feat/quickfix-title: fixed problem with compiling tiny version.
[vim_extended.git] / src / mbyte.c
blob7a3a27e06e937a7598860b7b0609ed127d86c207
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 six 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_ptr2len_len __ARGS((char_u *p, int size));
131 static int utf_ptr2cells_len __ARGS((char_u *p, int size));
132 static int dbcs_char2cells __ARGS((int c));
133 static int dbcs_ptr2cells_len __ARGS((char_u *p, int size));
134 static int dbcs_ptr2char __ARGS((char_u *p));
137 * Lookup table to quickly get the length in bytes of a UTF-8 character from
138 * the first byte of a UTF-8 string.
139 * Bytes which are illegal when used as the first byte have a 1.
140 * The NUL byte has length 1.
142 static char utf8len_tab[256] =
144 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,
145 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,
146 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,
147 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,
148 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,
149 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,
150 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,
151 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,
155 * Like utf8len_tab above, but using a zero for illegal lead bytes.
157 static char utf8len_tab_zero[256] =
159 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,
160 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,
161 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,
162 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,
163 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
164 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
165 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,
166 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6,0,0,
170 * XIM often causes trouble. Define XIM_DEBUG to get a log of XIM callbacks
171 * in the "xim.log" file.
173 /* #define XIM_DEBUG */
174 #ifdef XIM_DEBUG
175 static void
176 xim_log(char *s, ...)
178 va_list arglist;
179 static FILE *fd = NULL;
181 if (fd == (FILE *)-1)
182 return;
183 if (fd == NULL)
185 fd = mch_fopen("xim.log", "w");
186 if (fd == NULL)
188 EMSG("Cannot open xim.log");
189 fd = (FILE *)-1;
190 return;
194 va_start(arglist, s);
195 vfprintf(fd, s, arglist);
196 va_end(arglist);
198 #endif
200 #endif
202 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO)
204 * Canonical encoding names and their properties.
205 * "iso-8859-n" is handled by enc_canonize() directly.
207 static struct
208 { char *name; int prop; int codepage;}
209 enc_canon_table[] =
211 #define IDX_LATIN_1 0
212 {"latin1", ENC_8BIT + ENC_LATIN1, 1252},
213 #define IDX_ISO_2 1
214 {"iso-8859-2", ENC_8BIT, 0},
215 #define IDX_ISO_3 2
216 {"iso-8859-3", ENC_8BIT, 0},
217 #define IDX_ISO_4 3
218 {"iso-8859-4", ENC_8BIT, 0},
219 #define IDX_ISO_5 4
220 {"iso-8859-5", ENC_8BIT, 0},
221 #define IDX_ISO_6 5
222 {"iso-8859-6", ENC_8BIT, 0},
223 #define IDX_ISO_7 6
224 {"iso-8859-7", ENC_8BIT, 0},
225 #define IDX_ISO_8 7
226 {"iso-8859-8", ENC_8BIT, 0},
227 #define IDX_ISO_9 8
228 {"iso-8859-9", ENC_8BIT, 0},
229 #define IDX_ISO_10 9
230 {"iso-8859-10", ENC_8BIT, 0},
231 #define IDX_ISO_11 10
232 {"iso-8859-11", ENC_8BIT, 0},
233 #define IDX_ISO_13 11
234 {"iso-8859-13", ENC_8BIT, 0},
235 #define IDX_ISO_14 12
236 {"iso-8859-14", ENC_8BIT, 0},
237 #define IDX_ISO_15 13
238 {"iso-8859-15", ENC_8BIT + ENC_LATIN9, 0},
239 #define IDX_KOI8_R 14
240 {"koi8-r", ENC_8BIT, 0},
241 #define IDX_KOI8_U 15
242 {"koi8-u", ENC_8BIT, 0},
243 #define IDX_UTF8 16
244 {"utf-8", ENC_UNICODE, 0},
245 #define IDX_UCS2 17
246 {"ucs-2", ENC_UNICODE + ENC_ENDIAN_B + ENC_2BYTE, 0},
247 #define IDX_UCS2LE 18
248 {"ucs-2le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2BYTE, 0},
249 #define IDX_UTF16 19
250 {"utf-16", ENC_UNICODE + ENC_ENDIAN_B + ENC_2WORD, 0},
251 #define IDX_UTF16LE 20
252 {"utf-16le", ENC_UNICODE + ENC_ENDIAN_L + ENC_2WORD, 0},
253 #define IDX_UCS4 21
254 {"ucs-4", ENC_UNICODE + ENC_ENDIAN_B + ENC_4BYTE, 0},
255 #define IDX_UCS4LE 22
256 {"ucs-4le", ENC_UNICODE + ENC_ENDIAN_L + ENC_4BYTE, 0},
258 /* For debugging DBCS encoding on Unix. */
259 #define IDX_DEBUG 23
260 {"debug", ENC_DBCS, DBCS_DEBUG},
261 #define IDX_EUC_JP 24
262 {"euc-jp", ENC_DBCS, DBCS_JPNU},
263 #define IDX_SJIS 25
264 {"sjis", ENC_DBCS, DBCS_JPN},
265 #define IDX_EUC_KR 26
266 {"euc-kr", ENC_DBCS, DBCS_KORU},
267 #define IDX_EUC_CN 27
268 {"euc-cn", ENC_DBCS, DBCS_CHSU},
269 #define IDX_EUC_TW 28
270 {"euc-tw", ENC_DBCS, DBCS_CHTU},
271 #define IDX_BIG5 29
272 {"big5", ENC_DBCS, DBCS_CHT},
274 /* MS-DOS and MS-Windows codepages are included here, so that they can be
275 * used on Unix too. Most of them are similar to ISO-8859 encodings, but
276 * not exactly the same. */
277 #define IDX_CP437 30
278 {"cp437", ENC_8BIT, 437}, /* like iso-8859-1 */
279 #define IDX_CP737 31
280 {"cp737", ENC_8BIT, 737}, /* like iso-8859-7 */
281 #define IDX_CP775 32
282 {"cp775", ENC_8BIT, 775}, /* Baltic */
283 #define IDX_CP850 33
284 {"cp850", ENC_8BIT, 850}, /* like iso-8859-4 */
285 #define IDX_CP852 34
286 {"cp852", ENC_8BIT, 852}, /* like iso-8859-1 */
287 #define IDX_CP855 35
288 {"cp855", ENC_8BIT, 855}, /* like iso-8859-2 */
289 #define IDX_CP857 36
290 {"cp857", ENC_8BIT, 857}, /* like iso-8859-5 */
291 #define IDX_CP860 37
292 {"cp860", ENC_8BIT, 860}, /* like iso-8859-9 */
293 #define IDX_CP861 38
294 {"cp861", ENC_8BIT, 861}, /* like iso-8859-1 */
295 #define IDX_CP862 39
296 {"cp862", ENC_8BIT, 862}, /* like iso-8859-1 */
297 #define IDX_CP863 40
298 {"cp863", ENC_8BIT, 863}, /* like iso-8859-8 */
299 #define IDX_CP865 41
300 {"cp865", ENC_8BIT, 865}, /* like iso-8859-1 */
301 #define IDX_CP866 42
302 {"cp866", ENC_8BIT, 866}, /* like iso-8859-5 */
303 #define IDX_CP869 43
304 {"cp869", ENC_8BIT, 869}, /* like iso-8859-7 */
305 #define IDX_CP874 44
306 {"cp874", ENC_8BIT, 874}, /* Thai */
307 #define IDX_CP932 45
308 {"cp932", ENC_DBCS, DBCS_JPN},
309 #define IDX_CP936 46
310 {"cp936", ENC_DBCS, DBCS_CHS},
311 #define IDX_CP949 47
312 {"cp949", ENC_DBCS, DBCS_KOR},
313 #define IDX_CP950 48
314 {"cp950", ENC_DBCS, DBCS_CHT},
315 #define IDX_CP1250 49
316 {"cp1250", ENC_8BIT, 1250}, /* Czech, Polish, etc. */
317 #define IDX_CP1251 50
318 {"cp1251", ENC_8BIT, 1251}, /* Cyrillic */
319 /* cp1252 is considered to be equal to latin1 */
320 #define IDX_CP1253 51
321 {"cp1253", ENC_8BIT, 1253}, /* Greek */
322 #define IDX_CP1254 52
323 {"cp1254", ENC_8BIT, 1254}, /* Turkish */
324 #define IDX_CP1255 53
325 {"cp1255", ENC_8BIT, 1255}, /* Hebrew */
326 #define IDX_CP1256 54
327 {"cp1256", ENC_8BIT, 1256}, /* Arabic */
328 #define IDX_CP1257 55
329 {"cp1257", ENC_8BIT, 1257}, /* Baltic */
330 #define IDX_CP1258 56
331 {"cp1258", ENC_8BIT, 1258}, /* Vietnamese */
333 #define IDX_MACROMAN 57
334 {"macroman", ENC_8BIT + ENC_MACROMAN, 0}, /* Mac OS */
335 #define IDX_DECMCS 58
336 {"dec-mcs", ENC_8BIT, 0}, /* DEC MCS */
337 #define IDX_HPROMAN8 59
338 {"hp-roman8", ENC_8BIT, 0}, /* HP Roman8 */
339 #define IDX_COUNT 60
343 * Aliases for encoding names.
345 static struct
346 { char *name; int canon;}
347 enc_alias_table[] =
349 {"ansi", IDX_LATIN_1},
350 {"iso-8859-1", IDX_LATIN_1},
351 {"latin2", IDX_ISO_2},
352 {"latin3", IDX_ISO_3},
353 {"latin4", IDX_ISO_4},
354 {"cyrillic", IDX_ISO_5},
355 {"arabic", IDX_ISO_6},
356 {"greek", IDX_ISO_7},
357 #ifdef WIN3264
358 {"hebrew", IDX_CP1255},
359 #else
360 {"hebrew", IDX_ISO_8},
361 #endif
362 {"latin5", IDX_ISO_9},
363 {"turkish", IDX_ISO_9}, /* ? */
364 {"latin6", IDX_ISO_10},
365 {"nordic", IDX_ISO_10}, /* ? */
366 {"thai", IDX_ISO_11}, /* ? */
367 {"latin7", IDX_ISO_13},
368 {"latin8", IDX_ISO_14},
369 {"latin9", IDX_ISO_15},
370 {"utf8", IDX_UTF8},
371 {"unicode", IDX_UCS2},
372 {"ucs2", IDX_UCS2},
373 {"ucs2be", IDX_UCS2},
374 {"ucs-2be", IDX_UCS2},
375 {"ucs2le", IDX_UCS2LE},
376 {"utf16", IDX_UTF16},
377 {"utf16be", IDX_UTF16},
378 {"utf-16be", IDX_UTF16},
379 {"utf16le", IDX_UTF16LE},
380 {"ucs4", IDX_UCS4},
381 {"ucs4be", IDX_UCS4},
382 {"ucs-4be", IDX_UCS4},
383 {"ucs4le", IDX_UCS4LE},
384 {"utf32", IDX_UCS4},
385 {"utf-32", IDX_UCS4},
386 {"utf32be", IDX_UCS4},
387 {"utf-32be", IDX_UCS4},
388 {"utf32le", IDX_UCS4LE},
389 {"utf-32le", IDX_UCS4LE},
390 {"932", IDX_CP932},
391 {"949", IDX_CP949},
392 {"936", IDX_CP936},
393 {"gbk", IDX_CP936},
394 {"950", IDX_CP950},
395 {"eucjp", IDX_EUC_JP},
396 {"unix-jis", IDX_EUC_JP},
397 {"ujis", IDX_EUC_JP},
398 {"shift-jis", IDX_SJIS},
399 {"euckr", IDX_EUC_KR},
400 {"5601", IDX_EUC_KR}, /* Sun: KS C 5601 */
401 {"euccn", IDX_EUC_CN},
402 {"gb2312", IDX_EUC_CN},
403 {"euctw", IDX_EUC_TW},
404 #if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS)
405 {"japan", IDX_CP932},
406 {"korea", IDX_CP949},
407 {"prc", IDX_CP936},
408 {"chinese", IDX_CP936},
409 {"taiwan", IDX_CP950},
410 {"big5", IDX_CP950},
411 #else
412 {"japan", IDX_EUC_JP},
413 {"korea", IDX_EUC_KR},
414 {"prc", IDX_EUC_CN},
415 {"chinese", IDX_EUC_CN},
416 {"taiwan", IDX_EUC_TW},
417 {"cp950", IDX_BIG5},
418 {"950", IDX_BIG5},
419 #endif
420 {"mac", IDX_MACROMAN},
421 {"mac-roman", IDX_MACROMAN},
422 {NULL, 0}
425 #ifndef CP_UTF8
426 # define CP_UTF8 65001 /* magic number from winnls.h */
427 #endif
430 * Find encoding "name" in the list of canonical encoding names.
431 * Returns -1 if not found.
433 static int
434 enc_canon_search(name)
435 char_u *name;
437 int i;
439 for (i = 0; i < IDX_COUNT; ++i)
440 if (STRCMP(name, enc_canon_table[i].name) == 0)
441 return i;
442 return -1;
445 #endif
447 #if defined(FEAT_MBYTE) || defined(PROTO)
450 * Find canonical encoding "name" in the list and return its properties.
451 * Returns 0 if not found.
454 enc_canon_props(name)
455 char_u *name;
457 int i;
459 i = enc_canon_search(name);
460 if (i >= 0)
461 return enc_canon_table[i].prop;
462 #ifdef WIN3264
463 if (name[0] == 'c' && name[1] == 'p' && VIM_ISDIGIT(name[2]))
465 CPINFO cpinfo;
467 /* Get info on this codepage to find out what it is. */
468 if (GetCPInfo(atoi(name + 2), &cpinfo) != 0)
470 if (cpinfo.MaxCharSize == 1) /* some single-byte encoding */
471 return ENC_8BIT;
472 if (cpinfo.MaxCharSize == 2
473 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0))
474 /* must be a DBCS encoding */
475 return ENC_DBCS;
477 return 0;
479 #endif
480 if (STRNCMP(name, "2byte-", 6) == 0)
481 return ENC_DBCS;
482 if (STRNCMP(name, "8bit-", 5) == 0 || STRNCMP(name, "iso-8859-", 9) == 0)
483 return ENC_8BIT;
484 return 0;
488 * Set up for using multi-byte characters.
489 * Called in three cases:
490 * - by main() to initialize (p_enc == NULL)
491 * - by set_init_1() after 'encoding' was set to its default.
492 * - by do_set() when 'encoding' has been set.
493 * p_enc must have been passed through enc_canonize() already.
494 * Sets the "enc_unicode", "enc_utf8", "enc_dbcs" and "has_mbyte" flags.
495 * Fills mb_bytelen_tab[] and returns NULL when there are no problems.
496 * When there is something wrong: Returns an error message and doesn't change
497 * anything.
499 char_u *
500 mb_init()
502 int i;
503 int idx;
504 int n;
505 int enc_dbcs_new = 0;
506 #if defined(USE_ICONV) && !defined(WIN3264) && !defined(WIN32UNIX) \
507 && !defined(MACOS)
508 # define LEN_FROM_CONV
509 vimconv_T vimconv;
510 char_u *p;
511 #endif
513 if (p_enc == NULL)
515 /* Just starting up: set the whole table to one's. */
516 for (i = 0; i < 256; ++i)
517 mb_bytelen_tab[i] = 1;
518 input_conv.vc_type = CONV_NONE;
519 input_conv.vc_factor = 1;
520 output_conv.vc_type = CONV_NONE;
521 return NULL;
524 #ifdef WIN3264
525 if (p_enc[0] == 'c' && p_enc[1] == 'p' && VIM_ISDIGIT(p_enc[2]))
527 CPINFO cpinfo;
529 /* Get info on this codepage to find out what it is. */
530 if (GetCPInfo(atoi(p_enc + 2), &cpinfo) != 0)
532 if (cpinfo.MaxCharSize == 1)
534 /* some single-byte encoding */
535 enc_unicode = 0;
536 enc_utf8 = FALSE;
538 else if (cpinfo.MaxCharSize == 2
539 && (cpinfo.LeadByte[0] != 0 || cpinfo.LeadByte[1] != 0))
541 /* must be a DBCS encoding, check below */
542 enc_dbcs_new = atoi(p_enc + 2);
544 else
545 goto codepage_invalid;
547 else if (GetLastError() == ERROR_INVALID_PARAMETER)
549 codepage_invalid:
550 return (char_u *)N_("E543: Not a valid codepage");
553 #endif
554 else if (STRNCMP(p_enc, "8bit-", 5) == 0
555 || STRNCMP(p_enc, "iso-8859-", 9) == 0)
557 /* Accept any "8bit-" or "iso-8859-" name. */
558 enc_unicode = 0;
559 enc_utf8 = FALSE;
561 else if (STRNCMP(p_enc, "2byte-", 6) == 0)
563 #ifdef WIN3264
564 /* Windows: accept only valid codepage numbers, check below. */
565 if (p_enc[6] != 'c' || p_enc[7] != 'p'
566 || (enc_dbcs_new = atoi(p_enc + 8)) == 0)
567 return e_invarg;
568 #else
569 /* Unix: accept any "2byte-" name, assume current locale. */
570 enc_dbcs_new = DBCS_2BYTE;
571 #endif
573 else if ((idx = enc_canon_search(p_enc)) >= 0)
575 i = enc_canon_table[idx].prop;
576 if (i & ENC_UNICODE)
578 /* Unicode */
579 enc_utf8 = TRUE;
580 if (i & (ENC_2BYTE | ENC_2WORD))
581 enc_unicode = 2;
582 else if (i & ENC_4BYTE)
583 enc_unicode = 4;
584 else
585 enc_unicode = 0;
587 else if (i & ENC_DBCS)
589 /* 2byte, handle below */
590 enc_dbcs_new = enc_canon_table[idx].codepage;
592 else
594 /* Must be 8-bit. */
595 enc_unicode = 0;
596 enc_utf8 = FALSE;
599 else /* Don't know what encoding this is, reject it. */
600 return e_invarg;
602 if (enc_dbcs_new != 0)
604 #ifdef WIN3264
605 /* Check if the DBCS code page is OK. */
606 if (!IsValidCodePage(enc_dbcs_new))
607 goto codepage_invalid;
608 #endif
609 enc_unicode = 0;
610 enc_utf8 = FALSE;
612 enc_dbcs = enc_dbcs_new;
613 has_mbyte = (enc_dbcs != 0 || enc_utf8);
615 #ifdef WIN3264
616 enc_codepage = encname2codepage(p_enc);
617 enc_latin9 = (STRCMP(p_enc, "iso-8859-15") == 0);
618 #endif
620 /* Detect an encoding that uses latin1 characters. */
621 enc_latin1like = (enc_utf8 || STRCMP(p_enc, "latin1") == 0
622 || STRCMP(p_enc, "iso-8859-15") == 0);
625 * Set the function pointers.
627 if (enc_utf8)
629 mb_ptr2len = utfc_ptr2len;
630 mb_ptr2len_len = utfc_ptr2len_len;
631 mb_char2len = utf_char2len;
632 mb_char2bytes = utf_char2bytes;
633 mb_ptr2cells = utf_ptr2cells;
634 mb_ptr2cells_len = utf_ptr2cells_len;
635 mb_char2cells = utf_char2cells;
636 mb_off2cells = utf_off2cells;
637 mb_ptr2char = utf_ptr2char;
638 mb_head_off = utf_head_off;
640 else if (enc_dbcs != 0)
642 mb_ptr2len = dbcs_ptr2len;
643 mb_ptr2len_len = dbcs_ptr2len_len;
644 mb_char2len = dbcs_char2len;
645 mb_char2bytes = dbcs_char2bytes;
646 mb_ptr2cells = dbcs_ptr2cells;
647 mb_ptr2cells_len = dbcs_ptr2cells_len;
648 mb_char2cells = dbcs_char2cells;
649 mb_off2cells = dbcs_off2cells;
650 mb_ptr2char = dbcs_ptr2char;
651 mb_head_off = dbcs_head_off;
653 else
655 mb_ptr2len = latin_ptr2len;
656 mb_ptr2len_len = latin_ptr2len_len;
657 mb_char2len = latin_char2len;
658 mb_char2bytes = latin_char2bytes;
659 mb_ptr2cells = latin_ptr2cells;
660 mb_ptr2cells_len = latin_ptr2cells_len;
661 mb_char2cells = latin_char2cells;
662 mb_off2cells = latin_off2cells;
663 mb_ptr2char = latin_ptr2char;
664 mb_head_off = latin_head_off;
668 * Fill the mb_bytelen_tab[] for MB_BYTE2LEN().
670 #ifdef LEN_FROM_CONV
671 /* When 'encoding' is different from the current locale mblen() won't
672 * work. Use conversion to "utf-8" instead. */
673 vimconv.vc_type = CONV_NONE;
674 if (enc_dbcs)
676 p = enc_locale();
677 if (p == NULL || STRCMP(p, p_enc) != 0)
679 convert_setup(&vimconv, p_enc, (char_u *)"utf-8");
680 vimconv.vc_fail = TRUE;
682 vim_free(p);
684 #endif
686 for (i = 0; i < 256; ++i)
688 /* Our own function to reliably check the length of UTF-8 characters,
689 * independent of mblen(). */
690 if (enc_utf8)
691 n = utf8len_tab[i];
692 else if (enc_dbcs == 0)
693 n = 1;
694 else
696 #if defined(WIN3264) || defined(WIN32UNIX)
697 /* enc_dbcs is set by setting 'fileencoding'. It becomes a Windows
698 * CodePage identifier, which we can pass directly in to Windows
699 * API */
700 n = IsDBCSLeadByteEx(enc_dbcs, (BYTE)i) ? 2 : 1;
701 #else
702 # if defined(MACOS) || defined(__amigaos4__)
704 * if mblen() is not available, character which MSB is turned on
705 * are treated as leading byte character. (note : This assumption
706 * is not always true.)
708 n = (i & 0x80) ? 2 : 1;
709 # else
710 char buf[MB_MAXBYTES];
711 # ifdef X_LOCALE
712 # ifndef mblen
713 # define mblen _Xmblen
714 # endif
715 # endif
716 if (i == NUL) /* just in case mblen() can't handle "" */
717 n = 1;
718 else
720 buf[0] = i;
721 buf[1] = 0;
722 #ifdef LEN_FROM_CONV
723 if (vimconv.vc_type != CONV_NONE)
726 * string_convert() should fail when converting the first
727 * byte of a double-byte character.
729 p = string_convert(&vimconv, (char_u *)buf, NULL);
730 if (p != NULL)
732 vim_free(p);
733 n = 1;
735 else
736 n = 2;
738 else
739 #endif
742 * mblen() should return -1 for invalid (means the leading
743 * multibyte) character. However there are some platforms
744 * where mblen() returns 0 for invalid character.
745 * Therefore, following condition includes 0.
747 ignored = mblen(NULL, 0); /* First reset the state. */
748 if (mblen(buf, (size_t)1) <= 0)
749 n = 2;
750 else
751 n = 1;
754 # endif
755 #endif
758 mb_bytelen_tab[i] = n;
761 #ifdef LEN_FROM_CONV
762 convert_setup(&vimconv, NULL, NULL);
763 #endif
765 /* The cell width depends on the type of multi-byte characters. */
766 (void)init_chartab();
768 /* When enc_utf8 is set or reset, (de)allocate ScreenLinesUC[] */
769 screenalloc(FALSE);
771 /* When using Unicode, set default for 'fileencodings'. */
772 if (enc_utf8 && !option_was_set((char_u *)"fencs"))
773 set_string_option_direct((char_u *)"fencs", -1,
774 (char_u *)"ucs-bom,utf-8,default,latin1", OPT_FREE, 0);
776 #if defined(HAVE_BIND_TEXTDOMAIN_CODESET) && defined(FEAT_GETTEXT)
777 /* GNU gettext 0.10.37 supports this feature: set the codeset used for
778 * translated messages independently from the current locale. */
779 (void)bind_textdomain_codeset(VIMPACKAGE,
780 enc_utf8 ? "utf-8" : (char *)p_enc);
781 #endif
783 #ifdef WIN32
784 /* When changing 'encoding' while starting up, then convert the command
785 * line arguments from the active codepage to 'encoding'. */
786 if (starting != 0)
787 fix_arg_enc();
788 #endif
790 #ifdef FEAT_AUTOCMD
791 /* Fire an autocommand to let people do custom font setup. This must be
792 * after Vim has been setup for the new encoding. */
793 apply_autocmds(EVENT_ENCODINGCHANGED, NULL, (char_u *)"", FALSE, curbuf);
794 #endif
796 #ifdef FEAT_SPELL
797 /* Need to reload spell dictionaries */
798 spell_reload();
799 #endif
801 return NULL;
805 * Return the size of the BOM for the current buffer:
806 * 0 - no BOM
807 * 2 - UCS-2 or UTF-16 BOM
808 * 4 - UCS-4 BOM
809 * 3 - UTF-8 BOM
812 bomb_size()
814 int n = 0;
816 if (curbuf->b_p_bomb && !curbuf->b_p_bin)
818 if (*curbuf->b_p_fenc == NUL)
820 if (enc_utf8)
822 if (enc_unicode != 0)
823 n = enc_unicode;
824 else
825 n = 3;
828 else if (STRCMP(curbuf->b_p_fenc, "utf-8") == 0)
829 n = 3;
830 else if (STRNCMP(curbuf->b_p_fenc, "ucs-2", 5) == 0
831 || STRNCMP(curbuf->b_p_fenc, "utf-16", 6) == 0)
832 n = 2;
833 else if (STRNCMP(curbuf->b_p_fenc, "ucs-4", 5) == 0)
834 n = 4;
836 return n;
840 * Get class of pointer:
841 * 0 for blank or NUL
842 * 1 for punctuation
843 * 2 for an (ASCII) word character
844 * >2 for other word characters
847 mb_get_class(p)
848 char_u *p;
850 if (MB_BYTE2LEN(p[0]) == 1)
852 if (p[0] == NUL || vim_iswhite(p[0]))
853 return 0;
854 if (vim_iswordc(p[0]))
855 return 2;
856 return 1;
858 if (enc_dbcs != 0 && p[0] != NUL && p[1] != NUL)
859 return dbcs_class(p[0], p[1]);
860 if (enc_utf8)
861 return utf_class(utf_ptr2char(p));
862 return 0;
866 * Get class of a double-byte character. This always returns 3 or bigger.
867 * TODO: Should return 1 for punctuation.
870 dbcs_class(lead, trail)
871 unsigned lead;
872 unsigned trail;
874 switch (enc_dbcs)
876 /* please add classfy routine for your language in here */
878 case DBCS_JPNU: /* ? */
879 case DBCS_JPN:
881 /* JIS code classification */
882 unsigned char lb = lead;
883 unsigned char tb = trail;
885 /* convert process code to JIS */
886 # if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS)
887 /* process code is SJIS */
888 if (lb <= 0x9f)
889 lb = (lb - 0x81) * 2 + 0x21;
890 else
891 lb = (lb - 0xc1) * 2 + 0x21;
892 if (tb <= 0x7e)
893 tb -= 0x1f;
894 else if (tb <= 0x9e)
895 tb -= 0x20;
896 else
898 tb -= 0x7e;
899 lb += 1;
901 # else
903 * XXX: Code page identification can not use with all
904 * system! So, some other encoding information
905 * will be needed.
906 * In japanese: SJIS,EUC,UNICODE,(JIS)
907 * Note that JIS-code system don't use as
908 * process code in most system because it uses
909 * escape sequences(JIS is context depend encoding).
