Enable compilation without FEAT_MBYTE.
[MacVim/KaoriYa.git] / src / mbyte.c
blobb2303fdae35dfae8ad5a4b593a924b154bdc1cba
1 /* vi:set ts=8 sts=4 sw=4:
3 * VIM - Vi IMproved by Bram Moolenaar
4 * Multibyte extensions partly by Sung-Hoon Baek
6 * Do ":help uganda" in Vim to read copying and usage conditions.
7 * Do ":help credits" in Vim to see a list of people who contributed.
8 * See README.txt for an overview of the Vim source code.
9 */
11 * mbyte.c: Code specifically for handling multi-byte characters.
13 * The encoding used in the core is set with 'encoding'. When 'encoding' is
14 * changed, the following four variables are set (for speed).
15 * Currently these types of character encodings are supported:
17 * "enc_dbcs" When non-zero it tells the type of double byte character
18 * encoding (Chinese, Korean, Japanese, etc.).
19 * The cell width on the display is equal to the number of
20 * bytes. (exception: DBCS_JPNU with first byte 0x8e)
21 * Recognizing the first or second byte is difficult, it
22 * requires checking a byte sequence from the start.
23 * "enc_utf8" When TRUE use Unicode characters in UTF-8 encoding.
24 * The cell width on the display needs to be determined from
25 * the character value.
26 * Recognizing bytes is easy: 0xxx.xxxx is a single-byte
27 * char, 10xx.xxxx is a trailing byte, 11xx.xxxx is a leading
28 * byte of a multi-byte character.
29 * To make things complicated, up to two composing characters
30 * are allowed. These are drawn on top of the first char.
31 * For most editing the sequence of bytes with composing
32 * characters included is considered to be one character.
33 * "enc_unicode" When 2 use 16-bit Unicode characters (or UTF-16).
34 * When 4 use 32-but Unicode characters.
35 * Internally characters are stored in UTF-8 encoding to
36 * avoid NUL bytes. Conversion happens when doing I/O.
37 * "enc_utf8" will also be TRUE.
39 * "has_mbyte" is set when "enc_dbcs" or "enc_utf8" is non-zero.
41 * If none of these is TRUE, 8-bit bytes are used for a character. The
42 * encoding isn't currently specified (TODO).
44 * 'encoding' specifies the encoding used in the core. This is in registers,
45 * text manipulation, buffers, etc. Conversion has to be done when characters
46 * in another encoding are received or send:
48 * clipboard
49 * ^
50 * | (2)
51 * V
52 * +---------------+
53 * (1) | | (3)
54 * keyboard ----->| core |-----> display
55 * | |
56 * +---------------+
57 * ^
58 * | (4)
59 * V
60 * file
62 * (1) Typed characters arrive in the current locale. Conversion is to be
63 * done when 'encoding' is different from 'termencoding'.
64 * (2) Text will be made available with the encoding specified with
65 * 'encoding'. If this is not sufficient, system-specific conversion
66 * might be required.
67 * (3) For the GUI the correct font must be selected, no conversion done.
68 * Otherwise, conversion is to be done when 'encoding' differs from
69 * 'termencoding'. (Different in the GTK+ 2 port -- 'termencoding'
70 * is always used for both input and output and must always be set to
71 * "utf-8". gui_mch_init() does this automatically.)
72 * (4) The encoding of the file is specified with 'fileencoding'. Conversion
73 * is to be done when it's different from 'encoding'.
75 * The viminfo file is a special case: Only text is converted, not file names.
76 * Vim scripts may contain an ":encoding" command. This has an effect for
77 * some commands, like ":menutrans"
80 #include "vim.h"
82 #ifdef WIN32UNIX
83 # ifndef WIN32_LEAN_AND_MEAN
84 # define WIN32_LEAN_AND_MEAN
85 # endif
86 # include <windows.h>
87 # ifdef WIN32
88 # undef WIN32 /* Some windows.h define WIN32, we don't want that here. */
89 # endif
90 #endif
92 #if (defined(WIN3264) || defined(WIN32UNIX)) && !defined(__MINGW32__)
93 # include <winnls.h>
94 #endif
96 #ifdef FEAT_GUI_X11
97 # include <X11/Intrinsic.h>
98 #endif
99 #ifdef X_LOCALE
100 #include <X11/Xlocale.h>
101 #endif
103 #if defined(FEAT_GUI_GTK) && defined(FEAT_XIM) && defined(HAVE_GTK2)
104 # include <gdk/gdkkeysyms.h>
105 # ifdef WIN3264
106 # include <gdk/gdkwin32.h>
107 # else
108 # include <gdk/gdkx.h>
109 # endif
110 #endif
112 #ifdef HAVE_WCHAR_H
113 # include <wchar.h>
114 #endif
116 #if 0
117 /* This has been disabled, because several people reported problems with the
118 * wcwidth() and iswprint() library functions, esp. for Hebrew. */
119 # ifdef __STDC_ISO_10646__
120 # define USE_WCHAR_FUNCTIONS
121 # endif
122 #endif
124 #if defined(FEAT_MBYTE) || defined(PROTO)
126 static int enc_canon_search __ARGS((char_u *name));
127 static int dbcs_char2len __ARGS((int c));
128 static int dbcs_char2bytes __ARGS((int c, char_u *buf));
129 static int dbcs_ptr2len __ARGS((char_u *p));
130 static int dbcs_char2cells __ARGS((int c));
131 static int dbcs_ptr2char __ARGS((char_u *p));
133 /* Lookup table to quickly get the length in bytes of a UTF-8 character from
134 * the first byte of a UTF-8 string. Bytes which are illegal when used as the
135 * first byte have a one, because these will be used separately. */
136 static char utf8len_tab[256] =
138 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
139 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
140 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
141 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
142 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/
143 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/
144 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,
145 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6,1,1,
149 * XIM often causes trouble. Define XIM_DEBUG to get a log of XIM callbacks
150 * in the "xim.log" file.
152 /* #define XIM_DEBUG */
153 #ifdef XIM_DEBUG
154 static void
155 xim_log(char *s, ...)
157 va_list arglist;
158 static FILE *fd = NULL;
160 if (fd == (FILE *)-1)
161 return;
162 if (fd == NULL)
164 fd = mch_fopen("xim.log", "w");
165 if (fd == NULL)
167 EMSG("Cannot open xim.log");
168 fd = (FILE *)-1;
169 return;
173 va_start(arglist, s);
174 vfprintf(fd, s, arglist);
175 va_end(arglist);
177 #endif
179 #endif
181 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO)
183 * Canonical encoding names and their properties.
184 * "iso-8859-n" is handled by enc_canonize() directly.
186 static struct
187 { char *name; int prop; int codepage;}
188 enc_canon_table[] =
190 #define IDX_LATIN_1 0
191 {"latin1", ENC_8BIT + ENC_LATIN1, 1252},
192 #define IDX_ISO_2 1
193 {"iso-8859-2", ENC_8BIT, 0},
194 #define IDX_ISO_3 2
195 {"iso-8859-3", ENC_8BIT, 0},
196 #define IDX_ISO_4 3
197 {"iso-8859-4", ENC_8BIT, 0},
198 #define IDX_ISO_5 4
199 {"iso-8859-5", ENC_8BIT, 0},
200 #define IDX_ISO_6 5
201 {"iso-8859-6", ENC_8BIT, 0},
202 #define IDX_ISO_7 6
203 {"iso-8859-7", ENC_8BIT, 0},
204 #define IDX_ISO_8 7
205 {"iso-8859-8", ENC_8BIT, 0},
206 #define IDX_ISO_9 8
207 {"iso-8859-9", ENC_8BIT, 0},
208 #define IDX_ISO_10 9
209 {"iso-8859-10", ENC_8BIT, 0},
210 #define IDX_ISO_11 10
211 {"iso-8859-11", ENC_8BIT, 0},
212 #define IDX_ISO_13 11
213 {"iso-8859-13", ENC_8BIT, 0},
214 #define IDX_ISO_14 12
215 {"iso-8859-14", ENC_8BIT, 0},
216 #define IDX_ISO_15 13
217 {"iso-8859-15", ENC_8BIT + ENC_LATIN9, 0},
218 #define IDX_KOI8_R 14
219 {"koi8-r", ENC_8BIT, 0},
220 #define IDX_KOI8_U 15
221 {"koi8-u", ENC_8BIT, 0},
222 #define IDX_UTF8 16
223 {"utf-8", ENC_UNICODE, 0},
224 #define IDX_UCS2 17
225 {"ucs-2", ENC_UNICODE + ENC_ENDIAN_B + ENC_2BYTE, 0},
226 #define IDX_UCS2LE 18
227 {"ucs-2le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2BYTE, 0},
228 #define IDX_UTF16 19
229 {"utf-16", ENC_UNICODE + ENC_ENDIAN_B + ENC_2WORD, 0},
230 #define IDX_UTF16LE 20
231 {"utf-16le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2WORD, 0},
232 #define IDX_UCS4 21
233 {"ucs-4", ENC_UNICODE + ENC_ENDIAN_B + ENC_4BYTE, 0},
234 #define IDX_UCS4LE 22
235 {"ucs-4le", ENC_UNICODE + ENC_ENDIAN_L + ENC_4BYTE, 0},
237 /* For debugging DBCS encoding on Unix. */
238 #define IDX_DEBUG 23
239 {"debug", ENC_DBCS, DBCS_DEBUG},
240 #define IDX_EUC_JP 24
241 {"euc-jp", ENC_DBCS, DBCS_JPNU},
242 #define IDX_SJIS 25
243 {"sjis", ENC_DBCS, DBCS_JPN},
244 #define IDX_EUC_KR 26
245 {"euc-kr", ENC_DBCS, DBCS_KORU},
246 #define IDX_EUC_CN 27
247 {"euc-cn", ENC_DBCS, DBCS_CHSU},
248 #define IDX_EUC_TW 28
249 {"euc-tw", ENC_DBCS, DBCS_CHTU},
250 #define IDX_BIG5 29
251 {"big5", ENC_DBCS, DBCS_CHT},
253 /* MS-DOS and MS-Windows codepages are included here, so that they can be
254 * used on Unix too. Most of them are similar to ISO-8859 encodings, but
255 * not exactly the same. */
256 #define IDX_CP437 30
257 {"cp437", ENC_8BIT, 437}, /* like iso-8859-1 */
258 #define IDX_CP737 31
259 {"cp737", ENC_8BIT, 737}, /* like iso-8859-7 */
260 #define IDX_CP775 32
261 {"cp775", ENC_8BIT, 775}, /* Baltic */
262 #define IDX_CP850 33
263 {"cp850", ENC_8BIT, 850}, /* like iso-8859-4 */
264 #define IDX_CP852 34
265 {"cp852", ENC_8BIT, 852}, /* like iso-8859-1 */
266 #define IDX_CP855 35
267 {"cp855", ENC_8BIT, 855}, /* like iso-8859-2 */
268 #define IDX_CP857 36
269 {"cp857", ENC_8BIT, 857}, /* like iso-8859-5 */
270 #define IDX_CP860 37
271 {"cp860", ENC_8BIT, 860}, /* like iso-8859-9 */
272 #define IDX_CP861 38
273 {"cp861", ENC_8BIT, 861}, /* like iso-8859-1 */
274 #define IDX_CP862 39
275 {"cp862", ENC_8BIT, 862}, /* like iso-8859-1 */
276 #define IDX_CP863 40
277 {"cp863", ENC_8BIT, 863}, /* like iso-8859-8 */
278 #define IDX_CP865 41
279 {"cp865", ENC_8BIT, 865}, /* like iso-8859-1 */
280 #define IDX_CP866 42
281 {"cp866", ENC_8BIT, 866}, /* like iso-8859-5 */
282 #define IDX_CP869 43
283 {"cp869", ENC_8BIT, 869}, /* like iso-8859-7 */
284 #define IDX_CP874 44
285 {"cp874", ENC_8BIT, 874}, /* Thai */
286 #define IDX_CP932 45
287 {"cp932", ENC_DBCS, DBCS_JPN},
288 #define IDX_CP936 46
289 {"cp936", ENC_DBCS, DBCS_CHS},
290 #define IDX_CP949 47
291 {"cp949", ENC_DBCS, DBCS_KOR},
292 #define IDX_CP950 48
293 {"cp950", ENC_DBCS, DBCS_CHT},
294 #define IDX_CP1250 49
295 {"cp1250", ENC_8BIT, 1250}, /* Czech, Polish, etc. */
296 #define IDX_CP1251 50
297 {"cp1251", ENC_8BIT, 1251}, /* Cyrillic */
298 /* cp1252 is considered to be equal to latin1 */
299 #define IDX_CP1253 51
300 {"cp1253", ENC_8BIT, 1253}, /* Greek */
301 #define IDX_CP1254 52
302 {"cp1254", ENC_8BIT, 1254}, /* Turkish */
303 #define IDX_CP1255 53
304 {"cp1255", ENC_8BIT, 1255}, /* Hebrew */
305 #define IDX_CP1256 54
306 {"cp1256", ENC_8BIT, 1256}, /* Arabic */
307 #define IDX_CP1257 55
308 {"cp1257", ENC_8BIT, 1257}, /* Baltic */
309 #define IDX_CP1258 56
310 {"cp1258", ENC_8BIT, 1258}, /* Vietnamese */
312 #define IDX_MACROMAN 57
313 {"macroman", ENC_8BIT + ENC_MACROMAN, 0}, /* Mac OS */
314 #define IDX_DECMCS 58
315 {"dec-mcs", ENC_8BIT, 0}, /* DEC MCS */
316 #define IDX_HPROMAN8 59
317 {"hp-roman8", ENC_8BIT, 0}, /* HP Roman8 */
318 #define IDX_COUNT 60
322 * Aliases for encoding names.
324 static struct
325 { char *name; int canon;}
326 enc_alias_table[] =
328 {"ansi", IDX_LATIN_1},
329 {"iso-8859-1", IDX_LATIN_1},
330 {"latin2", IDX_ISO_2},
331 {"latin3", IDX_ISO_3},
332 {"latin4", IDX_ISO_4},
333 {"cyrillic", IDX_ISO_5},
334 {"arabic", IDX_ISO_6},
335 {"greek", IDX_ISO_7},
336 #ifdef WIN3264
337 {"hebrew", IDX_CP1255},
338 #else
339 {"hebrew", IDX_ISO_8},
340 #endif
341 {"latin5", IDX_ISO_9},
342 {"turkish", IDX_ISO_9}, /* ? */
343 {"latin6", IDX_ISO_10},
344 {"nordic", IDX_ISO_10}, /* ? */
345 {"thai", IDX_ISO_11}, /* ? */
346 {"latin7", IDX_ISO_13},
347 {"latin8", IDX_ISO_14},
348 {"latin9", IDX_ISO_15},
349 {"utf8", IDX_UTF8},
350 {"unicode", IDX_UCS2},
351 {"ucs2", IDX_UCS2},
352 {"ucs2be", IDX_UCS2},
353 {"ucs-2be", IDX_UCS2},
354 {"ucs2le", IDX_UCS2LE},
355 {"utf16", IDX_UTF16},
356 {"utf16be", IDX_UTF16},
357 {"utf-16be", IDX_UTF16},
358 {"utf16le", IDX_UTF16LE},
359 {"ucs4", IDX_UCS4},
360 {"ucs4be", IDX_UCS4},
361 {"ucs-4be", IDX_UCS4},
362 {"ucs4le", IDX_UCS4LE},
363 {"932", IDX_CP932},
364 {"949", IDX_CP949},
365 {"936", IDX_CP936},
366 {"gbk", IDX_CP936},
367 {"950", IDX_CP950},
368 {"eucjp", IDX_EUC_JP},
369 {"unix-jis", IDX_EUC_JP},
370 {"ujis", IDX_EUC_JP},
371 {"shift-jis", IDX_SJIS},
372 {"euckr", IDX_EUC_KR},
373 {"5601", IDX_EUC_KR}, /* Sun: KS C 5601 */
374 {"euccn", IDX_EUC_CN},
375 {"gb2312", IDX_EUC_CN},
376 {"euctw", IDX_EUC_TW},
377 #if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS)
378 {"japan", IDX_CP932},
379 {"korea", IDX_CP949},
380 {"prc", IDX_CP936},
381 {"chinese", IDX_CP936},
382 {"taiwan", IDX_CP950},
383 {"big5", IDX_CP950},
384 #else
385 {"japan", IDX_EUC_JP},
386 {"korea", IDX_EUC_KR},
387 {"prc", IDX_EUC_CN},
388 {"chinese", IDX_EUC_CN},
389 {"taiwan", IDX_EUC_TW},
390 {"cp950", IDX_BIG5},
391 {"950", IDX_BIG5},
392 #endif
393 {"mac", IDX_MACROMAN},
394 {"mac-roman", IDX_MACROMAN},
395 {NULL, 0}
398 #ifndef CP_UTF8
399 # define CP_UTF8 65001 /* magic number from winnls.h */
400 #endif
403 * Find encoding "name" in the list of canonical encoding names.
404 * Returns -1 if not found.
406 static int
407 enc_canon_search(name)
408 char_u *name;
410 int i;
412 for (i = 0; i < IDX_COUNT; ++i)
413 if (STRCMP(name, enc_canon_table[i].name) == 0)
414 return i;
415 return -1;
418 #endif
420 #if defined(FEAT_MBYTE) || defined(PROTO)
423 * Find canonical encoding "name" in the list and return its properties.
424 * Returns 0 if not found.
427 enc_canon_props(name)
428 char_u *name;
430 int i;
432 i = enc_canon_search(name);
433 if (i >= 0)
434 return enc_canon_table[i].prop;
435 #ifdef WIN3264
436 if (name[0] == 'c' && name[1] == 'p' && VIM_ISDIGIT(name[2]))
438 CPINFO cpinfo;
440 /* Get info on this codepage to find out what it is. */
441 if (GetCPInfo(atoi(name + 2), &cpinfo) != 0)
443 if (cpinfo.MaxCharSize == 1) /* some single-byte encoding */
444 return ENC_8BIT;
445 if (cpinfo.MaxCharSize == 2
446 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0))
447 /* must be a DBCS encoding */
448 return ENC_DBCS;
450 return 0;
452 #endif
453 if (STRNCMP(name, "2byte-", 6) == 0)
454 return ENC_DBCS;
455 if (STRNCMP(name, "8bit-", 5) == 0 || STRNCMP(name, "iso-8859-", 9) == 0)
456 return ENC_8BIT;
457 return 0;
461 * Set up for using multi-byte characters.
462 * Called in three cases:
463 * - by main() to initialize (p_enc == NULL)
464 * - by set_init_1() after 'encoding' was set to its default.
465 * - by do_set() when 'encoding' has been set.
466 * p_enc must have been passed through enc_canonize() already.
467 * Sets the "enc_unicode", "enc_utf8", "enc_dbcs" and "has_mbyte" flags.
468 * Fills mb_bytelen_tab[] and returns NULL when there are no problems.
469 * When there is something wrong: Returns an error message and doesn't change
470 * anything.
472 char_u *
473 mb_init()
475 int i;
476 int idx;
477 int n;
478 int enc_dbcs_new = 0;
479 #if defined(USE_ICONV) && !defined(WIN3264) && !defined(WIN32UNIX) \
480 && !defined(MACOS)
481 # define LEN_FROM_CONV
482 vimconv_T vimconv;
483 char_u *p;
484 #endif
486 if (p_enc == NULL)
488 /* Just starting up: set the whole table to one's. */
489 for (i = 0; i < 256; ++i)
490 mb_bytelen_tab[i] = 1;
491 input_conv.vc_type = CONV_NONE;
492 input_conv.vc_factor = 1;
493 output_conv.vc_type = CONV_NONE;
494 return NULL;
497 #ifdef WIN3264
498 if (p_enc[0] == 'c' && p_enc[1] == 'p' && VIM_ISDIGIT(p_enc[2]))
500 CPINFO cpinfo;
502 /* Get info on this codepage to find out what it is. */
503 if (GetCPInfo(atoi(p_enc + 2), &cpinfo) != 0)
505 if (cpinfo.MaxCharSize == 1)
507 /* some single-byte encoding */
508 enc_unicode = 0;
509 enc_utf8 = FALSE;
511 else if (cpinfo.MaxCharSize == 2
512 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0))
514 /* must be a DBCS encoding, check below */
515 enc_dbcs_new = atoi(p_enc + 2);
517 else
518 goto codepage_invalid;
520 else if (GetLastError() == ERROR_INVALID_PARAMETER)
522 codepage_invalid:
523 return (char_u *)N_("E543: Not a valid codepage");
526 #endif
527 else if (STRNCMP(p_enc, "8bit-", 5) == 0
528 || STRNCMP(p_enc, "iso-8859-", 9) == 0)
530 /* Accept any "8bit-" or "iso-8859-" name. */
531 enc_unicode = 0;
532 enc_utf8 = FALSE;
534 else if (STRNCMP(p_enc, "2byte-", 6) == 0)
536 #ifdef WIN3264
537 /* Windows: accept only valid codepage numbers, check below. */
538 if (p_enc[6] != 'c' || p_enc[7] != 'p'
539 || (enc_dbcs_new = atoi(p_enc + 8)) == 0)
540 return e_invarg;
541 #else
542 /* Unix: accept any "2byte-" name, assume current locale. */
543 enc_dbcs_new = DBCS_2BYTE;
544 #endif
546 else if ((idx = enc_canon_search(p_enc)) >= 0)
548 i = enc_canon_table[idx].prop;
549 if (i & ENC_UNICODE)
551 /* Unicode */
552 enc_utf8 = TRUE;
553 if (i & (ENC_2BYTE | ENC_2WORD))
554 enc_unicode = 2;
555 else if (i & ENC_4BYTE)
556 enc_unicode = 4;
557 else
558 enc_unicode = 0;
560 else if (i & ENC_DBCS)
562 /* 2byte, handle below */
563 enc_dbcs_new = enc_canon_table[idx].codepage;
565 else
567 /* Must be 8-bit. */
568 enc_unicode = 0;
569 enc_utf8 = FALSE;
572 else /* Don't know what encoding this is, reject it. */
573 return e_invarg;
575 if (enc_dbcs_new != 0)
577 #ifdef WIN3264
578 /* Check if the DBCS code page is OK. */
579 if (!IsValidCodePage(enc_dbcs_new))
580 goto codepage_invalid;
581 #endif
582 enc_unicode = 0;
583 enc_utf8 = FALSE;
585 enc_dbcs = enc_dbcs_new;
586 has_mbyte = (enc_dbcs != 0 || enc_utf8);
588 #ifdef WIN3264
589 enc_codepage = encname2codepage(p_enc);
590 enc_latin9 = (STRCMP(p_enc, "iso-8859-15") == 0);
591 #endif
593 /* Detect an encoding that uses latin1 characters. */
594 enc_latin1like = (enc_utf8 || STRCMP(p_enc, "latin1") == 0
595 || STRCMP(p_enc, "iso-8859-15") == 0);
598 * Set the function pointers.
600 if (enc_utf8)
602 mb_ptr2len = utfc_ptr2len;
603 mb_char2len = utf_char2len;
604 mb_char2bytes = utf_char2bytes;
605 mb_ptr2cells = utf_ptr2cells;
606 mb_char2cells = utf_char2cells;
607 mb_off2cells = utf_off2cells;
608 mb_ptr2char = utf_ptr2char;
609 mb_head_off = utf_head_off;
611 else if (enc_dbcs != 0)
613 mb_ptr2len = dbcs_ptr2len;
614 mb_char2len = dbcs_char2len;
615 mb_char2bytes = dbcs_char2bytes;
616 mb_ptr2cells = dbcs_ptr2cells;
617 mb_char2cells = dbcs_char2cells;
618 mb_off2cells = dbcs_off2cells;
619 mb_ptr2char = dbcs_ptr2char;
620 mb_head_off = dbcs_head_off;
622 else
624 mb_ptr2len = latin_ptr2len;
625 mb_char2len = latin_char2len;
626 mb_char2bytes = latin_char2bytes;
627 mb_ptr2cells = latin_ptr2cells;
628 mb_char2cells = latin_char2cells;
629 mb_off2cells = latin_off2cells;
630 mb_ptr2char = latin_ptr2char;
631 mb_head_off = latin_head_off;
635 * Fill the mb_bytelen_tab[] for MB_BYTE2LEN().
