->bar.l.size = BAR_L_LEN;
133	134		win->bar.r.size = BAR_R_LEN;
134		−	win->bar.l.buf = emalloc(win->bar.l.size);
135		−	win->bar.r.buf = emalloc(win->bar.r.size);
	135	+	/* 3 padding bytes needed by utf8_decode */
	136	+	win->bar.l.buf = emalloc(win->bar.l.size + 3);
	137	+	win->bar.r.buf = emalloc(win->bar.r.size + 3);
136	138		win->bar.h = options->hide_bar ? 0 : barheight;
137	139
138	140		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
371	373		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)
372	374		{
373	375		size_t len = 0;
374		−	int xshift = 0, newshift;
375		−	long codep;
	376	+	int err, xshift = 0, newshift;
	377	+	uint32_t codep;
376	378		char p, nextp;
377	379		FcCharSet* fccharset;
378	380		XftFont* fallback = NULL;
379	381
380	382		for (p = text; *p && (len < maxlen); p = nextp, len++) {
381		−	nextp = utf8codepoint(p, &codep);
	383	+	nextp = utf8_decode(p, &codep, &err);
382	384		if (!XftCharExists(win->env.dpy, font, codep)) {
383	385		fccharset = FcCharSetCreate();
384	386		FcCharSetAddChar(fccharset, codep);

nsxiv/3c7d6f3

Replace utf8codepoint with Chris Wellons' utf8_decode

nsxiv3c7d6f3