911 /* assume process code is JAPANESE-EUC */
912 lb &= 0x7f;
913 tb &= 0x7f;
914 # endif
915 /* exceptions */
916 switch (lb << 8 | tb)
918 case 0x2121: /* ZENKAKU space */
919 return 0;
920 case 0x2122: /* KU-TEN (Japanese comma) */
921 case 0x2123: /* TOU-TEN (Japanese period) */
922 case 0x2124: /* ZENKAKU comma */
923 case 0x2125: /* ZENKAKU period */
924 return 1;
925 case 0x213c: /* prolongedsound handled as KATAKANA */
926 return 13;
928 /* sieved by KU code */
929 switch (lb)
931 case 0x21:
932 case 0x22:
933 /* special symbols */
934 return 10;
935 case 0x23:
936 /* alpha-numeric */
937 return 11;
938 case 0x24:
939 /* hiragana */
940 return 12;
941 case 0x25:
942 /* katakana */
943 return 13;
944 case 0x26:
945 /* greek */
946 return 14;
947 case 0x27:
948 /* russian */
949 return 15;
950 case 0x28:
951 /* lines */
952 return 16;
953 default:
954 /* kanji */
955 return 17;
959 case DBCS_KORU: /* ? */
960 case DBCS_KOR:
962 /* KS code classification */
963 unsigned char c1 = lead;
964 unsigned char c2 = trail;
967 * 20 : Hangul
968 * 21 : Hanja
969 * 22 : Symbols
970 * 23 : Alpha-numeric/Roman Letter (Full width)
971 * 24 : Hangul Letter(Alphabet)
972 * 25 : Roman Numeral/Greek Letter
973 * 26 : Box Drawings
974 * 27 : Unit Symbols
975 * 28 : Circled/Parenthesized Letter
976 * 29 : Hirigana/Katakana
977 * 30 : Cyrillic Letter
980 if (c1 >= 0xB0 && c1 <= 0xC8)
981 /* Hangul */
982 return 20;
983 #if defined(WIN3264) || defined(WIN32UNIX)
984 else if (c1 <= 0xA0 || c2 <= 0xA0)
985 /* Extended Hangul Region : MS UHC(Unified Hangul Code) */
986 /* c1: 0x81-0xA0 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xFE
987 * c1: 0xA1-0xC6 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xA0
989 return 20;
990 #endif
992 else if (c1 >= 0xCA && c1 <= 0xFD)
993 /* Hanja */
994 return 21;
995 else switch (c1)
997 case 0xA1:
998 case 0xA2:
999 /* Symbols */
1000 return 22;
1001 case 0xA3:
1002 /* Alpha-numeric */
1003 return 23;
1004 case 0xA4:
1005 /* Hangul Letter(Alphabet) */
1006 return 24;
1007 case 0xA5:
1008 /* Roman Numeral/Greek Letter */
1009 return 25;
1010 case 0xA6:
1011 /* Box Drawings */
1012 return 26;
1013 case 0xA7:
1014 /* Unit Symbols */
1015 return 27;
1016 case 0xA8:
1017 case 0xA9:
1018 if (c2 <= 0xAF)
1019 return 25; /* Roman Letter */
1020 else if (c2 >= 0xF6)
1021 return 22; /* Symbols */
1022 else
1023 /* Circled/Parenthesized Letter */
1024 return 28;
1025 case 0xAA:
1026 case 0xAB:
1027 /* Hirigana/Katakana */
1028 return 29;
1029 case 0xAC:
1030 /* Cyrillic Letter */
1031 return 30;
1034 default:
1035 break;
1037 return 3;
1041 * mb_char2len() function pointer.
1042 * Return length in bytes of character "c".
1043 * Returns 1 for a single-byte character.
1046 latin_char2len(c)
1047 int c UNUSED;
1049 return 1;
1052 static int
1053 dbcs_char2len(c)
1054 int c;
1056 if (c >= 0x100)
1057 return 2;
1058 return 1;
1062 * mb_char2bytes() function pointer.
1063 * Convert a character to its bytes.
1064 * Returns the length in bytes.
1067 latin_char2bytes(c, buf)
1068 int c;
1069 char_u *buf;
1071 buf[0] = c;
1072 return 1;
1075 static int
1076 dbcs_char2bytes(c, buf)
1077 int c;
1078 char_u *buf;
1080 if (c >= 0x100)
1082 buf[0] = (unsigned)c >> 8;
1083 buf[1] = c;
1084 /* Never use a NUL byte, it causes lots of trouble. It's an invalid
1085 * character anyway. */
1086 if (buf[1] == NUL)
1087 buf[1] = '\n';
1088 return 2;
1090 buf[0] = c;
1091 return 1;
1095 * mb_ptr2len() function pointer.
1096 * Get byte length of character at "*p" but stop at a NUL.
1097 * For UTF-8 this includes following composing characters.
1098 * Returns 0 when *p is NUL.
1101 latin_ptr2len(p)
1102 char_u *p;
1104 return MB_BYTE2LEN(*p);
1107 static int
1108 dbcs_ptr2len(p)
1109 char_u *p;
1111 int len;
1113 /* Check if second byte is not missing. */
1114 len = MB_BYTE2LEN(*p);
1115 if (len == 2 && p[1] == NUL)
1116 len = 1;
1117 return len;
1121 * mb_ptr2len_len() function pointer.
1122 * Like mb_ptr2len(), but limit to read "size" bytes.
1123 * Returns 0 for an empty string.
1124 * Returns 1 for an illegal char or an incomplete byte sequence.
1127 latin_ptr2len_len(p, size)
1128 char_u *p;
1129 int size;
1131 if (size < 1 || *p == NUL)
1132 return 0;
1133 return 1;
1136 static int
1137 dbcs_ptr2len_len(p, size)
1138 char_u *p;
1139 int size;
1141 int len;
1143 if (size < 1 || *p == NUL)
1144 return 0;
1145 if (size == 1)
1146 return 1;
1147 /* Check that second byte is not missing. */
1148 len = MB_BYTE2LEN(*p);
1149 if (len == 2 && p[1] == NUL)
1150 len = 1;
1151 return len;
1154 struct interval
1156 long first;
1157 long last;
1159 static int intable __ARGS((struct interval *table, size_t size, int c));
1162 * Return TRUE if "c" is in "table[size / sizeof(struct interval)]".
1164 static int
1165 intable(table, size, c)
1166 struct interval *table;
1167 size_t size;
1168 int c;
1170 int mid, bot, top;
1172 /* first quick check for Latin1 etc. characters */
1173 if (c < table[0].first)
1174 return FALSE;
1176 /* binary search in table */
1177 bot = 0;
1178 top = (int)(size / sizeof(struct interval) - 1);
1179 while (top >= bot)
1181 mid = (bot + top) / 2;
1182 if (table[mid].last < c)
1183 bot = mid + 1;
1184 else if (table[mid].first > c)
1185 top = mid - 1;
1186 else
1187 return TRUE;
1189 return FALSE;
1193 * For UTF-8 character "c" return 2 for a double-width character, 1 for others.
1194 * Returns 4 or 6 for an unprintable character.
1195 * Is only correct for characters >= 0x80.
1196 * When p_ambw is "double", return 2 for a character with East Asian Width
1197 * class 'A'(mbiguous).
1200 utf_char2cells(c)
1201 int c;
1203 /* Sorted list of non-overlapping intervals of East Asian double width
1204 * characters, generated with ../runtime/tools/unicode.vim. */
1205 static struct interval doublewidth[] =
1207 {0x1100, 0x115f},
1208 {0x11a3, 0x11a7},
1209 {0x11fa, 0x11ff},
1210 {0x2329, 0x232a},
1211 {0x2e80, 0x2e99},
1212 {0x2e9b, 0x2ef3},
1213 {0x2f00, 0x2fd5},
1214 {0x2ff0, 0x2ffb},
1215 {0x3000, 0x3029},
1216 {0x3030, 0x303e},
1217 {0x3041, 0x3096},
1218 {0x309b, 0x30ff},
1219 {0x3105, 0x312d},
1220 {0x3131, 0x318e},
1221 {0x3190, 0x31b7},
1222 {0x31c0, 0x31e3},
1223 {0x31f0, 0x321e},
1224 {0x3220, 0x3247},
1225 {0x3250, 0x32fe},
1226 {0x3300, 0x4dbf},
1227 {0x4e00, 0xa48c},
1228 {0xa490, 0xa4c6},
1229 {0xa960, 0xa97c},
1230 {0xac00, 0xd7a3},
1231 {0xd7b0, 0xd7c6},
1232 {0xd7cb, 0xd7fb},
1233 {0xf900, 0xfaff},
1234 {0xfe10, 0xfe19},
1235 {0xfe30, 0xfe52},
1236 {0xfe54, 0xfe66},
1237 {0xfe68, 0xfe6b},
1238 {0xff01, 0xff60},
1239 {0xffe0, 0xffe6},
1240 {0x1f200, 0x1f200},
1241 {0x1f210, 0x1f231},
1242 {0x1f240, 0x1f248},
1243 {0x20000, 0x2fffd},
1244 {0x30000, 0x3fffd}
1246 /* Sorted list of non-overlapping intervals of East Asian Ambiguous
1247 * characters, generated with ../runtime/tools/unicode.vim. */
1248 static struct interval ambiguous[] =
1250 {0x00a1, 0x00a1},
1251 {0x00a4, 0x00a4},
1252 {0x00a7, 0x00a8},
1253 {0x00aa, 0x00aa},
1254 {0x00ad, 0x00ae},
1255 {0x00b0, 0x00b4},
1256 {0x00b6, 0x00ba},
1257 {0x00bc, 0x00bf},
1258 {0x00c6, 0x00c6},
1259 {0x00d0, 0x00d0},
1260 {0x00d7, 0x00d8},
1261 {0x00de, 0x00e1},
1262 {0x00e6, 0x00e6},
1263 {0x00e8, 0x00ea},
1264 {0x00ec, 0x00ed},
1265 {0x00f0, 0x00f0},
1266 {0x00f2, 0x00f3},
1267 {0x00f7, 0x00fa},
1268 {0x00fc, 0x00fc},
1269 {0x00fe, 0x00fe},
1270 {0x0101, 0x0101},
1271 {0x0111, 0x0111},
1272 {0x0113, 0x0113},
1273 {0x011b, 0x011b},
1274 {0x0126, 0x0127},
1275 {0x012b, 0x012b},
1276 {0x0131, 0x0133},
1277 {0x0138, 0x0138},
1278 {0x013f, 0x0142},
1279 {0x0144, 0x0144},
1280 {0x0148, 0x014b},
1281 {0x014d, 0x014d},
1282 {0x0152, 0x0153},
1283 {0x0166, 0x0167},
1284 {0x016b, 0x016b},
1285 {0x01ce, 0x01ce},
1286 {0x01d0, 0x01d0},
1287 {0x01d2, 0x01d2},
1288 {0x01d4, 0x01d4},
1289 {0x01d6, 0x01d6},
1290 {0x01d8, 0x01d8},
1291 {0x01da, 0x01da},
1292 {0x01dc, 0x01dc},
1293 {0x0251, 0x0251},
1294 {0x0261, 0x0261},
1295 {0x02c4, 0x02c4},
1296 {0x02c7, 0x02c7},
1297 {0x02c9, 0x02cb},
1298 {0x02cd, 0x02cd},
1299 {0x02d0, 0x02d0},
1300 {0x02d8, 0x02db},
1301 {0x02dd, 0x02dd},
1302 {0x02df, 0x02df},
1303 {0x0391, 0x03a1},
1304 {0x03a3, 0x03a9},
1305 {0x03b1, 0x03c1},
1306 {0x03c3, 0x03c9},
1307 {0x0401, 0x0401},
1308 {0x0410, 0x044f},
1309 {0x0451, 0x0451},
1310 {0x2010, 0x2010},
1311 {0x2013, 0x2016},
1312 {0x2018, 0x2019},
1313 {0x201c, 0x201d},
1314 {0x2020, 0x2022},
1315 {0x2024, 0x2027},
1316 {0x2030, 0x2030},
1317 {0x2032, 0x2033},
1318 {0x2035, 0x2035},
1319 {0x203b, 0x203b},
1320 {0x203e, 0x203e},
1321 {0x2074, 0x2074},
1322 {0x207f, 0x207f},
1323 {0x2081, 0x2084},
1324 {0x20ac, 0x20ac},
1325 {0x2103, 0x2103},
1326 {0x2105, 0x2105},
1327 {0x2109, 0x2109},
1328 {0x2113, 0x2113},
1329 {0x2116, 0x2116},
1330 {0x2121, 0x2122},
1331 {0x2126, 0x2126},
1332 {0x212b, 0x212b},
1333 {0x2153, 0x2154},
1334 {0x215b, 0x215e},
1335 {0x2160, 0x216b},
1336 {0x2170, 0x2179},
1337 {0x2189, 0x2189},
1338 {0x2190, 0x2199},
1339 {0x21b8, 0x21b9},
1340 {0x21d2, 0x21d2},
1341 {0x21d4, 0x21d4},
1342 {0x21e7, 0x21e7},
1343 {0x2200, 0x2200},
1344 {0x2202, 0x2203},
1345 {0x2207, 0x2208},
1346 {0x220b, 0x220b},
1347 {0x220f, 0x220f},
1348 {0x2211, 0x2211},
1349 {0x2215, 0x2215},
1350 {0x221a, 0x221a},
1351 {0x221d, 0x2220},
1352 {0x2223, 0x2223},
1353 {0x2225, 0x2225},
1354 {0x2227, 0x222c},
1355 {0x222e, 0x222e},
1356 {0x2234, 0x2237},
1357 {0x223c, 0x223d},
1358 {0x2248, 0x2248},
1359 {0x224c, 0x224c},
1360 {0x2252, 0x2252},
1361 {0x2260, 0x2261},
1362 {0x2264, 0x2267},
1363 {0x226a, 0x226b},
1364 {0x226e, 0x226f},
1365 {0x2282, 0x2283},
1366 {0x2286, 0x2287},
1367 {0x2295, 0x2295},
1368 {0x2299, 0x2299},
1369 {0x22a5, 0x22a5},
1370 {0x22bf, 0x22bf},
1371 {0x2312, 0x2312},
1372 {0x2460, 0x24e9},
1373 {0x24eb, 0x254b},
1374 {0x2550, 0x2573},
1375 {0x2580, 0x258f},
1376 {0x2592, 0x2595},
1377 {0x25a0, 0x25a1},
1378 {0x25a3, 0x25a9},
1379 {0x25b2, 0x25b3},
1380 {0x25b6, 0x25b7},
1381 {0x25bc, 0x25bd},
1382 {0x25c0, 0x25c1},
1383 {0x25c6, 0x25c8},
1384 {0x25cb, 0x25cb},
1385 {0x25ce, 0x25d1},
1386 {0x25e2, 0x25e5},
1387 {0x25ef, 0x25ef},
1388 {0x2605, 0x2606},
1389 {0x2609, 0x2609},
1390 {0x260e, 0x260f},
1391 {0x2614, 0x2615},
1392 {0x261c, 0x261c},
1393 {0x261e, 0x261e},
1394 {0x2640, 0x2640},
1395 {0x2642, 0x2642},
1396 {0x2660, 0x2661},
1397 {0x2663, 0x2665},
1398 {0x2667, 0x266a},
1399 {0x266c, 0x266d},
1400 {0x266f, 0x266f},
1401 {0x269e, 0x269f},
1402 {0x26be, 0x26bf},
1403 {0x26c4, 0x26cd},
1404 {0x26cf, 0x26e1},
1405 {0x26e3, 0x26e3},
1406 {0x26e8, 0x26ff},
1407 {0x273d, 0x273d},
1408 {0x2757, 0x2757},
1409 {0x2776, 0x277f},
1410 {0x2b55, 0x2b59},
1411 {0x3248, 0x324f},
1412 {0xe000, 0xf8ff},
1413 {0xfffd, 0xfffd},
1414 {0x1f100, 0x1f10a},
1415 {0x1f110, 0x1f12d},
1416 {0x1f131, 0x1f131},
1417 {0x1f13d, 0x1f13d},
1418 {0x1f13f, 0x1f13f},
1419 {0x1f142, 0x1f142},
1420 {0x1f146, 0x1f146},
1421 {0x1f14a, 0x1f14e},
1422 {0x1f157, 0x1f157},
1423 {0x1f15f, 0x1f15f},
1424 {0x1f179, 0x1f179},
1425 {0x1f17b, 0x1f17c},
1426 {0x1f17f, 0x1f17f},
1427 {0x1f18a, 0x1f18d},
1428 {0x1f190, 0x1f190},
1429 {0xf0000, 0xffffd},
1430 {0x100000, 0x10fffd}
1433 if (c >= 0x100)
1435 #ifdef USE_WCHAR_FUNCTIONS
1437 * Assume the library function wcwidth() works better than our own
1438 * stuff. It should return 1 for ambiguous width chars!
1440 int n = wcwidth(c);
1442 if (n < 0)
1443 return 6; /* unprintable, displays <xxxx> */
1444 if (n > 1)
1445 return n;
1446 #else
1447 if (!utf_printable(c))
1448 return 6; /* unprintable, displays <xxxx> */
1449 if (intable(doublewidth, sizeof(doublewidth), c))
1450 return 2;
1451 #endif
1454 /* Characters below 0x100 are influenced by 'isprint' option */
1455 else if (c >= 0x80 && !vim_isprintc(c))
1456 return 4; /* unprintable, displays <xx> */
1458 if (c >= 0x80 && *p_ambw == 'd' && intable(ambiguous, sizeof(ambiguous), c))
1459 return 2;
1461 return 1;
1465 * mb_ptr2cells() function pointer.
1466 * Return the number of display cells character at "*p" occupies.
1467 * This doesn't take care of unprintable characters, use ptr2cells() for that.
1470 latin_ptr2cells(p)
1471 char_u *p UNUSED;
1473 return 1;
1477 utf_ptr2cells(p)
1478 char_u *p;
1480 int c;
1482 /* Need to convert to a wide character. */
1483 if (*p >= 0x80)
1485 c = utf_ptr2char(p);
1486 /* An illegal byte is displayed as <xx>. */
1487 if (utf_ptr2len(p) == 1 || c == NUL)
1488 return 4;
1489 /* If the char is ASCII it must be an overlong sequence. */
1490 if (c < 0x80)
1491 return char2cells(c);
1492 return utf_char2cells(c);
1494 return 1;
1498 dbcs_ptr2cells(p)
1499 char_u *p;
1501 /* Number of cells is equal to number of bytes, except for euc-jp when
1502 * the first byte is 0x8e. */
1503 if (enc_dbcs == DBCS_JPNU && *p == 0x8e)
1504 return 1;
1505 return MB_BYTE2LEN(*p);
1509 * mb_ptr2cells_len() function pointer.
1510 * Like mb_ptr2cells(), but limit string length to "size".
1511 * For an empty string or truncated character returns 1.
1514 latin_ptr2cells_len(p, size)
1515 char_u *p UNUSED;
1516 int size UNUSED;
1518 return 1;
1521 static int
1522 utf_ptr2cells_len(p, size)
1523 char_u *p;
1524 int size;
1526 int c;
1528 /* Need to convert to a wide character. */
1529 if (size > 0 && *p >= 0x80)
1531 if (utf_ptr2len_len(p, size) < utf8len_tab[*p])
1532 return 1; /* truncated */
1533 c = utf_ptr2char(p);
1534 /* An illegal byte is displayed as <xx>. */
1535 if (utf_ptr2len(p) == 1 || c == NUL)
1536 return 4;
1537 /* If the char is ASCII it must be an overlong sequence. */
1538 if (c < 0x80)
1539 return char2cells(c);
1540 return utf_char2cells(c);
1542 return 1;
1545 static int
1546 dbcs_ptr2cells_len(p, size)
1547 char_u *p;
1548 int size;
1550 /* Number of cells is equal to number of bytes, except for euc-jp when
1551 * the first byte is 0x8e. */
1552 if (size <= 1 || (enc_dbcs == DBCS_JPNU && *p == 0x8e))
1553 return 1;
1554 return MB_BYTE2LEN(*p);
1558 * mb_char2cells() function pointer.
1559 * Return the number of display cells character "c" occupies.
1560 * Only takes care of multi-byte chars, not "^C" and such.
1563 latin_char2cells(c)
1564 int c UNUSED;
1566 return 1;
1569 static int
1570 dbcs_char2cells(c)
1571 int c;
1573 /* Number of cells is equal to number of bytes, except for euc-jp when
1574 * the first byte is 0x8e. */
1575 if (enc_dbcs == DBCS_JPNU && ((unsigned)c >> 8) == 0x8e)
1576 return 1;
1577 /* use the first byte */
1578 return MB_BYTE2LEN((unsigned)c >> 8);
1582 * mb_off2cells() function pointer.
1583 * Return number of display cells for char at ScreenLines[off].
1584 * We make sure that the offset used is less than "max_off".
1587 latin_off2cells(off, max_off)
1588 unsigned off UNUSED;
1589 unsigned max_off UNUSED;
1591 return 1;
1595 dbcs_off2cells(off, max_off)
1596 unsigned off;
1597 unsigned max_off;
1599 /* never check beyond end of the line */
1600 if (off >= max_off)
1601 return 1;
1603 /* Number of cells is equal to number of bytes, except for euc-jp when
1604 * the first byte is 0x8e. */
1605 if (enc_dbcs == DBCS_JPNU && ScreenLines[off] == 0x8e)
1606 return 1;
1607 return MB_BYTE2LEN(ScreenLines[off]);
1611 utf_off2cells(off, max_off)
1612 unsigned off;
1613 unsigned max_off;
1615 return (off + 1 < max_off && ScreenLines[off + 1] == 0) ? 2 : 1;
1619 * mb_ptr2char() function pointer.
1620 * Convert a byte sequence into a character.
1623 latin_ptr2char(p)
1624 char_u *p;
1626 return *p;
1629 static int
1630 dbcs_ptr2char(p)
1631 char_u *p;
1633 if (MB_BYTE2LEN(*p) > 1 && p[1] != NUL)
1634 return (p[0] << 8) + p[1];
1635 return *p;
1639 * Convert a UTF-8 byte sequence to a wide character.
1640 * If the sequence is illegal or truncated by a NUL the first byte is
1641 * returned.
1642 * Does not include composing characters, of course.
1645 utf_ptr2char(p)
1646 char_u *p;
1648 int len;
1650 if (p[0] < 0x80) /* be quick for ASCII */
1651 return p[0];
1653 len = utf8len_tab_zero[p[0]];
1654 if (len > 1 && (p[1] & 0xc0) == 0x80)
1656 if (len == 2)
1657 return ((p[0] & 0x1f) << 6) + (p[1] & 0x3f);
1658 if ((p[2] & 0xc0) == 0x80)
1660 if (len == 3)
1661 return ((p[0] & 0x0f) << 12) + ((p[1] & 0x3f) << 6)
1662 + (p[2] & 0x3f);
1663 if ((p[3] & 0xc0) == 0x80)
1665 if (len == 4)
1666 return ((p[0] & 0x07) << 18) + ((p[1] & 0x3f) << 12)
1667 + ((p[2] & 0x3f) << 6) + (p[3] & 0x3f);
1668 if ((p[4] & 0xc0) == 0x80)
1670 if (len == 5)
1671 return ((p[0] & 0x03) << 24) + ((p[1] & 0x3f) << 18)
1672 + ((p[2] & 0x3f) << 12) + ((p[3] & 0x3f) << 6)
1673 + (p[4] & 0x3f);
1674 if ((p[5] & 0xc0) == 0x80 && len == 6)
1675 return ((p[0] & 0x01) << 30) + ((p[1] & 0x3f) << 24)
1676 + ((p[2] & 0x3f) << 18) + ((p[3] & 0x3f) << 12)
1677 + ((p[4] & 0x3f) << 6) + (p[5] & 0x3f);
1682 /* Illegal value, just return the first byte */
1683 return p[0];
1687 * Get character at **pp and advance *pp to the next character.
1688 * Note: composing characters are skipped!
1691 mb_ptr2char_adv(pp)
1692 char_u **pp;
1694 int c;
1696 c = (*mb_ptr2char)(*pp);
1697 *pp += (*mb_ptr2len)(*pp);
1698 return c;
1702 * Get character at **pp and advance *pp to the next character.
1703 * Note: composing characters are returned as separate characters.
1706 mb_cptr2char_adv(pp)
1707 char_u **pp;
1709 int c;
1711 c = (*mb_ptr2char)(*pp);
1712 if (enc_utf8)
1713 *pp += utf_ptr2len(*pp);
1714 else
1715 *pp += (*mb_ptr2len)(*pp);
1716 return c;
1719 #if defined(FEAT_ARABIC) || defined(PROTO)
1721 * Check whether we are dealing with Arabic combining characters.
1722 * Note: these are NOT really composing characters!
1725 arabic_combine(one, two)
1726 int one; /* first character */
1727 int two; /* character just after "one" */
1729 if (one == a_LAM)
1730 return arabic_maycombine(two);
1731 return FALSE;
1735 * Check whether we are dealing with a character that could be regarded as an
1736 * Arabic combining character, need to check the character before this.
1739 arabic_maycombine(two)
1740 int two;
1742 if (p_arshape && !p_tbidi)
1743 return (two == a_ALEF_MADDA
1744 || two == a_ALEF_HAMZA_ABOVE
1745 || two == a_ALEF_HAMZA_BELOW
1746 || two == a_ALEF);
1747 return FALSE;
1751 * Check if the character pointed to by "p2" is a composing character when it
1752 * comes after "p1". For Arabic sometimes "ab" is replaced with "c", which
1753 * behaves like a composing character.
1756 utf_composinglike(p1, p2)
1757 char_u *p1;
1758 char_u *p2;
1760 int c2;
1762 c2 = utf_ptr2char(p2);
1763 if (utf_iscomposing(c2))
1764 return TRUE;
1765 if (!arabic_maycombine(c2))
1766 return FALSE;
1767 return arabic_combine(utf_ptr2char(p1), c2);
1769 #endif
1772 * Convert a UTF-8 byte string to a wide character. Also get up to MAX_MCO
1773 * composing characters.
1776 utfc_ptr2char(p, pcc)
1777 char_u *p;
1778 int *pcc; /* return: composing chars, last one is 0 */
1780 int len;
1781 int c;
1782 int cc;
1783 int i = 0;
1785 c = utf_ptr2char(p);
1786 len = utf_ptr2len(p);
1788 /* Only accept a composing char when the first char isn't illegal. */
1789 if ((len > 1 || *p < 0x80)
1790 && p[len] >= 0x80
1791 && UTF_COMPOSINGLIKE(p, p + len))
1793 cc = utf_ptr2char(p + len);
1794 for (;;)
1796 pcc[i++] = cc;
1797 if (i == MAX_MCO)
1798 break;
1799 len += utf_ptr2len(p + len);
1800 if (p[len] < 0x80 || !utf_iscomposing(cc = utf_ptr2char(p + len)))
1801 break;
1805 if (i < MAX_MCO) /* last composing char must be 0 */
1806 pcc[i] = 0;
1808 return c;
1812 * Convert a UTF-8 byte string to a wide character. Also get up to MAX_MCO
1813 * composing characters. Use no more than p[maxlen].