637 #ifdef LEN_FROM_CONV
638 /* When 'encoding' is different from the current locale mblen() won't
639 * work. Use conversion to "utf-8" instead. */
640 vimconv.vc_type = CONV_NONE;
641 if (enc_dbcs)
643 p = enc_locale();
644 if (p == NULL || STRCMP(p, p_enc) != 0)
646 convert_setup(&vimconv, p_enc, (char_u *)"utf-8");
647 vimconv.vc_fail = TRUE;
649 vim_free(p);
651 #endif
653 for (i = 0; i < 256; ++i)
655 /* Our own function to reliably check the length of UTF-8 characters,
656 * independent of mblen(). */
657 if (enc_utf8)
658 n = utf8len_tab[i];
659 else if (enc_dbcs == 0)
660 n = 1;
661 else
663 #if defined(WIN3264) || defined(WIN32UNIX)
664 /* enc_dbcs is set by setting 'fileencoding'. It becomes a Windows
665 * CodePage identifier, which we can pass directly in to Windows
666 * API */
667 n = IsDBCSLeadByteEx(enc_dbcs, (BYTE)i) ? 2 : 1;
668 #else
669 # if defined(MACOS) || defined(__amigaos4__)
671 * if mblen() is not available, character which MSB is turned on
672 * are treated as leading byte character. (note : This assumption
673 * is not always true.)
675 n = (i & 0x80) ? 2 : 1;
676 # else
677 char buf[MB_MAXBYTES];
678 # ifdef X_LOCALE
679 # ifndef mblen
680 # define mblen _Xmblen
681 # endif
682 # endif
683 if (i == NUL) /* just in case mblen() can't handle "" */
684 n = 1;
685 else
687 buf[0] = i;
688 buf[1] = 0;
689 #ifdef LEN_FROM_CONV
690 if (vimconv.vc_type != CONV_NONE)
693 * string_convert() should fail when converting the first
694 * byte of a double-byte character.
696 p = string_convert(&vimconv, (char_u *)buf, NULL);
697 if (p != NULL)
699 vim_free(p);
700 n = 1;
702 else
703 n = 2;
705 else
706 #endif
709 * mblen() should return -1 for invalid (means the leading
710 * multibyte) character. However there are some platforms
711 * where mblen() returns 0 for invalid character.
712 * Therefore, following condition includes 0.
714 (void)mblen(NULL, 0); /* First reset the state. */
715 if (mblen(buf, (size_t)1) <= 0)
716 n = 2;
717 else
718 n = 1;
721 # endif
722 #endif
725 mb_bytelen_tab[i] = n;
728 #ifdef LEN_FROM_CONV
729 convert_setup(&vimconv, NULL, NULL);
730 #endif
732 /* The cell width depends on the type of multi-byte characters. */
733 (void)init_chartab();
735 /* When enc_utf8 is set or reset, (de)allocate ScreenLinesUC[] */
736 screenalloc(FALSE);
738 /* When using Unicode, set default for 'fileencodings'. */
739 if (enc_utf8 && !option_was_set((char_u *)"fencs"))
740 set_string_option_direct((char_u *)"fencs", -1,
741 (char_u *)"ucs-bom,utf-8,default,latin1", OPT_FREE, 0);
743 #if defined(HAVE_BIND_TEXTDOMAIN_CODESET) && defined(FEAT_GETTEXT)
744 /* GNU gettext 0.10.37 supports this feature: set the codeset used for
745 * translated messages independently from the current locale. */
746 (void)bind_textdomain_codeset(VIMPACKAGE,
747 enc_utf8 ? "utf-8" : (char *)p_enc);
748 #endif
750 #ifdef WIN32
751 /* When changing 'encoding' while starting up, then convert the command
752 * line arguments from the active codepage to 'encoding'. */
753 if (starting != 0)
754 fix_arg_enc();
755 #endif
757 #ifdef FEAT_AUTOCMD
758 /* Fire an autocommand to let people do custom font setup. This must be
759 * after Vim has been setup for the new encoding. */
760 apply_autocmds(EVENT_ENCODINGCHANGED, NULL, (char_u *)"", FALSE, curbuf);
761 #endif
763 #ifdef FEAT_SPELL
764 /* Need to reload spell dictionaries */
765 spell_reload();
766 #endif
768 return NULL;
772 * Return the size of the BOM for the current buffer:
773 * 0 - no BOM
774 * 2 - UCS-2 or UTF-16 BOM
775 * 4 - UCS-4 BOM
776 * 3 - UTF-8 BOM
779 bomb_size()
781 int n = 0;
783 if (curbuf->b_p_bomb && !curbuf->b_p_bin)
785 if (*curbuf->b_p_fenc == NUL)
787 if (enc_utf8)
789 if (enc_unicode != 0)
790 n = enc_unicode;
791 else
792 n = 3;
795 else if (STRCMP(curbuf->b_p_fenc, "utf-8") == 0)
796 n = 3;
797 else if (STRNCMP(curbuf->b_p_fenc, "ucs-2", 5) == 0
798 || STRNCMP(curbuf->b_p_fenc, "utf-16", 6) == 0)
799 n = 2;
800 else if (STRNCMP(curbuf->b_p_fenc, "ucs-4", 5) == 0)
801 n = 4;
803 return n;
807 * Get class of pointer:
808 * 0 for blank or NUL
809 * 1 for punctuation
810 * 2 for an (ASCII) word character
811 * >2 for other word characters
814 mb_get_class(p)
815 char_u *p;
817 if (MB_BYTE2LEN(p[0]) == 1)
819 if (p[0] == NUL || vim_iswhite(p[0]))
820 return 0;
821 if (vim_iswordc(p[0]))
822 return 2;
823 return 1;
825 if (enc_dbcs != 0 && p[0] != NUL && p[1] != NUL)
826 return dbcs_class(p[0], p[1]);
827 if (enc_utf8)
828 return utf_class(utf_ptr2char(p));
829 return 0;
833 * Get class of a double-byte character. This always returns 3 or bigger.
834 * TODO: Should return 1 for punctuation.
837 dbcs_class(lead, trail)
838 unsigned lead;
839 unsigned trail;
841 switch (enc_dbcs)
843 /* please add classfy routine for your language in here */
845 case DBCS_JPNU: /* ? */
846 case DBCS_JPN:
848 /* JIS code classification */
849 unsigned char lb = lead;
850 unsigned char tb = trail;
852 /* convert process code to JIS */
853 # if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS)
854 /* process code is SJIS */
855 if (lb <= 0x9f)
856 lb = (lb - 0x81) * 2 + 0x21;
857 else
858 lb = (lb - 0xc1) * 2 + 0x21;
859 if (tb <= 0x7e)
860 tb -= 0x1f;
861 else if (tb <= 0x9e)
862 tb -= 0x20;
863 else
865 tb -= 0x7e;
866 lb += 1;
868 # else
870 * XXX: Code page identification can not use with all
871 * system! So, some other encoding information
872 * will be needed.
873 * In japanese: SJIS,EUC,UNICODE,(JIS)
874 * Note that JIS-code system don't use as
875 * process code in most system because it uses
876 * escape sequences(JIS is context depend encoding).
878 /* assume process code is JAPANESE-EUC */
879 lb &= 0x7f;
880 tb &= 0x7f;
881 # endif
882 /* exceptions */
883 switch (lb << 8 | tb)
885 case 0x2121: /* ZENKAKU space */
886 return 0;
887 case 0x2122: /* KU-TEN (Japanese comma) */
888 case 0x2123: /* TOU-TEN (Japanese period) */
889 case 0x2124: /* ZENKAKU comma */
890 case 0x2125: /* ZENKAKU period */
891 return 1;
892 case 0x213c: /* prolongedsound handled as KATAKANA */
893 return 13;
895 /* sieved by KU code */
896 switch (lb)
898 case 0x21:
899 case 0x22:
900 /* special symbols */
901 return 10;
902 case 0x23:
903 /* alpha-numeric */
904 return 11;
905 case 0x24:
906 /* hiragana */
907 return 12;
908 case 0x25:
909 /* katakana */
910 return 13;
911 case 0x26:
912 /* greek */
913 return 14;
914 case 0x27:
915 /* russian */
916 return 15;
917 case 0x28:
918 /* lines */
919 return 16;
920 default:
921 /* kanji */
922 return 17;
926 case DBCS_KORU: /* ? */
927 case DBCS_KOR:
929 /* KS code classification */
930 unsigned char c1 = lead;
931 unsigned char c2 = trail;
934 * 20 : Hangul
935 * 21 : Hanja
936 * 22 : Symbols
937 * 23 : Alpha-numeric/Roman Letter (Full width)
938 * 24 : Hangul Letter(Alphabet)
939 * 25 : Roman Numeral/Greek Letter
940 * 26 : Box Drawings
941 * 27 : Unit Symbols
942 * 28 : Circled/Parenthesized Letter
943 * 29 : Hirigana/Katakana
944 * 30 : Cyrillic Letter
947 if (c1 >= 0xB0 && c1 <= 0xC8)
948 /* Hangul */
949 return 20;
950 #if defined(WIN3264) || defined(WIN32UNIX)
951 else if (c1 <= 0xA0 || c2 <= 0xA0)
952 /* Extended Hangul Region : MS UHC(Unified Hangul Code) */
953 /* c1: 0x81-0xA0 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xFE
954 * c1: 0xA1-0xC6 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xA0
956 return 20;
957 #endif
959 else if (c1 >= 0xCA && c1 <= 0xFD)
960 /* Hanja */
961 return 21;
962 else switch (c1)
964 case 0xA1:
965 case 0xA2:
966 /* Symbols */
967 return 22;
968 case 0xA3:
969 /* Alpha-numeric */
970 return 23;
971 case 0xA4:
972 /* Hangul Letter(Alphabet) */
973 return 24;
974 case 0xA5:
975 /* Roman Numeral/Greek Letter */
976 return 25;
977 case 0xA6:
978 /* Box Drawings */
979 return 26;
980 case 0xA7:
981 /* Unit Symbols */
982 return 27;
983 case 0xA8:
984 case 0xA9:
985 if (c2 <= 0xAF)
986 return 25; /* Roman Letter */
987 else if (c2 >= 0xF6)
988 return 22; /* Symbols */
989 else
990 /* Circled/Parenthesized Letter */
991 return 28;
992 case 0xAA:
993 case 0xAB:
994 /* Hirigana/Katakana */
995 return 29;
996 case 0xAC:
997 /* Cyrillic Letter */
998 return 30;
1001 default:
1002 break;
1004 return 3;
1008 * mb_char2len() function pointer.
1009 * Return length in bytes of character "c".
1010 * Returns 1 for a single-byte character.
1012 /* ARGSUSED */
1014 latin_char2len(c)
1015 int c;
1017 return 1;
1020 static int
1021 dbcs_char2len(c)
1022 int c;
1024 if (c >= 0x100)
1025 return 2;
1026 return 1;
1030 * mb_char2bytes() function pointer.
1031 * Convert a character to its bytes.
1032 * Returns the length in bytes.
1035 latin_char2bytes(c, buf)
1036 int c;
1037 char_u *buf;
1039 buf[0] = c;
1040 return 1;
1043 static int
1044 dbcs_char2bytes(c, buf)
1045 int c;
1046 char_u *buf;
1048 if (c >= 0x100)
1050 buf[0] = (unsigned)c >> 8;
1051 buf[1] = c;
1052 /* Never use a NUL byte, it causes lots of trouble. It's an invalid
1053 * character anyway. */
1054 if (buf[1] == NUL)
1055 buf[1] = '\n';
1056 return 2;
1058 buf[0] = c;
1059 return 1;
1063 * mb_ptr2len() function pointer.
1064 * Get byte length of character at "*p" but stop at a NUL.
1065 * For UTF-8 this includes following composing characters.
1066 * Returns 0 when *p is NUL.
1070 latin_ptr2len(p)
1071 char_u *p;
1073 return MB_BYTE2LEN(*p);
1076 static int
1077 dbcs_ptr2len(p)
1078 char_u *p;
1080 int len;
1082 /* Check if second byte is not missing. */
1083 len = MB_BYTE2LEN(*p);
1084 if (len == 2 && p[1] == NUL)
1085 len = 1;
1086 return len;
1089 struct interval
1091 unsigned short first;
1092 unsigned short last;
1094 static int intable __ARGS((struct interval *table, size_t size, int c));
1097 * Return TRUE if "c" is in "table[size / sizeof(struct interval)]".
1099 static int
1100 intable(table, size, c)
1101 struct interval *table;
1102 size_t size;
1103 int c;
1105 int mid, bot, top;
1107 /* first quick check for Latin1 etc. characters */
1108 if (c < table[0].first)
1109 return FALSE;
1111 /* binary search in table */
1112 bot = 0;
1113 top = (int)(size / sizeof(struct interval) - 1);
1114 while (top >= bot)
1116 mid = (bot + top) / 2;
1117 if (table[mid].last < c)
1118 bot = mid + 1;
1119 else if (table[mid].first > c)
1120 top = mid - 1;
1121 else
1122 return TRUE;
1124 return FALSE;
1128 * For UTF-8 character "c" return 2 for a double-width character, 1 for others.
1129 * Returns 4 or 6 for an unprintable character.
1130 * Is only correct for characters >= 0x80.
1131 * When p_ambw is "double", return 2 for a character with East Asian Width
1132 * class 'A'(mbiguous).
1135 utf_char2cells(c)
1136 int c;
1138 /* sorted list of non-overlapping intervals of East Asian Ambiguous
1139 * characters, generated with:
1140 * "uniset +WIDTH-A -cat=Me -cat=Mn -cat=Cf c" */
1141 static struct interval ambiguous[] = {
1142 {0x00A1, 0x00A1}, {0x00A4, 0x00A4}, {0x00A7, 0x00A8},
1143 {0x00AA, 0x00AA}, {0x00AE, 0x00AE}, {0x00B0, 0x00B4},
1144 {0x00B6, 0x00BA}, {0x00BC, 0x00BF}, {0x00C6, 0x00C6},
1145 {0x00D0, 0x00D0}, {0x00D7, 0x00D8}, {0x00DE, 0x00E1},
1146 {0x00E6, 0x00E6}, {0x00E8, 0x00EA}, {0x00EC, 0x00ED},
1147 {0x00F0, 0x00F0}, {0x00F2, 0x00F3}, {0x00F7, 0x00FA},
1148 {0x00FC, 0x00FC}, {0x00FE, 0x00FE}, {0x0101, 0x0101},
1149 {0x0111, 0x0111}, {0x0113, 0x0113}, {0x011B, 0x011B},
1150 {0x0126, 0x0127}, {0x012B, 0x012B}, {0x0131, 0x0133},
1151 {0x0138, 0x0138}, {0x013F, 0x0142}, {0x0144, 0x0144},
1152 {0x0148, 0x014B}, {0x014D, 0x014D}, {0x0152, 0x0153},
1153 {0x0166, 0x0167}, {0x016B, 0x016B}, {0x01CE, 0x01CE},
1154 {0x01D0, 0x01D0}, {0x01D2, 0x01D2}, {0x01D4, 0x01D4},
1155 {0x01D6, 0x01D6}, {0x01D8, 0x01D8}, {0x01DA, 0x01DA},
1156 {0x01DC, 0x01DC}, {0x0251, 0x0251}, {0x0261, 0x0261},
1157 {0x02C4, 0x02C4}, {0x02C7, 0x02C7}, {0x02C9, 0x02CB},
1158 {0x02CD, 0x02CD}, {0x02D0, 0x02D0}, {0x02D8, 0x02DB},
1159 {0x02DD, 0x02DD}, {0x02DF, 0x02DF}, {0x0391, 0x03A1},
1160 {0x03A3, 0x03A9}, {0x03B1, 0x03C1}, {0x03C3, 0x03C9},
1161 {0x0401, 0x0401}, {0x0410, 0x044F}, {0x0451, 0x0451},
1162 {0x2010, 0x2010}, {0x2013, 0x2016}, {0x2018, 0x2019},
1163 {0x201C, 0x201D}, {0x2020, 0x2022}, {0x2024, 0x2027},
1164 {0x2030, 0x2030}, {0x2032, 0x2033}, {0x2035, 0x2035},
1165 {0x203B, 0x203B}, {0x203E, 0x203E}, {0x2074, 0x2074},
1166 {0x207F, 0x207F}, {0x2081, 0x2084}, {0x20AC, 0x20AC},
1167 {0x2103, 0x2103}, {0x2105, 0x2105}, {0x2109, 0x2109},
1168 {0x2113, 0x2113}, {0x2116, 0x2116}, {0x2121, 0x2122},
1169 {0x2126, 0x2126}, {0x212B, 0x212B}, {0x2153, 0x2154},
1170 {0x215B, 0x215E}, {0x2160, 0x216B}, {0x2170, 0x2179},
1171 {0x2190, 0x2199}, {0x21B8, 0x21B9}, {0x21D2, 0x21D2},
1172 {0x21D4, 0x21D4}, {0x21E7, 0x21E7}, {0x2200, 0x2200},
1173 {0x2202, 0x2203}, {0x2207, 0x2208}, {0x220B, 0x220B},
1174 {0x220F, 0x220F}, {0x2211, 0x2211}, {0x2215, 0x2215},
1175 {0x221A, 0x221A}, {0x221D, 0x2220}, {0x2223, 0x2223},
1176 {0x2225, 0x2225}, {0x2227, 0x222C}, {0x222E, 0x222E},
1177 {0x2234, 0x2237}, {0x223C, 0x223D}, {0x2248, 0x2248},
1178 {0x224C, 0x224C}, {0x2252, 0x2252}, {0x2260, 0x2261},
1179 {0x2264, 0x2267}, {0x226A, 0x226B}, {0x226E, 0x226F},
1180 {0x2282, 0x2283}, {0x2286, 0x2287}, {0x2295, 0x2295},
1181 {0x2299, 0x2299}, {0x22A5, 0x22A5}, {0x22BF, 0x22BF},
1182 {0x2312, 0x2312}, {0x2460, 0x24E9}, {0x24EB, 0x254B},
1183 {0x2550, 0x2573}, {0x2580, 0x258F}, {0x2592, 0x2595},
1184 {0x25A0, 0x25A1}, {0x25A3, 0x25A9}, {0x25B2, 0x25B3},
1185 {0x25B6, 0x25B7}, {0x25BC, 0x25BD}, {0x25C0, 0x25C1},
1186 {0x25C6, 0x25C8}, {0x25CB, 0x25CB}, {0x25CE, 0x25D1},
1187 {0x25E2, 0x25E5}, {0x25EF, 0x25EF}, {0x2605, 0x2606},
1188 {0x2609, 0x2609}, {0x260E, 0x260F}, {0x2614, 0x2615},
1189 {0x261C, 0x261C}, {0x261E, 0x261E}, {0x2640, 0x2640},
1190 {0x2642, 0x2642}, {0x2660, 0x2661}, {0x2663, 0x2665},
1191 {0x2667, 0x266A}, {0x266C, 0x266D}, {0x266F, 0x266F},
1192 {0x273D, 0x273D}, {0x2776, 0x277F}, {0xE000, 0xF8FF},
1193 {0xFFFD, 0xFFFD}, /* {0xF0000, 0xFFFFD}, {0x100000, 0x10FFFD} */
1196 if (c >= 0x100)
1198 #ifdef USE_WCHAR_FUNCTIONS
1200 * Assume the library function wcwidth() works better than our own
1201 * stuff. It should return 1 for ambiguous width chars!
1203 int n = wcwidth(c);
1205 if (n < 0)
1206 return 6; /* unprintable, displays <xxxx> */
1207 if (n > 1)
1208 return n;
1209 #else
1210 if (!utf_printable(c))
1211 return 6; /* unprintable, displays <xxxx> */
1212 if (c >= 0x1100
1213 && (c <= 0x115f /* Hangul Jamo */
1214 || c == 0x2329
1215 || c == 0x232a
1216 || (c >= 0x2e80 && c <= 0xa4cf
1217 && c != 0x303f) /* CJK ... Yi */
1218 || (c >= 0xac00 && c <= 0xd7a3) /* Hangul Syllables */
1219 || (c >= 0xf900 && c <= 0xfaff) /* CJK Compatibility
1220 Ideographs */
1221 || (c >= 0xfe30 && c <= 0xfe6f) /* CJK Compatibility Forms */
1222 || (c >= 0xff00 && c <= 0xff60) /* Fullwidth Forms */
1223 || (c >= 0xffe0 && c <= 0xffe6)
1224 || (c >= 0x20000 && c <= 0x2fffd)
1225 || (c >= 0x30000 && c <= 0x3fffd)))
1226 return 2;
1227 #endif
1230 /* Characters below 0x100 are influenced by 'isprint' option */
1231 else if (c >= 0x80 && !vim_isprintc(c))
1232 return 4; /* unprintable, displays <xx> */
1234 if (c >= 0x80 && *p_ambw == 'd' && intable(ambiguous, sizeof(ambiguous), c))
1235 return 2;
1237 return 1;
1241 * mb_ptr2cells() function pointer.
1242 * Return the number of display cells character at "*p" occupies.
1243 * This doesn't take care of unprintable characters, use ptr2cells() for that.
1245 /*ARGSUSED*/
1247 latin_ptr2cells(p)
1248 char_u *p;
1250 return 1;
1254 utf_ptr2cells(p)
1255 char_u *p;
1257 int c;
1259 /* Need to convert to a wide character. */
1260 if (*p >= 0x80)
1262 c = utf_ptr2char(p);
1263 /* An illegal byte is displayed as <xx>. */
1264 if (utf_ptr2len(p) == 1 || c == NUL)
1265 return 4;
1266 /* If the char is ASCII it must be an overlong sequence. */
1267 if (c < 0x80)
1268 return char2cells(c);
1269 return utf_char2cells(c);
1271 return 1;
1275 dbcs_ptr2cells(p)
1276 char_u *p;
1278 /* Number of cells is equal to number of bytes, except for euc-jp when
1279 * the first byte is 0x8e. */
1280 if (enc_dbcs == DBCS_JPNU && *p == 0x8e)
1281 return 1;
1282 return MB_BYTE2LEN(*p);
1286 * mb_char2cells() function pointer.
1287 * Return the number of display cells character "c" occupies.
1288 * Only takes care of multi-byte chars, not "^C" and such.
1290 /*ARGSUSED*/
1292 latin_char2cells(c)
1293 int c;
1295 return 1;
1298 static int
1299 dbcs_char2cells(c)
1300 int c;
1302 /* Number of cells is equal to number of bytes, except for euc-jp when
1303 * the first byte is 0x8e. */
1304 if (enc_dbcs == DBCS_JPNU && ((unsigned)c >> 8) == 0x8e)
1305 return 1;
1306 /* use the first byte */
1307 return MB_BYTE2LEN((unsigned)c >> 8);
1311 * mb_off2cells() function pointer.
1312 * Return number of display cells for char at ScreenLines[off].
1313 * We make sure that the offset used is less than "max_off".
1315 /*ARGSUSED*/
1317 latin_off2cells(off, max_off)
1318 unsigned off;
1319 unsigned max_off;
1321 return 1;
1325 dbcs_off2cells(off, max_off)
1326 unsigned off;
1327 unsigned max_off;
1329 /* never check beyond end of the line */
1330 if (off >= max_off)
1331 return 1;
1333 /* Number of cells is equal to number of bytes, except for euc-jp when
1334 * the first byte is 0x8e. */
1335 if (enc_dbcs == DBCS_JPNU && ScreenLines[off] == 0x8e)
1336 return 1;
1337 return MB_BYTE2LEN(ScreenLines[off]);
1341 utf_off2cells(off, max_off)
1342 unsigned off;
1343 unsigned max_off;
1345 return (off + 1 < max_off && ScreenLines[off + 1] == 0) ? 2 : 1;
1349 * mb_ptr2char() function pointer.
1350 * Convert a byte sequence into a character.
1353 latin_ptr2char(p)
1354 char_u *p;
1356 return *p;
1359 static int
1360 dbcs_ptr2char(p)
1361 char_u *p;
1363 if (MB_BYTE2LEN(*p) > 1 && p[1] != NUL)
1364 return (p[0] << 8) + p[1];
1365 return *p;
1369 * Convert a UTF-8 byte sequence to a wide character.
1370 * If the sequence is illegal or truncated by a NUL the first byte is
1371 * returned.
1372 * Does not include composing characters, of course.
1375 utf_ptr2char(p)
1376 char_u *p;
1378 int len;
1380 if (p[0] < 0x80) /* be quick for ASCII */
1381 return p[0];
1383 len = utf8len_tab[p[0]];
1384 if ((p[1] & 0xc0) == 0x80)
1386 if (len == 2)
1387 return ((p[0] & 0x1f) << 6) + (p[1] & 0x3f);
1388 if ((p[2] & 0xc0) == 0x80)
1390 if (len == 3)
1391 return ((p[0] & 0x0f) << 12) + ((p[1] & 0x3f) << 6)
1392 + (p[2] & 0x3f);
1393 if ((p[3] & 0xc0) == 0x80)
1395 if (len == 4)
1396 return ((p[0] & 0x07) << 18) + ((p[1] & 0x3f) << 12)
1397 + ((p[2] & 0x3f) << 6) + (p[3] & 0x3f);
1398 if ((p[4] & 0xc0) == 0x80)
1400 if (len == 5)
1401 return ((p[0] & 0x03) << 24) + ((p[1] & 0x3f) << 18)
1402 + ((p[2] & 0x3f) << 12) + ((p[3] & 0x3f) << 6)
1403 + (p[4] & 0x3f);
1404 if ((p[5] & 0xc0) == 0x80 && len == 6)
1405 return ((p[0] & 0x01) << 30) + ((p[1] & 0x3f) << 24)
1406 + ((p[2] & 0x3f) << 18) + ((p[3] & 0x3f) << 12)
1407 + ((p[4] & 0x3f) << 6) + (p[5] & 0x3f);
1412 /* Illegal value, just return the first byte */
1413 return p[0];
1417 * Get character at **pp and advance *pp to the next character.
1418 * Note: composing characters are skipped!
1421 mb_ptr2char_adv(pp)
1422 char_u **pp;
1424 int c;
1426 c = (*mb_ptr2char)(*pp);
1427 *pp += (*mb_ptr2len)(*pp);
1428 return c;
1432 * Get character at **pp and advance *pp to the next character.
1433 * Note: composing characters are returned as separate characters.
1436 mb_cptr2char_adv(pp)
1437 char_u **pp;
1439 int c;
1441 c = (*mb_ptr2char)(*pp);
1442 if (enc_utf8)
1443 *pp += utf_ptr2len(*pp);
1444 else
1445 *pp += (*mb_ptr2len)(*pp);
1446 return c;
1449 #if defined(FEAT_ARABIC) || defined(PROTO)
1451 * Check whether we are dealing with Arabic combining characters.
1452 * Note: these are NOT really composing characters!
1455 arabic_combine(one, two)
1456 int one; /* first character */
1457 int two; /* character just after "one" */
1459 if (one == a_LAM)
1460 return arabic_maycombine(two);
1461 return FALSE;
1465 * Check whether we are dealing with a character that could be regarded as an
1466 * Arabic combining character, need to check the character before this.
1469 arabic_maycombine(two)
1470 int two;
1472 if (p_arshape && !p_tbidi)
1473 return (two == a_ALEF_MADDA
1474 || two == a_ALEF_HAMZA_ABOVE
1475 || two == a_ALEF_HAMZA_BELOW
1476 || two == a_ALEF);
1477 return FALSE;
1481 * Check if the character pointed to by "p2" is a composing character when it
1482 * comes after "p1". For Arabic sometimes "ab" is replaced with "c", which
1483 * behaves like a composing character.
1486 utf_composinglike(p1, p2)
1487 char_u *p1;
1488 char_u *p2;
1490 int c2;
1492 c2 = utf_ptr2char(p2);
1493 if (utf_iscomposing(c2))
1494 return TRUE;
1495 if (!arabic_maycombine(c2))
1496 return FALSE;
1497 return arabic_combine(utf_ptr2char(p1), c2);
1499 #endif
1502 * Convert a UTF-8 byte string to a wide character. Also get up to MAX_MCO
1503 * composing characters.
1506 utfc_ptr2char(p, pcc)
1507 char_u *p;
1508 int *pcc; /* return: composing chars, last one is 0 */
1510 int len;
1511 int c;
1512 int cc;
1513 int i = 0;
1515 c = utf_ptr2char(p);
1516 len = utf_ptr2len(p);
1518 /* Only accept a composing char when the first char isn't illegal. */
1519 if ((len > 1 || *p < 0x80)
1520 && p[len] >= 0x80
1521 && UTF_COMPOSINGLIKE(p, p + len))
1523 cc = utf_ptr2char(p + len);
1524 for (;;)
1526 pcc[i++] = cc;
1527 if (i == MAX_MCO)
1528 break;
1529 len += utf_ptr2len(p + len);
1530 if (p[len] < 0x80 || !utf_iscomposing(cc = utf_ptr2char(p + len)))
1531 break;
1535 if (i < MAX_MCO) /* last composing char must be 0 */
1536 pcc[i] = 0;
1538 return c;
1542 * Convert a UTF-8 byte string to a wide character. Also get up to MAX_MCO
1543 * composing characters. Use no more than p[maxlen].
1546 utfc_ptr2char_len(p, pcc, maxlen)
1547 char_u *p;
1548 int *pcc; /* return: composing chars, last one is 0 */
1549 int maxlen;
1551 int len;
1552 int c;
1553 int cc;
1554 int i = 0;
1556 c = utf_ptr2char(p);
1557 len = utf_ptr2len_len(p, maxlen);
1558 /* Only accept a composing char when the first char isn't illegal. */
1559 if ((len > 1 || *p < 0x80)
1560 && len < maxlen
1561 && p[len] >= 0x80
1562 && UTF_COMPOSINGLIKE(p, p + len))
1564 cc = utf_ptr2char(p + len);
1565 for (;;)
1567 pcc[i++] = cc;
1568 if (i == MAX_MCO)
1569 break;
1570 len += utf_ptr2len_len(p + len, maxlen - len);
1571 if (len >= maxlen
1572 || p[len] < 0x80
1573 || !utf_iscomposing(cc = utf_ptr2char(p + len)))
1574 break;
1578 if (i < MAX_MCO) /* last composing char must be 0 */
1579 pcc[i] = 0;
1581 return c;
1585 * Convert the character at screen position "off" to a sequence of bytes.
1586 * Includes the composing characters.
1587 * "buf" must at least have the length MB_MAXBYTES.
1588 * Returns the produced number of bytes.
1591 utfc_char2bytes(off, buf)
1592 int off;
1593 char_u *buf;
1595 int len;
1596 int i;
1598 len = utf_char2bytes(ScreenLinesUC[off], buf);
1599 for (i = 0; i < Screen_mco; ++i)
1601 if (ScreenLinesC[i][off] == 0)
1602 break;
1603 len += utf_char2bytes(ScreenLinesC[i][off], buf + len);
1605 return len;
1609 * Get the length of a UTF-8 byte sequence, not including any following
1610 * composing characters.
1611 * Returns 0 for "".
1612 * Returns 1 for an illegal byte sequence.
1615 utf_ptr2len(p)
1616 char_u *p;
1618 int len;
1619 int i;
1621 if (*p == NUL)
1622 return 0;
1623 len = utf8len_tab[*p];
1624 for (i = 1; i < len; ++i)
1625 if ((p[i] & 0xc0) != 0x80)
1626 return 1;
1627 return len;
1631 * Return length of UTF-8 character, obtained from the first byte.
1632 * "b" must be between 0 and 255!
1635 utf_byte2len(b)
1636 int b;
1638 return utf8len_tab[b];
1642 * Get the length of UTF-8 byte sequence "p[size]". Does not include any
1643 * following composing characters.
1644 * Returns 1 for "".
1645 * Returns 1 for an illegal byte sequence.
1646 * Returns number > "size" for an incomplete byte sequence.
1649 utf_ptr2len_len(p, size)
1650 char_u *p;
1651 int size;
1653 int len;
1654 int i;
1656 if (*p == NUL)
1657 return 1;
1658 len = utf8len_tab[*p];
1659 if (len > size)
1660 return len; /* incomplete byte sequence. */
1661 for (i = 1; i < len; ++i)
1662 if ((p[i] & 0xc0) != 0x80)
1663 return 1;
1664 return len;
1668 * Return the number of bytes the UTF-8 encoding of the character at "p" takes.
1669 * This includes following composing characters.
1672 utfc_ptr2len(p)
1673 char_u *p;
1675 int len;
1676 int b0 = *p;
1677 #ifdef FEAT_ARABIC
1678 int prevlen;
1679 #endif
1681 if (b0 == NUL)
1682 return 0;
1683 if (b0 < 0x80 && p[1] < 0x80) /* be quick for ASCII */
1684 return 1;
1686 /* Skip over first UTF-8 char, stopping at a NUL byte. */
1687 len = utf_ptr2len(p);
1689 /* Check for illegal byte. */
1690 if (len == 1 && b0 >= 0x80)
1691 return 1;
1694 * Check for composing characters. We can handle only the first two, but
1695 * skip all of them (otherwise the cursor would get stuck).
1697 #ifdef FEAT_ARABIC
1698 prevlen = 0;
1699 #endif
1700 for (;;)
1702 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len))
1703 return len;
1705 /* Skip over composing char */
1706 #ifdef FEAT_ARABIC
1707 prevlen = len;
1708 #endif
1709 len += utf_ptr2len(p + len);
1714 * Return the number of bytes the UTF-8 encoding of the character at "p[size]"
1715 * takes. This includes following composing characters.
1716 * Returns 1 for an illegal char or an incomplete byte sequence.
1719 utfc_ptr2len_len(p, size)
1720 char_u *p;
1721 int size;
1723 int len;
1724 #ifdef FEAT_ARABIC
1725 int prevlen;
1726 #endif
1728 if (*p == NUL)
1729 return 0;
1730 if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */
1731 return 1;
1733 /* Skip over first UTF-8 char, stopping at a NUL byte. */
1734 len = utf_ptr2len_len(p, size);
1736 /* Check for illegal byte and incomplete byte sequence. */
1737 if ((len == 1 && p[0] >= 0x80) || len > size)
1738 return 1;
1741 * Check for composing characters. We can handle only the first two, but
1742 * skip all of them (otherwise the cursor would get stuck).
1744 #ifdef FEAT_ARABIC
1745 prevlen = 0;
1746 #endif
1747 while (len < size)
1749 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len))
1750 break;
1752 /* Skip over composing char */
1753 #ifdef FEAT_ARABIC
1754 prevlen = len;
1755 #endif
1756 len += utf_ptr2len_len(p + len, size - len);
1758 return len;
1762 * Return the number of bytes the UTF-8 encoding of character "c" takes.
1763 * This does not include composing characters.
1766 utf_char2len(c)
1767 int c;
1769 if (c < 0x80)
1770 return 1;
1771 if (c < 0x800)
1772 return 2;
1773 if (c < 0x10000)
1774 return 3;
1775 if (c < 0x200000)
1776 return 4;
1777 if (c < 0x4000000)
1778 return 5;
1779 return 6;
1783 * Convert Unicode character "c" to UTF-8 string in "buf[]".
1784 * Returns the number of bytes.
1785 * This does not include composing characters.
1788 utf_char2bytes(c, buf)
1789 int c;
1790 char_u *buf;
1792 if (c < 0x80) /* 7 bits */
1794 buf[0] = c;
1795 return 1;
1797 if (c < 0x800) /* 11 bits */
1799 buf[0] = 0xc0 + ((unsigned)c >> 6);
1800 buf[1] = 0x80 + (c & 0x3f);
1801 return 2;
1803 if (c < 0x10000) /* 16 bits */
1805 buf[0] = 0xe0 + ((unsigned)c >> 12);
1806 buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f);
1807 buf[2] = 0x80 + (c & 0x3f);
1808 return 3;
1810 if (c < 0x200000) /* 21 bits */
1812 buf[0] = 0xf0 + ((unsigned)c >> 18);
1813 buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f);
1814 buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f);
1815 buf[3] = 0x80 + (c & 0x3f);
1816 return 4;
1818 if (c < 0x4000000) /* 26 bits */
1820 buf[0] = 0xf8 + ((unsigned)c >> 24);
1821 buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f);
1822 buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f);
1823 buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f);
1824 buf[4] = 0x80 + (c & 0x3f);
1825 return 5;
1827 /* 31 bits */
1828 buf[0] = 0xfc + ((unsigned)c >> 30);
1829 buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f);
1830 buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f);
1831 buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f);
1832 buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f);
1833 buf[5] = 0x80 + (c & 0x3f);
1834 return 6;
1838 * Return TRUE if "c" is a composing UTF-8 character. This means it will be
1839 * drawn on top of the preceding character.
1840 * Based on code from Markus Kuhn.
1843 utf_iscomposing(c)
1844 int c;
1846 /* sorted list of non-overlapping intervals */
1847 static struct interval combining[] =
1849 {0x0300, 0x034f}, {0x0360, 0x036f}, {0x0483, 0x0486}, {0x0488, 0x0489},
1850 {0x0591, 0x05a1}, {0x05a3, 0x05b9}, {0x05bb, 0x05bd}, {0x05bf, 0x05bf},
1851 {0x05c1, 0x05c2}, {0x05c4, 0x05c4}, {0x0610, 0x0615}, {0x064b, 0x0658},
1852 {0x0670, 0x0670}, {0x06d6, 0x06dc}, {0x06de, 0x06e4}, {0x06e7, 0x06e8},
1853 {0x06ea, 0x06ed}, {0x0711, 0x0711}, {0x0730, 0x074a}, {0x07a6, 0x07b0},
1854 {0x0901, 0x0903}, {0x093c, 0x093c}, {0x093e, 0x094d}, {0x0951, 0x0954},
1855 {0x0962, 0x0963}, {0x0981, 0x0983}, {0x09bc, 0x09bc}, {0x09be, 0x09c4},
1856 {0x09c7, 0x09c8}, {0x09cb, 0x09cd}, {0x09d7, 0x09d7}, {0x09e2, 0x09e3},
1857 {0x0a01, 0x0a03}, {0x0a3c, 0x0a3c}, {0x0a3e, 0x0a42}, {0x0a47, 0x0a48},
1858 {0x0a4b, 0x0a4d}, {0x0a70, 0x0a71}, {0x0a81, 0x0a83}, {0x0abc, 0x0abc},
1859 {0x0abe, 0x0ac5}, {0x0ac7, 0x0ac9}, {0x0acb, 0x0acd}, {0x0ae2, 0x0ae3},
1860 {0x0b01, 0x0b03}, {0x0b3c, 0x0b3c}, {0x0b3e, 0x0b43}, {0x0b47, 0x0b48},
1861 {0x0b4b, 0x0b4d}, {0x0b56, 0x0b57}, {0x0b82, 0x0b82}, {0x0bbe, 0x0bc2},
1862 {0x0bc6, 0x0bc8}, {0x0bca, 0x0bcd}, {0x0bd7, 0x0bd7}, {0x0c01, 0x0c03},
1863 {0x0c3e, 0x0c44}, {0x0c46, 0x0c48}, {0x0c4a, 0x0c4d}, {0x0c55, 0x0c56},
1864 {0x0c82, 0x0c83}, {0x0cbc, 0x0cbc}, {0x0cbe, 0x0cc4}, {0x0cc6, 0x0cc8},
1865 {0x0cca, 0x0ccd}, {0x0cd5, 0x0cd6}, {0x0d02, 0x0d03}, {0x0d3e, 0x0d43},
1866 {0x0d46, 0x0d48}, {0x0d4a, 0x0d4d}, {0x0d57, 0x0d57}, {0x0d82, 0x0d83},
1867 {0x0dca, 0x0dca}, {0x0dcf, 0x0dd4}, {0x0dd6, 0x0dd6}, {0x0dd8, 0x0ddf},
1868 {0x0df2, 0x0df3}, {0x0e31, 0x0e31}, {0x0e34, 0x0e3a}, {0x0e47, 0x0e4e},
1869 {0x0eb1, 0x0eb1}, {0x0eb4, 0x0eb9}, {0x0ebb, 0x0ebc}, {0x0ec8, 0x0ecd},
1870 {0x0f18, 0x0f19}, {0x0f35, 0x0f35}, {0x0f37, 0x0f37}, {0x0f39, 0x0f39},
1871 {0x0f3e, 0x0f3f}, {0x0f71, 0x0f84}, {0x0f86, 0x0f87}, {0x0f90, 0x0f97},
1872 {0x0f99, 0x0fbc}, {0x0fc6, 0x0fc6}, {0x102c, 0x1032}, {0x1036, 0x1039},
1873 {0x1056, 0x1059}, {0x1712, 0x1714}, {0x1732, 0x1734}, {0x1752, 0x1753},
1874 {0x1772, 0x1773}, {0x17b6, 0x17d3}, {0x17dd, 0x17dd}, {0x180b, 0x180d},
1875 {0x18a9, 0x18a9}, {0x1920, 0x192b}, {0x1930, 0x193b}, {0x20d0, 0x20ea},
1876 {0x302a, 0x302f}, {0x3099, 0x309a}, {0xfb1e, 0xfb1e}, {0xfe00, 0xfe0f},
1877 {0xfe20, 0xfe23},
1880 return intable(combining, sizeof(combining), c);
1884 * Return TRUE for characters that can be displayed in a normal way.
1885 * Only for characters of 0x100 and above!
1888 utf_printable(c)
1889 int c;
1891 #ifdef USE_WCHAR_FUNCTIONS
1893 * Assume the iswprint() library function works better than our own stuff.
1895 return iswprint(c);
1896 #else
1897 /* Sorted list of non-overlapping intervals.
1898 * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */
1899 static struct interval nonprint[] =
1901 {0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e},
1902 {0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb},
1903 {0xfffe, 0xffff}
1906 return !intable(nonprint, sizeof(nonprint), c);
1907 #endif
1911 * Get class of a Unicode character.
1912 * 0: white space
1913 * 1: punctuation
1914 * 2 or bigger: some class of word character.
1917 utf_class(c)
1918 int c;
1920 /* sorted list of non-overlapping intervals */
1921 static struct clinterval
1923 unsigned short first;
1924 unsigned short last;
1925 unsigned short class;
1926 } classes[] =
1928 {0x037e, 0x037e, 1}, /* Greek question mark */
1929 {0x0387, 0x0387, 1}, /* Greek ano teleia */
1930 {0x055a, 0x055f, 1}, /* Armenian punctuation */
1931 {0x0589, 0x0589, 1}, /* Armenian full stop */
1932 {0x05be, 0x05be, 1},
1933 {0x05c0, 0x05c0, 1},
1934 {0x05c3, 0x05c3, 1},
1935 {0x05f3, 0x05f4, 1},
1936 {0x060c, 0x060c, 1},
1937 {0x061b, 0x061b, 1},
1938 {0x061f, 0x061f, 1},
1939 {0x066a, 0x066d, 1},
1940 {0x06d4, 0x06d4, 1},
1941 {0x0700, 0x070d, 1}, /* Syriac punctuation */
1942 {0x0964, 0x0965, 1},
1943 {0x0970, 0x0970, 1},
1944 {0x0df4, 0x0df4, 1},
1945 {0x0e4f, 0x0e4f, 1},
1946 {0x0e5a, 0x0e5b, 1},
1947 {0x0f04, 0x0f12, 1},
1948 {0x0f3a, 0x0f3d, 1},
1949 {0x0f85, 0x0f85, 1},
1950 {0x104a, 0x104f, 1}, /* Myanmar punctuation */
1951 {0x10fb, 0x10fb, 1}, /* Georgian punctuation */
1952 {0x1361, 0x1368, 1}, /* Ethiopic punctuation */
1953 {0x166d, 0x166e, 1}, /* Canadian Syl. punctuation */
1954 {0x1680, 0x1680, 0},
1955 {0x169b, 0x169c, 1},
1956 {0x16eb, 0x16ed, 1},
1957 {0x1735, 0x1736, 1},
1958 {0x17d4, 0x17dc, 1}, /* Khmer punctuation */
1959 {0x1800, 0x180a, 1}, /* Mongolian punctuation */
1960 {0x2000, 0x200b, 0}, /* spaces */
1961 {0x200c, 0x2027, 1}, /* punctuation and symbols */
1962 {0x2028, 0x2029, 0},
1963 {0x202a, 0x202e, 1}, /* punctuation and symbols */
1964 {0x202f, 0x202f, 0},
1965 {0x2030, 0x205e, 1}, /* punctuation and symbols */
1966 {0x205f, 0x205f, 0},
1967 {0x2060, 0x27ff, 1}, /* punctuation and symbols */
1968 {0x2070, 0x207f, 0x2070}, /* superscript */
1969 {0x2080, 0x208f, 0x2080}, /* subscript */
1970 {0x2983, 0x2998, 1},
1971 {0x29d8, 0x29db, 1},
1972 {0x29fc, 0x29fd, 1},
1973 {0x3000, 0x3000, 0}, /* ideographic space */
1974 {0x3001, 0x3020, 1}, /* ideographic punctuation */
1975 {0x3030, 0x3030, 1},
1976 {0x303d, 0x303d, 1},
1977 {0x3040, 0x309f, 0x3040}, /* Hiragana */
1978 {0x30a0, 0x30ff, 0x30a0}, /* Katakana */
1979 {0x3300, 0x9fff, 0x4e00}, /* CJK Ideographs */
1980 {0xac00, 0xd7a3, 0xac00}, /* Hangul Syllables */
1981 {0xf900, 0xfaff, 0x4e00}, /* CJK Ideographs */
1982 {0xfd3e, 0xfd3f, 1},
1983 {0xfe30, 0xfe6b, 1}, /* punctuation forms */
1984 {0xff00, 0xff0f, 1}, /* half/fullwidth ASCII */
1985 {0xff1a, 0xff20, 1}, /* half/fullwidth ASCII */
1986 {0xff3b, 0xff40, 1}, /* half/fullwidth ASCII */
1987 {0xff5b, 0xff65, 1}, /* half/fullwidth ASCII */
1989 int bot = 0;
1990 int top = sizeof(classes) / sizeof(struct clinterval) - 1;
1991 int mid;
1993 /* First quick check for Latin1 characters, use 'iskeyword'. */
1994 if (c < 0x100)
1996 if (c == ' ' || c == '\t' || c == NUL || c == 0xa0)
1997 return 0; /* blank */
1998 if (vim_iswordc(c))
1999 return 2; /* word character */
2000 return 1; /* punctuation */
2003 /* binary search in table */
2004 while (top >= bot)
2006 mid = (bot + top) / 2;
2007 if (classes[mid].last < c)
2008 bot = mid + 1;
2009 else if (classes[mid].first > c)
2010 top = mid - 1;
2011 else
2012 return (int)classes[mid].class;
2015 /* most other characters are "word" characters */
2016 return 2;
2020 * Code for Unicode case-dependent operations. Based on notes in
2021 * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt
2022 * This code uses simple case folding, not full case folding.
2026 * The following table is built by foldExtract.pl < CaseFolding.txt .
2027 * It must be in numeric order, because we use binary search on it.
2028 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range
2029 * from 0x41 to 0x5a inclusive, stepping by 1, are folded by adding 32.
2032 typedef struct
2034 int rangeStart;
2035 int rangeEnd;
2036 int step;
2037 int offset;
2038 } convertStruct;
2040 static convertStruct foldCase[] =
2042 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32},
2043 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1},
2044 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121},
2045 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1},
2046 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205},
2047 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202},
2048 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205},
2049 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209},
2050 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213},
2051 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218},
2052 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1},
2053 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217},
2054 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1},
2055 {0x1c4,0x1c4,-1,2}, {0x1c5,0x1c5,-1,1}, {0x1c7,0x1c7,-1,2},
2056 {0x1c8,0x1c8,-1,1}, {0x1ca,0x1ca,-1,2}, {0x1cb,0x1db,2,1},
2057 {0x1de,0x1ee,2,1}, {0x1f1,0x1f1,-1,2}, {0x1f2,0x1f4,2,1},
2058 {0x1f6,0x1f6,-1,-97}, {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1},
2059 {0x220,0x220,-1,-130}, {0x222,0x232,2,1}, {0x386,0x386,-1,38},
2060 {0x388,0x38a,1,37}, {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63},
2061 {0x391,0x3a1,1,32}, {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1},
2062 {0x3f4,0x3f4,-1,-60}, {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7},
2063 {0x3fa,0x3fa,-1,1}, {0x400,0x40f,1,80}, {0x410,0x42f,1,32},
2064 {0x460,0x480,2,1}, {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1},
2065 {0x4d0,0x4f4,2,1}, {0x4f8,0x500,8,1}, {0x502,0x50e,2,1},
2066 {0x531,0x556,1,48}, {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1},
2067 {0x1f08,0x1f0f,1,-8}, {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8},
2068 {0x1f38,0x1f3f,1,-8}, {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8},
2069 {0x1f68,0x1f6f,1,-8}, {0x1f88,0x1f8f,1,-8}, {0x1f98,0x1f9f,1,-8},
2070 {0x1fa8,0x1faf,1,-8}, {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74},
2071 {0x1fbc,0x1fbc,-1,-9}, {0x1fc8,0x1fcb,1,-86}, {0x1fcc,0x1fcc,-1,-9},
2072 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8},
2073 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128},
2074 {0x1ffa,0x1ffb,1,-126}, {0x1ffc,0x1ffc,-1,-9}, {0x2126,0x2126,-1,-7517},
2075 {0x212a,0x212a,-1,-8383}, {0x212b,0x212b,-1,-8262},
2076 {0x2160,0x216f,1,16}, {0x24b6,0x24cf,1,26}, {0xff21,0xff3a,1,32},
2077 {0x10400,0x10427,1,40}
2080 static int utf_convert(int a, convertStruct table[], int tableSize);
2083 * Generic conversion function for case operations.
2084 * Return the converted equivalent of "a", which is a UCS-4 character. Use
2085 * the given conversion "table". Uses binary search on "table".