1816 utfc_ptr2char_len(p, pcc, maxlen)
1817 char_u *p;
1818 int *pcc; /* return: composing chars, last one is 0 */
1819 int maxlen;
1821 int len;
1822 int c;
1823 int cc;
1824 int i = 0;
1826 c = utf_ptr2char(p);
1827 len = utf_ptr2len_len(p, maxlen);
1828 /* Only accept a composing char when the first char isn't illegal. */
1829 if ((len > 1 || *p < 0x80)
1830 && len < maxlen
1831 && p[len] >= 0x80
1832 && UTF_COMPOSINGLIKE(p, p + len))
1834 cc = utf_ptr2char(p + len);
1835 for (;;)
1837 pcc[i++] = cc;
1838 if (i == MAX_MCO)
1839 break;
1840 len += utf_ptr2len_len(p + len, maxlen - len);
1841 if (len >= maxlen
1842 || p[len] < 0x80
1843 || !utf_iscomposing(cc = utf_ptr2char(p + len)))
1844 break;
1848 if (i < MAX_MCO) /* last composing char must be 0 */
1849 pcc[i] = 0;
1851 return c;
1855 * Convert the character at screen position "off" to a sequence of bytes.
1856 * Includes the composing characters.
1857 * "buf" must at least have the length MB_MAXBYTES.
1858 * Returns the produced number of bytes.
1861 utfc_char2bytes(off, buf)
1862 int off;
1863 char_u *buf;
1865 int len;
1866 int i;
1868 len = utf_char2bytes(ScreenLinesUC[off], buf);
1869 for (i = 0; i < Screen_mco; ++i)
1871 if (ScreenLinesC[i][off] == 0)
1872 break;
1873 len += utf_char2bytes(ScreenLinesC[i][off], buf + len);
1875 return len;
1879 * Get the length of a UTF-8 byte sequence, not including any following
1880 * composing characters.
1881 * Returns 0 for "".
1882 * Returns 1 for an illegal byte sequence.
1885 utf_ptr2len(p)
1886 char_u *p;
1888 int len;
1889 int i;
1891 if (*p == NUL)
1892 return 0;
1893 len = utf8len_tab[*p];
1894 for (i = 1; i < len; ++i)
1895 if ((p[i] & 0xc0) != 0x80)
1896 return 1;
1897 return len;
1901 * Return length of UTF-8 character, obtained from the first byte.
1902 * "b" must be between 0 and 255!
1903 * Returns 1 for an invalid first byte value.
1906 utf_byte2len(b)
1907 int b;
1909 return utf8len_tab[b];
1913 * Get the length of UTF-8 byte sequence "p[size]". Does not include any
1914 * following composing characters.
1915 * Returns 1 for "".
1916 * Returns 1 for an illegal byte sequence (also in incomplete byte seq.).
1917 * Returns number > "size" for an incomplete byte sequence.
1918 * Never returns zero.
1921 utf_ptr2len_len(p, size)
1922 char_u *p;
1923 int size;
1925 int len;
1926 int i;
1927 int m;
1929 len = utf8len_tab[*p];
1930 if (len == 1)
1931 return 1; /* NUL, ascii or illegal lead byte */
1932 if (len > size)
1933 m = size; /* incomplete byte sequence. */
1934 else
1935 m = len;
1936 for (i = 1; i < m; ++i)
1937 if ((p[i] & 0xc0) != 0x80)
1938 return 1;
1939 return len;
1943 * Return the number of bytes the UTF-8 encoding of the character at "p" takes.
1944 * This includes following composing characters.
1947 utfc_ptr2len(p)
1948 char_u *p;
1950 int len;
1951 int b0 = *p;
1952 #ifdef FEAT_ARABIC
1953 int prevlen;
1954 #endif
1956 if (b0 == NUL)
1957 return 0;
1958 if (b0 < 0x80 && p[1] < 0x80) /* be quick for ASCII */
1959 return 1;
1961 /* Skip over first UTF-8 char, stopping at a NUL byte. */
1962 len = utf_ptr2len(p);
1964 /* Check for illegal byte. */
1965 if (len == 1 && b0 >= 0x80)
1966 return 1;
1969 * Check for composing characters. We can handle only the first six, but
1970 * skip all of them (otherwise the cursor would get stuck).
1972 #ifdef FEAT_ARABIC
1973 prevlen = 0;
1974 #endif
1975 for (;;)
1977 if (p[len] < 0x80 || !UTF_COMPOSINGLIKE(p + prevlen, p + len))
1978 return len;
1980 /* Skip over composing char */
1981 #ifdef FEAT_ARABIC
1982 prevlen = len;
1983 #endif
1984 len += utf_ptr2len(p + len);
1989 * Return the number of bytes the UTF-8 encoding of the character at "p[size]"
1990 * takes. This includes following composing characters.
1991 * Returns 0 for an empty string.
1992 * Returns 1 for an illegal char or an incomplete byte sequence.
1995 utfc_ptr2len_len(p, size)
1996 char_u *p;
1997 int size;
1999 int len;
2000 #ifdef FEAT_ARABIC
2001 int prevlen;
2002 #endif
2004 if (size < 1 || *p == NUL)
2005 return 0;
2006 if (p[0] < 0x80 && (size == 1 || p[1] < 0x80)) /* be quick for ASCII */
2007 return 1;
2009 /* Skip over first UTF-8 char, stopping at a NUL byte. */
2010 len = utf_ptr2len_len(p, size);
2012 /* Check for illegal byte and incomplete byte sequence. */
2013 if ((len == 1 && p[0] >= 0x80) || len > size)
2014 return 1;
2017 * Check for composing characters. We can handle only the first six, but
2018 * skip all of them (otherwise the cursor would get stuck).
2020 #ifdef FEAT_ARABIC
2021 prevlen = 0;
2022 #endif
2023 while (len < size)
2025 int len_next_char;
2027 if (p[len] < 0x80)
2028 break;
2031 * Next character length should not go beyond size to ensure that
2032 * UTF_COMPOSINGLIKE(...) does not read beyond size.
2034 len_next_char = utf_ptr2len_len(p + len, size - len);
2035 if (len_next_char > size - len)
2036 break;
2038 if (!UTF_COMPOSINGLIKE(p + prevlen, p + len))
2039 break;
2041 /* Skip over composing char */
2042 #ifdef FEAT_ARABIC
2043 prevlen = len;
2044 #endif
2045 len += len_next_char;
2047 return len;
2051 * Return the number of bytes the UTF-8 encoding of character "c" takes.
2052 * This does not include composing characters.
2055 utf_char2len(c)
2056 int c;
2058 if (c < 0x80)
2059 return 1;
2060 if (c < 0x800)
2061 return 2;
2062 if (c < 0x10000)
2063 return 3;
2064 if (c < 0x200000)
2065 return 4;
2066 if (c < 0x4000000)
2067 return 5;
2068 return 6;
2072 * Convert Unicode character "c" to UTF-8 string in "buf[]".
2073 * Returns the number of bytes.
2074 * This does not include composing characters.
2077 utf_char2bytes(c, buf)
2078 int c;
2079 char_u *buf;
2081 if (c < 0x80) /* 7 bits */
2083 buf[0] = c;
2084 return 1;
2086 if (c < 0x800) /* 11 bits */
2088 buf[0] = 0xc0 + ((unsigned)c >> 6);
2089 buf[1] = 0x80 + (c & 0x3f);
2090 return 2;
2092 if (c < 0x10000) /* 16 bits */
2094 buf[0] = 0xe0 + ((unsigned)c >> 12);
2095 buf[1] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2096 buf[2] = 0x80 + (c & 0x3f);
2097 return 3;
2099 if (c < 0x200000) /* 21 bits */
2101 buf[0] = 0xf0 + ((unsigned)c >> 18);
2102 buf[1] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2103 buf[2] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2104 buf[3] = 0x80 + (c & 0x3f);
2105 return 4;
2107 if (c < 0x4000000) /* 26 bits */
2109 buf[0] = 0xf8 + ((unsigned)c >> 24);
2110 buf[1] = 0x80 + (((unsigned)c >> 18) & 0x3f);
2111 buf[2] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2112 buf[3] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2113 buf[4] = 0x80 + (c & 0x3f);
2114 return 5;
2116 /* 31 bits */
2117 buf[0] = 0xfc + ((unsigned)c >> 30);
2118 buf[1] = 0x80 + (((unsigned)c >> 24) & 0x3f);
2119 buf[2] = 0x80 + (((unsigned)c >> 18) & 0x3f);
2120 buf[3] = 0x80 + (((unsigned)c >> 12) & 0x3f);
2121 buf[4] = 0x80 + (((unsigned)c >> 6) & 0x3f);
2122 buf[5] = 0x80 + (c & 0x3f);
2123 return 6;
2127 * Return TRUE if "c" is a composing UTF-8 character. This means it will be
2128 * drawn on top of the preceding character.
2129 * Based on code from Markus Kuhn.
2132 utf_iscomposing(c)
2133 int c;
2135 /* Sorted list of non-overlapping intervals.
2136 * Generated by ../runtime/tools/unicode.vim. */
2137 static struct interval combining[] =
2139 {0x0300, 0x036f},
2140 {0x0483, 0x0489},
2141 {0x0591, 0x05bd},
2142 {0x05bf, 0x05bf},
2143 {0x05c1, 0x05c2},
2144 {0x05c4, 0x05c5},
2145 {0x05c7, 0x05c7},
2146 {0x0610, 0x061a},
2147 {0x064b, 0x065e},
2148 {0x0670, 0x0670},
2149 {0x06d6, 0x06dc},
2150 {0x06de, 0x06e4},
2151 {0x06e7, 0x06e8},
2152 {0x06ea, 0x06ed},
2153 {0x0711, 0x0711},
2154 {0x0730, 0x074a},
2155 {0x07a6, 0x07b0},
2156 {0x07eb, 0x07f3},
2157 {0x0816, 0x0819},
2158 {0x081b, 0x0823},
2159 {0x0825, 0x0827},
2160 {0x0829, 0x082d},
2161 {0x0900, 0x0903},
2162 {0x093c, 0x093c},
2163 {0x093e, 0x094e},
2164 {0x0951, 0x0955},
2165 {0x0962, 0x0963},
2166 {0x0981, 0x0983},
2167 {0x09bc, 0x09bc},
2168 {0x09be, 0x09c4},
2169 {0x09c7, 0x09c8},
2170 {0x09cb, 0x09cd},
2171 {0x09d7, 0x09d7},
2172 {0x09e2, 0x09e3},
2173 {0x0a01, 0x0a03},
2174 {0x0a3c, 0x0a3c},
2175 {0x0a3e, 0x0a42},
2176 {0x0a47, 0x0a48},
2177 {0x0a4b, 0x0a4d},
2178 {0x0a51, 0x0a51},
2179 {0x0a70, 0x0a71},
2180 {0x0a75, 0x0a75},
2181 {0x0a81, 0x0a83},
2182 {0x0abc, 0x0abc},
2183 {0x0abe, 0x0ac5},
2184 {0x0ac7, 0x0ac9},
2185 {0x0acb, 0x0acd},
2186 {0x0ae2, 0x0ae3},
2187 {0x0b01, 0x0b03},
2188 {0x0b3c, 0x0b3c},
2189 {0x0b3e, 0x0b44},
2190 {0x0b47, 0x0b48},
2191 {0x0b4b, 0x0b4d},
2192 {0x0b56, 0x0b57},
2193 {0x0b62, 0x0b63},
2194 {0x0b82, 0x0b82},
2195 {0x0bbe, 0x0bc2},
2196 {0x0bc6, 0x0bc8},
2197 {0x0bca, 0x0bcd},
2198 {0x0bd7, 0x0bd7},
2199 {0x0c01, 0x0c03},
2200 {0x0c3e, 0x0c44},
2201 {0x0c46, 0x0c48},
2202 {0x0c4a, 0x0c4d},
2203 {0x0c55, 0x0c56},
2204 {0x0c62, 0x0c63},
2205 {0x0c82, 0x0c83},
2206 {0x0cbc, 0x0cbc},
2207 {0x0cbe, 0x0cc4},
2208 {0x0cc6, 0x0cc8},
2209 {0x0cca, 0x0ccd},
2210 {0x0cd5, 0x0cd6},
2211 {0x0ce2, 0x0ce3},
2212 {0x0d02, 0x0d03},
2213 {0x0d3e, 0x0d44},
2214 {0x0d46, 0x0d48},
2215 {0x0d4a, 0x0d4d},
2216 {0x0d57, 0x0d57},
2217 {0x0d62, 0x0d63},
2218 {0x0d82, 0x0d83},
2219 {0x0dca, 0x0dca},
2220 {0x0dcf, 0x0dd4},
2221 {0x0dd6, 0x0dd6},
2222 {0x0dd8, 0x0ddf},
2223 {0x0df2, 0x0df3},
2224 {0x0e31, 0x0e31},
2225 {0x0e34, 0x0e3a},
2226 {0x0e47, 0x0e4e},
2227 {0x0eb1, 0x0eb1},
2228 {0x0eb4, 0x0eb9},
2229 {0x0ebb, 0x0ebc},
2230 {0x0ec8, 0x0ecd},
2231 {0x0f18, 0x0f19},
2232 {0x0f35, 0x0f35},
2233 {0x0f37, 0x0f37},
2234 {0x0f39, 0x0f39},
2235 {0x0f3e, 0x0f3f},
2236 {0x0f71, 0x0f84},
2237 {0x0f86, 0x0f87},
2238 {0x0f90, 0x0f97},
2239 {0x0f99, 0x0fbc},
2240 {0x0fc6, 0x0fc6},
2241 {0x102b, 0x103e},
2242 {0x1056, 0x1059},
2243 {0x105e, 0x1060},
2244 {0x1062, 0x1064},
2245 {0x1067, 0x106d},
2246 {0x1071, 0x1074},
2247 {0x1082, 0x108d},
2248 {0x108f, 0x108f},
2249 {0x109a, 0x109d},
2250 {0x135f, 0x135f},
2251 {0x1712, 0x1714},
2252 {0x1732, 0x1734},
2253 {0x1752, 0x1753},
2254 {0x1772, 0x1773},
2255 {0x17b6, 0x17d3},
2256 {0x17dd, 0x17dd},
2257 {0x180b, 0x180d},
2258 {0x18a9, 0x18a9},
2259 {0x1920, 0x192b},
2260 {0x1930, 0x193b},
2261 {0x19b0, 0x19c0},
2262 {0x19c8, 0x19c9},
2263 {0x1a17, 0x1a1b},
2264 {0x1a55, 0x1a5e},
2265 {0x1a60, 0x1a7c},
2266 {0x1a7f, 0x1a7f},
2267 {0x1b00, 0x1b04},
2268 {0x1b34, 0x1b44},
2269 {0x1b6b, 0x1b73},
2270 {0x1b80, 0x1b82},
2271 {0x1ba1, 0x1baa},
2272 {0x1c24, 0x1c37},
2273 {0x1cd0, 0x1cd2},
2274 {0x1cd4, 0x1ce8},
2275 {0x1ced, 0x1ced},
2276 {0x1cf2, 0x1cf2},
2277 {0x1dc0, 0x1de6},
2278 {0x1dfd, 0x1dff},
2279 {0x20d0, 0x20f0},
2280 {0x2cef, 0x2cf1},
2281 {0x2de0, 0x2dff},
2282 {0x302a, 0x302f},
2283 {0x3099, 0x309a},
2284 {0xa66f, 0xa672},
2285 {0xa67c, 0xa67d},
2286 {0xa6f0, 0xa6f1},
2287 {0xa802, 0xa802},
2288 {0xa806, 0xa806},
2289 {0xa80b, 0xa80b},
2290 {0xa823, 0xa827},
2291 {0xa880, 0xa881},
2292 {0xa8b4, 0xa8c4},
2293 {0xa8e0, 0xa8f1},
2294 {0xa926, 0xa92d},
2295 {0xa947, 0xa953},
2296 {0xa980, 0xa983},
2297 {0xa9b3, 0xa9c0},
2298 {0xaa29, 0xaa36},
2299 {0xaa43, 0xaa43},
2300 {0xaa4c, 0xaa4d},
2301 {0xaa7b, 0xaa7b},
2302 {0xaab0, 0xaab0},
2303 {0xaab2, 0xaab4},
2304 {0xaab7, 0xaab8},
2305 {0xaabe, 0xaabf},
2306 {0xaac1, 0xaac1},
2307 {0xabe3, 0xabea},
2308 {0xabec, 0xabed},
2309 {0xfb1e, 0xfb1e},
2310 {0xfe00, 0xfe0f},
2311 {0xfe20, 0xfe26},
2312 {0x101fd, 0x101fd},
2313 {0x10a01, 0x10a03},
2314 {0x10a05, 0x10a06},
2315 {0x10a0c, 0x10a0f},
2316 {0x10a38, 0x10a3a},
2317 {0x10a3f, 0x10a3f},
2318 {0x11080, 0x11082},
2319 {0x110b0, 0x110ba},
2320 {0x1d165, 0x1d169},
2321 {0x1d16d, 0x1d172},
2322 {0x1d17b, 0x1d182},
2323 {0x1d185, 0x1d18b},
2324 {0x1d1aa, 0x1d1ad},
2325 {0x1d242, 0x1d244},
2326 {0xe0100, 0xe01ef}
2329 return intable(combining, sizeof(combining), c);
2333 * Return TRUE for characters that can be displayed in a normal way.
2334 * Only for characters of 0x100 and above!
2337 utf_printable(c)
2338 int c;
2340 #ifdef USE_WCHAR_FUNCTIONS
2342 * Assume the iswprint() library function works better than our own stuff.
2344 return iswprint(c);
2345 #else
2346 /* Sorted list of non-overlapping intervals.
2347 * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */
2348 static struct interval nonprint[] =
2350 {0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e},
2351 {0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb},
2352 {0xfffe, 0xffff}
2355 return !intable(nonprint, sizeof(nonprint), c);
2356 #endif
2360 * Get class of a Unicode character.
2361 * 0: white space
2362 * 1: punctuation
2363 * 2 or bigger: some class of word character.
2366 utf_class(c)
2367 int c;
2369 /* sorted list of non-overlapping intervals */
2370 static struct clinterval
2372 unsigned short first;
2373 unsigned short last;
2374 unsigned short class;
2375 } classes[] =
2377 {0x037e, 0x037e, 1}, /* Greek question mark */
2378 {0x0387, 0x0387, 1}, /* Greek ano teleia */
2379 {0x055a, 0x055f, 1}, /* Armenian punctuation */
2380 {0x0589, 0x0589, 1}, /* Armenian full stop */
2381 {0x05be, 0x05be, 1},
2382 {0x05c0, 0x05c0, 1},
2383 {0x05c3, 0x05c3, 1},
2384 {0x05f3, 0x05f4, 1},
2385 {0x060c, 0x060c, 1},
2386 {0x061b, 0x061b, 1},
2387 {0x061f, 0x061f, 1},
2388 {0x066a, 0x066d, 1},
2389 {0x06d4, 0x06d4, 1},
2390 {0x0700, 0x070d, 1}, /* Syriac punctuation */
2391 {0x0964, 0x0965, 1},
2392 {0x0970, 0x0970, 1},
2393 {0x0df4, 0x0df4, 1},
2394 {0x0e4f, 0x0e4f, 1},
2395 {0x0e5a, 0x0e5b, 1},
2396 {0x0f04, 0x0f12, 1},
2397 {0x0f3a, 0x0f3d, 1},
2398 {0x0f85, 0x0f85, 1},
2399 {0x104a, 0x104f, 1}, /* Myanmar punctuation */
2400 {0x10fb, 0x10fb, 1}, /* Georgian punctuation */
2401 {0x1361, 0x1368, 1}, /* Ethiopic punctuation */
2402 {0x166d, 0x166e, 1}, /* Canadian Syl. punctuation */
2403 {0x1680, 0x1680, 0},
2404 {0x169b, 0x169c, 1},
2405 {0x16eb, 0x16ed, 1},
2406 {0x1735, 0x1736, 1},
2407 {0x17d4, 0x17dc, 1}, /* Khmer punctuation */
2408 {0x1800, 0x180a, 1}, /* Mongolian punctuation */
2409 {0x2000, 0x200b, 0}, /* spaces */
2410 {0x200c, 0x2027, 1}, /* punctuation and symbols */
2411 {0x2028, 0x2029, 0},
2412 {0x202a, 0x202e, 1}, /* punctuation and symbols */
2413 {0x202f, 0x202f, 0},
2414 {0x2030, 0x205e, 1}, /* punctuation and symbols */
2415 {0x205f, 0x205f, 0},
2416 {0x2060, 0x27ff, 1}, /* punctuation and symbols */
2417 {0x2070, 0x207f, 0x2070}, /* superscript */
2418 {0x2080, 0x2094, 0x2080}, /* subscript */
2419 {0x20a0, 0x27ff, 1}, /* all kinds of symbols */
2420 {0x2800, 0x28ff, 0x2800}, /* braille */
2421 {0x2900, 0x2998, 1}, /* arrows, brackets, etc. */
2422 {0x29d8, 0x29db, 1},
2423 {0x29fc, 0x29fd, 1},
2424 {0x3000, 0x3000, 0}, /* ideographic space */
2425 {0x3001, 0x3020, 1}, /* ideographic punctuation */
2426 {0x3030, 0x3030, 1},
2427 {0x303d, 0x303d, 1},
2428 {0x3040, 0x309f, 0x3040}, /* Hiragana */
2429 {0x30a0, 0x30ff, 0x30a0}, /* Katakana */
2430 {0x3300, 0x9fff, 0x4e00}, /* CJK Ideographs */
2431 {0xac00, 0xd7a3, 0xac00}, /* Hangul Syllables */
2432 {0xf900, 0xfaff, 0x4e00}, /* CJK Ideographs */
2433 {0xfd3e, 0xfd3f, 1},
2434 {0xfe30, 0xfe6b, 1}, /* punctuation forms */
2435 {0xff00, 0xff0f, 1}, /* half/fullwidth ASCII */
2436 {0xff1a, 0xff20, 1}, /* half/fullwidth ASCII */
2437 {0xff3b, 0xff40, 1}, /* half/fullwidth ASCII */
2438 {0xff5b, 0xff65, 1}, /* half/fullwidth ASCII */
2440 int bot = 0;
2441 int top = sizeof(classes) / sizeof(struct clinterval) - 1;
2442 int mid;
2444 /* First quick check for Latin1 characters, use 'iskeyword'. */
2445 if (c < 0x100)
2447 if (c == ' ' || c == '\t' || c == NUL || c == 0xa0)
2448 return 0; /* blank */
2449 if (vim_iswordc(c))
2450 return 2; /* word character */
2451 return 1; /* punctuation */
2454 /* binary search in table */
2455 while (top >= bot)
2457 mid = (bot + top) / 2;
2458 if (classes[mid].last < c)
2459 bot = mid + 1;
2460 else if (classes[mid].first > c)
2461 top = mid - 1;
2462 else
2463 return (int)classes[mid].class;
2466 /* most other characters are "word" characters */
2467 return 2;
2471 * Code for Unicode case-dependent operations. Based on notes in
2472 * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt
2473 * This code uses simple case folding, not full case folding.
2474 * Last updated for Unicode 5.2.
2478 * The following tables are built by ../runtime/tools/unicode.vim.
2479 * They must be in numeric order, because we use binary search.
2480 * An entry such as {0x41,0x5a,1,32} means that Unicode characters in the
2481 * range from 0x41 to 0x5a inclusive, stepping by 1, are changed to
2482 * folded/upper/lower by adding 32.
2484 typedef struct
2486 int rangeStart;
2487 int rangeEnd;
2488 int step;
2489 int offset;
2490 } convertStruct;
2492 static convertStruct foldCase[] =
2494 {0x41,0x5a,1,32},
2495 {0xb5,0xb5,-1,775},
2496 {0xc0,0xd6,1,32},
2497 {0xd8,0xde,1,32},
2498 {0x100,0x12e,2,1},
2499 {0x132,0x136,2,1},
2500 {0x139,0x147,2,1},
2501 {0x14a,0x176,2,1},
2502 {0x178,0x178,-1,-121},
2503 {0x179,0x17d,2,1},
2504 {0x17f,0x17f,-1,-268},
2505 {0x181,0x181,-1,210},
2506 {0x182,0x184,2,1},
2507 {0x186,0x186,-1,206},
2508 {0x187,0x187,-1,1},
2509 {0x189,0x18a,1,205},
2510 {0x18b,0x18b,-1,1},
2511 {0x18e,0x18e,-1,79},
2512 {0x18f,0x18f,-1,202},
2513 {0x190,0x190,-1,203},
2514 {0x191,0x191,-1,1},
2515 {0x193,0x193,-1,205},
2516 {0x194,0x194,-1,207},
2517 {0x196,0x196,-1,211},
2518 {0x197,0x197,-1,209},
2519 {0x198,0x198,-1,1},
2520 {0x19c,0x19c,-1,211},
2521 {0x19d,0x19d,-1,213},
2522 {0x19f,0x19f,-1,214},
2523 {0x1a0,0x1a4,2,1},
2524 {0x1a6,0x1a6,-1,218},
2525 {0x1a7,0x1a7,-1,1},
2526 {0x1a9,0x1a9,-1,218},
2527 {0x1ac,0x1ac,-1,1},
2528 {0x1ae,0x1ae,-1,218},
2529 {0x1af,0x1af,-1,1},
2530 {0x1b1,0x1b2,1,217},
2531 {0x1b3,0x1b5,2,1},
2532 {0x1b7,0x1b7,-1,219},
2533 {0x1b8,0x1bc,4,1},
2534 {0x1c4,0x1c4,-1,2},
2535 {0x1c5,0x1c5,-1,1},
2536 {0x1c7,0x1c7,-1,2},
2537 {0x1c8,0x1c8,-1,1},
2538 {0x1ca,0x1ca,-1,2},
2539 {0x1cb,0x1db,2,1},
2540 {0x1de,0x1ee,2,1},
2541 {0x1f1,0x1f1,-1,2},
2542 {0x1f2,0x1f4,2,1},
2543 {0x1f6,0x1f6,-1,-97},
2544 {0x1f7,0x1f7,-1,-56},
2545 {0x1f8,0x21e,2,1},
2546 {0x220,0x220,-1,-130},
2547 {0x222,0x232,2,1},
2548 {0x23a,0x23a,-1,10795},
2549 {0x23b,0x23b,-1,1},
2550 {0x23d,0x23d,-1,-163},
2551 {0x23e,0x23e,-1,10792},
2552 {0x241,0x241,-1,1},
2553 {0x243,0x243,-1,-195},
2554 {0x244,0x244,-1,69},
2555 {0x245,0x245,-1,71},
2556 {0x246,0x24e,2,1},
2557 {0x345,0x345,-1,116},
2558 {0x370,0x372,2,1},
2559 {0x376,0x376,-1,1},
2560 {0x386,0x386,-1,38},
2561 {0x388,0x38a,1,37},
2562 {0x38c,0x38c,-1,64},
2563 {0x38e,0x38f,1,63},
2564 {0x391,0x3a1,1,32},
2565 {0x3a3,0x3ab,1,32},
2566 {0x3c2,0x3c2,-1,1},
2567 {0x3cf,0x3cf,-1,8},
2568 {0x3d0,0x3d0,-1,-30},
2569 {0x3d1,0x3d1,-1,-25},
2570 {0x3d5,0x3d5,-1,-15},
2571 {0x3d6,0x3d6,-1,-22},
2572 {0x3d8,0x3ee,2,1},
2573 {0x3f0,0x3f0,-1,-54},
2574 {0x3f1,0x3f1,-1,-48},
2575 {0x3f4,0x3f4,-1,-60},
2576 {0x3f5,0x3f5,-1,-64},
2577 {0x3f7,0x3f7,-1,1},
2578 {0x3f9,0x3f9,-1,-7},
2579 {0x3fa,0x3fa,-1,1},
2580 {0x3fd,0x3ff,1,-130},
2581 {0x400,0x40f,1,80},
2582 {0x410,0x42f,1,32},
2583 {0x460,0x480,2,1},
2584 {0x48a,0x4be,2,1},
2585 {0x4c0,0x4c0,-1,15},
2586 {0x4c1,0x4cd,2,1},
2587 {0x4d0,0x524,2,1},
2588 {0x531,0x556,1,48},
2589 {0x10a0,0x10c5,1,7264},
2590 {0x1e00,0x1e94,2,1},
2591 {0x1e9b,0x1e9b,-1,-58},
2592 {0x1e9e,0x1e9e,-1,-7615},
2593 {0x1ea0,0x1efe,2,1},
2594 {0x1f08,0x1f0f,1,-8},
2595 {0x1f18,0x1f1d,1,-8},
2596 {0x1f28,0x1f2f,1,-8},
2597 {0x1f38,0x1f3f,1,-8},
2598 {0x1f48,0x1f4d,1,-8},
2599 {0x1f59,0x1f5f,2,-8},
2600 {0x1f68,0x1f6f,1,-8},
2601 {0x1f88,0x1f8f,1,-8},
2602 {0x1f98,0x1f9f,1,-8},
2603 {0x1fa8,0x1faf,1,-8},
2604 {0x1fb8,0x1fb9,1,-8},
2605 {0x1fba,0x1fbb,1,-74},
2606 {0x1fbc,0x1fbc,-1,-9},
2607 {0x1fbe,0x1fbe,-1,-7173},
2608 {0x1fc8,0x1fcb,1,-86},
2609 {0x1fcc,0x1fcc,-1,-9},
2610 {0x1fd8,0x1fd9,1,-8},
2611 {0x1fda,0x1fdb,1,-100},
2612 {0x1fe8,0x1fe9,1,-8},
2613 {0x1fea,0x1feb,1,-112},
2614 {0x1fec,0x1fec,-1,-7},
2615 {0x1ff8,0x1ff9,1,-128},
2616 {0x1ffa,0x1ffb,1,-126},
2617 {0x1ffc,0x1ffc,-1,-9},
2618 {0x2126,0x2126,-1,-7517},
2619 {0x212a,0x212a,-1,-8383},
2620 {0x212b,0x212b,-1,-8262},
2621 {0x2132,0x2132,-1,28},
2622 {0x2160,0x216f,1,16},
2623 {0x2183,0x2183,-1,1},
2624 {0x24b6,0x24cf,1,26},
2625 {0x2c00,0x2c2e,1,48},
2626 {0x2c60,0x2c60,-1,1},
2627 {0x2c62,0x2c62,-1,-10743},
2628 {0x2c63,0x2c63,-1,-3814},
2629 {0x2c64,0x2c64,-1,-10727},
2630 {0x2c67,0x2c6b,2,1},
2631 {0x2c6d,0x2c6d,-1,-10780},
2632 {0x2c6e,0x2c6e,-1,-10749},
2633 {0x2c6f,0x2c6f,-1,-10783},
2634 {0x2c70,0x2c70,-1,-10782},
2635 {0x2c72,0x2c75,3,1},
2636 {0x2c7e,0x2c7f,1,-10815},
2637 {0x2c80,0x2ce2,2,1},
2638 {0x2ceb,0x2ced,2,1},
2639 {0xa640,0xa65e,2,1},
2640 {0xa662,0xa66c,2,1},
2641 {0xa680,0xa696,2,1},
2642 {0xa722,0xa72e,2,1},
2643 {0xa732,0xa76e,2,1},
2644 {0xa779,0xa77b,2,1},
2645 {0xa77d,0xa77d,-1,-35332},
2646 {0xa77e,0xa786,2,1},
2647 {0xa78b,0xa78b,-1,1},
2648 {0xff21,0xff3a,1,32},
2649 {0x10400,0x10427,1,40}
2652 static int utf_convert(int a, convertStruct table[], int tableSize);
2655 * Generic conversion function for case operations.