2087 static int
2088 utf_convert(a, table, tableSize)
2089 int a;
2090 convertStruct table[];
2091 int tableSize;
2093 int start, mid, end; /* indices into table */
2095 start = 0;
2096 end = tableSize / sizeof(convertStruct);
2097 while (start < end)
2099 /* need to search further */
2100 mid = (end + start) /2;
2101 if (table[mid].rangeEnd < a)
2102 start = mid + 1;
2103 else
2104 end = mid;
2106 if (table[start].rangeStart <= a && a <= table[start].rangeEnd
2107 && (a - table[start].rangeStart) % table[start].step == 0)
2108 return (a + table[start].offset);
2109 else
2110 return a;
2114 * Return the folded-case equivalent of "a", which is a UCS-4 character. Uses
2115 * simple case folding.
2118 utf_fold(a)
2119 int a;
2121 return utf_convert(a, foldCase, sizeof(foldCase));
2125 * The following tables are built by upperLowerExtract.pl < UnicodeData.txt .
2126 * They must be in numeric order, because we use binary search on them.
2127 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range
2128 * from 0x41 to 0x5a inclusive, stepping by 1, are switched to lower (for
2129 * example) by adding 32.
2131 static convertStruct toLower[] =
2133 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32},
2134 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1},
2135 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121},
2136 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1},
2137 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205},
2138 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202},
2139 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205},
2140 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209},
2141 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213},
2142 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218},
2143 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1},
2144 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217},
2145 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1},
2146 {0x1c4,0x1ca,3,2}, {0x1cd,0x1db,2,1}, {0x1de,0x1ee,2,1},
2147 {0x1f1,0x1f1,-1,2}, {0x1f4,0x1f4,-1,1}, {0x1f6,0x1f6,-1,-97},
2148 {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, {0x220,0x220,-1,-130},
2149 {0x222,0x232,2,1}, {0x386,0x386,-1,38}, {0x388,0x38a,1,37},
2150 {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, {0x391,0x3a1,1,32},
2151 {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, {0x3f4,0x3f4,-1,-60},
2152 {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, {0x3fa,0x3fa,-1,1},
2153 {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, {0x460,0x480,2,1},
2154 {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, {0x4d0,0x4f4,2,1},
2155 {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, {0x531,0x556,1,48},
2156 {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, {0x1f08,0x1f0f,1,-8},
2157 {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, {0x1f38,0x1f3f,1,-8},
2158 {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, {0x1f68,0x1f6f,1,-8},
2159 {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, {0x1fc8,0x1fcb,1,-86},
2160 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8},
2161 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128},
2162 {0x1ffa,0x1ffb,1,-126}, {0x2126,0x2126,-1,-7517}, {0x212a,0x212a,-1,-8383},
2163 {0x212b,0x212b,-1,-8262}, {0xff21,0xff3a,1,32}, {0x10400,0x10427,1,40}
2166 static convertStruct toUpper[] =
2168 {0x61,0x7a,1,-32}, {0xb5,0xb5,-1,743}, {0xe0,0xf6,1,-32},
2169 {0xf8,0xfe,1,-32}, {0xff,0xff,-1,121}, {0x101,0x12f,2,-1},
2170 {0x131,0x131,-1,-232}, {0x133,0x137,2,-1}, {0x13a,0x148,2,-1},
2171 {0x14b,0x177,2,-1}, {0x17a,0x17e,2,-1}, {0x17f,0x17f,-1,-300},
2172 {0x183,0x185,2,-1}, {0x188,0x18c,4,-1}, {0x192,0x192,-1,-1},
2173 {0x195,0x195,-1,97}, {0x199,0x199,-1,-1}, {0x19e,0x19e,-1,130},
2174 {0x1a1,0x1a5,2,-1}, {0x1a8,0x1ad,5,-1}, {0x1b0,0x1b4,4,-1},
2175 {0x1b6,0x1b9,3,-1}, {0x1bd,0x1bd,-1,-1}, {0x1bf,0x1bf,-1,56},
2176 {0x1c5,0x1c6,1,-1}, {0x1c8,0x1c9,1,-1}, {0x1cb,0x1cc,1,-1},
2177 {0x1ce,0x1dc,2,-1}, {0x1dd,0x1dd,-1,-79}, {0x1df,0x1ef,2,-1},
2178 {0x1f2,0x1f3,1,-1}, {0x1f5,0x1f9,4,-1}, {0x1fb,0x21f,2,-1},
2179 {0x223,0x233,2,-1}, {0x253,0x253,-1,-210}, {0x254,0x254,-1,-206},
2180 {0x256,0x257,1,-205}, {0x259,0x259,-1,-202}, {0x25b,0x25b,-1,-203},
2181 {0x260,0x260,-1,-205}, {0x263,0x263,-1,-207}, {0x268,0x268,-1,-209},
2182 {0x269,0x26f,6,-211}, {0x272,0x272,-1,-213}, {0x275,0x275,-1,-214},
2183 {0x280,0x283,3,-218}, {0x288,0x288,-1,-218}, {0x28a,0x28b,1,-217},
2184 {0x292,0x292,-1,-219}, {0x3ac,0x3ac,-1,-38}, {0x3ad,0x3af,1,-37},
2185 {0x3b1,0x3c1,1,-32}, {0x3c2,0x3c2,-1,-31}, {0x3c3,0x3cb,1,-32},
2186 {0x3cc,0x3cc,-1,-64}, {0x3cd,0x3ce,1,-63}, {0x3d0,0x3d0,-1,-62},
2187 {0x3d1,0x3d1,-1,-57}, {0x3d5,0x3d5,-1,-47}, {0x3d6,0x3d6,-1,-54},
2188 {0x3d9,0x3ef,2,-1}, {0x3f0,0x3f0,-1,-86}, {0x3f1,0x3f1,-1,-80},
2189 {0x3f2,0x3f2,-1,7}, {0x3f5,0x3f5,-1,-96}, {0x3f8,0x3fb,3,-1},
2190 {0x430,0x44f,1,-32}, {0x450,0x45f,1,-80}, {0x461,0x481,2,-1},
2191 {0x48b,0x4bf,2,-1}, {0x4c2,0x4ce,2,-1}, {0x4d1,0x4f5,2,-1},
2192 {0x4f9,0x501,8,-1}, {0x503,0x50f,2,-1}, {0x561,0x586,1,-48},
2193 {0x1e01,0x1e95,2,-1}, {0x1e9b,0x1e9b,-1,-59}, {0x1ea1,0x1ef9,2,-1},
2194 {0x1f00,0x1f07,1,8}, {0x1f10,0x1f15,1,8}, {0x1f20,0x1f27,1,8},
2195 {0x1f30,0x1f37,1,8}, {0x1f40,0x1f45,1,8}, {0x1f51,0x1f57,2,8},
2196 {0x1f60,0x1f67,1,8}, {0x1f70,0x1f71,1,74}, {0x1f72,0x1f75,1,86},
2197 {0x1f76,0x1f77,1,100}, {0x1f78,0x1f79,1,128}, {0x1f7a,0x1f7b,1,112},
2198 {0x1f7c,0x1f7d,1,126}, {0x1f80,0x1f87,1,8}, {0x1f90,0x1f97,1,8},
2199 {0x1fa0,0x1fa7,1,8}, {0x1fb0,0x1fb1,1,8}, {0x1fb3,0x1fb3,-1,9},
2200 {0x1fbe,0x1fbe,-1,-7205}, {0x1fc3,0x1fc3,-1,9}, {0x1fd0,0x1fd1,1,8},
2201 {0x1fe0,0x1fe1,1,8}, {0x1fe5,0x1fe5,-1,7}, {0x1ff3,0x1ff3,-1,9},
2202 {0xff41,0xff5a,1,-32}, {0x10428,0x1044f,1,-40}
2206 * Return the upper-case equivalent of "a", which is a UCS-4 character. Use
2207 * simple case folding.
2210 utf_toupper(a)
2211 int a;
2213 /* If 'casemap' contains "keepascii" use ASCII style toupper(). */
2214 if (a < 128 && (cmp_flags & CMP_KEEPASCII))
2215 return TOUPPER_ASC(a);
2217 #if defined(HAVE_TOWUPPER) && defined(__STDC_ISO_10646__)
2218 /* If towupper() is available and handles Unicode, use it. */
2219 if (!(cmp_flags & CMP_INTERNAL))
2220 return towupper(a);
2221 #endif
2223 /* For characters below 128 use locale sensitive toupper(). */
2224 if (a < 128)
2225 return TOUPPER_LOC(a);
2227 /* For any other characters use the above mapping table. */
2228 return utf_convert(a, toUpper, sizeof(toUpper));
2232 utf_islower(a)
2233 int a;
2235 return (utf_toupper(a) != a);
2239 * Return the lower-case equivalent of "a", which is a UCS-4 character. Use
2240 * simple case folding.
2243 utf_tolower(a)
2244 int a;
2246 /* If 'casemap' contains "keepascii" use ASCII style tolower(). */
2247 if (a < 128 && (cmp_flags & CMP_KEEPASCII))
2248 return TOLOWER_ASC(a);
2250 #if defined(HAVE_TOWLOWER) && defined(__STDC_ISO_10646__)
2251 /* If towlower() is available and handles Unicode, use it. */
2252 if (!(cmp_flags & CMP_INTERNAL))
2253 return towlower(a);
2254 #endif
2256 /* For characters below 128 use locale sensitive tolower(). */
2257 if (a < 128)
2258 return TOLOWER_LOC(a);
2260 /* For any other characters use the above mapping table. */
2261 return utf_convert(a, toLower, sizeof(toLower));
2265 utf_isupper(a)
2266 int a;
2268 return (utf_tolower(a) != a);
2272 * Version of strnicmp() that handles multi-byte characters.
2273 * Needed for Big5, Sjift-JIS and UTF-8 encoding. Other DBCS encodings can
2274 * probably use strnicmp(), because there are no ASCII characters in the
2275 * second byte.
2276 * Returns zero if s1 and s2 are equal (ignoring case), the difference between
2277 * two characters otherwise.
2280 mb_strnicmp(s1, s2, nn)
2281 char_u *s1, *s2;
2282 size_t nn;
2284 int i, j, l;
2285 int cdiff;
2286 int incomplete = FALSE;
2287 int n = (int)nn;
2289 for (i = 0; i < n; i += l)
2291 if (s1[i] == NUL && s2[i] == NUL) /* both strings end */
2292 return 0;
2293 if (enc_utf8)
2295 l = utf_byte2len(s1[i]);
2296 if (l > n - i)
2298 l = n - i; /* incomplete character */
2299 incomplete = TRUE;
2301 /* Check directly first, it's faster. */
2302 for (j = 0; j < l; ++j)
2304 if (s1[i + j] != s2[i + j])
2305 break;
2306 if (s1[i + j] == 0)
2307 /* Both stings have the same bytes but are incomplete or
2308 * have illegal bytes, accept them as equal. */
2309 l = j;
2311 if (j < l)
2313 /* If one of the two characters is incomplete return -1. */
2314 if (incomplete || i + utf_byte2len(s2[i]) > n)
2315 return -1;
2316 cdiff = utf_fold(utf_ptr2char(s1 + i))
2317 - utf_fold(utf_ptr2char(s2 + i));
2318 if (cdiff != 0)
2319 return cdiff;
2322 else
2324 l = (*mb_ptr2len)(s1 + i);
2325 if (l <= 1)
2327 /* Single byte: first check normally, then with ignore case. */
2328 if (s1[i] != s2[i])
2330 cdiff = MB_TOLOWER(s1[i]) - MB_TOLOWER(s2[i]);
2331 if (cdiff != 0)
2332 return cdiff;
2335 else
2337 /* For non-Unicode multi-byte don't ignore case. */
2338 if (l > n - i)
2339 l = n - i;
2340 cdiff = STRNCMP(s1 + i, s2 + i, l);
2341 if (cdiff != 0)
2342 return cdiff;
2346 return 0;
2350 * "g8": show bytes of the UTF-8 char under the cursor. Doesn't matter what
2351 * 'encoding' has been set to.
2353 void
2354 show_utf8()
2356 int len;
2357 int rlen = 0;
2358 char_u *line;
2359 int clen;
2360 int i;
2362 /* Get the byte length of the char under the cursor, including composing
2363 * characters. */
2364 line = ml_get_cursor();
2365 len = utfc_ptr2len(line);
2366 if (len == 0)
2368 MSG("NUL");
2369 return;
2372 clen = 0;
2373 for (i = 0; i < len; ++i)
2375 if (clen == 0)
2377 /* start of (composing) character, get its length */
2378 if (i > 0)
2380 STRCPY(IObuff + rlen, "+ ");
2381 rlen += 2;
2383 clen = utf_ptr2len(line + i);
2385 sprintf((char *)IObuff + rlen, "%02x ", line[i]);
2386 --clen;
2387 rlen += (int)STRLEN(IObuff + rlen);
2388 if (rlen > IOSIZE - 20)
2389 break;
2392 msg(IObuff);
2396 * mb_head_off() function pointer.
2397 * Return offset from "p" to the first byte of the character it points into.
2398 * Returns 0 when already at the first byte of a character.
2400 /*ARGSUSED*/
2402 latin_head_off(base, p)
2403 char_u *base;
2404 char_u *p;
2406 return 0;
2410 dbcs_head_off(base, p)
2411 char_u *base;
2412 char_u *p;
2414 char_u *q;
2416 /* It can't be a trailing byte when not using DBCS, at the start of the
2417 * string or the previous byte can't start a double-byte. */
2418 if (p <= base || MB_BYTE2LEN(p[-1]) == 1)
2419 return 0;
2421 /* This is slow: need to start at the base and go forward until the
2422 * byte we are looking for. Return 1 when we went past it, 0 otherwise. */
2423 q = base;
2424 while (q < p)
2425 q += dbcs_ptr2len(q);
2426 return (q == p) ? 0 : 1;
2429 #if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \
2430 || defined(PROTO)
2432 * Special version of dbcs_head_off() that works for ScreenLines[], where
2433 * single-width DBCS_JPNU characters are stored separately.
2436 dbcs_screen_head_off(base, p)
2437 char_u *base;
2438 char_u *p;
2440 char_u *q;
2442 /* It can't be a trailing byte when not using DBCS, at the start of the
2443 * string or the previous byte can't start a double-byte.
2444 * For euc-jp an 0x8e byte in the previous cell always means we have a
2445 * lead byte in the current cell. */
2446 if (p <= base
2447 || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e)
2448 || MB_BYTE2LEN(p[-1]) == 1)
2449 return 0;
2451 /* This is slow: need to start at the base and go forward until the
2452 * byte we are looking for. Return 1 when we went past it, 0 otherwise.
2453 * For DBCS_JPNU look out for 0x8e, which means the second byte is not
2454 * stored as the next byte. */
2455 q = base;
2456 while (q < p)
2458 if (enc_dbcs == DBCS_JPNU && *q == 0x8e)
2459 ++q;
2460 else
2461 q += dbcs_ptr2len(q);
2463 return (q == p) ? 0 : 1;
2465 #endif
2468 utf_head_off(base, p)
2469 char_u *base;
2470 char_u *p;
2472 char_u *q;
2473 char_u *s;
2474 int c;
2475 #ifdef FEAT_ARABIC
2476 char_u *j;
2477 #endif
2479 if (*p < 0x80) /* be quick for ASCII */
2480 return 0;
2482 /* Skip backwards over trailing bytes: 10xx.xxxx
2483 * Skip backwards again if on a composing char. */
2484 for (q = p; ; --q)
2486 /* Move s to the last byte of this char. */
2487 for (s = q; (s[1] & 0xc0) == 0x80; ++s)
2489 /* Move q to the first byte of this char. */
2490 while (q > base && (*q & 0xc0) == 0x80)
2491 --q;
2492 /* Check for illegal sequence. Do allow an illegal byte after where we
2493 * started. */
2494 if (utf8len_tab[*q] != (int)(s - q + 1)
2495 && utf8len_tab[*q] != (int)(p - q + 1))
2496 return 0;
2498 if (q <= base)
2499 break;
2501 c = utf_ptr2char(q);
2502 if (utf_iscomposing(c))
2503 continue;
2505 #ifdef FEAT_ARABIC
2506 if (arabic_maycombine(c))
2508 /* Advance to get a sneak-peak at the next char */
2509 j = q;
2510 --j;
2511 /* Move j to the first byte of this char. */
2512 while (j > base && (*j & 0xc0) == 0x80)
2513 --j;
2514 if (arabic_combine(utf_ptr2char(j), c))
2515 continue;
2517 #endif
2518 break;
2521 return (int)(p - q);
2524 #if defined(FEAT_EVAL) || defined(PROTO)
2526 * Copy a character from "*fp" to "*tp" and advance the pointers.
2528 void
2529 mb_copy_char(fp, tp)
2530 char_u **fp;
2531 char_u **tp;
2533 int l = (*mb_ptr2len)(*fp);
2535 mch_memmove(*tp, *fp, (size_t)l);
2536 *tp += l;
2537 *fp += l;
2539 #endif
2542 * Return the offset from "p" to the first byte of a character. When "p" is
2543 * at the start of a character 0 is returned, otherwise the offset to the next
2544 * character. Can start anywhere in a stream of bytes.
2547 mb_off_next(base, p)
2548 char_u *base;
2549 char_u *p;
2551 int i;
2552 int j;
2554 if (enc_utf8)
2556 if (*p < 0x80) /* be quick for ASCII */
2557 return 0;
2559 /* Find the next character that isn't 10xx.xxxx */
2560 for (i = 0; (p[i] & 0xc0) == 0x80; ++i)
2562 if (i > 0)
2564 /* Check for illegal sequence. */
2565 for (j = 0; p - j > base; ++j)
2566 if ((p[-j] & 0xc0) != 0x80)
2567 break;
2568 if (utf8len_tab[p[-j]] != i + j)
2569 return 0;
2571 return i;
2574 /* Only need to check if we're on a trail byte, it doesn't matter if we
2575 * want the offset to the next or current character. */
2576 return (*mb_head_off)(base, p);
2580 * Return the offset from "p" to the last byte of the character it points
2581 * into. Can start anywhere in a stream of bytes.
2584 mb_tail_off(base, p)
2585 char_u *base;
2586 char_u *p;
2588 int i;
2589 int j;
2591 if (*p == NUL)
2592 return 0;
2594 if (enc_utf8)
2596 /* Find the last character that is 10xx.xxxx */
2597 for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i)
2599 /* Check for illegal sequence. */
2600 for (j = 0; p - j > base; ++j)
2601 if ((p[-j] & 0xc0) != 0x80)
2602 break;
2603 if (utf8len_tab[p[-j]] != i + j + 1)
2604 return 0;
2605 return i;
2608 /* It can't be the first byte if a double-byte when not using DBCS, at the
2609 * end of the string or the byte can't start a double-byte. */
2610 if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1)
2611 return 0;
2613 /* Return 1 when on the lead byte, 0 when on the tail byte. */
2614 return 1 - dbcs_head_off(base, p);
2618 * Find the next illegal byte sequence.
2620 void
2621 utf_find_illegal()
2623 pos_T pos = curwin->w_cursor;
2624 char_u *p;
2625 int len;
2626 vimconv_T vimconv;
2627 char_u *tofree = NULL;
2629 vimconv.vc_type = CONV_NONE;
2630 if (enc_utf8 && (enc_canon_props(curbuf->b_p_fenc) & ENC_8BIT))
2632 /* 'encoding' is "utf-8" but we are editing a 8-bit encoded file,
2633 * possibly a utf-8 file with illegal bytes. Setup for conversion
2634 * from utf-8 to 'fileencoding'. */
2635 convert_setup(&vimconv, p_enc, curbuf->b_p_fenc);
2638 #ifdef FEAT_VIRTUALEDIT
2639 curwin->w_cursor.coladd = 0;
2640 #endif
2641 for (;;)
2643 p = ml_get_cursor();
2644 if (vimconv.vc_type != CONV_NONE)
2646 vim_free(tofree);
2647 tofree = string_convert(&vimconv, p, NULL);
2648 if (tofree == NULL)
2649 break;
2650 p = tofree;
2653 while (*p != NUL)
2655 /* Illegal means that there are not enough trail bytes (checked by
2656 * utf_ptr2len()) or too many of them (overlong sequence). */
2657 len = utf_ptr2len(p);
2658 if (*p >= 0x80 && (len == 1
2659 || utf_char2len(utf_ptr2char(p)) != len))
2661 if (vimconv.vc_type == CONV_NONE)
2662 curwin->w_cursor.col += (colnr_T)(p - ml_get_cursor());
2663 else
2665 int l;
2667 len = (int)(p - tofree);
2668 for (p = ml_get_cursor(); *p != NUL && len-- > 0; p += l)
2670 l = utf_ptr2len(p);
2671 curwin->w_cursor.col += l;
2674 goto theend;
2676 p += len;
2678 if (curwin->w_cursor.lnum == curbuf->b_ml.ml_line_count)
2679 break;
2680 ++curwin->w_cursor.lnum;
2681 curwin->w_cursor.col = 0;
2684 /* didn't find it: don't move and beep */
2685 curwin->w_cursor = pos;
2686 beep_flush();
2688 theend:
2689 vim_free(tofree);
2690 convert_setup(&vimconv, NULL, NULL);
2693 #if defined(HAVE_GTK2) || defined(PROTO)
2695 * Return TRUE if string "s" is a valid utf-8 string.
2696 * When "end" is NULL stop at the first NUL.
2697 * When "end" is positive stop there.
2700 utf_valid_string(s, end)
2701 char_u *s;
2702 char_u *end;
2704 int l;
2705 char_u *p = s;
2707 while (end == NULL ? *p != NUL : p < end)
2709 if ((*p & 0xc0) == 0x80)
2710 return FALSE; /* invalid lead byte */
2711 l = utf8len_tab[*p];
2712 if (end != NULL && p + l > end)
2713 return FALSE; /* incomplete byte sequence */
2714 ++p;
2715 while (--l > 0)
2716 if ((*p++ & 0xc0) != 0x80)
2717 return FALSE; /* invalid trail byte */
2719 return TRUE;
2721 #endif
2723 #if defined(FEAT_GUI) || defined(PROTO)
2725 * Special version of mb_tail_off() for use in ScreenLines[].
2728 dbcs_screen_tail_off(base, p)
2729 char_u *base;
2730 char_u *p;
2732 /* It can't be the first byte if a double-byte when not using DBCS, at the
2733 * end of the string or the byte can't start a double-byte.
2734 * For euc-jp an 0x8e byte always means we have a lead byte in the current
2735 * cell. */
2736 if (*p == NUL || p[1] == NUL
2737 || (enc_dbcs == DBCS_JPNU && *p == 0x8e)
2738 || MB_BYTE2LEN(*p) == 1)
2739 return 0;
2741 /* Return 1 when on the lead byte, 0 when on the tail byte. */
2742 return 1 - dbcs_screen_head_off(base, p);
2744 #endif
2747 * If the cursor moves on an trail byte, set the cursor on the lead byte.
2748 * Thus it moves left if necessary.
2749 * Return TRUE when the cursor was adjusted.
2751 void
2752 mb_adjust_cursor()
2754 mb_adjustpos(&curwin->w_cursor);
2758 * Adjust position "*lp" to point to the first byte of a multi-byte character.
2759 * If it points to a tail byte it's moved backwards to the head byte.
2761 void
2762 mb_adjustpos(lp)
2763 pos_T *lp;
2765 char_u *p;
2767 if (lp->col > 0
2768 #ifdef FEAT_VIRTUALEDIT
2769 || lp->coladd > 1
2770 #endif
2773 p = ml_get(lp->lnum);
2774 lp->col -= (*mb_head_off)(p, p + lp->col);
2775 #ifdef FEAT_VIRTUALEDIT
2776 /* Reset "coladd" when the cursor would be on the right half of a
2777 * double-wide character. */
2778 if (lp->coladd == 1
2779 && p[lp->col] != TAB
2780 && vim_isprintc((*mb_ptr2char)(p + lp->col))
2781 && ptr2cells(p + lp->col) > 1)
2782 lp->coladd = 0;
2783 #endif
2788 * Return a pointer to the character before "*p", if there is one.
2790 char_u *
2791 mb_prevptr(line, p)
2792 char_u *line; /* start of the string */
2793 char_u *p;
2795 if (p > line)
2796 mb_ptr_back(line, p);
2797 return p;
2801 * Return the character length of "str". Each multi-byte character (with
2802 * following composing characters) counts as one.
2805 mb_charlen(str)
2806 char_u *str;
2808 char_u *p = str;
2809 int count;
2811 if (p == NULL)
2812 return 0;
2814 for (count = 0; *p != NUL; count++)
2815 p += (*mb_ptr2len)(p);
2817 return count;
2820 #if defined(FEAT_SPELL) || defined(PROTO)
2822 * Like mb_charlen() but for a string with specified length.
2825 mb_charlen_len(str, len)
2826 char_u *str;
2827 int len;
2829 char_u *p = str;
2830 int count;
2832 for (count = 0; *p != NUL && p < str + len; count++)
2833 p += (*mb_ptr2len)(p);
2835 return count;
2837 #endif
2840 * Try to un-escape a multi-byte character.
2841 * Used for the "to" and "from" part of a mapping.
2842 * Return the un-escaped string if it is a multi-byte character, and advance
2843 * "pp" to just after the bytes that formed it.
2844 * Return NULL if no multi-byte char was found.
2846 char_u *
2847 mb_unescape(pp)
2848 char_u **pp;
2850 static char_u buf[MB_MAXBYTES + 1];
2851 int n, m = 0;
2852 char_u *str = *pp;
2854 /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI
2855 * KS_EXTRA KE_CSI to CSI. */
2856 for (n = 0; str[n] != NUL && m <= MB_MAXBYTES; ++n)
2858 if (str[n] == K_SPECIAL
2859 && str[n + 1] == KS_SPECIAL
2860 && str[n + 2] == KE_FILLER)
2862 buf[m++] = K_SPECIAL;
2863 n += 2;
2865 # ifdef FEAT_GUI
2866 else if (str[n] == CSI
2867 && str[n + 1] == KS_EXTRA
2868 && str[n + 2] == (int)KE_CSI)
2870 buf[m++] = CSI;
2871 n += 2;
2873 # endif
2874 else if (str[n] == K_SPECIAL
2875 # ifdef FEAT_GUI
2876 || str[n] == CSI
2877 # endif
2879 break; /* a special key can't be a multibyte char */
2880 else
2881 buf[m++] = str[n];
2882 buf[m] = NUL;
2884 /* Return a multi-byte character if it's found. An illegal sequence
2885 * will result in a 1 here. */
2886 if ((*mb_ptr2len)(buf) > 1)
2888 *pp = str + n + 1;
2889 return buf;
2892 return NULL;
2896 * Return TRUE if the character at "row"/"col" on the screen is the left side
2897 * of a double-width character.
2898 * Caller must make sure "row" and "col" are not invalid!
2901 mb_lefthalve(row, col)
2902 int row;
2903 int col;
2905 #ifdef FEAT_HANGULIN
2906 if (composing_hangul)
2907 return TRUE;
2908 #endif
2909 return (*mb_off2cells)(LineOffset[row] + col,
2910 LineOffset[row] + screen_Columns) > 1;
2913 # if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \
2914 || defined(PROTO)
2916 * Correct a position on the screen, if it's the right halve of a double-wide
2917 * char move it to the left halve. Returns the corrected column.
2920 mb_fix_col(col, row)
2921 int col;
2922 int row;
2924 col = check_col(col);
2925 row = check_row(row);
2926 if (has_mbyte && ScreenLines != NULL && col > 0
2927 && ((enc_dbcs
2928 && ScreenLines[LineOffset[row] + col] != NUL
2929 && dbcs_screen_head_off(ScreenLines + LineOffset[row],
2930 ScreenLines + LineOffset[row] + col))
2931 || (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0)))
2932 --col;
2933 return col;
2935 # endif
2936 #endif
2938 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO)
2939 static int enc_alias_search __ARGS((char_u *name));
2942 * Skip the Vim specific head of a 'encoding' name.
2944 char_u *
2945 enc_skip(p)
2946 char_u *p;
2948 if (STRNCMP(p, "2byte-", 6) == 0)
2949 return p + 6;
2950 if (STRNCMP(p, "8bit-", 5) == 0)
2951 return p + 5;
2952 return p;
2956 * Find the canonical name for encoding "enc".
2957 * When the name isn't recognized, returns "enc" itself, but with all lower
2958 * case characters and '_' replaced with '-'.
2959 * Returns an allocated string. NULL for out-of-memory.
2961 char_u *
2962 enc_canonize(enc)
2963 char_u *enc;
2965 char_u *r;
2966 char_u *p, *s;
2967 int i;
2969 # ifdef FEAT_MBYTE
2970 if (STRCMP(enc, "default") == 0)
2972 /* Use the default encoding as it's found by set_init_1(). */
2973 r = get_encoding_default();
2974 if (r == NULL)
2975 r = (char_u *)"latin1";
2976 return vim_strsave(r);
2978 # endif
2980 /* copy "enc" to allocated memory, with room for two '-' */
2981 r = alloc((unsigned)(STRLEN(enc) + 3));
2982 if (r != NULL)
2984 /* Make it all lower case and replace '_' with '-'. */
2985 p = r;
2986 for (s = enc; *s != NUL; ++s)
2988 if (*s == '_')
2989 *p++ = '-';
2990 else
2991 *p++ = TOLOWER_ASC(*s);
2993 *p = NUL;
2995 /* Skip "2byte-" and "8bit-". */
2996 p = enc_skip(r);
2998 /* Change "microsoft-cp" to "cp". Used in some spell files. */
2999 if (STRNCMP(p, "microsoft-cp", 12) == 0)
3000 mch_memmove(p, p + 10, STRLEN(p + 10) + 1);
3002 /* "iso8859" -> "iso-8859" */
3003 if (STRNCMP(p, "iso8859", 7) == 0)
3005 mch_memmove(p + 4, p + 3, STRLEN(p + 2));
3006 p[3] = '-';
3009 /* "iso-8859n" -> "iso-8859-n" */
3010 if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-')
3012 mch_memmove(p + 9, p + 8, STRLEN(p + 7));
3013 p[8] = '-';
3016 /* "latin-N" -> "latinN" */
3017 if (STRNCMP(p, "latin-", 6) == 0)
3018 mch_memmove(p + 5, p + 6, STRLEN(p + 5));
3020 if (enc_canon_search(p) >= 0)
3022 /* canonical name can be used unmodified */
3023 if (p != r)
3024 mch_memmove(r, p, STRLEN(p) + 1);
3026 else if ((i = enc_alias_search(p)) >= 0)
3028 /* alias recognized, get canonical name */
3029 vim_free(r);
3030 r = vim_strsave((char_u *)enc_canon_table[i].name);
3033 return r;
3037 * Search for an encoding alias of "name".
3038 * Returns -1 when not found.
3040 static int
3041 enc_alias_search(name)
3042 char_u *name;
3044 int i;
3046 for (i = 0; enc_alias_table[i].name != NULL; ++i)
3047 if (STRCMP(name, enc_alias_table[i].name) == 0)
3048 return enc_alias_table[i].canon;
3049 return -1;
3051 #endif
3053 #if defined(FEAT_MBYTE) || defined(PROTO)
3055 #ifdef HAVE_LANGINFO_H
3056 # include <langinfo.h>
3057 #endif
3060 * Get the canonicalized encoding of the current locale.
3061 * Returns an allocated string when successful, NULL when not.
3063 char_u *
3064 enc_locale()
3066 #ifndef WIN3264
3067 char *s;
3068 char *p;
3069 int i;
3070 #endif
3071 char buf[50];
3072 #ifdef WIN3264
3073 long acp = GetACP();
3075 if (acp == 1200)
3076 STRCPY(buf, "ucs-2le");
3077 else if (acp == 1252) /* cp1252 is used as latin1 */
3078 STRCPY(buf, "latin1");
3079 else
3080 sprintf(buf, "cp%ld", acp);
3081 #else
3082 # ifdef HAVE_NL_LANGINFO_CODESET
3083 if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL)
3084 # endif
3085 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE)
3086 if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL)
3087 # endif
3088 if ((s = getenv("LC_ALL")) == NULL || *s == NUL)
3089 if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL)
3090 s = getenv("LANG");
3092 if (s == NULL || *s == NUL)
3093 return FAIL;
3095 /* The most generic locale format is:
3096 * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]]
3097 * If there is a '.' remove the part before it.
3098 * if there is something after the codeset, remove it.
3099 * Make the name lowercase and replace '_' with '-'.
3100 * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn",
3101 * "ko_KR.EUC" == "euc-kr"
3103 if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL)
3105 if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0
3106 && !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_')
3108 /* copy "XY.EUC" to "euc-XY" to buf[10] */
3109 STRCPY(buf + 10, "euc-");
3110 buf[14] = p[-2];
3111 buf[15] = p[-1];
3112 buf[16] = 0;
3113 s = buf + 10;
3115 else
3116 s = p + 1;
3118 for (i = 0; s[i] != NUL && s + i < buf + sizeof(buf) - 1; ++i)
3120 if (s[i] == '_' || s[i] == '-')
3121 buf[i] = '-';
3122 else if (isalnum((int)s[i]))
3123 buf[i] = TOLOWER_ASC(s[i]);
3124 else
3125 break;
3127 buf[i] = NUL;
3128 #endif
3130 return enc_canonize((char_u *)buf);
3133 #if defined(WIN3264) || defined(PROTO)
3135 * Convert an encoding name to an MS-Windows codepage.
3136 * Returns zero if no codepage can be figured out.
3139 encname2codepage(name)
3140 char_u *name;
3142 int cp;
3143 char_u *p = name;
3144 int idx;
3146 if (STRNCMP(p, "8bit-", 5) == 0)
3147 p += 5;
3148 else if (STRNCMP(p_enc, "2byte-", 6) == 0)
3149 p += 6;
3151 if (p[0] == 'c' && p[1] == 'p')
3152 cp = atoi(p + 2);
3153 else if ((idx = enc_canon_search(p)) >= 0)
3154 cp = enc_canon_table[idx].codepage;
3155 else
3156 return 0;
3157 if (IsValidCodePage(cp))
3158 return cp;
3159 return 0;
3161 #endif
3163 # if defined(USE_ICONV) || defined(PROTO)
3165 static char_u *iconv_string __ARGS((vimconv_T *vcp, char_u *str, int slen, int *unconvlenp));
3168 * Call iconv_open() with a check if iconv() works properly (there are broken
3169 * versions).
3170 * Returns (void *)-1 if failed.
3171 * (should return iconv_t, but that causes problems with prototypes).
3173 void *
3174 my_iconv_open(to, from)
3175 char_u *to;
3176 char_u *from;
3178 iconv_t fd;
3179 #define ICONV_TESTLEN 400
3180 char_u tobuf[ICONV_TESTLEN];
3181 char *p;
3182 size_t tolen;
3183 static int iconv_ok = -1;
3185 if (iconv_ok == FALSE)
3186 return (void *)-1; /* detected a broken iconv() previously */
3188 #ifdef DYNAMIC_ICONV
3189 /* Check if the iconv.dll can be found. */
3190 if (!iconv_enabled(TRUE))
3191 return (void *)-1;
3192 #endif
3194 fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from));
3196 if (fd != (iconv_t)-1 && iconv_ok == -1)
3199 * Do a dummy iconv() call to check if it actually works. There is a
3200 * version of iconv() on Linux that is broken. We can't ignore it,
3201 * because it's wide-spread. The symptoms are that after outputting
3202 * the initial shift state the "to" pointer is NULL and conversion
3203 * stops for no apparent reason after about 8160 characters.
3205 p = (char *)tobuf;
3206 tolen = ICONV_TESTLEN;
3207 (void)iconv(fd, NULL, NULL, &p, &tolen);
3208 if (p == NULL)
3210 iconv_ok = FALSE;
3211 iconv_close(fd);
3212 fd = (iconv_t)-1;
3214 else
3215 iconv_ok = TRUE;
3218 return (void *)fd;
3222 * Convert the string "str[slen]" with iconv().
3223 * If "unconvlenp" is not NULL handle the string ending in an incomplete
3224 * sequence and set "*unconvlenp" to the length of it.
3225 * Returns the converted string in allocated memory. NULL for an error.
3227 static char_u *
3228 iconv_string(vcp, str, slen, unconvlenp)
3229 vimconv_T *vcp;
3230 char_u *str;
3231 int slen;
3232 int *unconvlenp;
3234 const char *from;
3235 size_t fromlen;
3236 char *to;
3237 size_t tolen;
3238 size_t len = 0;
3239 size_t done = 0;
3240 char_u *result = NULL;
3241 char_u *p;
3242 int l;
3244 from = (char *)str;
3245 fromlen = slen;
3246 for (;;)
3248 if (len == 0 || ICONV_ERRNO == ICONV_E2BIG)
3250 /* Allocate enough room for most conversions. When re-allocating
3251 * increase the buffer size. */
3252 len = len + fromlen * 2 + 40;
3253 p = alloc((unsigned)len);
3254 if (p != NULL && done > 0)
3255 mch_memmove(p, result, done);
3256 vim_free(result);
3257 result = p;
3258 if (result == NULL) /* out of memory */
3259 break;
3262 to = (char *)result + done;
3263 tolen = len - done - 2;
3264 /* Avoid a warning for systems with a wrong iconv() prototype by
3265 * casting the second argument to void *. */
3266 if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen)
3267 != (size_t)-1)
3269 /* Finished, append a NUL. */
3270 *to = NUL;
3271 break;
3274 /* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded
3275 * iconv library may use one of them. */
3276 if (!vcp->vc_fail && unconvlenp != NULL
3277 && (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
3279 /* Handle an incomplete sequence at the end. */
3280 *to = NUL;
3281 *unconvlenp = (int)fromlen;
3282 break;
3285 /* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded
3286 * iconv library may use one of them. */
3287 else if (!vcp->vc_fail
3288 && (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ
3289 || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
3291 /* Can't convert: insert a '?' and skip a character. This assumes
3292 * conversion from 'encoding' to something else. In other
3293 * situations we don't know what to skip anyway. */
3294 *to++ = '?';
3295 if ((*mb_ptr2cells)((char_u *)from) > 1)
3296 *to++ = '?';
3297 if (enc_utf8)
3298 l = utfc_ptr2len_len((char_u *)from, (int)fromlen);
3299 else
3301 l = (*mb_ptr2len)((char_u *)from);
3302 if (l > (int)fromlen)
3303 l = (int)fromlen;
3305 from += l;
3306 fromlen -= l;
3308 else if (ICONV_ERRNO != ICONV_E2BIG)
3310 /* conversion failed */
3311 vim_free(result);
3312 result = NULL;
3313 break;
3315 /* Not enough room or skipping illegal sequence. */
3316 done = to - (char *)result;
3318 return result;
3321 # if defined(DYNAMIC_ICONV) || defined(PROTO)
3323 * Dynamically load the "iconv.dll" on Win32.
3326 #ifndef DYNAMIC_ICONV /* just generating prototypes */
3327 # define HINSTANCE int
3328 #endif
3329 static HINSTANCE hIconvDLL = 0;
3330 static HINSTANCE hMsvcrtDLL = 0;
3332 # ifndef DYNAMIC_ICONV_DLL
3333 # define DYNAMIC_ICONV_DLL "iconv.dll"
3334 # define DYNAMIC_ICONV_DLL_ALT "libiconv.dll"
3335 # endif
3336 # ifndef DYNAMIC_MSVCRT_DLL
3337 # define DYNAMIC_MSVCRT_DLL "msvcrt.dll"
3338 # endif
3341 * Try opening the iconv.dll and return TRUE if iconv() can be used.
3344 iconv_enabled(verbose)
3345 int verbose;
3347 if (hIconvDLL != 0 && hMsvcrtDLL != 0)
3348 return TRUE;
3349 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL);
3350 if (hIconvDLL == 0) /* sometimes it's called libiconv.dll */
3351 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL_ALT);
3352 if (hIconvDLL != 0)
3353 hMsvcrtDLL = LoadLibrary(DYNAMIC_MSVCRT_DLL);
3354 if (hIconvDLL == 0 || hMsvcrtDLL == 0)
3356 /* Only give the message when 'verbose' is set, otherwise it might be
3357 * done whenever a conversion is attempted. */
3358 if (verbose && p_verbose > 0)
3360 verbose_enter();
3361 EMSG2(_(e_loadlib),
3362 hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL);
3363 verbose_leave();
3365 iconv_end();
3366 return FALSE;
3369 iconv = (void *)GetProcAddress(hIconvDLL, "libiconv");
3370 iconv_open = (void *)GetProcAddress(hIconvDLL, "libiconv_open");
3371 iconv_close = (void *)GetProcAddress(hIconvDLL, "libiconv_close");
3372 iconvctl = (void *)GetProcAddress(hIconvDLL, "libiconvctl");
3373 iconv_errno = (void *)GetProcAddress(hMsvcrtDLL, "_errno");
3374 if (iconv == NULL || iconv_open == NULL || iconv_close == NULL
3375 || iconvctl == NULL || iconv_errno == NULL)
3377 iconv_end();
3378 if (verbose && p_verbose > 0)
3380 verbose_enter();
3381 EMSG2(_(e_loadfunc), "for libiconv");
3382 verbose_leave();
3384 return FALSE;
3386 return TRUE;
3389 void
3390 iconv_end()
3392 /* Don't use iconv() when inputting or outputting characters. */
3393 if (input_conv.vc_type == CONV_ICONV)
3394 convert_setup(&input_conv, NULL, NULL);
3395 if (output_conv.vc_type == CONV_ICONV)
3396 convert_setup(&output_conv, NULL, NULL);
3398 if (hIconvDLL != 0)
3399 FreeLibrary(hIconvDLL);
3400 if (hMsvcrtDLL != 0)
3401 FreeLibrary(hMsvcrtDLL);
3402 hIconvDLL = 0;
3403 hMsvcrtDLL = 0;
3405 # endif /* DYNAMIC_ICONV */
3406 # endif /* USE_ICONV */
3408 #endif /* FEAT_MBYTE */
3410 #if defined(FEAT_XIM) || defined(PROTO)
3412 # ifdef FEAT_GUI_GTK
3413 static int xim_has_preediting INIT(= FALSE); /* IM current status */
3416 * Set preedit_start_col to the current cursor position.
3418 static void
3419 init_preedit_start_col(void)
3421 if (State & CMDLINE)
3422 preedit_start_col = cmdline_getvcol_cursor();
3423 else if (curwin != NULL)
3424 getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL);
3425 /* Prevent that preediting marks the buffer as changed. */
3426 xim_changed_while_preediting = curbuf->b_changed;
3428 # endif
3430 # if defined(HAVE_GTK2) && !defined(PROTO)
3432 static int im_is_active = FALSE; /* IM is enabled for current mode */
3433 static int im_preedit_cursor = 0; /* cursor offset in characters */
3434 static int im_preedit_trailing = 0; /* number of characters after cursor */
3436 static unsigned long im_commit_handler_id = 0;
3437 static unsigned int im_activatekey_keyval = GDK_VoidSymbol;
3438 static unsigned int im_activatekey_state = 0;
3440 void
3441 im_set_active(int active)
3443 int was_active;
3445 was_active = !!im_is_active;
3446 im_is_active = (active && !p_imdisable);
3448 if (im_is_active != was_active)
3449 xim_reset();
3452 void
3453 xim_set_focus(int focus)
3455 if (xic != NULL)
3457 if (focus)
3458 gtk_im_context_focus_in(xic);
3459 else
3460 gtk_im_context_focus_out(xic);
3464 void
3465 im_set_position(int row, int col)
3467 if (xic != NULL)
3469 GdkRectangle area;
3471 area.x = FILL_X(col);
3472 area.y = FILL_Y(row);
3473 area.width = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1);
3474 area.height = gui.char_height;
3476 gtk_im_context_set_cursor_location(xic, &area);
3480 # if 0 || defined(PROTO) /* apparently only used in gui_x11.c */
3481 void
3482 xim_set_preedit(void)
3484 im_set_position(gui.row, gui.col);
3486 # endif
3488 static void
3489 im_add_to_input(char_u *str, int len)
3491 /* Convert from 'termencoding' (always "utf-8") to 'encoding' */
3492 if (input_conv.vc_type != CONV_NONE)
3494 str = string_convert(&input_conv, str, &len);
3495 g_return_if_fail(str != NULL);
3498 add_to_input_buf_csi(str, len);
3500 if (input_conv.vc_type != CONV_NONE)
3501 vim_free(str);
3503 if (p_mh) /* blank out the pointer if necessary */
3504 gui_mch_mousehide(TRUE);
3507 static void
3508 im_delete_preedit(void)
3510 char_u bskey[] = {CSI, 'k', 'b'};
3511 char_u delkey[] = {CSI, 'k', 'D'};
3513 if (State & NORMAL)
3515 im_preedit_cursor = 0;
3516 return;
3518 for (; im_preedit_cursor > 0; --im_preedit_cursor)
3519 add_to_input_buf(bskey, (int)sizeof(bskey));
3521 for (; im_preedit_trailing > 0; --im_preedit_trailing)
3522 add_to_input_buf(delkey, (int)sizeof(delkey));
3526 * Move the cursor left by "num_move_back" characters.
3527 * Note that ins_left() checks im_is_preediting() to avoid breaking undo for
3528 * these K_LEFT keys.
3530 static void
3531 im_correct_cursor(int num_move_back)
3533 char_u backkey[] = {CSI, 'k', 'l'};
3535 if (State & NORMAL)
3536 return;
3537 # ifdef FEAT_RIGHTLEFT
3538 if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl)
3539 backkey[2] = 'r';
3540 # endif
3541 for (; num_move_back > 0; --num_move_back)
3542 add_to_input_buf(backkey, (int)sizeof(backkey));
3545 static int xim_expected_char = NUL;
3546 static int xim_ignored_char = FALSE;
3549 * Update the mode and cursor while in an IM callback.
3551 static void
3552 im_show_info(void)
3554 int old_vgetc_busy;
3556 old_vgetc_busy = vgetc_busy;
3557 vgetc_busy = TRUE;
3558 showmode();
3559 vgetc_busy = old_vgetc_busy;
3560 setcursor();
3561 out_flush();
3565 * Callback invoked when the user finished preediting.
3566 * Put the final string into the input buffer.
3568 /*ARGSUSED0*/
3569 static void
3570 im_commit_cb(GtkIMContext *context, const gchar *str, gpointer data)
3572 int slen = (int)STRLEN(str);
3573 int add_to_input = TRUE;
3574 int clen;
3575 int len = slen;
3576 int commit_with_preedit = TRUE;
3577 char_u *im_str, *p;
3579 #ifdef XIM_DEBUG
3580 xim_log("im_commit_cb(): %s\n", str);
3581 #endif
3583 /* The imhangul module doesn't reset the preedit string before
3584 * committing. Call im_delete_preedit() to work around that. */
3585 im_delete_preedit();
3587 /* Indicate that preediting has finished. */
3588 if (preedit_start_col == MAXCOL)
3590 init_preedit_start_col();
3591 commit_with_preedit = FALSE;
3594 /* The thing which setting "preedit_start_col" to MAXCOL means that
3595 * "preedit_start_col" will be set forcely when calling
3596 * preedit_changed_cb() next time.
3597 * "preedit_start_col" should not reset with MAXCOL on this part. Vim
3598 * is simulating the preediting by using add_to_input_str(). when
3599 * preedit begin immediately before committed, the typebuf is not
3600 * flushed to screen, then it can't get correct "preedit_start_col".
3601 * Thus, it should calculate the cells by adding cells of the committed
3602 * string. */
3603 if (input_conv.vc_type != CONV_NONE)
3605 im_str = string_convert(&input_conv, (char_u *)str, &len);
3606 g_return_if_fail(im_str != NULL);
3608 else
3609 im_str = (char_u *)str;
3610 clen = 0;
3611 for (p = im_str; p < im_str + len; p += (*mb_ptr2len)(p))
3612 clen += (*mb_ptr2cells)(p);
3613 if (input_conv.vc_type != CONV_NONE)
3614 vim_free(im_str);
3615 preedit_start_col += clen;
3617 /* Is this a single character that matches a keypad key that's just
3618 * been pressed? If so, we don't want it to be entered as such - let
3619 * us carry on processing the raw keycode so that it may be used in
3620 * mappings as <kSomething>. */
3621 if (xim_expected_char != NUL)
3623 /* We're currently processing a keypad or other special key */
3624 if (slen == 1 && str[0] == xim_expected_char)
3626 /* It's a match - don't do it here */
3627 xim_ignored_char = TRUE;
3628 add_to_input = FALSE;
3630 else
3632 /* Not a match */
3633 xim_ignored_char = FALSE;
3637 if (add_to_input)
3638 im_add_to_input((char_u *)str, slen);
3640 /* Inserting chars while "im_is_active" is set does not cause a change of
3641 * buffer. When the chars are committed the buffer must be marked as
3642 * changed. */
3643 if (!commit_with_preedit)
3644 preedit_start_col = MAXCOL;
3646 /* This flag is used in changed() at next call. */
3647 xim_changed_while_preediting = TRUE;
3649 if (gtk_main_level() > 0)
3650 gtk_main_quit();
3654 * Callback invoked after start to the preedit.
3656 /*ARGSUSED*/
3657 static void
3658 im_preedit_start_cb(GtkIMContext *context, gpointer data)
3660 #ifdef XIM_DEBUG
3661 xim_log("im_preedit_start_cb()\n");
3662 #endif
3664 im_is_active = TRUE;
3665 gui_update_cursor(TRUE, FALSE);
3669 * Callback invoked after end to the preedit.