2656 * Return the converted equivalent of "a", which is a UCS-4 character. Use
2657 * the given conversion "table". Uses binary search on "table".
2659 static int
2660 utf_convert(a, table, tableSize)
2661 int a;
2662 convertStruct table[];
2663 int tableSize;
2665 int start, mid, end; /* indices into table */
2667 start = 0;
2668 end = tableSize / sizeof(convertStruct);
2669 while (start < end)
2671 /* need to search further */
2672 mid = (end + start) /2;
2673 if (table[mid].rangeEnd < a)
2674 start = mid + 1;
2675 else
2676 end = mid;
2678 if (table[start].rangeStart <= a && a <= table[start].rangeEnd
2679 && (a - table[start].rangeStart) % table[start].step == 0)
2680 return (a + table[start].offset);
2681 else
2682 return a;
2686 * Return the folded-case equivalent of "a", which is a UCS-4 character. Uses
2687 * simple case folding.
2690 utf_fold(a)
2691 int a;
2693 return utf_convert(a, foldCase, sizeof(foldCase));
2696 static convertStruct toLower[] =
2698 {0x41,0x5a,1,32},
2699 {0xc0,0xd6,1,32},
2700 {0xd8,0xde,1,32},
2701 {0x100,0x12e,2,1},
2702 {0x130,0x130,-1,-199},
2703 {0x132,0x136,2,1},
2704 {0x139,0x147,2,1},
2705 {0x14a,0x176,2,1},
2706 {0x178,0x178,-1,-121},
2707 {0x179,0x17d,2,1},
2708 {0x181,0x181,-1,210},
2709 {0x182,0x184,2,1},
2710 {0x186,0x186,-1,206},
2711 {0x187,0x187,-1,1},
2712 {0x189,0x18a,1,205},
2713 {0x18b,0x18b,-1,1},
2714 {0x18e,0x18e,-1,79},
2715 {0x18f,0x18f,-1,202},
2716 {0x190,0x190,-1,203},
2717 {0x191,0x191,-1,1},
2718 {0x193,0x193,-1,205},
2719 {0x194,0x194,-1,207},
2720 {0x196,0x196,-1,211},
2721 {0x197,0x197,-1,209},
2722 {0x198,0x198,-1,1},
2723 {0x19c,0x19c,-1,211},
2724 {0x19d,0x19d,-1,213},
2725 {0x19f,0x19f,-1,214},
2726 {0x1a0,0x1a4,2,1},
2727 {0x1a6,0x1a6,-1,218},
2728 {0x1a7,0x1a7,-1,1},
2729 {0x1a9,0x1a9,-1,218},
2730 {0x1ac,0x1ac,-1,1},
2731 {0x1ae,0x1ae,-1,218},
2732 {0x1af,0x1af,-1,1},
2733 {0x1b1,0x1b2,1,217},
2734 {0x1b3,0x1b5,2,1},
2735 {0x1b7,0x1b7,-1,219},
2736 {0x1b8,0x1bc,4,1},
2737 {0x1c4,0x1c4,-1,2},
2738 {0x1c5,0x1c5,-1,1},
2739 {0x1c7,0x1c7,-1,2},
2740 {0x1c8,0x1c8,-1,1},
2741 {0x1ca,0x1ca,-1,2},
2742 {0x1cb,0x1db,2,1},
2743 {0x1de,0x1ee,2,1},
2744 {0x1f1,0x1f1,-1,2},
2745 {0x1f2,0x1f4,2,1},
2746 {0x1f6,0x1f6,-1,-97},
2747 {0x1f7,0x1f7,-1,-56},
2748 {0x1f8,0x21e,2,1},
2749 {0x220,0x220,-1,-130},
2750 {0x222,0x232,2,1},
2751 {0x23a,0x23a,-1,10795},
2752 {0x23b,0x23b,-1,1},
2753 {0x23d,0x23d,-1,-163},
2754 {0x23e,0x23e,-1,10792},
2755 {0x241,0x241,-1,1},
2756 {0x243,0x243,-1,-195},
2757 {0x244,0x244,-1,69},
2758 {0x245,0x245,-1,71},
2759 {0x246,0x24e,2,1},
2760 {0x370,0x372,2,1},
2761 {0x376,0x376,-1,1},
2762 {0x386,0x386,-1,38},
2763 {0x388,0x38a,1,37},
2764 {0x38c,0x38c,-1,64},
2765 {0x38e,0x38f,1,63},
2766 {0x391,0x3a1,1,32},
2767 {0x3a3,0x3ab,1,32},
2768 {0x3cf,0x3cf,-1,8},
2769 {0x3d8,0x3ee,2,1},
2770 {0x3f4,0x3f4,-1,-60},
2771 {0x3f7,0x3f7,-1,1},
2772 {0x3f9,0x3f9,-1,-7},
2773 {0x3fa,0x3fa,-1,1},
2774 {0x3fd,0x3ff,1,-130},
2775 {0x400,0x40f,1,80},
2776 {0x410,0x42f,1,32},
2777 {0x460,0x480,2,1},
2778 {0x48a,0x4be,2,1},
2779 {0x4c0,0x4c0,-1,15},
2780 {0x4c1,0x4cd,2,1},
2781 {0x4d0,0x524,2,1},
2782 {0x531,0x556,1,48},
2783 {0x10a0,0x10c5,1,7264},
2784 {0x1e00,0x1e94,2,1},
2785 {0x1e9e,0x1e9e,-1,-7615},
2786 {0x1ea0,0x1efe,2,1},
2787 {0x1f08,0x1f0f,1,-8},
2788 {0x1f18,0x1f1d,1,-8},
2789 {0x1f28,0x1f2f,1,-8},
2790 {0x1f38,0x1f3f,1,-8},
2791 {0x1f48,0x1f4d,1,-8},
2792 {0x1f59,0x1f5f,2,-8},
2793 {0x1f68,0x1f6f,1,-8},
2794 {0x1f88,0x1f8f,1,-8},
2795 {0x1f98,0x1f9f,1,-8},
2796 {0x1fa8,0x1faf,1,-8},
2797 {0x1fb8,0x1fb9,1,-8},
2798 {0x1fba,0x1fbb,1,-74},
2799 {0x1fbc,0x1fbc,-1,-9},
2800 {0x1fc8,0x1fcb,1,-86},
2801 {0x1fcc,0x1fcc,-1,-9},
2802 {0x1fd8,0x1fd9,1,-8},
2803 {0x1fda,0x1fdb,1,-100},
2804 {0x1fe8,0x1fe9,1,-8},
2805 {0x1fea,0x1feb,1,-112},
2806 {0x1fec,0x1fec,-1,-7},
2807 {0x1ff8,0x1ff9,1,-128},
2808 {0x1ffa,0x1ffb,1,-126},
2809 {0x1ffc,0x1ffc,-1,-9},
2810 {0x2126,0x2126,-1,-7517},
2811 {0x212a,0x212a,-1,-8383},
2812 {0x212b,0x212b,-1,-8262},
2813 {0x2132,0x2132,-1,28},
2814 {0x2160,0x216f,1,16},
2815 {0x2183,0x2183,-1,1},
2816 {0x24b6,0x24cf,1,26},
2817 {0x2c00,0x2c2e,1,48},
2818 {0x2c60,0x2c60,-1,1},
2819 {0x2c62,0x2c62,-1,-10743},
2820 {0x2c63,0x2c63,-1,-3814},
2821 {0x2c64,0x2c64,-1,-10727},
2822 {0x2c67,0x2c6b,2,1},
2823 {0x2c6d,0x2c6d,-1,-10780},
2824 {0x2c6e,0x2c6e,-1,-10749},
2825 {0x2c6f,0x2c6f,-1,-10783},
2826 {0x2c70,0x2c70,-1,-10782},
2827 {0x2c72,0x2c75,3,1},
2828 {0x2c7e,0x2c7f,1,-10815},
2829 {0x2c80,0x2ce2,2,1},
2830 {0x2ceb,0x2ced,2,1},
2831 {0xa640,0xa65e,2,1},
2832 {0xa662,0xa66c,2,1},
2833 {0xa680,0xa696,2,1},
2834 {0xa722,0xa72e,2,1},
2835 {0xa732,0xa76e,2,1},
2836 {0xa779,0xa77b,2,1},
2837 {0xa77d,0xa77d,-1,-35332},
2838 {0xa77e,0xa786,2,1},
2839 {0xa78b,0xa78b,-1,1},
2840 {0xff21,0xff3a,1,32},
2841 {0x10400,0x10427,1,40}
2844 static convertStruct toUpper[] =
2846 {0x61,0x7a,1,-32},
2847 {0xb5,0xb5,-1,743},
2848 {0xe0,0xf6,1,-32},
2849 {0xf8,0xfe,1,-32},
2850 {0xff,0xff,-1,121},
2851 {0x101,0x12f,2,-1},
2852 {0x131,0x131,-1,-232},
2853 {0x133,0x137,2,-1},
2854 {0x13a,0x148,2,-1},
2855 {0x14b,0x177,2,-1},
2856 {0x17a,0x17e,2,-1},
2857 {0x17f,0x17f,-1,-300},
2858 {0x180,0x180,-1,195},
2859 {0x183,0x185,2,-1},
2860 {0x188,0x18c,4,-1},
2861 {0x192,0x192,-1,-1},
2862 {0x195,0x195,-1,97},
2863 {0x199,0x199,-1,-1},
2864 {0x19a,0x19a,-1,163},
2865 {0x19e,0x19e,-1,130},
2866 {0x1a1,0x1a5,2,-1},
2867 {0x1a8,0x1ad,5,-1},
2868 {0x1b0,0x1b4,4,-1},
2869 {0x1b6,0x1b9,3,-1},
2870 {0x1bd,0x1bd,-1,-1},
2871 {0x1bf,0x1bf,-1,56},
2872 {0x1c5,0x1c5,-1,-1},
2873 {0x1c6,0x1c6,-1,-2},
2874 {0x1c8,0x1c8,-1,-1},
2875 {0x1c9,0x1c9,-1,-2},
2876 {0x1cb,0x1cb,-1,-1},
2877 {0x1cc,0x1cc,-1,-2},
2878 {0x1ce,0x1dc,2,-1},
2879 {0x1dd,0x1dd,-1,-79},
2880 {0x1df,0x1ef,2,-1},
2881 {0x1f2,0x1f2,-1,-1},
2882 {0x1f3,0x1f3,-1,-2},
2883 {0x1f5,0x1f9,4,-1},
2884 {0x1fb,0x21f,2,-1},
2885 {0x223,0x233,2,-1},
2886 {0x23c,0x23c,-1,-1},
2887 {0x23f,0x240,1,10815},
2888 {0x242,0x247,5,-1},
2889 {0x249,0x24f,2,-1},
2890 {0x250,0x250,-1,10783},
2891 {0x251,0x251,-1,10780},
2892 {0x252,0x252,-1,10782},
2893 {0x253,0x253,-1,-210},
2894 {0x254,0x254,-1,-206},
2895 {0x256,0x257,1,-205},
2896 {0x259,0x259,-1,-202},
2897 {0x25b,0x25b,-1,-203},
2898 {0x260,0x260,-1,-205},
2899 {0x263,0x263,-1,-207},
2900 {0x268,0x268,-1,-209},
2901 {0x269,0x269,-1,-211},
2902 {0x26b,0x26b,-1,10743},
2903 {0x26f,0x26f,-1,-211},
2904 {0x271,0x271,-1,10749},
2905 {0x272,0x272,-1,-213},
2906 {0x275,0x275,-1,-214},
2907 {0x27d,0x27d,-1,10727},
2908 {0x280,0x283,3,-218},
2909 {0x288,0x288,-1,-218},
2910 {0x289,0x289,-1,-69},
2911 {0x28a,0x28b,1,-217},
2912 {0x28c,0x28c,-1,-71},
2913 {0x292,0x292,-1,-219},
2914 {0x345,0x345,-1,84},
2915 {0x371,0x373,2,-1},
2916 {0x377,0x377,-1,-1},
2917 {0x37b,0x37d,1,130},
2918 {0x3ac,0x3ac,-1,-38},
2919 {0x3ad,0x3af,1,-37},
2920 {0x3b1,0x3c1,1,-32},
2921 {0x3c2,0x3c2,-1,-31},
2922 {0x3c3,0x3cb,1,-32},
2923 {0x3cc,0x3cc,-1,-64},
2924 {0x3cd,0x3ce,1,-63},
2925 {0x3d0,0x3d0,-1,-62},
2926 {0x3d1,0x3d1,-1,-57},
2927 {0x3d5,0x3d5,-1,-47},
2928 {0x3d6,0x3d6,-1,-54},
2929 {0x3d7,0x3d7,-1,-8},
2930 {0x3d9,0x3ef,2,-1},
2931 {0x3f0,0x3f0,-1,-86},
2932 {0x3f1,0x3f1,-1,-80},
2933 {0x3f2,0x3f2,-1,7},
2934 {0x3f5,0x3f5,-1,-96},
2935 {0x3f8,0x3fb,3,-1},
2936 {0x430,0x44f,1,-32},
2937 {0x450,0x45f,1,-80},
2938 {0x461,0x481,2,-1},
2939 {0x48b,0x4bf,2,-1},
2940 {0x4c2,0x4ce,2,-1},
2941 {0x4cf,0x4cf,-1,-15},
2942 {0x4d1,0x525,2,-1},
2943 {0x561,0x586,1,-48},
2944 {0x1d79,0x1d79,-1,35332},
2945 {0x1d7d,0x1d7d,-1,3814},
2946 {0x1e01,0x1e95,2,-1},
2947 {0x1e9b,0x1e9b,-1,-59},
2948 {0x1ea1,0x1eff,2,-1},
2949 {0x1f00,0x1f07,1,8},
2950 {0x1f10,0x1f15,1,8},
2951 {0x1f20,0x1f27,1,8},
2952 {0x1f30,0x1f37,1,8},
2953 {0x1f40,0x1f45,1,8},
2954 {0x1f51,0x1f57,2,8},
2955 {0x1f60,0x1f67,1,8},
2956 {0x1f70,0x1f71,1,74},
2957 {0x1f72,0x1f75,1,86},
2958 {0x1f76,0x1f77,1,100},
2959 {0x1f78,0x1f79,1,128},
2960 {0x1f7a,0x1f7b,1,112},
2961 {0x1f7c,0x1f7d,1,126},
2962 {0x1f80,0x1f87,1,8},
2963 {0x1f90,0x1f97,1,8},
2964 {0x1fa0,0x1fa7,1,8},
2965 {0x1fb0,0x1fb1,1,8},
2966 {0x1fb3,0x1fb3,-1,9},
2967 {0x1fbe,0x1fbe,-1,-7205},
2968 {0x1fc3,0x1fc3,-1,9},
2969 {0x1fd0,0x1fd1,1,8},
2970 {0x1fe0,0x1fe1,1,8},
2971 {0x1fe5,0x1fe5,-1,7},
2972 {0x1ff3,0x1ff3,-1,9},
2973 {0x214e,0x214e,-1,-28},
2974 {0x2170,0x217f,1,-16},
2975 {0x2184,0x2184,-1,-1},
2976 {0x24d0,0x24e9,1,-26},
2977 {0x2c30,0x2c5e,1,-48},
2978 {0x2c61,0x2c61,-1,-1},
2979 {0x2c65,0x2c65,-1,-10795},
2980 {0x2c66,0x2c66,-1,-10792},
2981 {0x2c68,0x2c6c,2,-1},
2982 {0x2c73,0x2c76,3,-1},
2983 {0x2c81,0x2ce3,2,-1},
2984 {0x2cec,0x2cee,2,-1},
2985 {0x2d00,0x2d25,1,-7264},
2986 {0xa641,0xa65f,2,-1},
2987 {0xa663,0xa66d,2,-1},
2988 {0xa681,0xa697,2,-1},
2989 {0xa723,0xa72f,2,-1},
2990 {0xa733,0xa76f,2,-1},
2991 {0xa77a,0xa77c,2,-1},
2992 {0xa77f,0xa787,2,-1},
2993 {0xa78c,0xa78c,-1,-1},
2994 {0xff41,0xff5a,1,-32},
2995 {0x10428,0x1044f,1,-40}
2999 * Return the upper-case equivalent of "a", which is a UCS-4 character. Use
3000 * simple case folding.
3003 utf_toupper(a)
3004 int a;
3006 /* If 'casemap' contains "keepascii" use ASCII style toupper(). */
3007 if (a < 128 && (cmp_flags & CMP_KEEPASCII))
3008 return TOUPPER_ASC(a);
3010 #if defined(HAVE_TOWUPPER) && defined(__STDC_ISO_10646__)
3011 /* If towupper() is available and handles Unicode, use it. */
3012 if (!(cmp_flags & CMP_INTERNAL))
3013 return towupper(a);
3014 #endif
3016 /* For characters below 128 use locale sensitive toupper(). */
3017 if (a < 128)
3018 return TOUPPER_LOC(a);
3020 /* For any other characters use the above mapping table. */
3021 return utf_convert(a, toUpper, sizeof(toUpper));
3025 utf_islower(a)
3026 int a;
3028 return (utf_toupper(a) != a);
3032 * Return the lower-case equivalent of "a", which is a UCS-4 character. Use
3033 * simple case folding.
3036 utf_tolower(a)
3037 int a;
3039 /* If 'casemap' contains "keepascii" use ASCII style tolower(). */
3040 if (a < 128 && (cmp_flags & CMP_KEEPASCII))
3041 return TOLOWER_ASC(a);
3043 #if defined(HAVE_TOWLOWER) && defined(__STDC_ISO_10646__)
3044 /* If towlower() is available and handles Unicode, use it. */
3045 if (!(cmp_flags & CMP_INTERNAL))
3046 return towlower(a);
3047 #endif
3049 /* For characters below 128 use locale sensitive tolower(). */
3050 if (a < 128)
3051 return TOLOWER_LOC(a);
3053 /* For any other characters use the above mapping table. */
3054 return utf_convert(a, toLower, sizeof(toLower));
3058 utf_isupper(a)
3059 int a;
3061 return (utf_tolower(a) != a);
3065 * Version of strnicmp() that handles multi-byte characters.
3066 * Needed for Big5, Sjift-JIS and UTF-8 encoding. Other DBCS encodings can
3067 * probably use strnicmp(), because there are no ASCII characters in the
3068 * second byte.
3069 * Returns zero if s1 and s2 are equal (ignoring case), the difference between
3070 * two characters otherwise.
3073 mb_strnicmp(s1, s2, nn)
3074 char_u *s1, *s2;
3075 size_t nn;
3077 int i, j, l;
3078 int cdiff;
3079 int incomplete = FALSE;
3080 int n = (int)nn;
3082 for (i = 0; i < n; i += l)
3084 if (s1[i] == NUL && s2[i] == NUL) /* both strings end */
3085 return 0;
3086 if (enc_utf8)
3088 l = utf_byte2len(s1[i]);
3089 if (l > n - i)
3091 l = n - i; /* incomplete character */
3092 incomplete = TRUE;
3094 /* Check directly first, it's faster. */
3095 for (j = 0; j < l; ++j)
3097 if (s1[i + j] != s2[i + j])
3098 break;
3099 if (s1[i + j] == 0)
3100 /* Both stings have the same bytes but are incomplete or
3101 * have illegal bytes, accept them as equal. */
3102 l = j;
3104 if (j < l)
3106 /* If one of the two characters is incomplete return -1. */
3107 if (incomplete || i + utf_byte2len(s2[i]) > n)
3108 return -1;
3109 cdiff = utf_fold(utf_ptr2char(s1 + i))
3110 - utf_fold(utf_ptr2char(s2 + i));
3111 if (cdiff != 0)
3112 return cdiff;
3115 else
3117 l = (*mb_ptr2len)(s1 + i);
3118 if (l <= 1)
3120 /* Single byte: first check normally, then with ignore case. */
3121 if (s1[i] != s2[i])
3123 cdiff = MB_TOLOWER(s1[i]) - MB_TOLOWER(s2[i]);
3124 if (cdiff != 0)
3125 return cdiff;
3128 else
3130 /* For non-Unicode multi-byte don't ignore case. */
3131 if (l > n - i)
3132 l = n - i;
3133 cdiff = STRNCMP(s1 + i, s2 + i, l);
3134 if (cdiff != 0)
3135 return cdiff;
3139 return 0;
3143 * "g8": show bytes of the UTF-8 char under the cursor. Doesn't matter what
3144 * 'encoding' has been set to.
3146 void
3147 show_utf8()
3149 int len;
3150 int rlen = 0;
3151 char_u *line;
3152 int clen;
3153 int i;
3155 /* Get the byte length of the char under the cursor, including composing
3156 * characters. */
3157 line = ml_get_cursor();
3158 len = utfc_ptr2len(line);
3159 if (len == 0)
3161 MSG("NUL");
3162 return;
3165 clen = 0;
3166 for (i = 0; i < len; ++i)
3168 if (clen == 0)
3170 /* start of (composing) character, get its length */
3171 if (i > 0)
3173 STRCPY(IObuff + rlen, "+ ");
3174 rlen += 2;
3176 clen = utf_ptr2len(line + i);
3178 sprintf((char *)IObuff + rlen, "%02x ", line[i]);
3179 --clen;
3180 rlen += (int)STRLEN(IObuff + rlen);
3181 if (rlen > IOSIZE - 20)
3182 break;
3185 msg(IObuff);
3189 * mb_head_off() function pointer.
3190 * Return offset from "p" to the first byte of the character it points into.
3191 * If "p" points to the NUL at the end of the string return 0.
3192 * Returns 0 when already at the first byte of a character.
3195 latin_head_off(base, p)
3196 char_u *base UNUSED;
3197 char_u *p UNUSED;
3199 return 0;
3203 dbcs_head_off(base, p)
3204 char_u *base;
3205 char_u *p;
3207 char_u *q;
3209 /* It can't be a trailing byte when not using DBCS, at the start of the
3210 * string or the previous byte can't start a double-byte. */
3211 if (p <= base || MB_BYTE2LEN(p[-1]) == 1 || *p == NUL)
3212 return 0;
3214 /* This is slow: need to start at the base and go forward until the
3215 * byte we are looking for. Return 1 when we went past it, 0 otherwise. */
3216 q = base;
3217 while (q < p)
3218 q += dbcs_ptr2len(q);
3219 return (q == p) ? 0 : 1;
3223 * Special version of dbcs_head_off() that works for ScreenLines[], where
3224 * single-width DBCS_JPNU characters are stored separately.
3227 dbcs_screen_head_off(base, p)
3228 char_u *base;
3229 char_u *p;
3231 char_u *q;
3233 /* It can't be a trailing byte when not using DBCS, at the start of the
3234 * string or the previous byte can't start a double-byte.
3235 * For euc-jp an 0x8e byte in the previous cell always means we have a
3236 * lead byte in the current cell. */
3237 if (p <= base
3238 || (enc_dbcs == DBCS_JPNU && p[-1] == 0x8e)
3239 || MB_BYTE2LEN(p[-1]) == 1
3240 || *p == NUL)
3241 return 0;
3243 /* This is slow: need to start at the base and go forward until the
3244 * byte we are looking for. Return 1 when we went past it, 0 otherwise.