3671 /*ARGSUSED*/
3672 static void
3673 im_preedit_end_cb(GtkIMContext *context, gpointer data)
3675 #ifdef XIM_DEBUG
3676 xim_log("im_preedit_end_cb()\n");
3677 #endif
3678 im_delete_preedit();
3680 /* Indicate that preediting has finished */
3681 preedit_start_col = MAXCOL;
3682 xim_has_preediting = FALSE;
3684 im_is_active = FALSE;
3685 gui_update_cursor(TRUE, FALSE);
3686 im_show_info();
3690 * Callback invoked after changes to the preedit string. If the preedit
3691 * string was empty before, remember the preedit start column so we know
3692 * where to apply feedback attributes. Delete the previous preedit string
3693 * if there was one, save the new preedit cursor offset, and put the new
3694 * string into the input buffer.
3696 * TODO: The pragmatic "put into input buffer" approach used here has
3697 * several fundamental problems:
3699 * - The characters in the preedit string are subject to remapping.
3700 * That's broken, only the finally committed string should be remapped.
3702 * - There is a race condition involved: The retrieved value for the
3703 * current cursor position will be wrong if any unprocessed characters
3704 * are still queued in the input buffer.
3706 * - Due to the lack of synchronization between the file buffer in memory
3707 * and any typed characters, it's practically impossible to implement the
3708 * "retrieve_surrounding" and "delete_surrounding" signals reliably. IM
3709 * modules for languages such as Thai are likely to rely on this feature
3710 * for proper operation.
3712 * Conclusions: I think support for preediting needs to be moved to the
3713 * core parts of Vim. Ideally, until it has been committed, the preediting
3714 * string should only be displayed and not affect the buffer content at all.
3715 * The question how to deal with the synchronization issue still remains.
3716 * Circumventing the input buffer is probably not desirable. Anyway, I think
3717 * implementing "retrieve_surrounding" is the only hard problem.
3719 * One way to solve all of this in a clean manner would be to queue all key
3720 * press/release events "as is" in the input buffer, and apply the IM filtering
3721 * at the receiving end of the queue. This, however, would have a rather large
3722 * impact on the code base. If there is an easy way to force processing of all
3723 * remaining input from within the "retrieve_surrounding" signal handler, this
3724 * might not be necessary. Gotta ask on vim-dev for opinions.
3726 /*ARGSUSED1*/
3727 static void
3728 im_preedit_changed_cb(GtkIMContext *context, gpointer data)
3730 char *preedit_string = NULL;
3731 int cursor_index = 0;
3732 int num_move_back = 0;
3733 char_u *str;
3734 char_u *p;
3735 int i;
3737 gtk_im_context_get_preedit_string(context,
3738 &preedit_string, NULL,
3739 &cursor_index);
3741 #ifdef XIM_DEBUG
3742 xim_log("im_preedit_changed_cb(): %s\n", preedit_string);
3743 #endif
3745 g_return_if_fail(preedit_string != NULL); /* just in case */
3747 /* If preedit_start_col is MAXCOL set it to the current cursor position. */
3748 if (preedit_start_col == MAXCOL && preedit_string[0] != '\0')
3750 xim_has_preediting = TRUE;
3752 /* Urgh, this breaks if the input buffer isn't empty now */
3753 init_preedit_start_col();
3755 else if (cursor_index == 0 && preedit_string[0] == '\0')
3757 xim_has_preediting = FALSE;
3759 /* If at the start position (after typing backspace)
3760 * preedit_start_col must be reset. */
3761 preedit_start_col = MAXCOL;
3764 im_delete_preedit();
3767 * Compute the end of the preediting area: "preedit_end_col".
3768 * According to the documentation of gtk_im_context_get_preedit_string(),
3769 * the cursor_pos output argument returns the offset in bytes. This is
3770 * unfortunately not true -- real life shows the offset is in characters,
3771 * and the GTK+ source code agrees with me. Will file a bug later.
3773 if (preedit_start_col != MAXCOL)
3774 preedit_end_col = preedit_start_col;
3775 str = (char_u *)preedit_string;
3776 for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i)
3778 int is_composing;
3780 is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p)));
3782 * These offsets are used as counters when generating <BS> and <Del>
3783 * to delete the preedit string. So don't count composing characters
3784 * unless 'delcombine' is enabled.
3786 if (!is_composing || p_deco)
3788 if (i < cursor_index)
3789 ++im_preedit_cursor;
3790 else
3791 ++im_preedit_trailing;
3793 if (!is_composing && i >= cursor_index)
3795 /* This is essentially the same as im_preedit_trailing, except
3796 * composing characters are not counted even if p_deco is set. */
3797 ++num_move_back;
3799 if (preedit_start_col != MAXCOL)
3800 preedit_end_col += utf_ptr2cells(p);
3803 if (p > str)
3805 im_add_to_input(str, (int)(p - str));
3806 im_correct_cursor(num_move_back);
3809 g_free(preedit_string);
3811 if (gtk_main_level() > 0)
3812 gtk_main_quit();
3816 * Translate the Pango attributes at iter to Vim highlighting attributes.
3817 * Ignore attributes not supported by Vim highlighting. This shouldn't have
3818 * too much impact -- right now we handle even more attributes than necessary
3819 * for the IM modules I tested with.
3821 static int
3822 translate_pango_attributes(PangoAttrIterator *iter)
3824 PangoAttribute *attr;
3825 int char_attr = HL_NORMAL;
3827 attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE);
3828 if (attr != NULL && ((PangoAttrInt *)attr)->value
3829 != (int)PANGO_UNDERLINE_NONE)
3830 char_attr |= HL_UNDERLINE;
3832 attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT);
3833 if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD)
3834 char_attr |= HL_BOLD;
3836 attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE);
3837 if (attr != NULL && ((PangoAttrInt *)attr)->value
3838 != (int)PANGO_STYLE_NORMAL)
3839 char_attr |= HL_ITALIC;
3841 attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND);
3842 if (attr != NULL)
3844 const PangoColor *color = &((PangoAttrColor *)attr)->color;
3846 /* Assume inverse if black background is requested */
3847 if ((color->red | color->green | color->blue) == 0)
3848 char_attr |= HL_INVERSE;
3851 return char_attr;
3855 * Retrieve the highlighting attributes at column col in the preedit string.
3856 * Return -1 if not in preediting mode or if col is out of range.
3859 im_get_feedback_attr(int col)
3861 char *preedit_string = NULL;
3862 PangoAttrList *attr_list = NULL;
3863 int char_attr = -1;
3865 if (xic == NULL)
3866 return char_attr;
3868 gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL);
3870 if (preedit_string != NULL && attr_list != NULL)
3872 int idx;
3874 /* Get the byte index as used by PangoAttrIterator */
3875 for (idx = 0; col > 0 && preedit_string[idx] != '\0'; --col)
3876 idx += utfc_ptr2len((char_u *)preedit_string + idx);
3878 if (preedit_string[idx] != '\0')
3880 PangoAttrIterator *iter;
3881 int start, end;
3883 char_attr = HL_NORMAL;
3884 iter = pango_attr_list_get_iterator(attr_list);
3886 /* Extract all relevant attributes from the list. */
3889 pango_attr_iterator_range(iter, &start, &end);
3891 if (idx >= start && idx < end)
3892 char_attr |= translate_pango_attributes(iter);
3894 while (pango_attr_iterator_next(iter));
3896 pango_attr_iterator_destroy(iter);
3900 if (attr_list != NULL)
3901 pango_attr_list_unref(attr_list);
3902 g_free(preedit_string);
3904 return char_attr;
3907 void
3908 xim_init(void)
3910 #ifdef XIM_DEBUG
3911 xim_log("xim_init()\n");
3912 #endif
3914 g_return_if_fail(gui.drawarea != NULL);
3915 g_return_if_fail(gui.drawarea->window != NULL);
3917 xic = gtk_im_multicontext_new();
3918 g_object_ref(xic);
3920 im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit",
3921 G_CALLBACK(&im_commit_cb), NULL);
3922 g_signal_connect(G_OBJECT(xic), "preedit_changed",
3923 G_CALLBACK(&im_preedit_changed_cb), NULL);
3924 g_signal_connect(G_OBJECT(xic), "preedit_start",
3925 G_CALLBACK(&im_preedit_start_cb), NULL);
3926 g_signal_connect(G_OBJECT(xic), "preedit_end",
3927 G_CALLBACK(&im_preedit_end_cb), NULL);
3929 gtk_im_context_set_client_window(xic, gui.drawarea->window);
3932 void
3933 im_shutdown(void)
3935 #ifdef XIM_DEBUG
3936 xim_log("im_shutdown()\n");
3937 #endif
3939 if (xic != NULL)
3941 gtk_im_context_focus_out(xic);
3942 g_object_unref(xic);
3943 xic = NULL;
3945 im_is_active = FALSE;
3946 im_commit_handler_id = 0;
3947 preedit_start_col = MAXCOL;
3948 xim_has_preediting = FALSE;
3952 * Convert the string argument to keyval and state for GdkEventKey.
3953 * If str is valid return TRUE, otherwise FALSE.
3955 * See 'imactivatekey' for documentation of the format.
3957 static int
3958 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state)
3960 const char *mods_end;
3961 unsigned tmp_keyval;
3962 unsigned tmp_state = 0;
3964 mods_end = strrchr(str, '-');
3965 mods_end = (mods_end != NULL) ? mods_end + 1 : str;
3967 /* Parse modifier keys */
3968 while (str < mods_end)
3969 switch (*str++)
3971 case '-': break;
3972 case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK; break;
3973 case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK; break;
3974 case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break;
3975 case '1': tmp_state |= (unsigned)GDK_MOD1_MASK; break;
3976 case '2': tmp_state |= (unsigned)GDK_MOD2_MASK; break;
3977 case '3': tmp_state |= (unsigned)GDK_MOD3_MASK; break;
3978 case '4': tmp_state |= (unsigned)GDK_MOD4_MASK; break;
3979 case '5': tmp_state |= (unsigned)GDK_MOD5_MASK; break;
3980 default:
3981 return FALSE;
3984 tmp_keyval = gdk_keyval_from_name(str);
3986 if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol)
3987 return FALSE;
3989 if (keyval != NULL)
3990 *keyval = tmp_keyval;
3991 if (state != NULL)
3992 *state = tmp_state;
3994 return TRUE;
3998 * Return TRUE if p_imak is valid, otherwise FALSE. As a special case, an
3999 * empty string is also regarded as valid.
4001 * Note: The numerical key value of p_imak is cached if it was valid; thus
4002 * boldly assuming im_xim_isvalid_imactivate() will always be called whenever
4003 * 'imak' changes. This is currently the case but not obvious -- should
4004 * probably rename the function for clarity.
4007 im_xim_isvalid_imactivate(void)
4009 if (p_imak[0] == NUL)
4011 im_activatekey_keyval = GDK_VoidSymbol;
4012 im_activatekey_state = 0;
4013 return TRUE;
4016 return im_string_to_keyval((const char *)p_imak,
4017 &im_activatekey_keyval,
4018 &im_activatekey_state);
4021 static void
4022 im_synthesize_keypress(unsigned int keyval, unsigned int state)
4024 GdkEventKey *event;
4026 # ifdef HAVE_GTK_MULTIHEAD
4027 event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS);
4028 g_object_ref(gui.drawarea->window); /* unreffed by gdk_event_free() */
4029 # else
4030 event = (GdkEventKey *)g_malloc0((gulong)sizeof(GdkEvent));
4031 event->type = GDK_KEY_PRESS;
4032 # endif
4033 event->window = gui.drawarea->window;
4034 event->send_event = TRUE;
4035 event->time = GDK_CURRENT_TIME;
4036 event->state = state;
4037 event->keyval = keyval;
4038 event->hardware_keycode = /* needed for XIM */
4039 XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval);
4040 event->length = 0;
4041 event->string = NULL;
4043 gtk_im_context_filter_keypress(xic, event);
4045 /* For consistency, also send the corresponding release event. */
4046 event->type = GDK_KEY_RELEASE;
4047 event->send_event = FALSE;
4048 gtk_im_context_filter_keypress(xic, event);
4050 # ifdef HAVE_GTK_MULTIHEAD
4051 gdk_event_free((GdkEvent *)event);
4052 # else
4053 g_free(event);
4054 # endif
4057 void
4058 xim_reset(void)
4060 if (xic != NULL)
4063 * The third-party imhangul module (and maybe others too) ignores
4064 * gtk_im_context_reset() or at least doesn't reset the active state.
4065 * Thus sending imactivatekey would turn it off if it was on before,
4066 * which is clearly not what we want. Fortunately we can work around
4067 * that for imhangul by sending GDK_Escape, but I don't know if it
4068 * works with all IM modules that support an activation key :/
4070 * An alternative approach would be to destroy the IM context and
4071 * recreate it. But that means loading/unloading the IM module on
4072 * every mode switch, which causes a quite noticable delay even on
4073 * my rather fast box...
4075 * Moreover, there are some XIM which cannot respond to
4076 * im_synthesize_keypress(). we hope that they reset by
4077 * xim_shutdown().
4079 if (im_activatekey_keyval != GDK_VoidSymbol && im_is_active)
4080 im_synthesize_keypress(GDK_Escape, 0U);
4082 gtk_im_context_reset(xic);
4085 * HACK for Ami: This sequence of function calls makes Ami handle
4086 * the IM reset graciously, without breaking loads of other stuff.
4087 * It seems to force English mode as well, which is exactly what we
4088 * want because it makes the Ami status display work reliably.
4090 gtk_im_context_set_use_preedit(xic, FALSE);
4092 if (p_imdisable)
4093 im_shutdown();
4094 else
4096 gtk_im_context_set_use_preedit(xic, TRUE);
4097 xim_set_focus(gui.in_focus);
4099 if (im_activatekey_keyval != GDK_VoidSymbol)
4101 if (im_is_active)
4103 g_signal_handler_block(xic, im_commit_handler_id);
4104 im_synthesize_keypress(im_activatekey_keyval,
4105 im_activatekey_state);
4106 g_signal_handler_unblock(xic, im_commit_handler_id);
4109 else
4111 im_shutdown();
4112 xim_init();
4113 xim_set_focus(gui.in_focus);
4118 preedit_start_col = MAXCOL;
4119 xim_has_preediting = FALSE;
4123 xim_queue_key_press_event(GdkEventKey *event, int down)
4125 if (down)
4128 * Workaround GTK2 XIM 'feature' that always converts keypad keys to
4129 * chars., even when not part of an IM sequence (ref. feature of
4130 * gdk/gdkkeyuni.c).
4131 * Flag any keypad keys that might represent a single char.
4132 * If this (on its own - i.e., not part of an IM sequence) is
4133 * committed while we're processing one of these keys, we can ignore
4134 * that commit and go ahead & process it ourselves. That way we can
4135 * still distinguish keypad keys for use in mappings.
4137 switch (event->keyval)
4139 case GDK_KP_Add: xim_expected_char = '+'; break;
4140 case GDK_KP_Subtract: xim_expected_char = '-'; break;
4141 case GDK_KP_Divide: xim_expected_char = '/'; break;
4142 case GDK_KP_Multiply: xim_expected_char = '*'; break;
4143 case GDK_KP_Decimal: xim_expected_char = '.'; break;
4144 case GDK_KP_Equal: xim_expected_char = '='; break;
4145 case GDK_KP_0: xim_expected_char = '0'; break;
4146 case GDK_KP_1: xim_expected_char = '1'; break;
4147 case GDK_KP_2: xim_expected_char = '2'; break;
4148 case GDK_KP_3: xim_expected_char = '3'; break;
4149 case GDK_KP_4: xim_expected_char = '4'; break;
4150 case GDK_KP_5: xim_expected_char = '5'; break;
4151 case GDK_KP_6: xim_expected_char = '6'; break;
4152 case GDK_KP_7: xim_expected_char = '7'; break;
4153 case GDK_KP_8: xim_expected_char = '8'; break;
4154 case GDK_KP_9: xim_expected_char = '9'; break;
4155 default: xim_expected_char = NUL;
4157 xim_ignored_char = FALSE;
4161 * When typing fFtT, XIM may be activated. Thus it must pass
4162 * gtk_im_context_filter_keypress() in Normal mode.
4163 * And while doing :sh too.
4165 if (xic != NULL && !p_imdisable
4166 && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0)
4169 * Filter 'imactivatekey' and map it to CTRL-^. This way, Vim is
4170 * always aware of the current status of IM, and can even emulate
4171 * the activation key for modules that don't support one.
4173 if (event->keyval == im_activatekey_keyval
4174 && (event->state & im_activatekey_state) == im_activatekey_state)
4176 unsigned int state_mask;
4178 /* Require the state of the 3 most used modifiers to match exactly.
4179 * Otherwise e.g. <S-C-space> would be unusable for other purposes
4180 * if the IM activate key is <S-space>. */
4181 state_mask = im_activatekey_state;
4182 state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK
4183 | (int)GDK_MOD1_MASK);
4185 if ((event->state & state_mask) != im_activatekey_state)
4186 return FALSE;
4188 /* Don't send it a second time on GDK_KEY_RELEASE. */
4189 if (event->type != GDK_KEY_PRESS)
4190 return TRUE;
4192 if (map_to_exists_mode((char_u *)"", LANGMAP, FALSE))
4194 im_set_active(FALSE);
4196 /* ":lmap" mappings exists, toggle use of mappings. */
4197 State ^= LANGMAP;
4198 if (State & LANGMAP)
4200 curbuf->b_p_iminsert = B_IMODE_NONE;
4201 State &= ~LANGMAP;
4203 else
4205 curbuf->b_p_iminsert = B_IMODE_LMAP;
4206 State |= LANGMAP;
4208 return TRUE;
4211 return gtk_im_context_filter_keypress(xic, event);
4214 /* Don't filter events through the IM context if IM isn't active
4215 * right now. Unlike with GTK+ 1.2 we cannot rely on the IM module
4216 * not doing anything before the activation key was sent. */
4217 if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active)
4219 int imresult = gtk_im_context_filter_keypress(xic, event);
4221 /* Some XIM send following sequence:
4222 * 1. preedited string.
4223 * 2. committed string.
4224 * 3. line changed key.
4225 * 4. preedited string.
4226 * 5. remove preedited string.
4227 * if 3, Vim can't move back the above line for 5.
4228 * thus, this part should not parse the key. */
4229 if (!imresult && preedit_start_col != MAXCOL
4230 && event->keyval == GDK_Return)
4232 im_synthesize_keypress(GDK_Return, 0U);
4233 return FALSE;
4236 /* If XIM tried to commit a keypad key as a single char.,
4237 * ignore it so we can use the keypad key 'raw', for mappings. */
4238 if (xim_expected_char != NUL && xim_ignored_char)
4239 /* We had a keypad key, and XIM tried to thieve it */
4240 return FALSE;
4242 /* Normal processing */
4243 return imresult;
4247 return FALSE;
4251 im_get_status(void)
4253 return im_is_active;
4256 # else /* !HAVE_GTK2 */
4258 static int xim_is_active = FALSE; /* XIM should be active in the current
4259 mode */
4260 static int xim_has_focus = FALSE; /* XIM is really being used for Vim */
4261 #ifdef FEAT_GUI_X11
4262 static XIMStyle input_style;
4263 static int status_area_enabled = TRUE;
4264 #endif
4266 #ifdef FEAT_GUI_GTK
4267 # ifdef WIN3264
4268 # include <gdk/gdkwin32.h>
4269 # else
4270 # include <gdk/gdkx.h>
4271 # endif
4272 #else
4273 # ifdef PROTO
4274 /* Define a few things to be able to generate prototypes while not configured
4275 * for GTK. */
4276 # define GSList int
4277 # define gboolean int
4278 typedef int GdkEvent;
4279 typedef int GdkEventKey;
4280 # define GdkIC int
4281 # endif
4282 #endif
4284 #if defined(FEAT_GUI_GTK) || defined(PROTO)
4285 static int preedit_buf_len = 0;
4286 static int xim_can_preediting INIT(= FALSE); /* XIM in showmode() */
4287 static int xim_input_style;
4288 #ifndef FEAT_GUI_GTK
4289 # define gboolean int
4290 #endif
4291 static gboolean use_status_area = 0;
4293 static int im_xim_str2keycode __ARGS((unsigned int *code, unsigned int *state));
4294 static void im_xim_send_event_imactivate __ARGS((void));
4297 * Convert string to keycode and state for XKeyEvent.
4298 * When string is valid return OK, when invalid return FAIL.
4300 * See 'imactivatekey' documentation for the format.
4302 static int
4303 im_xim_str2keycode(code, state)
4304 unsigned int *code;
4305 unsigned int *state;
4307 int retval = OK;
4308 int len;
4309 unsigned keycode = 0, keystate = 0;
4310 Window window;
4311 Display *display;
4312 char_u *flag_end;
4313 char_u *str;
4315 if (*p_imak != NUL)
4317 len = STRLEN(p_imak);
4318 for (flag_end = p_imak + len - 1;
4319 flag_end > p_imak && *flag_end != '-'; --flag_end)
4322 /* Parse modifier keys */
4323 for (str = p_imak; str < flag_end; ++str)
4325 switch (*str)
4327 case 's': case 'S':
4328 keystate |= ShiftMask;
4329 break;
4330 case 'l': case 'L':
4331 keystate |= LockMask;
4332 break;
4333 case 'c': case 'C':
4334 keystate |= ControlMask;
4335 break;
4336 case '1':
4337 keystate |= Mod1Mask;
4338 break;
4339 case '2':
4340 keystate |= Mod2Mask;
4341 break;
4342 case '3':
4343 keystate |= Mod3Mask;
4344 break;
4345 case '4':
4346 keystate |= Mod4Mask;
4347 break;
4348 case '5':
4349 keystate |= Mod5Mask;
4350 break;
4351 case '-':
4352 break;
4353 default:
4354 retval = FAIL;
4357 if (*str == '-')
4358 ++str;
4360 /* Get keycode from string. */
4361 gui_get_x11_windis(&window, &display);
4362 if (display)
4363 keycode = XKeysymToKeycode(display, XStringToKeysym((char *)str));
4364 if (keycode == 0)
4365 retval = FAIL;
4367 if (code != NULL)
4368 *code = keycode;
4369 if (state != NULL)
4370 *state = keystate;
4372 return retval;
4375 static void
4376 im_xim_send_event_imactivate()
4378 /* Force turn on preedit state by symulate keypress event.
4379 * Keycode and state is specified by 'imactivatekey'.
4381 XKeyEvent ev;
4383 gui_get_x11_windis(&ev.window, &ev.display);
4384 ev.root = RootWindow(ev.display, DefaultScreen(ev.display));
4385 ev.subwindow = None;
4386 ev.time = CurrentTime;
4387 ev.x = 1;
4388 ev.y = 1;
4389 ev.x_root = 1;
4390 ev.y_root = 1;
4391 ev.same_screen = 1;
4392 ev.type = KeyPress;
4393 if (im_xim_str2keycode(&ev.keycode, &ev.state) == OK)
4394 XSendEvent(ev.display, ev.window, 1, KeyPressMask, (XEvent*)&ev);
4398 * Return TRUE if 'imactivatekey' has a valid value.
4401 im_xim_isvalid_imactivate()
4403 return im_xim_str2keycode(NULL, NULL) == OK;
4405 #endif /* FEAT_GUI_GTK */
4408 * Switch using XIM on/off. This is used by the code that changes "State".
4410 void
4411 im_set_active(active)
4412 int active;
4414 if (xic == NULL)
4415 return;
4417 /* If 'imdisable' is set, XIM is never active. */
4418 if (p_imdisable)
4419 active = FALSE;
4420 #if !defined (FEAT_GUI_GTK)
4421 else if (input_style & XIMPreeditPosition)
4422 /* There is a problem in switching XIM off when preediting is used,
4423 * and it is not clear how this can be solved. For now, keep XIM on
4424 * all the time, like it was done in Vim 5.8. */
4425 active = TRUE;
4426 #endif
4428 /* Remember the active state, it is needed when Vim gets keyboard focus. */
4429 xim_is_active = active;
4431 #ifdef FEAT_GUI_GTK
4432 /* When 'imactivatekey' has valid key-string, try to control XIM preedit
4433 * state. When 'imactivatekey' has no or invalid string, try old XIM
4434 * focus control.
4436 if (*p_imak != NUL)
4438 /* BASIC STRATEGY:
4439 * Destroy old Input Context (XIC), and create new one. New XIC
4440 * would have a state of preedit that is off. When argument:active
4441 * is false, that's all. Else argument:active is true, send a key
4442 * event specified by 'imactivatekey' to activate XIM preedit state.
4445 xim_is_active = TRUE; /* Disable old XIM focus control */
4446 /* If we can monitor preedit state with preedit callback functions,
4447 * try least creation of new XIC.
4449 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)
4451 if (xim_can_preediting && !active)
4453 /* Force turn off preedit state. With some IM
4454 * implementations, we cannot turn off preedit state by
4455 * symulate keypress event. It is why using such a method
4456 * that destroy old IC (input context), and create new one.
4457 * When create new IC, its preedit state is usually off.
4459 xim_reset();
4460 xim_set_focus(FALSE);
4461 gdk_ic_destroy(xic);
4462 xim_init();
4463 xim_can_preediting = FALSE;
4465 else if (!xim_can_preediting && active)
4466 im_xim_send_event_imactivate();
4468 else
4470 /* First, force destroy old IC, and create new one. It
4471 * symulates "turning off preedit state".
4473 xim_set_focus(FALSE);
4474 gdk_ic_destroy(xic);
4475 xim_init();
4476 xim_can_preediting = FALSE;
4478 /* 2nd, when requested to activate IM, symulate this by sending
4479 * the event.
4481 if (active)
4483 im_xim_send_event_imactivate();
4484 xim_can_preediting = TRUE;
4488 else
4490 # ifndef XIMPreeditUnKnown
4491 /* X11R5 doesn't have these, it looks safe enough to define here. */
4492 typedef unsigned long XIMPreeditState;
4493 # define XIMPreeditUnKnown 0L
4494 # define XIMPreeditEnable 1L
4495 # define XIMPreeditDisable (1L<<1)
4496 # define XNPreeditState "preeditState"
4497 # endif
4498 XIMPreeditState preedit_state = XIMPreeditUnKnown;
4499 XVaNestedList preedit_attr;
4500 XIC pxic;
4502 preedit_attr = XVaCreateNestedList(0,
4503 XNPreeditState, &preedit_state,
4504 NULL);
4505 pxic = ((GdkICPrivate *)xic)->xic;
4507 if (!XGetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL))
4509 XFree(preedit_attr);
4510 preedit_attr = XVaCreateNestedList(0,
4511 XNPreeditState,
4512 active ? XIMPreeditEnable : XIMPreeditDisable,
4513 NULL);
4514 XSetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL);
4515 xim_can_preediting = active;
4516 xim_is_active = active;
4518 XFree(preedit_attr);
4520 if (xim_input_style & XIMPreeditCallbacks)
4522 preedit_buf_len = 0;
4523 init_preedit_start_col();
4525 #else
4526 # if 0
4527 /* When had tested kinput2 + canna + Athena GUI version with
4528 * 'imactivatekey' is "s-space", im_xim_send_event_imactivate() did not
4529 * work correctly. It just inserted one space. I don't know why we
4530 * couldn't switch state of XIM preediting. This is reason why these
4531 * codes are commented out.
4533 /* First, force destroy old IC, and create new one. It symulates
4534 * "turning off preedit state".
4536 xim_set_focus(FALSE);
4537 XDestroyIC(xic);
4538 xic = NULL;
4539 xim_init();
4541 /* 2nd, when requested to activate IM, symulate this by sending the
4542 * event.
4544 if (active)
4545 im_xim_send_event_imactivate();
4546 # endif
4547 #endif
4548 xim_set_preedit();
4552 * Adjust using XIM for gaining or losing keyboard focus. Also called when
4553 * "xim_is_active" changes.
4555 void
4556 xim_set_focus(focus)
4557 int focus;
4559 if (xic == NULL)
4560 return;
4563 * XIM only gets focus when the Vim window has keyboard focus and XIM has
4564 * been set active for the current mode.
4566 if (focus && xim_is_active)
4568 if (!xim_has_focus)
4570 xim_has_focus = TRUE;
4571 #ifdef FEAT_GUI_GTK
4572 gdk_im_begin(xic, gui.drawarea->window);
4573 #else
4574 XSetICFocus(xic);
4575 #endif
4578 else
4580 if (xim_has_focus)
4582 xim_has_focus = FALSE;
4583 #ifdef FEAT_GUI_GTK
4584 gdk_im_end();
4585 #else
4586 XUnsetICFocus(xic);
4587 #endif
4592 /*ARGSUSED*/
4593 void
4594 im_set_position(row, col)
4595 int row;
4596 int col;
4598 xim_set_preedit();
4602 * Set the XIM to the current cursor position.
4604 void
4605 xim_set_preedit()
4607 if (xic == NULL)
4608 return;
4610 xim_set_focus(TRUE);
4612 #ifdef FEAT_GUI_GTK
4613 if (gdk_im_ready())
4615 int attrmask;
4616 GdkICAttr *attr;
4618 if (!xic_attr)
4619 return;
4621 attr = xic_attr;
4622 attrmask = 0;
4624 # ifdef FEAT_XFONTSET
4625 if ((xim_input_style & (int)GDK_IM_PREEDIT_POSITION)
4626 && gui.fontset != NOFONTSET
4627 && gui.fontset->type == GDK_FONT_FONTSET)
4629 if (!xim_has_focus)
4631 if (attr->spot_location.y >= 0)
4633 attr->spot_location.x = 0;
4634 attr->spot_location.y = -100;
4635 attrmask |= (int)GDK_IC_SPOT_LOCATION;
4638 else
4640 gint width, height;
4642 if (attr->spot_location.x != TEXT_X(gui.col)
4643 || attr->spot_location.y != TEXT_Y(gui.row))
4645 attr->spot_location.x = TEXT_X(gui.col);
4646 attr->spot_location.y = TEXT_Y(gui.row);
4647 attrmask |= (int)GDK_IC_SPOT_LOCATION;
4650 gdk_window_get_size(gui.drawarea->window, &width, &height);
4651 width -= 2 * gui.border_offset;
4652 height -= 2 * gui.border_offset;
4653 if (xim_input_style & (int)GDK_IM_STATUS_AREA)
4654 height -= gui.char_height;
4655 if (attr->preedit_area.width != width
4656 || attr->preedit_area.height != height)
4658 attr->preedit_area.x = gui.border_offset;
4659 attr->preedit_area.y = gui.border_offset;
4660 attr->preedit_area.width = width;
4661 attr->preedit_area.height = height;
4662 attrmask |= (int)GDK_IC_PREEDIT_AREA;
4665 if (attr->preedit_fontset != gui.current_font)
4667 attr->preedit_fontset = gui.current_font;
4668 attrmask |= (int)GDK_IC_PREEDIT_FONTSET;
4672 # endif /* FEAT_XFONTSET */
4674 if (xim_fg_color == INVALCOLOR)
4676 xim_fg_color = gui.def_norm_pixel;
4677 xim_bg_color = gui.def_back_pixel;
4679 if (attr->preedit_foreground.pixel != xim_fg_color)
4681 attr->preedit_foreground.pixel = xim_fg_color;
4682 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND;
4684 if (attr->preedit_background.pixel != xim_bg_color)
4686 attr->preedit_background.pixel = xim_bg_color;
4687 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND;
4690 if (attrmask != 0)
4691 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask);
4693 #else /* FEAT_GUI_GTK */
4695 XVaNestedList attr_list;
4696 XRectangle spot_area;
4697 XPoint over_spot;
4698 int line_space;
4700 if (!xim_has_focus)
4702 /* hide XIM cursor */
4703 over_spot.x = 0;
4704 over_spot.y = -100; /* arbitrary invisible position */
4705 attr_list = (XVaNestedList) XVaCreateNestedList(0,
4706 XNSpotLocation,
4707 &over_spot,
4708 NULL);
4709 XSetICValues(xic, XNPreeditAttributes, attr_list, NULL);
4710 XFree(attr_list);
4711 return;
4714 if (input_style & XIMPreeditPosition)
4716 if (xim_fg_color == INVALCOLOR)
4718 xim_fg_color = gui.def_norm_pixel;
4719 xim_bg_color = gui.def_back_pixel;
4721 over_spot.x = TEXT_X(gui.col);
4722 over_spot.y = TEXT_Y(gui.row);
4723 spot_area.x = 0;
4724 spot_area.y = 0;
4725 spot_area.height = gui.char_height * Rows;
4726 spot_area.width = gui.char_width * Columns;
4727 line_space = gui.char_height;
4728 attr_list = (XVaNestedList) XVaCreateNestedList(0,
4729 XNSpotLocation, &over_spot,
4730 XNForeground, (Pixel) xim_fg_color,
4731 XNBackground, (Pixel) xim_bg_color,
4732 XNArea, &spot_area,
4733 XNLineSpace, line_space,
4734 NULL);
4735 if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL))
4736 EMSG(_("E284: Cannot set IC values"));
4737 XFree(attr_list);
4740 #endif /* FEAT_GUI_GTK */
4744 * Set up the status area.
4746 * This should use a separate Widget, but that seems not possible, because
4747 * preedit_area and status_area should be set to the same window as for the
4748 * text input. Unfortunately this means the status area pollutes the text
4749 * window...
4751 void
4752 xim_set_status_area()
4754 if (xic == NULL)
4755 return;
4757 #ifdef FEAT_GUI_GTK
4758 # if defined(FEAT_XFONTSET)
4759 if (use_status_area)
4761 GdkICAttr *attr;
4762 int style;
4763 gint width, height;
4764 GtkWidget *widget;
4765 int attrmask;
4767 if (!xic_attr)
4768 return;
4770 attr = xic_attr;
4771 attrmask = 0;
4772 style = (int)gdk_ic_get_style(xic);
4773 if ((style & (int)GDK_IM_STATUS_MASK) == (int)GDK_IM_STATUS_AREA)
4775 if (gui.fontset != NOFONTSET
4776 && gui.fontset->type == GDK_FONT_FONTSET)
4778 widget = gui.mainwin;
4779 gdk_window_get_size(widget->window, &width, &height);
4781 attrmask |= (int)GDK_IC_STATUS_AREA;
4782 attr->status_area.x = 0;
4783 attr->status_area.y = height - gui.char_height - 1;
4784 attr->status_area.width = width;
4785 attr->status_area.height = gui.char_height;
4788 if (attrmask != 0)
4789 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask);
4791 # endif
4792 #else
4794 XVaNestedList preedit_list = 0, status_list = 0, list = 0;
4795 XRectangle pre_area, status_area;
4797 if (input_style & XIMStatusArea)
4799 if (input_style & XIMPreeditArea)
4801 XRectangle *needed_rect;
4803 /* to get status_area width */
4804 status_list = XVaCreateNestedList(0, XNAreaNeeded,
4805 &needed_rect, NULL);
4806 XGetICValues(xic, XNStatusAttributes, status_list, NULL);
4807 XFree(status_list);
4809 status_area.width = needed_rect->width;
4811 else
4812 status_area.width = gui.char_width * Columns;
4814 status_area.x = 0;
4815 status_area.y = gui.char_height * Rows + gui.border_offset;
4816 if (gui.which_scrollbars[SBAR_BOTTOM])
4817 status_area.y += gui.scrollbar_height;
4818 #ifdef FEAT_MENU
4819 if (gui.menu_is_active)
4820 status_area.y += gui.menu_height;
4821 #endif
4822 status_area.height = gui.char_height;
4823 status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL);
4825 else
4827 status_area.x = 0;
4828 status_area.y = gui.char_height * Rows + gui.border_offset;
4829 if (gui.which_scrollbars[SBAR_BOTTOM])
4830 status_area.y += gui.scrollbar_height;
4831 #ifdef FEAT_MENU
4832 if (gui.menu_is_active)
4833 status_area.y += gui.menu_height;
4834 #endif
4835 status_area.width = 0;
4836 status_area.height = gui.char_height;
4839 if (input_style & XIMPreeditArea) /* off-the-spot */
4841 pre_area.x = status_area.x + status_area.width;
4842 pre_area.y = gui.char_height * Rows + gui.border_offset;
4843 pre_area.width = gui.char_width * Columns - pre_area.x;
4844 if (gui.which_scrollbars[SBAR_BOTTOM])
4845 pre_area.y += gui.scrollbar_height;
4846 #ifdef FEAT_MENU
4847 if (gui.menu_is_active)
4848 pre_area.y += gui.menu_height;
4849 #endif
4850 pre_area.height = gui.char_height;
4851 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
4853 else if (input_style & XIMPreeditPosition) /* over-the-spot */
4855 pre_area.x = 0;
4856 pre_area.y = 0;
4857 pre_area.height = gui.char_height * Rows;
4858 pre_area.width = gui.char_width * Columns;
4859 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
4862 if (preedit_list && status_list)
4863 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
4864 XNStatusAttributes, status_list, NULL);
4865 else if (preedit_list)
4866 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
4867 NULL);
4868 else if (status_list)
4869 list = XVaCreateNestedList(0, XNStatusAttributes, status_list,
4870 NULL);
4871 else
4872 list = NULL;
4874 if (list)
4876 XSetICValues(xic, XNVaNestedList, list, NULL);
4877 XFree(list);
4879 if (status_list)
4880 XFree(status_list);
4881 if (preedit_list)
4882 XFree(preedit_list);
4884 #endif
4887 #if defined(FEAT_GUI_X11) || defined(FEAT_GUI_GTK)
4888 static char e_xim[] = N_("E285: Failed to create input context");
4889 #endif
4891 #if defined(FEAT_GUI_X11) || defined(PROTO)
4892 # if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(sun)
4893 # define USE_X11R6_XIM
4894 # endif
4896 static int xim_real_init __ARGS((Window x11_window, Display *x11_display));
4899 #ifdef USE_X11R6_XIM
4900 static void xim_instantiate_cb __ARGS((Display *display, XPointer client_data, XPointer call_data));
4901 static void xim_destroy_cb __ARGS((XIM im, XPointer client_data, XPointer call_data));
4903 /*ARGSUSED*/
4904 static void
4905 xim_instantiate_cb(display, client_data, call_data)
4906 Display *display;
4907 XPointer client_data;
4908 XPointer call_data;
4910 Window x11_window;
4911 Display *x11_display;
4913 #ifdef XIM_DEBUG
4914 xim_log("xim_instantiate_cb()\n");
4915 #endif
4917 gui_get_x11_windis(&x11_window, &x11_display);
4918 if (display != x11_display)
4919 return;
4921 xim_real_init(x11_window, x11_display);
4922 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
4923 if (xic != NULL)
4924 XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
4925 xim_instantiate_cb, NULL);
4928 /*ARGSUSED*/
4929 static void
4930 xim_destroy_cb(im, client_data, call_data)
4931 XIM im;
4932 XPointer client_data;
4933 XPointer call_data;
4935 Window x11_window;
4936 Display *x11_display;
4938 #ifdef XIM_DEBUG
4939 xim_log("xim_destroy_cb()\n");
4940 #endif
4941 gui_get_x11_windis(&x11_window, &x11_display);
4943 xic = NULL;
4944 status_area_enabled = FALSE;
4946 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
4948 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
4949 xim_instantiate_cb, NULL);
4951 #endif
4953 void
4954 xim_init()
4956 Window x11_window;
4957 Display *x11_display;
4959 #ifdef XIM_DEBUG
4960 xim_log("xim_init()\n");
4961 #endif
4963 gui_get_x11_windis(&x11_window, &x11_display);
4965 xic = NULL;
4967 if (xim_real_init(x11_window, x11_display))
4968 return;
4970 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
4972 #ifdef USE_X11R6_XIM
4973 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
4974 xim_instantiate_cb, NULL);
4975 #endif
4978 static int
4979 xim_real_init(x11_window, x11_display)
4980 Window x11_window;
4981 Display *x11_display;
4983 int i;
4984 char *p,
4986 *ns,
4987 *end,
4988 tmp[1024];
4989 #define IMLEN_MAX 40
4990 char buf[IMLEN_MAX + 7];
4991 XIM xim = NULL;
4992 XIMStyles *xim_styles;
4993 XIMStyle this_input_style = 0;
4994 Boolean found;
4995 XPoint over_spot;
4996 XVaNestedList preedit_list, status_list;
4998 input_style = 0;
4999 status_area_enabled = FALSE;
5001 if (xic != NULL)
5002 return FALSE;
5004 if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL)
5006 strcpy(tmp, gui.rsrc_input_method);
5007 for (ns = s = tmp; ns != NULL && *s != NUL;)
5009 s = (char *)skipwhite((char_u *)s);
5010 if (*s == NUL)
5011 break;
5012 if ((ns = end = strchr(s, ',')) == NULL)
5013 end = s + strlen(s);
5014 while (isspace(((char_u *)end)[-1]))
5015 end--;
5016 *end = NUL;
5018 if (strlen(s) <= IMLEN_MAX)
5020 strcpy(buf, "@im=");
5021 strcat(buf, s);
5022 if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL
5023 && (xim = XOpenIM(x11_display, NULL, NULL, NULL))
5024 != NULL)
5025 break;
5028 s = ns + 1;
5032 if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL)
5033 xim = XOpenIM(x11_display, NULL, NULL, NULL);
5035 /* This is supposed to be useful to obtain characters through
5036 * XmbLookupString() without really using a XIM. */
5037 if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL
5038 && *p != NUL)
5039 xim = XOpenIM(x11_display, NULL, NULL, NULL);
5041 if (xim == NULL)
5043 /* Only give this message when verbose is set, because too many people
5044 * got this message when they didn't want to use a XIM. */
5045 if (p_verbose > 0)
5047 verbose_enter();
5048 EMSG(_("E286: Failed to open input method"));
5049 verbose_leave();
5051 return FALSE;
5054 #ifdef USE_X11R6_XIM
5056 XIMCallback destroy_cb;
5058 destroy_cb.callback = xim_destroy_cb;
5059 destroy_cb.client_data = NULL;
5060 if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL))
5061 EMSG(_("E287: Warning: Could not set destroy callback to IM"));
5063 #endif
5065 if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles)
5067 EMSG(_("E288: input method doesn't support any style"));
5068 XCloseIM(xim);
5069 return FALSE;
5072 found = False;
5073 strcpy(tmp, gui.rsrc_preedit_type_name);
5074 for (s = tmp; s && !found; )
5076 while (*s && isspace((unsigned char)*s))
5077 s++;
5078 if (!*s)
5079 break;
5080 if ((ns = end = strchr(s, ',')) != 0)
5081 ns++;
5082 else
5083 end = s + strlen(s);
5084 while (isspace((unsigned char)*end))
5085 end--;
5086 *end = '\0';
5088 if (!strcmp(s, "OverTheSpot"))
5089 this_input_style = (XIMPreeditPosition | XIMStatusArea);
5090 else if (!strcmp(s, "OffTheSpot"))
5091 this_input_style = (XIMPreeditArea | XIMStatusArea);
5092 else if (!strcmp(s, "Root"))
5093 this_input_style = (XIMPreeditNothing | XIMStatusNothing);
5095 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
5097 if (this_input_style == xim_styles->supported_styles[i])
5099 found = True;
5100 break;
5103 if (!found)
5104 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
5106 if ((xim_styles->supported_styles[i] & this_input_style)
5107 == (this_input_style & ~XIMStatusArea))
5109 this_input_style &= ~XIMStatusArea;
5110 found = True;
5111 break;
5115 s = ns;
5117 XFree(xim_styles);
5119 if (!found)
5121 /* Only give this message when verbose is set, because too many people
5122 * got this message when they didn't want to use a XIM. */
5123 if (p_verbose > 0)
5125 verbose_enter();
5126 EMSG(_("E289: input method doesn't support my preedit type"));
5127 verbose_leave();
5129 XCloseIM(xim);
5130 return FALSE;
5133 over_spot.x = TEXT_X(gui.col);
5134 over_spot.y = TEXT_Y(gui.row);
5135 input_style = this_input_style;
5137 /* A crash was reported when trying to pass gui.norm_font as XNFontSet,
5138 * thus that has been removed. Hopefully the default works... */
5139 #ifdef FEAT_XFONTSET
5140 if (gui.fontset != NOFONTSET)
5142 preedit_list = XVaCreateNestedList(0,
5143 XNSpotLocation, &over_spot,
5144 XNForeground, (Pixel)gui.def_norm_pixel,
5145 XNBackground, (Pixel)gui.def_back_pixel,
5146 XNFontSet, (XFontSet)gui.fontset,
5147 NULL);
5148 status_list = XVaCreateNestedList(0,
5149 XNForeground, (Pixel)gui.def_norm_pixel,
5150 XNBackground, (Pixel)gui.def_back_pixel,
5151 XNFontSet, (XFontSet)gui.fontset,
5152 NULL);
5154 else
5155 #endif
5157 preedit_list = XVaCreateNestedList(0,
5158 XNSpotLocation, &over_spot,
5159 XNForeground, (Pixel)gui.def_norm_pixel,
5160 XNBackground, (Pixel)gui.def_back_pixel,
5161 NULL);
5162 status_list = XVaCreateNestedList(0,
5163 XNForeground, (Pixel)gui.def_norm_pixel,
5164 XNBackground, (Pixel)gui.def_back_pixel,
5165 NULL);
5168 xic = XCreateIC(xim,
5169 XNInputStyle, input_style,
5170 XNClientWindow, x11_window,
5171 XNFocusWindow, gui.wid,
5172 XNPreeditAttributes, preedit_list,
5173 XNStatusAttributes, status_list,
5174 NULL);
5175 XFree(status_list);
5176 XFree(preedit_list);
5177 if (xic != NULL)
5179 if (input_style & XIMStatusArea)
5181 xim_set_status_area();
5182 status_area_enabled = TRUE;
5184 else
5185 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
5187 else
5189 EMSG(_(e_xim));
5190 XCloseIM(xim);
5191 return FALSE;
5194 return TRUE;
5197 #endif /* FEAT_GUI_X11 */
5199 #if defined(FEAT_GUI_GTK) || defined(PROTO)
5201 # ifdef FEAT_XFONTSET
5202 static char e_overthespot[] = N_("E290: over-the-spot style requires fontset");
5203 # endif
5205 # ifdef PROTO
5206 typedef int GdkIC;
5207 # endif
5209 void
5210 xim_decide_input_style()
5212 /* GDK_IM_STATUS_CALLBACKS was disabled, enabled it to allow Japanese
5213 * OverTheSpot. */
5214 int supported_style = (int)GDK_IM_PREEDIT_NONE |
5215 (int)GDK_IM_PREEDIT_NOTHING |
5216 (int)GDK_IM_PREEDIT_POSITION |
5217 (int)GDK_IM_PREEDIT_CALLBACKS |
5218 (int)GDK_IM_STATUS_CALLBACKS |
5219 (int)GDK_IM_STATUS_AREA |
5220 (int)GDK_IM_STATUS_NONE |
5221 (int)GDK_IM_STATUS_NOTHING;
5223 #ifdef XIM_DEBUG
5224 xim_log("xim_decide_input_style()\n");
5225 #endif
5227 if (!gdk_im_ready())
5228 xim_input_style = 0;
5229 else
5231 if (gtk_major_version > 1
5232 || (gtk_major_version == 1
5233 && (gtk_minor_version > 2
5234 || (gtk_minor_version == 2 && gtk_micro_version >= 3))))
5235 use_status_area = TRUE;
5236 else
5238 EMSG(_("E291: Your GTK+ is older than 1.2.3. Status area disabled"));
5239 use_status_area = FALSE;
5241 #ifdef FEAT_XFONTSET
5242 if (gui.fontset == NOFONTSET || gui.fontset->type != GDK_FONT_FONTSET)
5243 #endif
5244 supported_style &= ~((int)GDK_IM_PREEDIT_POSITION
5245 | (int)GDK_IM_STATUS_AREA);
5246 if (!use_status_area)
5247 supported_style &= ~(int)GDK_IM_STATUS_AREA;
5248 xim_input_style = (int)gdk_im_decide_style((GdkIMStyle)supported_style);
5252 /*ARGSUSED*/
5253 static void
5254 preedit_start_cbproc(XIC xic, XPointer client_data, XPointer call_data)
5256 #ifdef XIM_DEBUG
5257 xim_log("xim_decide_input_style()\n");
5258 #endif
5260 draw_feedback = NULL;
5261 xim_can_preediting = TRUE;
5262 xim_has_preediting = TRUE;
5263 gui_update_cursor(TRUE, FALSE);
5264 if (showmode() > 0)
5266 setcursor();
5267 out_flush();
5271 static void
5272 xim_back_delete(int n)
5274 char_u str[3];
5276 str[0] = CSI;
5277 str[1] = 'k';
5278 str[2] = 'b';
5279 while (n-- > 0)
5280 add_to_input_buf(str, 3);
5283 static GSList *key_press_event_queue = NULL;
5284 static gboolean processing_queued_event = FALSE;
5286 /*ARGSUSED*/
5287 static void
5288 preedit_draw_cbproc(XIC xic, XPointer client_data, XPointer call_data)
5290 XIMPreeditDrawCallbackStruct *draw_data;
5291 XIMText *text;
5292 char *src;
5293 GSList *event_queue;
5295 #ifdef XIM_DEBUG
5296 xim_log("preedit_draw_cbproc()\n");
5297 #endif
5299 draw_data = (XIMPreeditDrawCallbackStruct *) call_data;
5300 text = (XIMText *) draw_data->text;
5302 if ((text == NULL && draw_data->chg_length == preedit_buf_len)
5303 || preedit_buf_len == 0)
5305 init_preedit_start_col();
5306 vim_free(draw_feedback);
5307 draw_feedback = NULL;
5309 if (draw_data->chg_length > 0)
5311 int bs_cnt;
5313 if (draw_data->chg_length > preedit_buf_len)
5314 bs_cnt = preedit_buf_len;
5315 else
5316 bs_cnt = draw_data->chg_length;
5317 xim_back_delete(bs_cnt);
5318 preedit_buf_len -= bs_cnt;
5320 if (text != NULL)
5322 int len;
5323 #ifdef FEAT_MBYTE
5324 char_u *buf = NULL;
5325 unsigned int nfeedback = 0;
5326 #endif
5327 char_u *ptr;
5329 src = text->string.multi_byte;
5330 if (src != NULL && !text->encoding_is_wchar)
5332 len = strlen(src);
5333 ptr = (char_u *)src;
5334 /* Avoid the enter for decision */
5335 if (*ptr == '\n')
5336 return;
5338 #ifdef FEAT_MBYTE
5339 if (input_conv.vc_type != CONV_NONE
5340 && (buf = string_convert(&input_conv,
5341 (char_u *)src, &len)) != NULL)
5343 /* Converted from 'termencoding' to 'encoding'. */
5344 add_to_input_buf_csi(buf, len);
5345 ptr = buf;
5347 else
5348 #endif
5349 add_to_input_buf_csi((char_u *)src, len);
5350 /* Add count of character to preedit_buf_len */
5351 while (*ptr != NUL)
5353 #ifdef FEAT_MBYTE
5354 if (draw_data->text->feedback != NULL)
5356 if (draw_feedback == NULL)
5357 draw_feedback = (char *)alloc(draw_data->chg_first
5358 + text->length);
5359 else
5360 draw_feedback = realloc(draw_feedback,
5361 draw_data->chg_first + text->length);
5362 if (draw_feedback != NULL)
5364 draw_feedback[nfeedback + draw_data->chg_first]
5365 = draw_data->text->feedback[nfeedback];
5366 nfeedback++;
5369 if (has_mbyte)
5370 ptr += (*mb_ptr2len)(ptr);
5371 else
5372 #endif
5373 ptr++;
5374 preedit_buf_len++;
5376 #ifdef FEAT_MBYTE
5377 vim_free(buf);
5378 #endif
5379 preedit_end_col = MAXCOL;
5382 if (text != NULL || draw_data->chg_length > 0)
5384 event_queue = key_press_event_queue;
5385 processing_queued_event = TRUE;
5386 while (event_queue != NULL && processing_queued_event)
5388 GdkEvent *ev = event_queue->data;
5390 gboolean *ret;
5391 gtk_signal_emit_by_name((GtkObject*)gui.mainwin, "key_press_event",
5392 ev, &ret);
5393 gdk_event_free(ev);
5394 event_queue = event_queue->next;
5396 processing_queued_event = FALSE;
5397 if (key_press_event_queue)
5399 g_slist_free(key_press_event_queue);
5400 key_press_event_queue = NULL;
5403 if (gtk_main_level() > 0)
5404 gtk_main_quit();
5408 * Retrieve the highlighting attributes at column col in the preedit string.