3245 * For DBCS_JPNU look out for 0x8e, which means the second byte is not
3246 * stored as the next byte. */
3247 q = base;
3248 while (q < p)
3250 if (enc_dbcs == DBCS_JPNU && *q == 0x8e)
3251 ++q;
3252 else
3253 q += dbcs_ptr2len(q);
3255 return (q == p) ? 0 : 1;
3259 utf_head_off(base, p)
3260 char_u *base;
3261 char_u *p;
3263 char_u *q;
3264 char_u *s;
3265 int c;
3266 int len;
3267 #ifdef FEAT_ARABIC
3268 char_u *j;
3269 #endif
3271 if (*p < 0x80) /* be quick for ASCII */
3272 return 0;
3274 /* Skip backwards over trailing bytes: 10xx.xxxx
3275 * Skip backwards again if on a composing char. */
3276 for (q = p; ; --q)
3278 /* Move s to the last byte of this char. */
3279 for (s = q; (s[1] & 0xc0) == 0x80; ++s)
3281 /* Move q to the first byte of this char. */
3282 while (q > base && (*q & 0xc0) == 0x80)
3283 --q;
3284 /* Check for illegal sequence. Do allow an illegal byte after where we
3285 * started. */
3286 len = utf8len_tab[*q];
3287 if (len != (int)(s - q + 1) && len != (int)(p - q + 1))
3288 return 0;
3290 if (q <= base)
3291 break;
3293 c = utf_ptr2char(q);
3294 if (utf_iscomposing(c))
3295 continue;
3297 #ifdef FEAT_ARABIC
3298 if (arabic_maycombine(c))
3300 /* Advance to get a sneak-peak at the next char */
3301 j = q;
3302 --j;
3303 /* Move j to the first byte of this char. */
3304 while (j > base && (*j & 0xc0) == 0x80)
3305 --j;
3306 if (arabic_combine(utf_ptr2char(j), c))
3307 continue;
3309 #endif
3310 break;
3313 return (int)(p - q);
3317 * Copy a character from "*fp" to "*tp" and advance the pointers.
3319 void
3320 mb_copy_char(fp, tp)
3321 char_u **fp;
3322 char_u **tp;
3324 int l = (*mb_ptr2len)(*fp);
3326 mch_memmove(*tp, *fp, (size_t)l);
3327 *tp += l;
3328 *fp += l;
3332 * Return the offset from "p" to the first byte of a character. When "p" is
3333 * at the start of a character 0 is returned, otherwise the offset to the next
3334 * character. Can start anywhere in a stream of bytes.
3337 mb_off_next(base, p)
3338 char_u *base;
3339 char_u *p;
3341 int i;
3342 int j;
3344 if (enc_utf8)
3346 if (*p < 0x80) /* be quick for ASCII */
3347 return 0;
3349 /* Find the next character that isn't 10xx.xxxx */
3350 for (i = 0; (p[i] & 0xc0) == 0x80; ++i)
3352 if (i > 0)
3354 /* Check for illegal sequence. */
3355 for (j = 0; p - j > base; ++j)
3356 if ((p[-j] & 0xc0) != 0x80)
3357 break;
3358 if (utf8len_tab[p[-j]] != i + j)
3359 return 0;
3361 return i;
3364 /* Only need to check if we're on a trail byte, it doesn't matter if we
3365 * want the offset to the next or current character. */
3366 return (*mb_head_off)(base, p);
3370 * Return the offset from "p" to the last byte of the character it points
3371 * into. Can start anywhere in a stream of bytes.
3374 mb_tail_off(base, p)
3375 char_u *base;
3376 char_u *p;
3378 int i;
3379 int j;
3381 if (*p == NUL)
3382 return 0;
3384 if (enc_utf8)
3386 /* Find the last character that is 10xx.xxxx */
3387 for (i = 0; (p[i + 1] & 0xc0) == 0x80; ++i)
3389 /* Check for illegal sequence. */
3390 for (j = 0; p - j > base; ++j)
3391 if ((p[-j] & 0xc0) != 0x80)
3392 break;
3393 if (utf8len_tab[p[-j]] != i + j + 1)
3394 return 0;
3395 return i;
3398 /* It can't be the first byte if a double-byte when not using DBCS, at the
3399 * end of the string or the byte can't start a double-byte. */
3400 if (enc_dbcs == 0 || p[1] == NUL || MB_BYTE2LEN(*p) == 1)
3401 return 0;
3403 /* Return 1 when on the lead byte, 0 when on the tail byte. */
3404 return 1 - dbcs_head_off(base, p);
3408 * Find the next illegal byte sequence.
3410 void
3411 utf_find_illegal()
3413 pos_T pos = curwin->w_cursor;
3414 char_u *p;
3415 int len;
3416 vimconv_T vimconv;
3417 char_u *tofree = NULL;
3419 vimconv.vc_type = CONV_NONE;
3420 if (enc_utf8 && (enc_canon_props(curbuf->b_p_fenc) & ENC_8BIT))
3422 /* 'encoding' is "utf-8" but we are editing a 8-bit encoded file,
3423 * possibly a utf-8 file with illegal bytes. Setup for conversion
3424 * from utf-8 to 'fileencoding'. */
3425 convert_setup(&vimconv, p_enc, curbuf->b_p_fenc);
3428 #ifdef FEAT_VIRTUALEDIT
3429 curwin->w_cursor.coladd = 0;
3430 #endif
3431 for (;;)
3433 p = ml_get_cursor();
3434 if (vimconv.vc_type != CONV_NONE)
3436 vim_free(tofree);
3437 tofree = string_convert(&vimconv, p, NULL);
3438 if (tofree == NULL)
3439 break;
3440 p = tofree;
3443 while (*p != NUL)
3445 /* Illegal means that there are not enough trail bytes (checked by
3446 * utf_ptr2len()) or too many of them (overlong sequence). */
3447 len = utf_ptr2len(p);
3448 if (*p >= 0x80 && (len == 1
3449 || utf_char2len(utf_ptr2char(p)) != len))
3451 if (vimconv.vc_type == CONV_NONE)
3452 curwin->w_cursor.col += (colnr_T)(p - ml_get_cursor());
3453 else
3455 int l;
3457 len = (int)(p - tofree);
3458 for (p = ml_get_cursor(); *p != NUL && len-- > 0; p += l)
3460 l = utf_ptr2len(p);
3461 curwin->w_cursor.col += l;
3464 goto theend;
3466 p += len;
3468 if (curwin->w_cursor.lnum == curbuf->b_ml.ml_line_count)
3469 break;
3470 ++curwin->w_cursor.lnum;
3471 curwin->w_cursor.col = 0;
3474 /* didn't find it: don't move and beep */
3475 curwin->w_cursor = pos;
3476 beep_flush();
3478 theend:
3479 vim_free(tofree);
3480 convert_setup(&vimconv, NULL, NULL);
3483 #if defined(HAVE_GTK2) || defined(PROTO)
3485 * Return TRUE if string "s" is a valid utf-8 string.
3486 * When "end" is NULL stop at the first NUL.
3487 * When "end" is positive stop there.
3490 utf_valid_string(s, end)
3491 char_u *s;
3492 char_u *end;
3494 int l;
3495 char_u *p = s;
3497 while (end == NULL ? *p != NUL : p < end)
3499 l = utf8len_tab_zero[*p];
3500 if (l == 0)
3501 return FALSE; /* invalid lead byte */
3502 if (end != NULL && p + l > end)
3503 return FALSE; /* incomplete byte sequence */
3504 ++p;
3505 while (--l > 0)
3506 if ((*p++ & 0xc0) != 0x80)
3507 return FALSE; /* invalid trail byte */
3509 return TRUE;
3511 #endif
3513 #if defined(FEAT_GUI) || defined(PROTO)
3515 * Special version of mb_tail_off() for use in ScreenLines[].
3518 dbcs_screen_tail_off(base, p)
3519 char_u *base;
3520 char_u *p;
3522 /* It can't be the first byte if a double-byte when not using DBCS, at the
3523 * end of the string or the byte can't start a double-byte.
3524 * For euc-jp an 0x8e byte always means we have a lead byte in the current
3525 * cell. */
3526 if (*p == NUL || p[1] == NUL
3527 || (enc_dbcs == DBCS_JPNU && *p == 0x8e)
3528 || MB_BYTE2LEN(*p) == 1)
3529 return 0;
3531 /* Return 1 when on the lead byte, 0 when on the tail byte. */
3532 return 1 - dbcs_screen_head_off(base, p);
3534 #endif
3537 * If the cursor moves on an trail byte, set the cursor on the lead byte.
3538 * Thus it moves left if necessary.
3539 * Return TRUE when the cursor was adjusted.
3541 void
3542 mb_adjust_cursor()
3544 mb_adjustpos(&curwin->w_cursor);
3548 * Adjust position "*lp" to point to the first byte of a multi-byte character.
3549 * If it points to a tail byte it's moved backwards to the head byte.
3551 void
3552 mb_adjustpos(lp)
3553 pos_T *lp;
3555 char_u *p;
3557 if (lp->col > 0
3558 #ifdef FEAT_VIRTUALEDIT
3559 || lp->coladd > 1
3560 #endif
3563 p = ml_get(lp->lnum);
3564 lp->col -= (*mb_head_off)(p, p + lp->col);
3565 #ifdef FEAT_VIRTUALEDIT
3566 /* Reset "coladd" when the cursor would be on the right half of a
3567 * double-wide character. */
3568 if (lp->coladd == 1
3569 && p[lp->col] != TAB
3570 && vim_isprintc((*mb_ptr2char)(p + lp->col))
3571 && ptr2cells(p + lp->col) > 1)
3572 lp->coladd = 0;
3573 #endif
3578 * Return a pointer to the character before "*p", if there is one.
3580 char_u *
3581 mb_prevptr(line, p)
3582 char_u *line; /* start of the string */
3583 char_u *p;
3585 if (p > line)
3586 mb_ptr_back(line, p);
3587 return p;
3591 * Return the character length of "str". Each multi-byte character (with
3592 * following composing characters) counts as one.
3595 mb_charlen(str)
3596 char_u *str;
3598 char_u *p = str;
3599 int count;
3601 if (p == NULL)
3602 return 0;
3604 for (count = 0; *p != NUL; count++)
3605 p += (*mb_ptr2len)(p);
3607 return count;
3610 #if defined(FEAT_SPELL) || defined(PROTO)
3612 * Like mb_charlen() but for a string with specified length.
3615 mb_charlen_len(str, len)
3616 char_u *str;
3617 int len;
3619 char_u *p = str;
3620 int count;
3622 for (count = 0; *p != NUL && p < str + len; count++)
3623 p += (*mb_ptr2len)(p);
3625 return count;
3627 #endif
3630 * Try to un-escape a multi-byte character.
3631 * Used for the "to" and "from" part of a mapping.
3632 * Return the un-escaped string if it is a multi-byte character, and advance
3633 * "pp" to just after the bytes that formed it.
3634 * Return NULL if no multi-byte char was found.
3636 char_u *
3637 mb_unescape(pp)
3638 char_u **pp;
3640 static char_u buf[MB_MAXBYTES + 1];
3641 int n, m = 0;
3642 char_u *str = *pp;
3644 /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI
3645 * KS_EXTRA KE_CSI to CSI. */
3646 for (n = 0; str[n] != NUL && m <= MB_MAXBYTES; ++n)
3648 if (str[n] == K_SPECIAL
3649 && str[n + 1] == KS_SPECIAL
3650 && str[n + 2] == KE_FILLER)
3652 buf[m++] = K_SPECIAL;
3653 n += 2;
3655 else if ((str[n] == K_SPECIAL
3656 # ifdef FEAT_GUI
3657 || str[n] == CSI
3658 # endif
3660 && str[n + 1] == KS_EXTRA
3661 && str[n + 2] == (int)KE_CSI)
3663 buf[m++] = CSI;
3664 n += 2;
3666 else if (str[n] == K_SPECIAL
3667 # ifdef FEAT_GUI
3668 || str[n] == CSI
3669 # endif
3671 break; /* a special key can't be a multibyte char */
3672 else
3673 buf[m++] = str[n];
3674 buf[m] = NUL;
3676 /* Return a multi-byte character if it's found. An illegal sequence
3677 * will result in a 1 here. */
3678 if ((*mb_ptr2len)(buf) > 1)
3680 *pp = str + n + 1;
3681 return buf;
3684 return NULL;
3688 * Return TRUE if the character at "row"/"col" on the screen is the left side
3689 * of a double-width character.
3690 * Caller must make sure "row" and "col" are not invalid!
3693 mb_lefthalve(row, col)
3694 int row;
3695 int col;
3697 #ifdef FEAT_HANGULIN
3698 if (composing_hangul)
3699 return TRUE;
3700 #endif
3701 return (*mb_off2cells)(LineOffset[row] + col,
3702 LineOffset[row] + screen_Columns) > 1;
3706 * Correct a position on the screen, if it's the right half of a double-wide
3707 * char move it to the left half. Returns the corrected column.
3710 mb_fix_col(col, row)
3711 int col;
3712 int row;
3714 col = check_col(col);
3715 row = check_row(row);
3716 if (has_mbyte && ScreenLines != NULL && col > 0
3717 && ((enc_dbcs
3718 && ScreenLines[LineOffset[row] + col] != NUL
3719 && dbcs_screen_head_off(ScreenLines + LineOffset[row],
3720 ScreenLines + LineOffset[row] + col))
3721 || (enc_utf8 && ScreenLines[LineOffset[row] + col] == 0)))
3722 return col - 1;
3723 return col;
3725 #endif
3727 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO)
3728 static int enc_alias_search __ARGS((char_u *name));
3731 * Skip the Vim specific head of a 'encoding' name.
3733 char_u *
3734 enc_skip(p)
3735 char_u *p;
3737 if (STRNCMP(p, "2byte-", 6) == 0)
3738 return p + 6;
3739 if (STRNCMP(p, "8bit-", 5) == 0)
3740 return p + 5;
3741 return p;
3745 * Find the canonical name for encoding "enc".
3746 * When the name isn't recognized, returns "enc" itself, but with all lower
3747 * case characters and '_' replaced with '-'.
3748 * Returns an allocated string. NULL for out-of-memory.
3750 char_u *
3751 enc_canonize(enc)
3752 char_u *enc;
3754 char_u *r;
3755 char_u *p, *s;
3756 int i;
3758 # ifdef FEAT_MBYTE
3759 if (STRCMP(enc, "default") == 0)
3761 /* Use the default encoding as it's found by set_init_1(). */
3762 r = get_encoding_default();
3763 if (r == NULL)
3764 r = (char_u *)"latin1";
3765 return vim_strsave(r);
3767 # endif
3769 /* copy "enc" to allocated memory, with room for two '-' */
3770 r = alloc((unsigned)(STRLEN(enc) + 3));
3771 if (r != NULL)
3773 /* Make it all lower case and replace '_' with '-'. */
3774 p = r;
3775 for (s = enc; *s != NUL; ++s)
3777 if (*s == '_')
3778 *p++ = '-';
3779 else
3780 *p++ = TOLOWER_ASC(*s);
3782 *p = NUL;
3784 /* Skip "2byte-" and "8bit-". */
3785 p = enc_skip(r);
3787 /* Change "microsoft-cp" to "cp". Used in some spell files. */
3788 if (STRNCMP(p, "microsoft-cp", 12) == 0)
3789 STRMOVE(p, p + 10);
3791 /* "iso8859" -> "iso-8859" */
3792 if (STRNCMP(p, "iso8859", 7) == 0)
3794 STRMOVE(p + 4, p + 3);
3795 p[3] = '-';
3798 /* "iso-8859n" -> "iso-8859-n" */
3799 if (STRNCMP(p, "iso-8859", 8) == 0 && p[8] != '-')
3801 STRMOVE(p + 9, p + 8);
3802 p[8] = '-';
3805 /* "latin-N" -> "latinN" */
3806 if (STRNCMP(p, "latin-", 6) == 0)
3807 STRMOVE(p + 5, p + 6);
3809 if (enc_canon_search(p) >= 0)
3811 /* canonical name can be used unmodified */
3812 if (p != r)
3813 STRMOVE(r, p);
3815 else if ((i = enc_alias_search(p)) >= 0)
3817 /* alias recognized, get canonical name */
3818 vim_free(r);
3819 r = vim_strsave((char_u *)enc_canon_table[i].name);
3822 return r;
3826 * Search for an encoding alias of "name".
3827 * Returns -1 when not found.
3829 static int
3830 enc_alias_search(name)
3831 char_u *name;
3833 int i;
3835 for (i = 0; enc_alias_table[i].name != NULL; ++i)
3836 if (STRCMP(name, enc_alias_table[i].name) == 0)
3837 return enc_alias_table[i].canon;
3838 return -1;
3840 #endif
3842 #if defined(FEAT_MBYTE) || defined(PROTO)
3844 #ifdef HAVE_LANGINFO_H
3845 # include <langinfo.h>
3846 #endif
3849 * Get the canonicalized encoding of the current locale.
3850 * Returns an allocated string when successful, NULL when not.
3852 char_u *
3853 enc_locale()
3855 #ifndef WIN3264
3856 char *s;
3857 char *p;
3858 int i;
3859 #endif
3860 char buf[50];
3861 #ifdef WIN3264
3862 long acp = GetACP();
3864 if (acp == 1200)
3865 STRCPY(buf, "ucs-2le");
3866 else if (acp == 1252) /* cp1252 is used as latin1 */
3867 STRCPY(buf, "latin1");
3868 else
3869 sprintf(buf, "cp%ld", acp);
3870 #else
3871 # ifdef HAVE_NL_LANGINFO_CODESET
3872 if ((s = nl_langinfo(CODESET)) == NULL || *s == NUL)
3873 # endif
3874 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE)
3875 if ((s = setlocale(LC_CTYPE, NULL)) == NULL || *s == NUL)
3876 # endif
3877 if ((s = getenv("LC_ALL")) == NULL || *s == NUL)
3878 if ((s = getenv("LC_CTYPE")) == NULL || *s == NUL)
3879 s = getenv("LANG");
3881 if (s == NULL || *s == NUL)
3882 return FAIL;
3884 /* The most generic locale format is:
3885 * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]]
3886 * If there is a '.' remove the part before it.
3887 * if there is something after the codeset, remove it.
3888 * Make the name lowercase and replace '_' with '-'.
3889 * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn",
3890 * "ko_KR.EUC" == "euc-kr"
3892 if ((p = (char *)vim_strchr((char_u *)s, '.')) != NULL)
3894 if (p > s + 2 && STRNICMP(p + 1, "EUC", 3) == 0
3895 && !isalnum((int)p[4]) && p[4] != '-' && p[-3] == '_')
3897 /* copy "XY.EUC" to "euc-XY" to buf[10] */
3898 STRCPY(buf + 10, "euc-");
3899 buf[14] = p[-2];
3900 buf[15] = p[-1];
3901 buf[16] = 0;
3902 s = buf + 10;
3904 else
3905 s = p + 1;
3907 for (i = 0; s[i] != NUL && i < (int)sizeof(buf) - 1; ++i)
3909 if (s[i] == '_' || s[i] == '-')
3910 buf[i] = '-';
3911 else if (isalnum((int)s[i]))
3912 buf[i] = TOLOWER_ASC(s[i]);
3913 else
3914 break;
3916 buf[i] = NUL;
3917 #endif
3919 return enc_canonize((char_u *)buf);
3922 #if defined(WIN3264) || defined(PROTO)
3924 * Convert an encoding name to an MS-Windows codepage.
3925 * Returns zero if no codepage can be figured out.
3928 encname2codepage(name)
3929 char_u *name;
3931 int cp;
3932 char_u *p = name;
3933 int idx;
3935 if (STRNCMP(p, "8bit-", 5) == 0)
3936 p += 5;
3937 else if (STRNCMP(p_enc, "2byte-", 6) == 0)
3938 p += 6;
3940 if (p[0] == 'c' && p[1] == 'p')
3941 cp = atoi(p + 2);
3942 else if ((idx = enc_canon_search(p)) >= 0)
3943 cp = enc_canon_table[idx].codepage;
3944 else
3945 return 0;
3946 if (IsValidCodePage(cp))
3947 return cp;
3948 return 0;
3950 #endif
3952 # if defined(USE_ICONV) || defined(PROTO)
3954 static char_u *iconv_string __ARGS((vimconv_T *vcp, char_u *str, int slen, int *unconvlenp, int *resultlenp));
3957 * Call iconv_open() with a check if iconv() works properly (there are broken
3958 * versions).
3959 * Returns (void *)-1 if failed.
3960 * (should return iconv_t, but that causes problems with prototypes).
3962 void *
3963 my_iconv_open(to, from)
3964 char_u *to;
3965 char_u *from;
3967 iconv_t fd;
3968 #define ICONV_TESTLEN 400
3969 char_u tobuf[ICONV_TESTLEN];
3970 char *p;
3971 size_t tolen;
3972 static int iconv_ok = -1;
3974 if (iconv_ok == FALSE)
3975 return (void *)-1; /* detected a broken iconv() previously */
3977 #ifdef DYNAMIC_ICONV
3978 /* Check if the iconv.dll can be found. */
3979 if (!iconv_enabled(TRUE))
3980 return (void *)-1;
3981 #endif
3983 fd = iconv_open((char *)enc_skip(to), (char *)enc_skip(from));
3985 if (fd != (iconv_t)-1 && iconv_ok == -1)
3988 * Do a dummy iconv() call to check if it actually works. There is a
3989 * version of iconv() on Linux that is broken. We can't ignore it,
3990 * because it's wide-spread. The symptoms are that after outputting
3991 * the initial shift state the "to" pointer is NULL and conversion
3992 * stops for no apparent reason after about 8160 characters.
3994 p = (char *)tobuf;
3995 tolen = ICONV_TESTLEN;
3996 (void)iconv(fd, NULL, NULL, &p, &tolen);
3997 if (p == NULL)
3999 iconv_ok = FALSE;
4000 iconv_close(fd);
4001 fd = (iconv_t)-1;
4003 else
4004 iconv_ok = TRUE;
4007 return (void *)fd;
4011 * Convert the string "str[slen]" with iconv().
4012 * If "unconvlenp" is not NULL handle the string ending in an incomplete
4013 * sequence and set "*unconvlenp" to the length of it.
4014 * Returns the converted string in allocated memory. NULL for an error.
4015 * If resultlenp is not NULL, sets it to the result length in bytes.
4017 static char_u *
4018 iconv_string(vcp, str, slen, unconvlenp, resultlenp)
4019 vimconv_T *vcp;
4020 char_u *str;
4021 int slen;
4022 int *unconvlenp;
4023 int *resultlenp;
4025 const char *from;
4026 size_t fromlen;
4027 char *to;
4028 size_t tolen;
4029 size_t len = 0;
4030 size_t done = 0;
4031 char_u *result = NULL;
4032 char_u *p;
4033 int l;
4035 from = (char *)str;
4036 fromlen = slen;
4037 for (;;)
4039 if (len == 0 || ICONV_ERRNO == ICONV_E2BIG)
4041 /* Allocate enough room for most conversions. When re-allocating
4042 * increase the buffer size. */
4043 len = len + fromlen * 2 + 40;
4044 p = alloc((unsigned)len);
4045 if (p != NULL && done > 0)
4046 mch_memmove(p, result, done);
4047 vim_free(result);
4048 result = p;
4049 if (result == NULL) /* out of memory */
4050 break;
4053 to = (char *)result + done;
4054 tolen = len - done - 2;
4055 /* Avoid a warning for systems with a wrong iconv() prototype by
4056 * casting the second argument to void *. */
4057 if (iconv(vcp->vc_fd, (void *)&from, &fromlen, &to, &tolen)
4058 != (size_t)-1)
4060 /* Finished, append a NUL. */
4061 *to = NUL;
4062 break;
4065 /* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded
4066 * iconv library may use one of them. */
4067 if (!vcp->vc_fail && unconvlenp != NULL
4068 && (ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
4070 /* Handle an incomplete sequence at the end. */
4071 *to = NUL;
4072 *unconvlenp = (int)fromlen;
4073 break;
4076 /* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded
4077 * iconv library may use one of them. */
4078 else if (!vcp->vc_fail
4079 && (ICONV_ERRNO == ICONV_EILSEQ || ICONV_ERRNO == EILSEQ
4080 || ICONV_ERRNO == ICONV_EINVAL || ICONV_ERRNO == EINVAL))
4082 /* Can't convert: insert a '?' and skip a character. This assumes
4083 * conversion from 'encoding' to something else. In other
4084 * situations we don't know what to skip anyway. */
4085 *to++ = '?';
4086 if ((*mb_ptr2cells)((char_u *)from) > 1)
4087 *to++ = '?';
4088 if (enc_utf8)
4089 l = utfc_ptr2len_len((char_u *)from, (int)fromlen);
4090 else
4092 l = (*mb_ptr2len)((char_u *)from);
4093 if (l > (int)fromlen)
4094 l = (int)fromlen;
4096 from += l;
4097 fromlen -= l;
4099 else if (ICONV_ERRNO != ICONV_E2BIG)
4101 /* conversion failed */
4102 vim_free(result);
4103 result = NULL;
4104 break;
4106 /* Not enough room or skipping illegal sequence. */
4107 done = to - (char *)result;
4110 if (resultlenp != NULL)
4111 *resultlenp = (int)(to - (char *)result);
4112 return result;
4115 # if defined(DYNAMIC_ICONV) || defined(PROTO)
4117 * Dynamically load the "iconv.dll" on Win32.
4120 #ifndef DYNAMIC_ICONV /* just generating prototypes */
4121 # define HINSTANCE int
4122 #endif
4123 static HINSTANCE hIconvDLL = 0;
4124 static HINSTANCE hMsvcrtDLL = 0;
4126 # ifndef DYNAMIC_ICONV_DLL
4127 # define DYNAMIC_ICONV_DLL "iconv.dll"
4128 # define DYNAMIC_ICONV_DLL_ALT "libiconv.dll"
4129 # endif
4130 # ifndef DYNAMIC_MSVCRT_DLL
4131 # define DYNAMIC_MSVCRT_DLL "msvcrt.dll"
4132 # endif
4135 * Try opening the iconv.dll and return TRUE if iconv() can be used.
4138 iconv_enabled(verbose)
4139 int verbose;
4141 if (hIconvDLL != 0 && hMsvcrtDLL != 0)
4142 return TRUE;
4143 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL);
4144 if (hIconvDLL == 0) /* sometimes it's called libiconv.dll */
4145 hIconvDLL = LoadLibrary(DYNAMIC_ICONV_DLL_ALT);
4146 if (hIconvDLL != 0)
4147 hMsvcrtDLL = LoadLibrary(DYNAMIC_MSVCRT_DLL);
4148 if (hIconvDLL == 0 || hMsvcrtDLL == 0)
4150 /* Only give the message when 'verbose' is set, otherwise it might be
4151 * done whenever a conversion is attempted. */
4152 if (verbose && p_verbose > 0)
4154 verbose_enter();
4155 EMSG2(_(e_loadlib),
4156 hIconvDLL == 0 ? DYNAMIC_ICONV_DLL : DYNAMIC_MSVCRT_DLL);
4157 verbose_leave();
4159 iconv_end();
4160 return FALSE;
4163 iconv = (void *)GetProcAddress(hIconvDLL, "libiconv");
4164 iconv_open = (void *)GetProcAddress(hIconvDLL, "libiconv_open");
4165 iconv_close = (void *)GetProcAddress(hIconvDLL, "libiconv_close");
4166 iconvctl = (void *)GetProcAddress(hIconvDLL, "libiconvctl");
4167 iconv_errno = (void *)GetProcAddress(hMsvcrtDLL, "_errno");
4168 if (iconv == NULL || iconv_open == NULL || iconv_close == NULL
4169 || iconvctl == NULL || iconv_errno == NULL)
4171 iconv_end();
4172 if (verbose && p_verbose > 0)
4174 verbose_enter();
4175 EMSG2(_(e_loadfunc), "for libiconv");
4176 verbose_leave();
4178 return FALSE;
4180 return TRUE;
4183 void
4184 iconv_end()
4186 /* Don't use iconv() when inputting or outputting characters. */
4187 if (input_conv.vc_type == CONV_ICONV)
4188 convert_setup(&input_conv, NULL, NULL);
4189 if (output_conv.vc_type == CONV_ICONV)
4190 convert_setup(&output_conv, NULL, NULL);
4192 if (hIconvDLL != 0)
4193 FreeLibrary(hIconvDLL);
4194 if (hMsvcrtDLL != 0)
4195 FreeLibrary(hMsvcrtDLL);
4196 hIconvDLL = 0;
4197 hMsvcrtDLL = 0;
4199 # endif /* DYNAMIC_ICONV */
4200 # endif /* USE_ICONV */
4202 #endif /* FEAT_MBYTE */
4204 #if defined(FEAT_XIM) || defined(PROTO)
4206 # ifdef FEAT_GUI_GTK
4207 static int xim_has_preediting INIT(= FALSE); /* IM current status */
4210 * Set preedit_start_col to the current cursor position.