5409 * Return -1 if not in preediting mode or if col is out of range.
5412 im_get_feedback_attr(int col)
5414 if (draw_feedback != NULL && col < preedit_buf_len)
5416 if (draw_feedback[col] & XIMReverse)
5417 return HL_INVERSE;
5418 else if (draw_feedback[col] & XIMUnderline)
5419 return HL_UNDERLINE;
5420 else
5421 return hl_attr(HLF_V);
5424 return -1;
5427 /*ARGSUSED*/
5428 static void
5429 preedit_caret_cbproc(XIC xic, XPointer client_data, XPointer call_data)
5431 #ifdef XIM_DEBUG
5432 xim_log("preedit_caret_cbproc()\n");
5433 #endif
5436 /*ARGSUSED*/
5437 static void
5438 preedit_done_cbproc(XIC xic, XPointer client_data, XPointer call_data)
5440 #ifdef XIM_DEBUG
5441 xim_log("preedit_done_cbproc()\n");
5442 #endif
5444 vim_free(draw_feedback);
5445 draw_feedback = NULL;
5446 xim_can_preediting = FALSE;
5447 xim_has_preediting = FALSE;
5448 gui_update_cursor(TRUE, FALSE);
5449 if (showmode() > 0)
5451 setcursor();
5452 out_flush();
5456 void
5457 xim_reset(void)
5459 char *text;
5461 #ifdef XIM_DEBUG
5462 xim_log("xim_reset()\n");
5463 #endif
5465 if (xic != NULL)
5467 text = XmbResetIC(((GdkICPrivate *)xic)->xic);
5468 if (text != NULL && !(xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS))
5469 add_to_input_buf_csi((char_u *)text, strlen(text));
5470 else
5471 preedit_buf_len = 0;
5472 if (text != NULL)
5473 XFree(text);
5477 /*ARGSUSED*/
5479 xim_queue_key_press_event(GdkEventKey *event, int down)
5481 #ifdef XIM_DEBUG
5482 xim_log("xim_queue_key_press_event()\n");
5483 #endif
5485 if (preedit_buf_len <= 0)
5486 return FALSE;
5487 if (processing_queued_event)
5488 processing_queued_event = FALSE;
5490 key_press_event_queue = g_slist_append(key_press_event_queue,
5491 gdk_event_copy((GdkEvent *)event));
5492 return TRUE;
5495 /*ARGSUSED*/
5496 static void
5497 preedit_callback_setup(GdkIC *ic)
5499 XIC xxic;
5500 XVaNestedList preedit_attr;
5501 XIMCallback preedit_start_cb;
5502 XIMCallback preedit_draw_cb;
5503 XIMCallback preedit_caret_cb;
5504 XIMCallback preedit_done_cb;
5506 xxic = ((GdkICPrivate*)xic)->xic;
5507 preedit_start_cb.callback = (XIMProc)preedit_start_cbproc;
5508 preedit_draw_cb.callback = (XIMProc)preedit_draw_cbproc;
5509 preedit_caret_cb.callback = (XIMProc)preedit_caret_cbproc;
5510 preedit_done_cb.callback = (XIMProc)preedit_done_cbproc;
5511 preedit_attr
5512 = XVaCreateNestedList (0,
5513 XNPreeditStartCallback, &preedit_start_cb,
5514 XNPreeditDrawCallback, &preedit_draw_cb,
5515 XNPreeditCaretCallback, &preedit_caret_cb,
5516 XNPreeditDoneCallback, &preedit_done_cb,
5518 XSetICValues (xxic, XNPreeditAttributes, preedit_attr, 0);
5519 XFree(preedit_attr);
5522 /*ARGSUSED*/
5523 static void
5524 reset_state_setup(GdkIC *ic)
5526 #ifdef USE_X11R6_XIM
5527 /* don't change the input context when we call reset */
5528 XSetICValues(((GdkICPrivate*)ic)->xic, XNResetState, XIMPreserveState, 0);
5529 #endif
5532 void
5533 xim_init(void)
5535 #ifdef XIM_DEBUG
5536 xim_log("xim_init()\n");
5537 #endif
5539 xic = NULL;
5540 xic_attr = NULL;
5542 if (!gdk_im_ready())
5544 if (p_verbose > 0)
5546 verbose_enter();
5547 EMSG(_("E292: Input Method Server is not running"));
5548 verbose_leave();
5550 return;
5552 if ((xic_attr = gdk_ic_attr_new()) != NULL)
5554 #ifdef FEAT_XFONTSET
5555 gint width, height;
5556 #endif
5557 int mask;
5558 GdkColormap *colormap;
5559 GdkICAttr *attr = xic_attr;
5560 int attrmask = (int)GDK_IC_ALL_REQ;
5561 GtkWidget *widget = gui.drawarea;
5563 attr->style = (GdkIMStyle)xim_input_style;
5564 attr->client_window = gui.mainwin->window;
5566 if ((colormap = gtk_widget_get_colormap(widget)) !=
5567 gtk_widget_get_default_colormap())
5569 attrmask |= (int)GDK_IC_PREEDIT_COLORMAP;
5570 attr->preedit_colormap = colormap;
5572 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND;
5573 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND;
5574 attr->preedit_foreground = widget->style->fg[GTK_STATE_NORMAL];
5575 attr->preedit_background = widget->style->base[GTK_STATE_NORMAL];
5577 #ifdef FEAT_XFONTSET
5578 if ((xim_input_style & (int)GDK_IM_PREEDIT_MASK)
5579 == (int)GDK_IM_PREEDIT_POSITION)
5581 if (gui.fontset == NOFONTSET
5582 || gui.fontset->type != GDK_FONT_FONTSET)
5584 EMSG(_(e_overthespot));
5586 else
5588 gdk_window_get_size(widget->window, &width, &height);
5590 attrmask |= (int)GDK_IC_PREEDIT_POSITION_REQ;
5591 attr->spot_location.x = TEXT_X(0);
5592 attr->spot_location.y = TEXT_Y(0);
5593 attr->preedit_area.x = gui.border_offset;
5594 attr->preedit_area.y = gui.border_offset;
5595 attr->preedit_area.width = width - 2*gui.border_offset;
5596 attr->preedit_area.height = height - 2*gui.border_offset;
5597 attr->preedit_fontset = gui.fontset;
5601 if ((xim_input_style & (int)GDK_IM_STATUS_MASK)
5602 == (int)GDK_IM_STATUS_AREA)
5604 if (gui.fontset == NOFONTSET
5605 || gui.fontset->type != GDK_FONT_FONTSET)
5607 EMSG(_(e_overthespot));
5609 else
5611 gdk_window_get_size(gui.mainwin->window, &width, &height);
5612 attrmask |= (int)GDK_IC_STATUS_AREA_REQ;
5613 attr->status_area.x = 0;
5614 attr->status_area.y = height - gui.char_height - 1;
5615 attr->status_area.width = width;
5616 attr->status_area.height = gui.char_height;
5617 attr->status_fontset = gui.fontset;
5620 else if ((xim_input_style & (int)GDK_IM_STATUS_MASK)
5621 == (int)GDK_IM_STATUS_CALLBACKS)
5623 /* FIXME */
5625 #endif
5627 xic = gdk_ic_new(attr, (GdkICAttributesType)attrmask);
5629 if (xic == NULL)
5630 EMSG(_(e_xim));
5631 else
5633 mask = (int)gdk_window_get_events(widget->window);
5634 mask |= (int)gdk_ic_get_events(xic);
5635 gdk_window_set_events(widget->window, (GdkEventMask)mask);
5636 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)
5637 preedit_callback_setup(xic);
5638 reset_state_setup(xic);
5643 void
5644 im_shutdown(void)
5646 #ifdef XIM_DEBUG
5647 xim_log("im_shutdown()\n");
5648 #endif
5650 if (xic != NULL)
5652 gdk_im_end();
5653 gdk_ic_destroy(xic);
5654 xic = NULL;
5656 xim_is_active = FALSE;
5657 xim_can_preediting = FALSE;
5658 preedit_start_col = MAXCOL;
5659 xim_has_preediting = FALSE;
5662 #endif /* FEAT_GUI_GTK */
5665 xim_get_status_area_height()
5667 #ifdef FEAT_GUI_GTK
5668 if (xim_input_style & (int)GDK_IM_STATUS_AREA)
5669 return gui.char_height;
5670 #else
5671 if (status_area_enabled)
5672 return gui.char_height;
5673 #endif
5674 return 0;
5678 * Get IM status. When IM is on, return TRUE. Else return FALSE.
5679 * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is
5680 * active, when not having focus XIM may still be active (e.g., when using a
5681 * tear-off menu item).
5684 im_get_status()
5686 # ifdef FEAT_GUI_GTK
5687 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)
5688 return xim_can_preediting;
5689 # endif
5690 return xim_has_focus;
5693 # endif /* !HAVE_GTK2 */
5695 # if defined(FEAT_GUI_GTK) || defined(PROTO)
5697 im_is_preediting()
5699 return xim_has_preediting;
5701 # endif
5702 #endif /* FEAT_XIM */
5704 #if defined(FEAT_MBYTE) || defined(PROTO)
5707 * Setup "vcp" for conversion from "from" to "to".
5708 * The names must have been made canonical with enc_canonize().
5709 * vcp->vc_type must have been initialized to CONV_NONE.
5710 * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8
5711 * instead).
5712 * Afterwards invoke with "from" and "to" equal to NULL to cleanup.
5713 * Return FAIL when conversion is not supported, OK otherwise.
5716 convert_setup(vcp, from, to)
5717 vimconv_T *vcp;
5718 char_u *from;
5719 char_u *to;
5721 int from_prop;
5722 int to_prop;
5724 /* Reset to no conversion. */
5725 # ifdef USE_ICONV
5726 if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1)
5727 iconv_close(vcp->vc_fd);
5728 # endif
5729 vcp->vc_type = CONV_NONE;
5730 vcp->vc_factor = 1;
5731 vcp->vc_fail = FALSE;
5733 /* No conversion when one of the names is empty or they are equal. */
5734 if (from == NULL || *from == NUL || to == NULL || *to == NUL
5735 || STRCMP(from, to) == 0)
5736 return OK;
5738 from_prop = enc_canon_props(from);
5739 to_prop = enc_canon_props(to);
5740 if ((from_prop & ENC_LATIN1) && (to_prop & ENC_UNICODE))
5742 /* Internal latin1 -> utf-8 conversion. */
5743 vcp->vc_type = CONV_TO_UTF8;
5744 vcp->vc_factor = 2; /* up to twice as long */
5746 else if ((from_prop & ENC_LATIN9) && (to_prop & ENC_UNICODE))
5748 /* Internal latin9 -> utf-8 conversion. */
5749 vcp->vc_type = CONV_9_TO_UTF8;
5750 vcp->vc_factor = 3; /* up to three as long (euro sign) */
5752 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN1))
5754 /* Internal utf-8 -> latin1 conversion. */
5755 vcp->vc_type = CONV_TO_LATIN1;
5757 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_LATIN9))
5759 /* Internal utf-8 -> latin9 conversion. */
5760 vcp->vc_type = CONV_TO_LATIN9;
5762 #ifdef WIN3264
5763 /* Win32-specific codepage <-> codepage conversion without iconv. */
5764 else if (((from_prop & ENC_UNICODE) || encname2codepage(from) > 0)
5765 && ((to_prop & ENC_UNICODE) || encname2codepage(to) > 0))
5767 vcp->vc_type = CONV_CODEPAGE;
5768 vcp->vc_factor = 2; /* up to twice as long */
5769 vcp->vc_cpfrom = (from_prop & ENC_UNICODE) ? 0 : encname2codepage(from);
5770 vcp->vc_cpto = (to_prop & ENC_UNICODE) ? 0 : encname2codepage(to);
5772 #endif
5773 #ifdef MACOS_X
5774 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1))
5776 vcp->vc_type = CONV_MAC_LATIN1;
5778 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_UNICODE))
5780 vcp->vc_type = CONV_MAC_UTF8;
5781 vcp->vc_factor = 2; /* up to twice as long */
5783 else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN))
5785 vcp->vc_type = CONV_LATIN1_MAC;
5787 else if ((from_prop & ENC_UNICODE) && (to_prop & ENC_MACROMAN))
5789 vcp->vc_type = CONV_UTF8_MAC;
5791 #endif
5792 # ifdef USE_ICONV
5793 else
5795 /* Use iconv() for conversion. */
5796 vcp->vc_fd = (iconv_t)my_iconv_open(
5797 (to_prop & ENC_UNICODE) ? (char_u *)"utf-8" : to,
5798 (from_prop & ENC_UNICODE) ? (char_u *)"utf-8" : from);
5799 if (vcp->vc_fd != (iconv_t)-1)
5801 vcp->vc_type = CONV_ICONV;
5802 vcp->vc_factor = 4; /* could be longer too... */
5805 # endif
5806 if (vcp->vc_type == CONV_NONE)
5807 return FAIL;
5809 return OK;
5812 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \
5813 || defined(MSDOS) || defined(PROTO)
5815 * Do conversion on typed input characters in-place.
5816 * The input and output are not NUL terminated!
5817 * Returns the length after conversion.
5820 convert_input(ptr, len, maxlen)
5821 char_u *ptr;
5822 int len;
5823 int maxlen;
5825 return convert_input_safe(ptr, len, maxlen, NULL, NULL);
5827 #endif
5830 * Like convert_input(), but when there is an incomplete byte sequence at the
5831 * end return that as an allocated string in "restp" and set "*restlenp" to
5832 * the length. If "restp" is NULL it is not used.
5835 convert_input_safe(ptr, len, maxlen, restp, restlenp)
5836 char_u *ptr;
5837 int len;
5838 int maxlen;
5839 char_u **restp;
5840 int *restlenp;
5842 char_u *d;
5843 int dlen = len;
5844 int unconvertlen = 0;
5846 d = string_convert_ext(&input_conv, ptr, &dlen,
5847 restp == NULL ? NULL : &unconvertlen);
5848 if (d != NULL)
5850 if (dlen <= maxlen)
5852 if (unconvertlen > 0)
5854 /* Move the unconverted characters to allocated memory. */
5855 *restp = alloc(unconvertlen);
5856 if (*restp != NULL)
5857 mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen);
5858 *restlenp = unconvertlen;
5860 mch_memmove(ptr, d, dlen);
5862 else
5863 /* result is too long, keep the unconverted text (the caller must
5864 * have done something wrong!) */
5865 dlen = len;
5866 vim_free(d);
5868 return dlen;
5872 * Convert text "ptr[*lenp]" according to "vcp".
5873 * Returns the result in allocated memory and sets "*lenp".
5874 * When "lenp" is NULL, use NUL terminated strings.
5875 * Illegal chars are often changed to "?", unless vcp->vc_fail is set.
5876 * When something goes wrong, NULL is returned and "*lenp" is unchanged.
5878 char_u *
5879 string_convert(vcp, ptr, lenp)
5880 vimconv_T *vcp;
5881 char_u *ptr;
5882 int *lenp;
5884 return string_convert_ext(vcp, ptr, lenp, NULL);
5888 * Like string_convert(), but when "unconvlenp" is not NULL and there are is
5889 * an incomplete sequence at the end it is not converted and "*unconvlenp" is
5890 * set to the number of remaining bytes.
5892 char_u *
5893 string_convert_ext(vcp, ptr, lenp, unconvlenp)
5894 vimconv_T *vcp;
5895 char_u *ptr;
5896 int *lenp;
5897 int *unconvlenp;
5899 char_u *retval = NULL;
5900 char_u *d;
5901 int len;
5902 int i;
5903 int l;
5904 int c;
5906 if (lenp == NULL)
5907 len = (int)STRLEN(ptr);
5908 else
5909 len = *lenp;
5910 if (len == 0)
5911 return vim_strsave((char_u *)"");
5913 switch (vcp->vc_type)
5915 case CONV_TO_UTF8: /* latin1 to utf-8 conversion */
5916 retval = alloc(len * 2 + 1);
5917 if (retval == NULL)
5918 break;
5919 d = retval;
5920 for (i = 0; i < len; ++i)
5922 c = ptr[i];
5923 if (c < 0x80)
5924 *d++ = c;
5925 else
5927 *d++ = 0xc0 + ((unsigned)c >> 6);
5928 *d++ = 0x80 + (c & 0x3f);
5931 *d = NUL;
5932 if (lenp != NULL)
5933 *lenp = (int)(d - retval);
5934 break;
5936 case CONV_9_TO_UTF8: /* latin9 to utf-8 conversion */
5937 retval = alloc(len * 3 + 1);
5938 if (retval == NULL)
5939 break;
5940 d = retval;
5941 for (i = 0; i < len; ++i)
5943 c = ptr[i];
5944 switch (c)
5946 case 0xa4: c = 0x20ac; break; /* euro */
5947 case 0xa6: c = 0x0160; break; /* S hat */
5948 case 0xa8: c = 0x0161; break; /* S -hat */
5949 case 0xb4: c = 0x017d; break; /* Z hat */
5950 case 0xb8: c = 0x017e; break; /* Z -hat */
5951 case 0xbc: c = 0x0152; break; /* OE */
5952 case 0xbd: c = 0x0153; break; /* oe */
5953 case 0xbe: c = 0x0178; break; /* Y */
5955 d += utf_char2bytes(c, d);
5957 *d = NUL;
5958 if (lenp != NULL)
5959 *lenp = (int)(d - retval);
5960 break;
5962 case CONV_TO_LATIN1: /* utf-8 to latin1 conversion */
5963 case CONV_TO_LATIN9: /* utf-8 to latin9 conversion */
5964 retval = alloc(len + 1);
5965 if (retval == NULL)
5966 break;
5967 d = retval;
5968 for (i = 0; i < len; ++i)
5970 l = utf_ptr2len(ptr + i);
5971 if (l == 0)
5972 *d++ = NUL;
5973 else if (l == 1)
5975 if (unconvlenp != NULL && utf8len_tab[ptr[i]] > len - i)
5977 /* Incomplete sequence at the end. */
5978 *unconvlenp = len - i;
5979 break;
5981 *d++ = ptr[i];
5983 else
5985 c = utf_ptr2char(ptr + i);
5986 if (vcp->vc_type == CONV_TO_LATIN9)
5987 switch (c)
5989 case 0x20ac: c = 0xa4; break; /* euro */
5990 case 0x0160: c = 0xa6; break; /* S hat */
5991 case 0x0161: c = 0xa8; break; /* S -hat */
5992 case 0x017d: c = 0xb4; break; /* Z hat */
5993 case 0x017e: c = 0xb8; break; /* Z -hat */
5994 case 0x0152: c = 0xbc; break; /* OE */
5995 case 0x0153: c = 0xbd; break; /* oe */
5996 case 0x0178: c = 0xbe; break; /* Y */
5997 case 0xa4:
5998 case 0xa6:
5999 case 0xa8:
6000 case 0xb4:
6001 case 0xb8:
6002 case 0xbc:
6003 case 0xbd:
6004 case 0xbe: c = 0x100; break; /* not in latin9 */
6006 if (!utf_iscomposing(c)) /* skip composing chars */
6008 if (c < 0x100)
6009 *d++ = c;
6010 else if (vcp->vc_fail)
6012 vim_free(retval);
6013 return NULL;
6015 else
6017 *d++ = 0xbf;
6018 if (utf_char2cells(c) > 1)
6019 *d++ = '?';
6022 i += l - 1;
6025 *d = NUL;
6026 if (lenp != NULL)
6027 *lenp = (int)(d - retval);
6028 break;
6030 # ifdef MACOS_CONVERT
6031 case CONV_MAC_LATIN1:
6032 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6033 'm', 'l', unconvlenp);
6034 break;
6036 case CONV_LATIN1_MAC:
6037 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6038 'l', 'm', unconvlenp);
6039 break;
6041 case CONV_MAC_UTF8:
6042 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6043 'm', 'u', unconvlenp);
6044 break;
6046 case CONV_UTF8_MAC:
6047 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6048 'u', 'm', unconvlenp);
6049 break;
6050 # endif
6052 # ifdef USE_ICONV
6053 case CONV_ICONV: /* conversion with output_conv.vc_fd */
6054 retval = iconv_string(vcp, ptr, len, unconvlenp);
6055 if (retval != NULL && lenp != NULL)
6056 *lenp = (int)STRLEN(retval);
6057 break;
6058 # endif
6059 # ifdef WIN3264
6060 case CONV_CODEPAGE: /* codepage -> codepage */
6062 int retlen;
6063 int tmp_len;
6064 short_u *tmp;
6066 /* 1. codepage/UTF-8 -> ucs-2. */
6067 if (vcp->vc_cpfrom == 0)
6068 tmp_len = utf8_to_ucs2(ptr, len, NULL, NULL);
6069 else
6070 tmp_len = MultiByteToWideChar(vcp->vc_cpfrom, 0,
6071 ptr, len, 0, 0);
6072 tmp = (short_u *)alloc(sizeof(short_u) * tmp_len);
6073 if (tmp == NULL)
6074 break;
6075 if (vcp->vc_cpfrom == 0)
6076 utf8_to_ucs2(ptr, len, tmp, unconvlenp);
6077 else
6078 MultiByteToWideChar(vcp->vc_cpfrom, 0, ptr, len, tmp, tmp_len);
6080 /* 2. ucs-2 -> codepage/UTF-8. */
6081 if (vcp->vc_cpto == 0)
6082 retlen = ucs2_to_utf8(tmp, tmp_len, NULL);
6083 else
6084 retlen = WideCharToMultiByte(vcp->vc_cpto, 0,
6085 tmp, tmp_len, 0, 0, 0, 0);
6086 retval = alloc(retlen + 1);
6087 if (retval != NULL)
6089 if (vcp->vc_cpto == 0)
6090 ucs2_to_utf8(tmp, tmp_len, retval);
6091 else
6092 WideCharToMultiByte(vcp->vc_cpto, 0,
6093 tmp, tmp_len, retval, retlen, 0, 0);
6094 retval[retlen] = NUL;
6095 if (lenp != NULL)
6096 *lenp = retlen;
6098 vim_free(tmp);
6099 break;
6101 # endif
6104 return retval;
6106 #endif