4212 static void
4213 init_preedit_start_col(void)
4215 if (State & CMDLINE)
4216 preedit_start_col = cmdline_getvcol_cursor();
4217 else if (curwin != NULL)
4218 getvcol(curwin, &curwin->w_cursor, &preedit_start_col, NULL, NULL);
4219 /* Prevent that preediting marks the buffer as changed. */
4220 xim_changed_while_preediting = curbuf->b_changed;
4222 # endif
4224 # if defined(HAVE_GTK2) && !defined(PROTO)
4226 static int im_is_active = FALSE; /* IM is enabled for current mode */
4227 static int preedit_is_active = FALSE;
4228 static int im_preedit_cursor = 0; /* cursor offset in characters */
4229 static int im_preedit_trailing = 0; /* number of characters after cursor */
4231 static unsigned long im_commit_handler_id = 0;
4232 static unsigned int im_activatekey_keyval = GDK_VoidSymbol;
4233 static unsigned int im_activatekey_state = 0;
4235 void
4236 im_set_active(int active)
4238 int was_active;
4240 was_active = !!im_is_active;
4241 im_is_active = (active && !p_imdisable);
4243 if (im_is_active != was_active)
4244 xim_reset();
4247 void
4248 xim_set_focus(int focus)
4250 if (xic != NULL)
4252 if (focus)
4253 gtk_im_context_focus_in(xic);
4254 else
4255 gtk_im_context_focus_out(xic);
4259 void
4260 im_set_position(int row, int col)
4262 if (xic != NULL)
4264 GdkRectangle area;
4266 area.x = FILL_X(col);
4267 area.y = FILL_Y(row);
4268 area.width = gui.char_width * (mb_lefthalve(row, col) ? 2 : 1);
4269 area.height = gui.char_height;
4271 gtk_im_context_set_cursor_location(xic, &area);
4275 # if 0 || defined(PROTO) /* apparently only used in gui_x11.c */
4276 void
4277 xim_set_preedit(void)
4279 im_set_position(gui.row, gui.col);
4281 # endif
4283 static void
4284 im_add_to_input(char_u *str, int len)
4286 /* Convert from 'termencoding' (always "utf-8") to 'encoding' */
4287 if (input_conv.vc_type != CONV_NONE)
4289 str = string_convert(&input_conv, str, &len);
4290 g_return_if_fail(str != NULL);
4293 add_to_input_buf_csi(str, len);
4295 if (input_conv.vc_type != CONV_NONE)
4296 vim_free(str);
4298 if (p_mh) /* blank out the pointer if necessary */
4299 gui_mch_mousehide(TRUE);
4302 static void
4303 im_delete_preedit(void)
4305 char_u bskey[] = {CSI, 'k', 'b'};
4306 char_u delkey[] = {CSI, 'k', 'D'};
4308 if (State & NORMAL)
4310 im_preedit_cursor = 0;
4311 return;
4313 for (; im_preedit_cursor > 0; --im_preedit_cursor)
4314 add_to_input_buf(bskey, (int)sizeof(bskey));
4316 for (; im_preedit_trailing > 0; --im_preedit_trailing)
4317 add_to_input_buf(delkey, (int)sizeof(delkey));
4321 * Move the cursor left by "num_move_back" characters.
4322 * Note that ins_left() checks im_is_preediting() to avoid breaking undo for
4323 * these K_LEFT keys.
4325 static void
4326 im_correct_cursor(int num_move_back)
4328 char_u backkey[] = {CSI, 'k', 'l'};
4330 if (State & NORMAL)
4331 return;
4332 # ifdef FEAT_RIGHTLEFT
4333 if ((State & CMDLINE) == 0 && curwin != NULL && curwin->w_p_rl)
4334 backkey[2] = 'r';
4335 # endif
4336 for (; num_move_back > 0; --num_move_back)
4337 add_to_input_buf(backkey, (int)sizeof(backkey));
4340 static int xim_expected_char = NUL;
4341 static int xim_ignored_char = FALSE;
4344 * Update the mode and cursor while in an IM callback.
4346 static void
4347 im_show_info(void)
4349 int old_vgetc_busy;
4351 old_vgetc_busy = vgetc_busy;
4352 vgetc_busy = TRUE;
4353 showmode();
4354 vgetc_busy = old_vgetc_busy;
4355 setcursor();
4356 out_flush();
4360 * Callback invoked when the user finished preediting.
4361 * Put the final string into the input buffer.
4363 static void
4364 im_commit_cb(GtkIMContext *context UNUSED,
4365 const gchar *str,
4366 gpointer data UNUSED)
4368 int slen = (int)STRLEN(str);
4369 int add_to_input = TRUE;
4370 int clen;
4371 int len = slen;
4372 int commit_with_preedit = TRUE;
4373 char_u *im_str, *p;
4375 #ifdef XIM_DEBUG
4376 xim_log("im_commit_cb(): %s\n", str);
4377 #endif
4379 /* The imhangul module doesn't reset the preedit string before
4380 * committing. Call im_delete_preedit() to work around that. */
4381 im_delete_preedit();
4383 /* Indicate that preediting has finished. */
4384 if (preedit_start_col == MAXCOL)
4386 init_preedit_start_col();
4387 commit_with_preedit = FALSE;
4390 /* The thing which setting "preedit_start_col" to MAXCOL means that
4391 * "preedit_start_col" will be set forcely when calling
4392 * preedit_changed_cb() next time.
4393 * "preedit_start_col" should not reset with MAXCOL on this part. Vim
4394 * is simulating the preediting by using add_to_input_str(). when
4395 * preedit begin immediately before committed, the typebuf is not
4396 * flushed to screen, then it can't get correct "preedit_start_col".
4397 * Thus, it should calculate the cells by adding cells of the committed
4398 * string. */
4399 if (input_conv.vc_type != CONV_NONE)
4401 im_str = string_convert(&input_conv, (char_u *)str, &len);
4402 g_return_if_fail(im_str != NULL);
4404 else
4405 im_str = (char_u *)str;
4406 clen = 0;
4407 for (p = im_str; p < im_str + len; p += (*mb_ptr2len)(p))
4408 clen += (*mb_ptr2cells)(p);
4409 if (input_conv.vc_type != CONV_NONE)
4410 vim_free(im_str);
4411 preedit_start_col += clen;
4413 /* Is this a single character that matches a keypad key that's just
4414 * been pressed? If so, we don't want it to be entered as such - let
4415 * us carry on processing the raw keycode so that it may be used in
4416 * mappings as <kSomething>. */
4417 if (xim_expected_char != NUL)
4419 /* We're currently processing a keypad or other special key */
4420 if (slen == 1 && str[0] == xim_expected_char)
4422 /* It's a match - don't do it here */
4423 xim_ignored_char = TRUE;
4424 add_to_input = FALSE;
4426 else
4428 /* Not a match */
4429 xim_ignored_char = FALSE;
4433 if (add_to_input)
4434 im_add_to_input((char_u *)str, slen);
4436 /* Inserting chars while "im_is_active" is set does not cause a change of
4437 * buffer. When the chars are committed the buffer must be marked as
4438 * changed. */
4439 if (!commit_with_preedit)
4440 preedit_start_col = MAXCOL;
4442 /* This flag is used in changed() at next call. */
4443 xim_changed_while_preediting = TRUE;
4445 if (gtk_main_level() > 0)
4446 gtk_main_quit();
4450 * Callback invoked after start to the preedit.
4452 static void
4453 im_preedit_start_cb(GtkIMContext *context UNUSED, gpointer data UNUSED)
4455 #ifdef XIM_DEBUG
4456 xim_log("im_preedit_start_cb()\n");
4457 #endif
4459 im_is_active = TRUE;
4460 preedit_is_active = TRUE;
4461 gui_update_cursor(TRUE, FALSE);
4462 im_show_info();
4466 * Callback invoked after end to the preedit.
4468 static void
4469 im_preedit_end_cb(GtkIMContext *context UNUSED, gpointer data UNUSED)
4471 #ifdef XIM_DEBUG
4472 xim_log("im_preedit_end_cb()\n");
4473 #endif
4474 im_delete_preedit();
4476 /* Indicate that preediting has finished */
4477 preedit_start_col = MAXCOL;
4478 xim_has_preediting = FALSE;
4480 #if 0
4481 /* Removal of this line suggested by Takuhiro Nishioka. Fixes that IM was
4482 * switched off unintentionally. We now use preedit_is_active (added by
4483 * SungHyun Nam). */
4484 im_is_active = FALSE;
4485 #endif
4486 preedit_is_active = FALSE;
4487 gui_update_cursor(TRUE, FALSE);
4488 im_show_info();
4492 * Callback invoked after changes to the preedit string. If the preedit
4493 * string was empty before, remember the preedit start column so we know
4494 * where to apply feedback attributes. Delete the previous preedit string
4495 * if there was one, save the new preedit cursor offset, and put the new
4496 * string into the input buffer.
4498 * TODO: The pragmatic "put into input buffer" approach used here has
4499 * several fundamental problems:
4501 * - The characters in the preedit string are subject to remapping.
4502 * That's broken, only the finally committed string should be remapped.
4504 * - There is a race condition involved: The retrieved value for the
4505 * current cursor position will be wrong if any unprocessed characters
4506 * are still queued in the input buffer.
4508 * - Due to the lack of synchronization between the file buffer in memory
4509 * and any typed characters, it's practically impossible to implement the
4510 * "retrieve_surrounding" and "delete_surrounding" signals reliably. IM
4511 * modules for languages such as Thai are likely to rely on this feature
4512 * for proper operation.
4514 * Conclusions: I think support for preediting needs to be moved to the
4515 * core parts of Vim. Ideally, until it has been committed, the preediting
4516 * string should only be displayed and not affect the buffer content at all.
4517 * The question how to deal with the synchronization issue still remains.
4518 * Circumventing the input buffer is probably not desirable. Anyway, I think
4519 * implementing "retrieve_surrounding" is the only hard problem.
4521 * One way to solve all of this in a clean manner would be to queue all key
4522 * press/release events "as is" in the input buffer, and apply the IM filtering
4523 * at the receiving end of the queue. This, however, would have a rather large
4524 * impact on the code base. If there is an easy way to force processing of all
4525 * remaining input from within the "retrieve_surrounding" signal handler, this
4526 * might not be necessary. Gotta ask on vim-dev for opinions.
4528 static void
4529 im_preedit_changed_cb(GtkIMContext *context, gpointer data UNUSED)
4531 char *preedit_string = NULL;
4532 int cursor_index = 0;
4533 int num_move_back = 0;
4534 char_u *str;
4535 char_u *p;
4536 int i;
4538 gtk_im_context_get_preedit_string(context,
4539 &preedit_string, NULL,
4540 &cursor_index);
4542 #ifdef XIM_DEBUG
4543 xim_log("im_preedit_changed_cb(): %s\n", preedit_string);
4544 #endif
4546 g_return_if_fail(preedit_string != NULL); /* just in case */
4548 /* If preedit_start_col is MAXCOL set it to the current cursor position. */
4549 if (preedit_start_col == MAXCOL && preedit_string[0] != '\0')
4551 xim_has_preediting = TRUE;
4553 /* Urgh, this breaks if the input buffer isn't empty now */
4554 init_preedit_start_col();
4556 else if (cursor_index == 0 && preedit_string[0] == '\0')
4558 xim_has_preediting = FALSE;
4560 /* If at the start position (after typing backspace)
4561 * preedit_start_col must be reset. */
4562 preedit_start_col = MAXCOL;
4565 im_delete_preedit();
4568 * Compute the end of the preediting area: "preedit_end_col".
4569 * According to the documentation of gtk_im_context_get_preedit_string(),
4570 * the cursor_pos output argument returns the offset in bytes. This is
4571 * unfortunately not true -- real life shows the offset is in characters,
4572 * and the GTK+ source code agrees with me. Will file a bug later.
4574 if (preedit_start_col != MAXCOL)
4575 preedit_end_col = preedit_start_col;
4576 str = (char_u *)preedit_string;
4577 for (p = str, i = 0; *p != NUL; p += utf_byte2len(*p), ++i)
4579 int is_composing;
4581 is_composing = ((*p & 0x80) != 0 && utf_iscomposing(utf_ptr2char(p)));
4583 * These offsets are used as counters when generating <BS> and <Del>
4584 * to delete the preedit string. So don't count composing characters
4585 * unless 'delcombine' is enabled.
4587 if (!is_composing || p_deco)
4589 if (i < cursor_index)
4590 ++im_preedit_cursor;
4591 else
4592 ++im_preedit_trailing;
4594 if (!is_composing && i >= cursor_index)
4596 /* This is essentially the same as im_preedit_trailing, except
4597 * composing characters are not counted even if p_deco is set. */
4598 ++num_move_back;
4600 if (preedit_start_col != MAXCOL)
4601 preedit_end_col += utf_ptr2cells(p);
4604 if (p > str)
4606 im_add_to_input(str, (int)(p - str));
4607 im_correct_cursor(num_move_back);
4610 g_free(preedit_string);
4612 if (gtk_main_level() > 0)
4613 gtk_main_quit();
4617 * Translate the Pango attributes at iter to Vim highlighting attributes.
4618 * Ignore attributes not supported by Vim highlighting. This shouldn't have
4619 * too much impact -- right now we handle even more attributes than necessary
4620 * for the IM modules I tested with.
4622 static int
4623 translate_pango_attributes(PangoAttrIterator *iter)
4625 PangoAttribute *attr;
4626 int char_attr = HL_NORMAL;
4628 attr = pango_attr_iterator_get(iter, PANGO_ATTR_UNDERLINE);
4629 if (attr != NULL && ((PangoAttrInt *)attr)->value
4630 != (int)PANGO_UNDERLINE_NONE)
4631 char_attr |= HL_UNDERLINE;
4633 attr = pango_attr_iterator_get(iter, PANGO_ATTR_WEIGHT);
4634 if (attr != NULL && ((PangoAttrInt *)attr)->value >= (int)PANGO_WEIGHT_BOLD)
4635 char_attr |= HL_BOLD;
4637 attr = pango_attr_iterator_get(iter, PANGO_ATTR_STYLE);
4638 if (attr != NULL && ((PangoAttrInt *)attr)->value
4639 != (int)PANGO_STYLE_NORMAL)
4640 char_attr |= HL_ITALIC;
4642 attr = pango_attr_iterator_get(iter, PANGO_ATTR_BACKGROUND);
4643 if (attr != NULL)
4645 const PangoColor *color = &((PangoAttrColor *)attr)->color;
4647 /* Assume inverse if black background is requested */
4648 if ((color->red | color->green | color->blue) == 0)
4649 char_attr |= HL_INVERSE;
4652 return char_attr;
4656 * Retrieve the highlighting attributes at column col in the preedit string.
4657 * Return -1 if not in preediting mode or if col is out of range.
4660 im_get_feedback_attr(int col)
4662 char *preedit_string = NULL;
4663 PangoAttrList *attr_list = NULL;
4664 int char_attr = -1;
4666 if (xic == NULL)
4667 return char_attr;
4669 gtk_im_context_get_preedit_string(xic, &preedit_string, &attr_list, NULL);
4671 if (preedit_string != NULL && attr_list != NULL)
4673 int idx;
4675 /* Get the byte index as used by PangoAttrIterator */
4676 for (idx = 0; col > 0 && preedit_string[idx] != '\0'; --col)
4677 idx += utfc_ptr2len((char_u *)preedit_string + idx);
4679 if (preedit_string[idx] != '\0')
4681 PangoAttrIterator *iter;
4682 int start, end;
4684 char_attr = HL_NORMAL;
4685 iter = pango_attr_list_get_iterator(attr_list);
4687 /* Extract all relevant attributes from the list. */
4690 pango_attr_iterator_range(iter, &start, &end);
4692 if (idx >= start && idx < end)
4693 char_attr |= translate_pango_attributes(iter);
4695 while (pango_attr_iterator_next(iter));
4697 pango_attr_iterator_destroy(iter);
4701 if (attr_list != NULL)
4702 pango_attr_list_unref(attr_list);
4703 g_free(preedit_string);
4705 return char_attr;
4708 void
4709 xim_init(void)
4711 #ifdef XIM_DEBUG
4712 xim_log("xim_init()\n");
4713 #endif
4715 g_return_if_fail(gui.drawarea != NULL);
4716 g_return_if_fail(gui.drawarea->window != NULL);
4718 xic = gtk_im_multicontext_new();
4719 g_object_ref(xic);
4721 im_commit_handler_id = g_signal_connect(G_OBJECT(xic), "commit",
4722 G_CALLBACK(&im_commit_cb), NULL);
4723 g_signal_connect(G_OBJECT(xic), "preedit_changed",
4724 G_CALLBACK(&im_preedit_changed_cb), NULL);
4725 g_signal_connect(G_OBJECT(xic), "preedit_start",
4726 G_CALLBACK(&im_preedit_start_cb), NULL);
4727 g_signal_connect(G_OBJECT(xic), "preedit_end",
4728 G_CALLBACK(&im_preedit_end_cb), NULL);
4730 gtk_im_context_set_client_window(xic, gui.drawarea->window);
4733 void
4734 im_shutdown(void)
4736 #ifdef XIM_DEBUG
4737 xim_log("im_shutdown()\n");
4738 #endif
4740 if (xic != NULL)
4742 gtk_im_context_focus_out(xic);
4743 g_object_unref(xic);
4744 xic = NULL;
4746 im_is_active = FALSE;
4747 im_commit_handler_id = 0;
4748 preedit_start_col = MAXCOL;
4749 xim_has_preediting = FALSE;
4753 * Convert the string argument to keyval and state for GdkEventKey.
4754 * If str is valid return TRUE, otherwise FALSE.
4756 * See 'imactivatekey' for documentation of the format.
4758 static int
4759 im_string_to_keyval(const char *str, unsigned int *keyval, unsigned int *state)
4761 const char *mods_end;
4762 unsigned tmp_keyval;
4763 unsigned tmp_state = 0;
4765 mods_end = strrchr(str, '-');
4766 mods_end = (mods_end != NULL) ? mods_end + 1 : str;
4768 /* Parse modifier keys */
4769 while (str < mods_end)
4770 switch (*str++)
4772 case '-': break;
4773 case 'S': case 's': tmp_state |= (unsigned)GDK_SHIFT_MASK; break;
4774 case 'L': case 'l': tmp_state |= (unsigned)GDK_LOCK_MASK; break;
4775 case 'C': case 'c': tmp_state |= (unsigned)GDK_CONTROL_MASK;break;
4776 case '1': tmp_state |= (unsigned)GDK_MOD1_MASK; break;
4777 case '2': tmp_state |= (unsigned)GDK_MOD2_MASK; break;
4778 case '3': tmp_state |= (unsigned)GDK_MOD3_MASK; break;
4779 case '4': tmp_state |= (unsigned)GDK_MOD4_MASK; break;
4780 case '5': tmp_state |= (unsigned)GDK_MOD5_MASK; break;
4781 default:
4782 return FALSE;
4785 tmp_keyval = gdk_keyval_from_name(str);
4787 if (tmp_keyval == 0 || tmp_keyval == GDK_VoidSymbol)
4788 return FALSE;
4790 if (keyval != NULL)
4791 *keyval = tmp_keyval;
4792 if (state != NULL)
4793 *state = tmp_state;
4795 return TRUE;
4799 * Return TRUE if p_imak is valid, otherwise FALSE. As a special case, an
4800 * empty string is also regarded as valid.
4802 * Note: The numerical key value of p_imak is cached if it was valid; thus
4803 * boldly assuming im_xim_isvalid_imactivate() will always be called whenever
4804 * 'imak' changes. This is currently the case but not obvious -- should
4805 * probably rename the function for clarity.
4808 im_xim_isvalid_imactivate(void)
4810 if (p_imak[0] == NUL)
4812 im_activatekey_keyval = GDK_VoidSymbol;
4813 im_activatekey_state = 0;
4814 return TRUE;
4817 return im_string_to_keyval((const char *)p_imak,
4818 &im_activatekey_keyval,
4819 &im_activatekey_state);
4822 static void
4823 im_synthesize_keypress(unsigned int keyval, unsigned int state)
4825 GdkEventKey *event;
4827 # ifdef HAVE_GTK_MULTIHEAD
4828 event = (GdkEventKey *)gdk_event_new(GDK_KEY_PRESS);
4829 g_object_ref(gui.drawarea->window); /* unreffed by gdk_event_free() */
4830 # else
4831 event = (GdkEventKey *)g_malloc0((gulong)sizeof(GdkEvent));
4832 event->type = GDK_KEY_PRESS;
4833 # endif
4834 event->window = gui.drawarea->window;
4835 event->send_event = TRUE;
4836 event->time = GDK_CURRENT_TIME;
4837 event->state = state;
4838 event->keyval = keyval;
4839 event->hardware_keycode = /* needed for XIM */
4840 XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event->window), (KeySym)keyval);
4841 event->length = 0;
4842 event->string = NULL;
4844 gtk_im_context_filter_keypress(xic, event);
4846 /* For consistency, also send the corresponding release event. */
4847 event->type = GDK_KEY_RELEASE;
4848 event->send_event = FALSE;
4849 gtk_im_context_filter_keypress(xic, event);
4851 # ifdef HAVE_GTK_MULTIHEAD
4852 gdk_event_free((GdkEvent *)event);
4853 # else
4854 g_free(event);
4855 # endif
4858 void
4859 xim_reset(void)
4861 if (xic != NULL)
4864 * The third-party imhangul module (and maybe others too) ignores
4865 * gtk_im_context_reset() or at least doesn't reset the active state.
4866 * Thus sending imactivatekey would turn it off if it was on before,
4867 * which is clearly not what we want. Fortunately we can work around
4868 * that for imhangul by sending GDK_Escape, but I don't know if it
4869 * works with all IM modules that support an activation key :/
4871 * An alternative approach would be to destroy the IM context and
4872 * recreate it. But that means loading/unloading the IM module on
4873 * every mode switch, which causes a quite noticable delay even on
4874 * my rather fast box...
4876 * Moreover, there are some XIM which cannot respond to
4877 * im_synthesize_keypress(). we hope that they reset by
4878 * xim_shutdown().
4880 if (im_activatekey_keyval != GDK_VoidSymbol && im_is_active)
4881 im_synthesize_keypress(GDK_Escape, 0U);
4883 gtk_im_context_reset(xic);
4886 * HACK for Ami: This sequence of function calls makes Ami handle
4887 * the IM reset graciously, without breaking loads of other stuff.
4888 * It seems to force English mode as well, which is exactly what we
4889 * want because it makes the Ami status display work reliably.
4891 gtk_im_context_set_use_preedit(xic, FALSE);
4893 if (p_imdisable)
4894 im_shutdown();
4895 else
4897 gtk_im_context_set_use_preedit(xic, TRUE);
4898 xim_set_focus(gui.in_focus);
4900 if (im_activatekey_keyval != GDK_VoidSymbol)
4902 if (im_is_active)
4904 g_signal_handler_block(xic, im_commit_handler_id);
4905 im_synthesize_keypress(im_activatekey_keyval,
4906 im_activatekey_state);
4907 g_signal_handler_unblock(xic, im_commit_handler_id);
4910 else
4912 im_shutdown();
4913 xim_init();
4914 xim_set_focus(gui.in_focus);
4919 preedit_start_col = MAXCOL;
4920 xim_has_preediting = FALSE;
4924 xim_queue_key_press_event(GdkEventKey *event, int down)
4926 if (down)
4929 * Workaround GTK2 XIM 'feature' that always converts keypad keys to
4930 * chars., even when not part of an IM sequence (ref. feature of
4931 * gdk/gdkkeyuni.c).
4932 * Flag any keypad keys that might represent a single char.
4933 * If this (on its own - i.e., not part of an IM sequence) is
4934 * committed while we're processing one of these keys, we can ignore
4935 * that commit and go ahead & process it ourselves. That way we can
4936 * still distinguish keypad keys for use in mappings.
4937 * Also add GDK_space to make <S-Space> work.
4939 switch (event->keyval)
4941 case GDK_KP_Add: xim_expected_char = '+'; break;
4942 case GDK_KP_Subtract: xim_expected_char = '-'; break;
4943 case GDK_KP_Divide: xim_expected_char = '/'; break;
4944 case GDK_KP_Multiply: xim_expected_char = '*'; break;
4945 case GDK_KP_Decimal: xim_expected_char = '.'; break;
4946 case GDK_KP_Equal: xim_expected_char = '='; break;
4947 case GDK_KP_0: xim_expected_char = '0'; break;
4948 case GDK_KP_1: xim_expected_char = '1'; break;
4949 case GDK_KP_2: xim_expected_char = '2'; break;
4950 case GDK_KP_3: xim_expected_char = '3'; break;
4951 case GDK_KP_4: xim_expected_char = '4'; break;
4952 case GDK_KP_5: xim_expected_char = '5'; break;
4953 case GDK_KP_6: xim_expected_char = '6'; break;
4954 case GDK_KP_7: xim_expected_char = '7'; break;
4955 case GDK_KP_8: xim_expected_char = '8'; break;
4956 case GDK_KP_9: xim_expected_char = '9'; break;
4957 case GDK_space: xim_expected_char = ' '; break;
4958 default: xim_expected_char = NUL;
4960 xim_ignored_char = FALSE;
4964 * When typing fFtT, XIM may be activated. Thus it must pass
4965 * gtk_im_context_filter_keypress() in Normal mode.
4966 * And while doing :sh too.
4968 if (xic != NULL && !p_imdisable
4969 && (State & (INSERT | CMDLINE | NORMAL | EXTERNCMD)) != 0)
4972 * Filter 'imactivatekey' and map it to CTRL-^. This way, Vim is
4973 * always aware of the current status of IM, and can even emulate
4974 * the activation key for modules that don't support one.
4976 if (event->keyval == im_activatekey_keyval
4977 && (event->state & im_activatekey_state) == im_activatekey_state)
4979 unsigned int state_mask;
4981 /* Require the state of the 3 most used modifiers to match exactly.
4982 * Otherwise e.g. <S-C-space> would be unusable for other purposes
4983 * if the IM activate key is <S-space>. */
4984 state_mask = im_activatekey_state;
4985 state_mask |= ((int)GDK_SHIFT_MASK | (int)GDK_CONTROL_MASK
4986 | (int)GDK_MOD1_MASK);
4988 if ((event->state & state_mask) != im_activatekey_state)
4989 return FALSE;
4991 /* Don't send it a second time on GDK_KEY_RELEASE. */
4992 if (event->type != GDK_KEY_PRESS)
4993 return TRUE;
4995 if (map_to_exists_mode((char_u *)"", LANGMAP, FALSE))
4997 im_set_active(FALSE);
4999 /* ":lmap" mappings exists, toggle use of mappings. */
5000 State ^= LANGMAP;
5001 if (State & LANGMAP)
5003 curbuf->b_p_iminsert = B_IMODE_NONE;
5004 State &= ~LANGMAP;
5006 else
5008 curbuf->b_p_iminsert = B_IMODE_LMAP;
5009 State |= LANGMAP;
5011 return TRUE;
5014 return gtk_im_context_filter_keypress(xic, event);
5017 /* Don't filter events through the IM context if IM isn't active
5018 * right now. Unlike with GTK+ 1.2 we cannot rely on the IM module
5019 * not doing anything before the activation key was sent. */
5020 if (im_activatekey_keyval == GDK_VoidSymbol || im_is_active)
5022 int imresult = gtk_im_context_filter_keypress(xic, event);
5024 /* Some XIM send following sequence:
5025 * 1. preedited string.
5026 * 2. committed string.
5027 * 3. line changed key.
5028 * 4. preedited string.
5029 * 5. remove preedited string.
5030 * if 3, Vim can't move back the above line for 5.
5031 * thus, this part should not parse the key. */
5032 if (!imresult && preedit_start_col != MAXCOL
5033 && event->keyval == GDK_Return)
5035 im_synthesize_keypress(GDK_Return, 0U);
5036 return FALSE;
5039 /* If XIM tried to commit a keypad key as a single char.,
5040 * ignore it so we can use the keypad key 'raw', for mappings. */
5041 if (xim_expected_char != NUL && xim_ignored_char)
5042 /* We had a keypad key, and XIM tried to thieve it */
5043 return FALSE;
5045 /* Normal processing */
5046 return imresult;
5050 return FALSE;
5054 im_get_status(void)
5056 return im_is_active;
5059 # else /* !HAVE_GTK2 */
5061 static int xim_is_active = FALSE; /* XIM should be active in the current
5062 mode */
5063 static int xim_has_focus = FALSE; /* XIM is really being used for Vim */
5064 #ifdef FEAT_GUI_X11
5065 static XIMStyle input_style;
5066 static int status_area_enabled = TRUE;
5067 #endif
5069 #ifdef FEAT_GUI_GTK
5070 # ifdef WIN3264
5071 # include <gdk/gdkwin32.h>
5072 # else
5073 # include <gdk/gdkx.h>
5074 # endif
5075 #else
5076 # ifdef PROTO
5077 /* Define a few things to be able to generate prototypes while not configured
5078 * for GTK. */
5079 # define GSList int
5080 # define gboolean int
5081 typedef int GdkEvent;
5082 typedef int GdkEventKey;
5083 # define GdkIC int
5084 # endif
5085 #endif
5087 #if defined(FEAT_GUI_GTK) || defined(PROTO)
5088 static int preedit_buf_len = 0;
5089 static int xim_can_preediting INIT(= FALSE); /* XIM in showmode() */
5090 static int xim_input_style;
5091 #ifndef FEAT_GUI_GTK
5092 # define gboolean int
5093 #endif
5094 static gboolean use_status_area = 0;
5096 static int im_xim_str2keycode __ARGS((unsigned int *code, unsigned int *state));
5097 static void im_xim_send_event_imactivate __ARGS((void));
5100 * Convert string to keycode and state for XKeyEvent.
5101 * When string is valid return OK, when invalid return FAIL.
5103 * See 'imactivatekey' documentation for the format.
5105 static int
5106 im_xim_str2keycode(code, state)
5107 unsigned int *code;
5108 unsigned int *state;
5110 int retval = OK;
5111 int len;
5112 unsigned keycode = 0, keystate = 0;
5113 Window window;
5114 Display *display;
5115 char_u *flag_end;
5116 char_u *str;
5118 if (*p_imak != NUL)
5120 len = STRLEN(p_imak);
5121 for (flag_end = p_imak + len - 1;
5122 flag_end > p_imak && *flag_end != '-'; --flag_end)
5125 /* Parse modifier keys */
5126 for (str = p_imak; str < flag_end; ++str)
5128 switch (*str)
5130 case 's': case 'S':
5131 keystate |= ShiftMask;
5132 break;
5133 case 'l': case 'L':
5134 keystate |= LockMask;
5135 break;
5136 case 'c': case 'C':
5137 keystate |= ControlMask;
5138 break;
5139 case '1':
5140 keystate |= Mod1Mask;
5141 break;
5142 case '2':
5143 keystate |= Mod2Mask;
5144 break;
5145 case '3':
5146 keystate |= Mod3Mask;
5147 break;
5148 case '4':
5149 keystate |= Mod4Mask;
5150 break;
5151 case '5':
5152 keystate |= Mod5Mask;
5153 break;
5154 case '-':
5155 break;
5156 default:
5157 retval = FAIL;
5160 if (*str == '-')
5161 ++str;
5163 /* Get keycode from string. */
5164 gui_get_x11_windis(&window, &display);
5165 if (display)
5166 keycode = XKeysymToKeycode(display, XStringToKeysym((char *)str));
5167 if (keycode == 0)
5168 retval = FAIL;
5170 if (code != NULL)
5171 *code = keycode;
5172 if (state != NULL)
5173 *state = keystate;
5175 return retval;
5178 static void
5179 im_xim_send_event_imactivate()
5181 /* Force turn on preedit state by symulate keypress event.
5182 * Keycode and state is specified by 'imactivatekey'.
5184 XKeyEvent ev;
5186 gui_get_x11_windis(&ev.window, &ev.display);
5187 ev.root = RootWindow(ev.display, DefaultScreen(ev.display));
5188 ev.subwindow = None;
5189 ev.time = CurrentTime;
5190 ev.x = 1;
5191 ev.y = 1;
5192 ev.x_root = 1;
5193 ev.y_root = 1;
5194 ev.same_screen = 1;
5195 ev.type = KeyPress;
5196 if (im_xim_str2keycode(&ev.keycode, &ev.state) == OK)
5197 XSendEvent(ev.display, ev.window, 1, KeyPressMask, (XEvent*)&ev);
5201 * Return TRUE if 'imactivatekey' has a valid value.
5204 im_xim_isvalid_imactivate()
5206 return im_xim_str2keycode(NULL, NULL) == OK;
5208 #endif /* FEAT_GUI_GTK */
5211 * Switch using XIM on/off. This is used by the code that changes "State".
5213 void
5214 im_set_active(active)
5215 int active;
5217 if (xic == NULL)
5218 return;
5220 /* If 'imdisable' is set, XIM is never active. */
5221 if (p_imdisable)
5222 active = FALSE;
5223 #if !defined (FEAT_GUI_GTK)
5224 else if (input_style & XIMPreeditPosition)
5225 /* There is a problem in switching XIM off when preediting is used,
5226 * and it is not clear how this can be solved. For now, keep XIM on
5227 * all the time, like it was done in Vim 5.8. */
5228 active = TRUE;
5229 #endif
5231 /* Remember the active state, it is needed when Vim gets keyboard focus. */
5232 xim_is_active = active;
5234 #ifdef FEAT_GUI_GTK
5235 /* When 'imactivatekey' has valid key-string, try to control XIM preedit
5236 * state. When 'imactivatekey' has no or invalid string, try old XIM
5237 * focus control.
5239 if (*p_imak != NUL)
5241 /* BASIC STRATEGY:
5242 * Destroy old Input Context (XIC), and create new one. New XIC
5243 * would have a state of preedit that is off. When argument:active
5244 * is false, that's all. Else argument:active is true, send a key
5245 * event specified by 'imactivatekey' to activate XIM preedit state.
5248 xim_is_active = TRUE; /* Disable old XIM focus control */
5249 /* If we can monitor preedit state with preedit callback functions,
5250 * try least creation of new XIC.
5252 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)
5254 if (xim_can_preediting && !active)
5256 /* Force turn off preedit state. With some IM
5257 * implementations, we cannot turn off preedit state by
5258 * symulate keypress event. It is why using such a method
5259 * that destroy old IC (input context), and create new one.
5260 * When create new IC, its preedit state is usually off.
5262 xim_reset();
5263 xim_set_focus(FALSE);
5264 gdk_ic_destroy(xic);
5265 xim_init();
5266 xim_can_preediting = FALSE;
5268 else if (!xim_can_preediting && active)
5269 im_xim_send_event_imactivate();
5271 else
5273 /* First, force destroy old IC, and create new one. It
5274 * symulates "turning off preedit state".
5276 xim_set_focus(FALSE);
5277 gdk_ic_destroy(xic);
5278 xim_init();
5279 xim_can_preediting = FALSE;
5281 /* 2nd, when requested to activate IM, symulate this by sending
5282 * the event.
5284 if (active)
5286 im_xim_send_event_imactivate();
5287 xim_can_preediting = TRUE;
5291 else
5293 # ifndef XIMPreeditUnKnown
5294 /* X11R5 doesn't have these, it looks safe enough to define here. */
5295 typedef unsigned long XIMPreeditState;
5296 # define XIMPreeditUnKnown 0L
5297 # define XIMPreeditEnable 1L
5298 # define XIMPreeditDisable (1L<<1)
5299 # define XNPreeditState "preeditState"
5300 # endif
5301 XIMPreeditState preedit_state = XIMPreeditUnKnown;
5302 XVaNestedList preedit_attr;
5303 XIC pxic;
5305 preedit_attr = XVaCreateNestedList(0,
5306 XNPreeditState, &preedit_state,
5307 NULL);
5308 pxic = ((GdkICPrivate *)xic)->xic;
5310 if (!XGetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL))
5312 XFree(preedit_attr);
5313 preedit_attr = XVaCreateNestedList(0,
5314 XNPreeditState,
5315 active ? XIMPreeditEnable : XIMPreeditDisable,
5316 NULL);
5317 XSetICValues(pxic, XNPreeditAttributes, preedit_attr, NULL);
5318 xim_can_preediting = active;
5319 xim_is_active = active;
5321 XFree(preedit_attr);
5323 if (xim_input_style & XIMPreeditCallbacks)
5325 preedit_buf_len = 0;
5326 init_preedit_start_col();
5328 #else
5329 # if 0
5330 /* When had tested kinput2 + canna + Athena GUI version with
5331 * 'imactivatekey' is "s-space", im_xim_send_event_imactivate() did not
5332 * work correctly. It just inserted one space. I don't know why we
5333 * couldn't switch state of XIM preediting. This is reason why these
5334 * codes are commented out.
5336 /* First, force destroy old IC, and create new one. It symulates
5337 * "turning off preedit state".
5339 xim_set_focus(FALSE);
5340 XDestroyIC(xic);
5341 xic = NULL;
5342 xim_init();
5344 /* 2nd, when requested to activate IM, symulate this by sending the
5345 * event.
5347 if (active)
5348 im_xim_send_event_imactivate();
5349 # endif
5350 #endif
5351 xim_set_preedit();
5355 * Adjust using XIM for gaining or losing keyboard focus. Also called when
5356 * "xim_is_active" changes.
5358 void
5359 xim_set_focus(focus)
5360 int focus;
5362 if (xic == NULL)
5363 return;
5366 * XIM only gets focus when the Vim window has keyboard focus and XIM has
5367 * been set active for the current mode.
5369 if (focus && xim_is_active)
5371 if (!xim_has_focus)
5373 xim_has_focus = TRUE;
5374 #ifdef FEAT_GUI_GTK
5375 gdk_im_begin(xic, gui.drawarea->window);
5376 #else
5377 XSetICFocus(xic);
5378 #endif
5381 else
5383 if (xim_has_focus)
5385 xim_has_focus = FALSE;
5386 #ifdef FEAT_GUI_GTK
5387 gdk_im_end();
5388 #else
5389 XUnsetICFocus(xic);
5390 #endif
5395 void
5396 im_set_position(row, col)
5397 int row UNUSED;
5398 int col UNUSED;
5400 xim_set_preedit();
5404 * Set the XIM to the current cursor position.
5406 void
5407 xim_set_preedit()
5409 if (xic == NULL)
5410 return;
5412 xim_set_focus(TRUE);
5414 #ifdef FEAT_GUI_GTK
5415 if (gdk_im_ready())
5417 int attrmask;
5418 GdkICAttr *attr;
5420 if (!xic_attr)
5421 return;
5423 attr = xic_attr;
5424 attrmask = 0;
5426 # ifdef FEAT_XFONTSET
5427 if ((xim_input_style & (int)GDK_IM_PREEDIT_POSITION)
5428 && gui.fontset != NOFONTSET
5429 && gui.fontset->type == GDK_FONT_FONTSET)
5431 if (!xim_has_focus)
5433 if (attr->spot_location.y >= 0)
5435 attr->spot_location.x = 0;
5436 attr->spot_location.y = -100;
5437 attrmask |= (int)GDK_IC_SPOT_LOCATION;
5440 else
5442 gint width, height;
5444 if (attr->spot_location.x != TEXT_X(gui.col)
5445 || attr->spot_location.y != TEXT_Y(gui.row))
5447 attr->spot_location.x = TEXT_X(gui.col);
5448 attr->spot_location.y = TEXT_Y(gui.row);
5449 attrmask |= (int)GDK_IC_SPOT_LOCATION;
5452 gdk_window_get_size(gui.drawarea->window, &width, &height);
5453 width -= 2 * gui.border_offset;
5454 height -= 2 * gui.border_offset;
5455 if (xim_input_style & (int)GDK_IM_STATUS_AREA)
5456 height -= gui.char_height;
5457 if (attr->preedit_area.width != width
5458 || attr->preedit_area.height != height)
5460 attr->preedit_area.x = gui.border_offset;
5461 attr->preedit_area.y = gui.border_offset;
5462 attr->preedit_area.width = width;
5463 attr->preedit_area.height = height;
5464 attrmask |= (int)GDK_IC_PREEDIT_AREA;
5467 if (attr->preedit_fontset != gui.current_font)
5469 attr->preedit_fontset = gui.current_font;
5470 attrmask |= (int)GDK_IC_PREEDIT_FONTSET;
5474 # endif /* FEAT_XFONTSET */
5476 if (xim_fg_color == INVALCOLOR)
5478 xim_fg_color = gui.def_norm_pixel;
5479 xim_bg_color = gui.def_back_pixel;
5481 if (attr->preedit_foreground.pixel != xim_fg_color)
5483 attr->preedit_foreground.pixel = xim_fg_color;
5484 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND;
5486 if (attr->preedit_background.pixel != xim_bg_color)
5488 attr->preedit_background.pixel = xim_bg_color;
5489 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND;
5492 if (attrmask != 0)
5493 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask);
5495 #else /* FEAT_GUI_GTK */
5497 XVaNestedList attr_list;
5498 XRectangle spot_area;
5499 XPoint over_spot;
5500 int line_space;
5502 if (!xim_has_focus)
5504 /* hide XIM cursor */
5505 over_spot.x = 0;
5506 over_spot.y = -100; /* arbitrary invisible position */
5507 attr_list = (XVaNestedList) XVaCreateNestedList(0,
5508 XNSpotLocation,
5509 &over_spot,
5510 NULL);
5511 XSetICValues(xic, XNPreeditAttributes, attr_list, NULL);
5512 XFree(attr_list);
5513 return;
5516 if (input_style & XIMPreeditPosition)
5518 if (xim_fg_color == INVALCOLOR)
5520 xim_fg_color = gui.def_norm_pixel;
5521 xim_bg_color = gui.def_back_pixel;
5523 over_spot.x = TEXT_X(gui.col);
5524 over_spot.y = TEXT_Y(gui.row);
5525 spot_area.x = 0;
5526 spot_area.y = 0;
5527 spot_area.height = gui.char_height * Rows;
5528 spot_area.width = gui.char_width * Columns;
5529 line_space = gui.char_height;
5530 attr_list = (XVaNestedList) XVaCreateNestedList(0,
5531 XNSpotLocation, &over_spot,
5532 XNForeground, (Pixel) xim_fg_color,
5533 XNBackground, (Pixel) xim_bg_color,
5534 XNArea, &spot_area,
5535 XNLineSpace, line_space,
5536 NULL);
5537 if (XSetICValues(xic, XNPreeditAttributes, attr_list, NULL))
5538 EMSG(_("E284: Cannot set IC values"));
5539 XFree(attr_list);
5542 #endif /* FEAT_GUI_GTK */
5546 * Set up the status area.
5548 * This should use a separate Widget, but that seems not possible, because
5549 * preedit_area and status_area should be set to the same window as for the
5550 * text input. Unfortunately this means the status area pollutes the text
5551 * window...
5553 void
5554 xim_set_status_area()
5556 if (xic == NULL)
5557 return;
5559 #ifdef FEAT_GUI_GTK
5560 # if defined(FEAT_XFONTSET)
5561 if (use_status_area)
5563 GdkICAttr *attr;
5564 int style;
5565 gint width, height;
5566 GtkWidget *widget;
5567 int attrmask;
5569 if (!xic_attr)
5570 return;
5572 attr = xic_attr;
5573 attrmask = 0;
5574 style = (int)gdk_ic_get_style(xic);
5575 if ((style & (int)GDK_IM_STATUS_MASK) == (int)GDK_IM_STATUS_AREA)
5577 if (gui.fontset != NOFONTSET
5578 && gui.fontset->type == GDK_FONT_FONTSET)
5580 widget = gui.mainwin;
5581 gdk_window_get_size(widget->window, &width, &height);
5583 attrmask |= (int)GDK_IC_STATUS_AREA;
5584 attr->status_area.x = 0;
5585 attr->status_area.y = height - gui.char_height - 1;
5586 attr->status_area.width = width;
5587 attr->status_area.height = gui.char_height;
5590 if (attrmask != 0)
5591 gdk_ic_set_attr(xic, attr, (GdkICAttributesType)attrmask);
5593 # endif
5594 #else
5596 XVaNestedList preedit_list = 0, status_list = 0, list = 0;
5597 XRectangle pre_area, status_area;
5599 if (input_style & XIMStatusArea)
5601 if (input_style & XIMPreeditArea)
5603 XRectangle *needed_rect;
5605 /* to get status_area width */
5606 status_list = XVaCreateNestedList(0, XNAreaNeeded,
5607 &needed_rect, NULL);
5608 XGetICValues(xic, XNStatusAttributes, status_list, NULL);
5609 XFree(status_list);
5611 status_area.width = needed_rect->width;
5613 else
5614 status_area.width = gui.char_width * Columns;
5616 status_area.x = 0;
5617 status_area.y = gui.char_height * Rows + gui.border_offset;
5618 if (gui.which_scrollbars[SBAR_BOTTOM])
5619 status_area.y += gui.scrollbar_height;
5620 #ifdef FEAT_MENU
5621 if (gui.menu_is_active)
5622 status_area.y += gui.menu_height;
5623 #endif
5624 status_area.height = gui.char_height;
5625 status_list = XVaCreateNestedList(0, XNArea, &status_area, NULL);
5627 else
5629 status_area.x = 0;
5630 status_area.y = gui.char_height * Rows + gui.border_offset;
5631 if (gui.which_scrollbars[SBAR_BOTTOM])
5632 status_area.y += gui.scrollbar_height;
5633 #ifdef FEAT_MENU
5634 if (gui.menu_is_active)
5635 status_area.y += gui.menu_height;
5636 #endif
5637 status_area.width = 0;
5638 status_area.height = gui.char_height;
5641 if (input_style & XIMPreeditArea) /* off-the-spot */
5643 pre_area.x = status_area.x + status_area.width;
5644 pre_area.y = gui.char_height * Rows + gui.border_offset;
5645 pre_area.width = gui.char_width * Columns - pre_area.x;
5646 if (gui.which_scrollbars[SBAR_BOTTOM])
5647 pre_area.y += gui.scrollbar_height;
5648 #ifdef FEAT_MENU
5649 if (gui.menu_is_active)
5650 pre_area.y += gui.menu_height;
5651 #endif
5652 pre_area.height = gui.char_height;
5653 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
5655 else if (input_style & XIMPreeditPosition) /* over-the-spot */
5657 pre_area.x = 0;
5658 pre_area.y = 0;
5659 pre_area.height = gui.char_height * Rows;
5660 pre_area.width = gui.char_width * Columns;
5661 preedit_list = XVaCreateNestedList(0, XNArea, &pre_area, NULL);
5664 if (preedit_list && status_list)
5665 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
5666 XNStatusAttributes, status_list, NULL);
5667 else if (preedit_list)
5668 list = XVaCreateNestedList(0, XNPreeditAttributes, preedit_list,
5669 NULL);
5670 else if (status_list)
5671 list = XVaCreateNestedList(0, XNStatusAttributes, status_list,
5672 NULL);
5673 else
5674 list = NULL;
5676 if (list)
5678 XSetICValues(xic, XNVaNestedList, list, NULL);
5679 XFree(list);
5681 if (status_list)
5682 XFree(status_list);
5683 if (preedit_list)
5684 XFree(preedit_list);
5686 #endif
5689 #if defined(FEAT_GUI_X11) || defined(FEAT_GUI_GTK)
5690 static char e_xim[] = N_("E285: Failed to create input context");
5691 #endif
5693 #if defined(FEAT_GUI_X11) || defined(PROTO)
5694 # if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(sun)
5695 # define USE_X11R6_XIM
5696 # endif
5698 static int xim_real_init __ARGS((Window x11_window, Display *x11_display));
5701 #ifdef USE_X11R6_XIM
5702 static void xim_instantiate_cb __ARGS((Display *display, XPointer client_data, XPointer call_data));
5703 static void xim_destroy_cb __ARGS((XIM im, XPointer client_data, XPointer call_data));
5705 static void
5706 xim_instantiate_cb(display, client_data, call_data)
5707 Display *display;
5708 XPointer client_data UNUSED;
5709 XPointer call_data UNUSED;
5711 Window x11_window;
5712 Display *x11_display;
5714 #ifdef XIM_DEBUG
5715 xim_log("xim_instantiate_cb()\n");
5716 #endif
5718 gui_get_x11_windis(&x11_window, &x11_display);
5719 if (display != x11_display)
5720 return;
5722 xim_real_init(x11_window, x11_display);
5723 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
5724 if (xic != NULL)
5725 XUnregisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
5726 xim_instantiate_cb, NULL);
5729 static void
5730 xim_destroy_cb(im, client_data, call_data)
5731 XIM im UNUSED;
5732 XPointer client_data UNUSED;
5733 XPointer call_data UNUSED;
5735 Window x11_window;
5736 Display *x11_display;
5738 #ifdef XIM_DEBUG
5739 xim_log("xim_destroy_cb()\n");
5740 #endif
5741 gui_get_x11_windis(&x11_window, &x11_display);
5743 xic = NULL;
5744 status_area_enabled = FALSE;
5746 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
5748 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
5749 xim_instantiate_cb, NULL);
5751 #endif
5753 void
5754 xim_init()
5756 Window x11_window;
5757 Display *x11_display;
5759 #ifdef XIM_DEBUG
5760 xim_log("xim_init()\n");
5761 #endif
5763 gui_get_x11_windis(&x11_window, &x11_display);
5765 xic = NULL;
5767 if (xim_real_init(x11_window, x11_display))
5768 return;
5770 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
5772 #ifdef USE_X11R6_XIM
5773 XRegisterIMInstantiateCallback(x11_display, NULL, NULL, NULL,
5774 xim_instantiate_cb, NULL);
5775 #endif
5778 static int
5779 xim_real_init(x11_window, x11_display)
5780 Window x11_window;
5781 Display *x11_display;
5783 int i;
5784 char *p,
5786 *ns,
5787 *end,
5788 tmp[1024];
5789 #define IMLEN_MAX 40
5790 char buf[IMLEN_MAX + 7];
5791 XIM xim = NULL;
5792 XIMStyles *xim_styles;
5793 XIMStyle this_input_style = 0;
5794 Boolean found;
5795 XPoint over_spot;
5796 XVaNestedList preedit_list, status_list;
5798 input_style = 0;
5799 status_area_enabled = FALSE;
5801 if (xic != NULL)
5802 return FALSE;
5804 if (gui.rsrc_input_method != NULL && *gui.rsrc_input_method != NUL)
5806 strcpy(tmp, gui.rsrc_input_method);
5807 for (ns = s = tmp; ns != NULL && *s != NUL;)
5809 s = (char *)skipwhite((char_u *)s);
5810 if (*s == NUL)
5811 break;
5812 if ((ns = end = strchr(s, ',')) == NULL)
5813 end = s + strlen(s);
5814 while (isspace(((char_u *)end)[-1]))
5815 end--;
5816 *end = NUL;
5818 if (strlen(s) <= IMLEN_MAX)
5820 strcpy(buf, "@im=");
5821 strcat(buf, s);
5822 if ((p = XSetLocaleModifiers(buf)) != NULL && *p != NUL
5823 && (xim = XOpenIM(x11_display, NULL, NULL, NULL))
5824 != NULL)
5825 break;
5828 s = ns + 1;
5832 if (xim == NULL && (p = XSetLocaleModifiers("")) != NULL && *p != NUL)
5833 xim = XOpenIM(x11_display, NULL, NULL, NULL);
5835 /* This is supposed to be useful to obtain characters through
5836 * XmbLookupString() without really using a XIM. */
5837 if (xim == NULL && (p = XSetLocaleModifiers("@im=none")) != NULL
5838 && *p != NUL)
5839 xim = XOpenIM(x11_display, NULL, NULL, NULL);
5841 if (xim == NULL)
5843 /* Only give this message when verbose is set, because too many people
5844 * got this message when they didn't want to use a XIM. */
5845 if (p_verbose > 0)
5847 verbose_enter();
5848 EMSG(_("E286: Failed to open input method"));
5849 verbose_leave();
5851 return FALSE;
5854 #ifdef USE_X11R6_XIM
5856 XIMCallback destroy_cb;
5858 destroy_cb.callback = xim_destroy_cb;
5859 destroy_cb.client_data = NULL;
5860 if (XSetIMValues(xim, XNDestroyCallback, &destroy_cb, NULL))
5861 EMSG(_("E287: Warning: Could not set destroy callback to IM"));
5863 #endif
5865 if (XGetIMValues(xim, XNQueryInputStyle, &xim_styles, NULL) || !xim_styles)
5867 EMSG(_("E288: input method doesn't support any style"));
5868 XCloseIM(xim);
5869 return FALSE;
5872 found = False;
5873 strcpy(tmp, gui.rsrc_preedit_type_name);
5874 for (s = tmp; s && !found; )
5876 while (*s && isspace((unsigned char)*s))
5877 s++;
5878 if (!*s)
5879 break;
5880 if ((ns = end = strchr(s, ',')) != 0)
5881 ns++;
5882 else
5883 end = s + strlen(s);
5884 while (isspace((unsigned char)*end))
5885 end--;
5886 *end = '\0';
5888 if (!strcmp(s, "OverTheSpot"))
5889 this_input_style = (XIMPreeditPosition | XIMStatusArea);
5890 else if (!strcmp(s, "OffTheSpot"))
5891 this_input_style = (XIMPreeditArea | XIMStatusArea);
5892 else if (!strcmp(s, "Root"))
5893 this_input_style = (XIMPreeditNothing | XIMStatusNothing);
5895 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
5897 if (this_input_style == xim_styles->supported_styles[i])
5899 found = True;
5900 break;
5903 if (!found)
5904 for (i = 0; (unsigned short)i < xim_styles->count_styles; i++)
5906 if ((xim_styles->supported_styles[i] & this_input_style)
5907 == (this_input_style & ~XIMStatusArea))
5909 this_input_style &= ~XIMStatusArea;
5910 found = True;
5911 break;
5915 s = ns;
5917 XFree(xim_styles);
5919 if (!found)
5921 /* Only give this message when verbose is set, because too many people
5922 * got this message when they didn't want to use a XIM. */
5923 if (p_verbose > 0)
5925 verbose_enter();
5926 EMSG(_("E289: input method doesn't support my preedit type"));
5927 verbose_leave();
5929 XCloseIM(xim);
5930 return FALSE;
5933 over_spot.x = TEXT_X(gui.col);
5934 over_spot.y = TEXT_Y(gui.row);
5935 input_style = this_input_style;
5937 /* A crash was reported when trying to pass gui.norm_font as XNFontSet,
5938 * thus that has been removed. Hopefully the default works... */
5939 #ifdef FEAT_XFONTSET
5940 if (gui.fontset != NOFONTSET)
5942 preedit_list = XVaCreateNestedList(0,
5943 XNSpotLocation, &over_spot,
5944 XNForeground, (Pixel)gui.def_norm_pixel,
5945 XNBackground, (Pixel)gui.def_back_pixel,
5946 XNFontSet, (XFontSet)gui.fontset,
5947 NULL);
5948 status_list = XVaCreateNestedList(0,
5949 XNForeground, (Pixel)gui.def_norm_pixel,
5950 XNBackground, (Pixel)gui.def_back_pixel,
5951 XNFontSet, (XFontSet)gui.fontset,
5952 NULL);
5954 else
5955 #endif
5957 preedit_list = XVaCreateNestedList(0,
5958 XNSpotLocation, &over_spot,
5959 XNForeground, (Pixel)gui.def_norm_pixel,
5960 XNBackground, (Pixel)gui.def_back_pixel,
5961 NULL);
5962 status_list = XVaCreateNestedList(0,
5963 XNForeground, (Pixel)gui.def_norm_pixel,
5964 XNBackground, (Pixel)gui.def_back_pixel,
5965 NULL);
5968 xic = XCreateIC(xim,
5969 XNInputStyle, input_style,
5970 XNClientWindow, x11_window,
5971 XNFocusWindow, gui.wid,
5972 XNPreeditAttributes, preedit_list,
5973 XNStatusAttributes, status_list,
5974 NULL);
5975 XFree(status_list);
5976 XFree(preedit_list);
5977 if (xic != NULL)
5979 if (input_style & XIMStatusArea)
5981 xim_set_status_area();
5982 status_area_enabled = TRUE;
5984 else
5985 gui_set_shellsize(FALSE, FALSE, RESIZE_BOTH);
5987 else
5989 EMSG(_(e_xim));
5990 XCloseIM(xim);
5991 return FALSE;
5994 return TRUE;
5997 #endif /* FEAT_GUI_X11 */
5999 #if defined(FEAT_GUI_GTK) || defined(PROTO)
6001 # ifdef FEAT_XFONTSET
6002 static char e_overthespot[] = N_("E290: over-the-spot style requires fontset");
6003 # endif
6005 # ifdef PROTO
6006 typedef int GdkIC;
6007 # endif
6009 void
6010 xim_decide_input_style()
6012 /* GDK_IM_STATUS_CALLBACKS was disabled, enabled it to allow Japanese
6013 * OverTheSpot. */
6014 int supported_style = (int)GDK_IM_PREEDIT_NONE |
6015 (int)GDK_IM_PREEDIT_NOTHING |
6016 (int)GDK_IM_PREEDIT_POSITION |
6017 (int)GDK_IM_PREEDIT_CALLBACKS |
6018 (int)GDK_IM_STATUS_CALLBACKS |
6019 (int)GDK_IM_STATUS_AREA |
6020 (int)GDK_IM_STATUS_NONE |
6021 (int)GDK_IM_STATUS_NOTHING;
6023 #ifdef XIM_DEBUG
6024 xim_log("xim_decide_input_style()\n");
6025 #endif
6027 if (!gdk_im_ready())
6028 xim_input_style = 0;
6029 else
6031 if (gtk_major_version > 1
6032 || (gtk_major_version == 1
6033 && (gtk_minor_version > 2
6034 || (gtk_minor_version == 2 && gtk_micro_version >= 3))))
6035 use_status_area = TRUE;
6036 else
6038 EMSG(_("E291: Your GTK+ is older than 1.2.3. Status area disabled"));
6039 use_status_area = FALSE;
6041 #ifdef FEAT_XFONTSET
6042 if (gui.fontset == NOFONTSET || gui.fontset->type != GDK_FONT_FONTSET)
6043 #endif
6044 supported_style &= ~((int)GDK_IM_PREEDIT_POSITION
6045 | (int)GDK_IM_STATUS_AREA);
6046 if (!use_status_area)
6047 supported_style &= ~(int)GDK_IM_STATUS_AREA;
6048 xim_input_style = (int)gdk_im_decide_style((GdkIMStyle)supported_style);
6052 static void
6053 preedit_start_cbproc(XIC thexic UNUSED,
6054 XPointer client_data UNUSED,
6055 XPointer call_data UNUSED)
6057 #ifdef XIM_DEBUG
6058 xim_log("xim_decide_input_style()\n");
6059 #endif
6061 draw_feedback = NULL;
6062 xim_can_preediting = TRUE;
6063 xim_has_preediting = TRUE;
6064 gui_update_cursor(TRUE, FALSE);
6065 if (showmode() > 0)
6067 setcursor();
6068 out_flush();
6072 static void
6073 xim_back_delete(int n)
6075 char_u str[3];
6077 str[0] = CSI;
6078 str[1] = 'k';
6079 str[2] = 'b';
6080 while (n-- > 0)
6081 add_to_input_buf(str, 3);
6084 static GSList *key_press_event_queue = NULL;
6085 static gboolean processing_queued_event = FALSE;
6087 static void
6088 preedit_draw_cbproc(XIC thexic UNUSED,
6089 XPointer client_data UNUSED,
6090 XPointer call_data)
6092 XIMPreeditDrawCallbackStruct *draw_data;
6093 XIMText *text;
6094 char *src;
6095 GSList *event_queue;
6097 #ifdef XIM_DEBUG
6098 xim_log("preedit_draw_cbproc()\n");
6099 #endif
6101 draw_data = (XIMPreeditDrawCallbackStruct *) call_data;
6102 text = (XIMText *) draw_data->text;
6104 if ((text == NULL && draw_data->chg_length == preedit_buf_len)
6105 || preedit_buf_len == 0)
6107 init_preedit_start_col();
6108 vim_free(draw_feedback);
6109 draw_feedback = NULL;
6111 if (draw_data->chg_length > 0)
6113 int bs_cnt;
6115 if (draw_data->chg_length > preedit_buf_len)
6116 bs_cnt = preedit_buf_len;
6117 else
6118 bs_cnt = draw_data->chg_length;
6119 xim_back_delete(bs_cnt);
6120 preedit_buf_len -= bs_cnt;
6122 if (text != NULL)
6124 int len;
6125 #ifdef FEAT_MBYTE
6126 char_u *buf = NULL;
6127 unsigned int nfeedback = 0;
6128 #endif
6129 char_u *ptr;
6131 src = text->string.multi_byte;
6132 if (src != NULL && !text->encoding_is_wchar)
6134 len = strlen(src);
6135 ptr = (char_u *)src;
6136 /* Avoid the enter for decision */
6137 if (*ptr == '\n')
6138 return;
6140 #ifdef FEAT_MBYTE
6141 if (input_conv.vc_type != CONV_NONE
6142 && (buf = string_convert(&input_conv,
6143 (char_u *)src, &len)) != NULL)
6145 /* Converted from 'termencoding' to 'encoding'. */
6146 add_to_input_buf_csi(buf, len);
6147 ptr = buf;
6149 else
6150 #endif
6151 add_to_input_buf_csi((char_u *)src, len);
6152 /* Add count of character to preedit_buf_len */
6153 while (*ptr != NUL)
6155 #ifdef FEAT_MBYTE
6156 if (draw_data->text->feedback != NULL)
6158 if (draw_feedback == NULL)
6159 draw_feedback = (char *)alloc(draw_data->chg_first
6160 + text->length);
6161 else
6162 draw_feedback = vim_realloc(draw_feedback,
6163 draw_data->chg_first + text->length);
6164 if (draw_feedback != NULL)
6166 draw_feedback[nfeedback + draw_data->chg_first]
6167 = draw_data->text->feedback[nfeedback];
6168 nfeedback++;
6171 if (has_mbyte)
6172 ptr += (*mb_ptr2len)(ptr);
6173 else
6174 #endif
6175 ptr++;
6176 preedit_buf_len++;
6178 #ifdef FEAT_MBYTE
6179 vim_free(buf);
6180 #endif
6181 preedit_end_col = MAXCOL;
6184 if (text != NULL || draw_data->chg_length > 0)
6186 event_queue = key_press_event_queue;
6187 processing_queued_event = TRUE;
6188 while (event_queue != NULL && processing_queued_event)
6190 GdkEvent *ev = event_queue->data;
6192 gboolean *ret;
6193 gtk_signal_emit_by_name((GtkObject*)gui.mainwin, "key_press_event",
6194 ev, &ret);
6195 gdk_event_free(ev);
6196 event_queue = event_queue->next;
6198 processing_queued_event = FALSE;
6199 if (key_press_event_queue)
6201 g_slist_free(key_press_event_queue);
6202 key_press_event_queue = NULL;
6205 if (gtk_main_level() > 0)
6206 gtk_main_quit();
6210 * Retrieve the highlighting attributes at column col in the preedit string.
6211 * Return -1 if not in preediting mode or if col is out of range.
6214 im_get_feedback_attr(int col)
6216 if (draw_feedback != NULL && col < preedit_buf_len)
6218 if (draw_feedback[col] & XIMReverse)
6219 return HL_INVERSE;
6220 else if (draw_feedback[col] & XIMUnderline)
6221 return HL_UNDERLINE;
6222 else
6223 return hl_attr(HLF_V);
6226 return -1;
6229 static void
6230 preedit_caret_cbproc(XIC thexic UNUSED,
6231 XPointer client_data UNUSED,
6232 XPointer call_data UNUSED)
6234 #ifdef XIM_DEBUG
6235 xim_log("preedit_caret_cbproc()\n");
6236 #endif
6239 static void
6240 preedit_done_cbproc(XIC thexic UNUSED,
6241 XPointer client_data UNUSED,
6242 XPointer call_data UNUSED)
6244 #ifdef XIM_DEBUG
6245 xim_log("preedit_done_cbproc()\n");
6246 #endif
6248 vim_free(draw_feedback);
6249 draw_feedback = NULL;
6250 xim_can_preediting = FALSE;
6251 xim_has_preediting = FALSE;
6252 gui_update_cursor(TRUE, FALSE);
6253 if (showmode() > 0)
6255 setcursor();
6256 out_flush();
6260 void
6261 xim_reset(void)
6263 char *text;
6265 #ifdef XIM_DEBUG
6266 xim_log("xim_reset()\n");
6267 #endif
6269 if (xic != NULL)
6271 text = XmbResetIC(((GdkICPrivate *)xic)->xic);
6272 if (text != NULL && !(xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS))
6273 add_to_input_buf_csi((char_u *)text, strlen(text));
6274 else
6275 preedit_buf_len = 0;
6276 if (text != NULL)
6277 XFree(text);
6282 xim_queue_key_press_event(GdkEventKey *event, int down UNUSED)
6284 #ifdef XIM_DEBUG
6285 xim_log("xim_queue_key_press_event()\n");
6286 #endif
6288 if (preedit_buf_len <= 0)
6289 return FALSE;
6290 if (processing_queued_event)
6291 processing_queued_event = FALSE;
6293 key_press_event_queue = g_slist_append(key_press_event_queue,
6294 gdk_event_copy((GdkEvent *)event));
6295 return TRUE;
6298 static void
6299 preedit_callback_setup(GdkIC *ic UNUSED)
6301 XIC xxic;
6302 XVaNestedList preedit_attr;
6303 XIMCallback preedit_start_cb;
6304 XIMCallback preedit_draw_cb;
6305 XIMCallback preedit_caret_cb;
6306 XIMCallback preedit_done_cb;
6308 xxic = ((GdkICPrivate*)xic)->xic;
6309 preedit_start_cb.callback = (XIMProc)preedit_start_cbproc;
6310 preedit_draw_cb.callback = (XIMProc)preedit_draw_cbproc;
6311 preedit_caret_cb.callback = (XIMProc)preedit_caret_cbproc;
6312 preedit_done_cb.callback = (XIMProc)preedit_done_cbproc;
6313 preedit_attr
6314 = XVaCreateNestedList(0,
6315 XNPreeditStartCallback, &preedit_start_cb,
6316 XNPreeditDrawCallback, &preedit_draw_cb,
6317 XNPreeditCaretCallback, &preedit_caret_cb,
6318 XNPreeditDoneCallback, &preedit_done_cb,
6319 NULL);
6320 XSetICValues(xxic, XNPreeditAttributes, preedit_attr, NULL);
6321 XFree(preedit_attr);
6324 static void
6325 reset_state_setup(GdkIC *ic UNUSED)
6327 #ifdef USE_X11R6_XIM
6328 /* don't change the input context when we call reset */
6329 XSetICValues(((GdkICPrivate *)ic)->xic, XNResetState, XIMPreserveState,
6330 NULL);
6331 #endif
6334 void
6335 xim_init(void)
6337 #ifdef XIM_DEBUG
6338 xim_log("xim_init()\n");
6339 #endif
6341 xic = NULL;
6342 xic_attr = NULL;
6344 if (!gdk_im_ready())
6346 if (p_verbose > 0)
6348 verbose_enter();
6349 EMSG(_("E292: Input Method Server is not running"));
6350 verbose_leave();
6352 return;
6354 if ((xic_attr = gdk_ic_attr_new()) != NULL)
6356 #ifdef FEAT_XFONTSET
6357 gint width, height;
6358 #endif
6359 int mask;
6360 GdkColormap *colormap;
6361 GdkICAttr *attr = xic_attr;
6362 int attrmask = (int)GDK_IC_ALL_REQ;
6363 GtkWidget *widget = gui.drawarea;
6365 attr->style = (GdkIMStyle)xim_input_style;
6366 attr->client_window = gui.mainwin->window;
6368 if ((colormap = gtk_widget_get_colormap(widget)) !=
6369 gtk_widget_get_default_colormap())
6371 attrmask |= (int)GDK_IC_PREEDIT_COLORMAP;
6372 attr->preedit_colormap = colormap;
6374 attrmask |= (int)GDK_IC_PREEDIT_FOREGROUND;
6375 attrmask |= (int)GDK_IC_PREEDIT_BACKGROUND;
6376 attr->preedit_foreground = widget->style->fg[GTK_STATE_NORMAL];
6377 attr->preedit_background = widget->style->base[GTK_STATE_NORMAL];
6379 #ifdef FEAT_XFONTSET
6380 if ((xim_input_style & (int)GDK_IM_PREEDIT_MASK)
6381 == (int)GDK_IM_PREEDIT_POSITION)
6383 if (gui.fontset == NOFONTSET
6384 || gui.fontset->type != GDK_FONT_FONTSET)
6386 EMSG(_(e_overthespot));
6388 else
6390 gdk_window_get_size(widget->window, &width, &height);
6392 attrmask |= (int)GDK_IC_PREEDIT_POSITION_REQ;
6393 attr->spot_location.x = TEXT_X(0);
6394 attr->spot_location.y = TEXT_Y(0);
6395 attr->preedit_area.x = gui.border_offset;
6396 attr->preedit_area.y = gui.border_offset;
6397 attr->preedit_area.width = width - 2*gui.border_offset;
6398 attr->preedit_area.height = height - 2*gui.border_offset;
6399 attr->preedit_fontset = gui.fontset;
6403 if ((xim_input_style & (int)GDK_IM_STATUS_MASK)
6404 == (int)GDK_IM_STATUS_AREA)
6406 if (gui.fontset == NOFONTSET
6407 || gui.fontset->type != GDK_FONT_FONTSET)
6409 EMSG(_(e_overthespot));
6411 else
6413 gdk_window_get_size(gui.mainwin->window, &width, &height);
6414 attrmask |= (int)GDK_IC_STATUS_AREA_REQ;
6415 attr->status_area.x = 0;
6416 attr->status_area.y = height - gui.char_height - 1;
6417 attr->status_area.width = width;
6418 attr->status_area.height = gui.char_height;
6419 attr->status_fontset = gui.fontset;
6422 else if ((xim_input_style & (int)GDK_IM_STATUS_MASK)
6423 == (int)GDK_IM_STATUS_CALLBACKS)
6425 /* FIXME */
6427 #endif
6429 xic = gdk_ic_new(attr, (GdkICAttributesType)attrmask);
6431 if (xic == NULL)
6432 EMSG(_(e_xim));
6433 else
6435 mask = (int)gdk_window_get_events(widget->window);
6436 mask |= (int)gdk_ic_get_events(xic);
6437 gdk_window_set_events(widget->window, (GdkEventMask)mask);
6438 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)
6439 preedit_callback_setup(xic);
6440 reset_state_setup(xic);
6445 void
6446 im_shutdown(void)
6448 #ifdef XIM_DEBUG
6449 xim_log("im_shutdown()\n");
6450 #endif
6452 if (xic != NULL)
6454 gdk_im_end();
6455 gdk_ic_destroy(xic);
6456 xic = NULL;
6458 xim_is_active = FALSE;
6459 xim_can_preediting = FALSE;
6460 preedit_start_col = MAXCOL;
6461 xim_has_preediting = FALSE;
6464 #endif /* FEAT_GUI_GTK */
6467 xim_get_status_area_height()
6469 #ifdef FEAT_GUI_GTK
6470 if (xim_input_style & (int)GDK_IM_STATUS_AREA)
6471 return gui.char_height;
6472 #else
6473 if (status_area_enabled)
6474 return gui.char_height;
6475 #endif
6476 return 0;
6480 * Get IM status. When IM is on, return TRUE. Else return FALSE.
6481 * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is
6482 * active, when not having focus XIM may still be active (e.g., when using a
6483 * tear-off menu item).
6486 im_get_status()
6488 # ifdef FEAT_GUI_GTK
6489 if (xim_input_style & (int)GDK_IM_PREEDIT_CALLBACKS)
6490 return xim_can_preediting;
6491 # endif
6492 return xim_has_focus;
6495 # endif /* !HAVE_GTK2 */
6497 # if defined(HAVE_GTK2) || defined(PROTO)
6499 preedit_get_status(void)
6501 return preedit_is_active;
6503 # endif
6505 # if defined(FEAT_GUI_GTK) || defined(PROTO)
6507 im_is_preediting()
6509 return xim_has_preediting;
6511 # endif
6512 #endif /* FEAT_XIM */
6514 #if defined(FEAT_MBYTE) || defined(PROTO)
6517 * Setup "vcp" for conversion from "from" to "to".
6518 * The names must have been made canonical with enc_canonize().
6519 * vcp->vc_type must have been initialized to CONV_NONE.
6520 * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8
6521 * instead).
6522 * Afterwards invoke with "from" and "to" equal to NULL to cleanup.
6523 * Return FAIL when conversion is not supported, OK otherwise.
6526 convert_setup(vcp, from, to)
6527 vimconv_T *vcp;
6528 char_u *from;
6529 char_u *to;
6531 return convert_setup_ext(vcp, from, TRUE, to, TRUE);
6535 * As convert_setup(), but only when from_unicode_is_utf8 is TRUE will all
6536 * "from" unicode charsets be considered utf-8. Same for "to".
6539 convert_setup_ext(vcp, from, from_unicode_is_utf8, to, to_unicode_is_utf8)
6540 vimconv_T *vcp;
6541 char_u *from;
6542 int from_unicode_is_utf8;
6543 char_u *to;
6544 int to_unicode_is_utf8;
6546 int from_prop;
6547 int to_prop;
6548 int from_is_utf8;
6549 int to_is_utf8;
6551 /* Reset to no conversion. */
6552 # ifdef USE_ICONV
6553 if (vcp->vc_type == CONV_ICONV && vcp->vc_fd != (iconv_t)-1)
6554 iconv_close(vcp->vc_fd);
6555 # endif
6556 vcp->vc_type = CONV_NONE;
6557 vcp->vc_factor = 1;
6558 vcp->vc_fail = FALSE;
6560 /* No conversion when one of the names is empty or they are equal. */
6561 if (from == NULL || *from == NUL || to == NULL || *to == NUL
6562 || STRCMP(from, to) == 0)
6563 return OK;
6565 from_prop = enc_canon_props(from);
6566 to_prop = enc_canon_props(to);
6567 if (from_unicode_is_utf8)
6568 from_is_utf8 = from_prop & ENC_UNICODE;
6569 else
6570 from_is_utf8 = from_prop == ENC_UNICODE;
6571 if (to_unicode_is_utf8)
6572 to_is_utf8 = to_prop & ENC_UNICODE;
6573 else
6574 to_is_utf8 = to_prop == ENC_UNICODE;
6576 if ((from_prop & ENC_LATIN1) && to_is_utf8)
6578 /* Internal latin1 -> utf-8 conversion. */
6579 vcp->vc_type = CONV_TO_UTF8;
6580 vcp->vc_factor = 2; /* up to twice as long */
6582 else if ((from_prop & ENC_LATIN9) && to_is_utf8)
6584 /* Internal latin9 -> utf-8 conversion. */
6585 vcp->vc_type = CONV_9_TO_UTF8;
6586 vcp->vc_factor = 3; /* up to three as long (euro sign) */
6588 else if (from_is_utf8 && (to_prop & ENC_LATIN1))
6590 /* Internal utf-8 -> latin1 conversion. */
6591 vcp->vc_type = CONV_TO_LATIN1;
6593 else if (from_is_utf8 && (to_prop & ENC_LATIN9))
6595 /* Internal utf-8 -> latin9 conversion. */
6596 vcp->vc_type = CONV_TO_LATIN9;
6598 #ifdef WIN3264
6599 /* Win32-specific codepage <-> codepage conversion without iconv. */
6600 else if ((from_is_utf8 || encname2codepage(from) > 0)
6601 && (to_is_utf8 || encname2codepage(to) > 0))
6603 vcp->vc_type = CONV_CODEPAGE;
6604 vcp->vc_factor = 2; /* up to twice as long */
6605 vcp->vc_cpfrom = from_is_utf8 ? 0 : encname2codepage(from);
6606 vcp->vc_cpto = to_is_utf8 ? 0 : encname2codepage(to);
6608 #endif
6609 #ifdef MACOS_X
6610 else if ((from_prop & ENC_MACROMAN) && (to_prop & ENC_LATIN1))
6612 vcp->vc_type = CONV_MAC_LATIN1;
6614 else if ((from_prop & ENC_MACROMAN) && to_is_utf8)
6616 vcp->vc_type = CONV_MAC_UTF8;
6617 vcp->vc_factor = 2; /* up to twice as long */
6619 else if ((from_prop & ENC_LATIN1) && (to_prop & ENC_MACROMAN))
6621 vcp->vc_type = CONV_LATIN1_MAC;
6623 else if (from_is_utf8 && (to_prop & ENC_MACROMAN))
6625 vcp->vc_type = CONV_UTF8_MAC;
6627 #endif
6628 # ifdef USE_ICONV
6629 else
6631 /* Use iconv() for conversion. */
6632 vcp->vc_fd = (iconv_t)my_iconv_open(
6633 to_is_utf8 ? (char_u *)"utf-8" : to,
6634 from_is_utf8 ? (char_u *)"utf-8" : from);
6635 if (vcp->vc_fd != (iconv_t)-1)
6637 vcp->vc_type = CONV_ICONV;
6638 vcp->vc_factor = 4; /* could be longer too... */
6641 # endif
6642 if (vcp->vc_type == CONV_NONE)
6643 return FAIL;
6645 return OK;
6648 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \
6649 || defined(MSDOS) || defined(PROTO)
6651 * Do conversion on typed input characters in-place.
6652 * The input and output are not NUL terminated!
6653 * Returns the length after conversion.
6656 convert_input(ptr, len, maxlen)
6657 char_u *ptr;
6658 int len;
6659 int maxlen;
6661 return convert_input_safe(ptr, len, maxlen, NULL, NULL);
6663 #endif
6666 * Like convert_input(), but when there is an incomplete byte sequence at the
6667 * end return that as an allocated string in "restp" and set "*restlenp" to
6668 * the length. If "restp" is NULL it is not used.
6671 convert_input_safe(ptr, len, maxlen, restp, restlenp)
6672 char_u *ptr;
6673 int len;
6674 int maxlen;
6675 char_u **restp;
6676 int *restlenp;
6678 char_u *d;
6679 int dlen = len;
6680 int unconvertlen = 0;
6682 d = string_convert_ext(&input_conv, ptr, &dlen,
6683 restp == NULL ? NULL : &unconvertlen);
6684 if (d != NULL)
6686 if (dlen <= maxlen)
6688 if (unconvertlen > 0)
6690 /* Move the unconverted characters to allocated memory. */
6691 *restp = alloc(unconvertlen);
6692 if (*restp != NULL)
6693 mch_memmove(*restp, ptr + len - unconvertlen, unconvertlen);
6694 *restlenp = unconvertlen;
6696 mch_memmove(ptr, d, dlen);
6698 else
6699 /* result is too long, keep the unconverted text (the caller must
6700 * have done something wrong!) */
6701 dlen = len;
6702 vim_free(d);
6704 return dlen;
6708 * Convert text "ptr[*lenp]" according to "vcp".
6709 * Returns the result in allocated memory and sets "*lenp".
6710 * When "lenp" is NULL, use NUL terminated strings.
6711 * Illegal chars are often changed to "?", unless vcp->vc_fail is set.
6712 * When something goes wrong, NULL is returned and "*lenp" is unchanged.
6714 char_u *
6715 string_convert(vcp, ptr, lenp)
6716 vimconv_T *vcp;
6717 char_u *ptr;
6718 int *lenp;
6720 return string_convert_ext(vcp, ptr, lenp, NULL);
6724 * Like string_convert(), but when "unconvlenp" is not NULL and there are is
6725 * an incomplete sequence at the end it is not converted and "*unconvlenp" is
6726 * set to the number of remaining bytes.
6728 char_u *
6729 string_convert_ext(vcp, ptr, lenp, unconvlenp)
6730 vimconv_T *vcp;
6731 char_u *ptr;
6732 int *lenp;
6733 int *unconvlenp;
6735 char_u *retval = NULL;
6736 char_u *d;
6737 int len;
6738 int i;
6739 int l;
6740 int c;
6742 if (lenp == NULL)
6743 len = (int)STRLEN(ptr);
6744 else
6745 len = *lenp;
6746 if (len == 0)
6747 return vim_strsave((char_u *)"");
6749 switch (vcp->vc_type)
6751 case CONV_TO_UTF8: /* latin1 to utf-8 conversion */
6752 retval = alloc(len * 2 + 1);
6753 if (retval == NULL)
6754 break;
6755 d = retval;
6756 for (i = 0; i < len; ++i)
6758 c = ptr[i];
6759 if (c < 0x80)
6760 *d++ = c;
6761 else
6763 *d++ = 0xc0 + ((unsigned)c >> 6);
6764 *d++ = 0x80 + (c & 0x3f);
6767 *d = NUL;
6768 if (lenp != NULL)
6769 *lenp = (int)(d - retval);
6770 break;
6772 case CONV_9_TO_UTF8: /* latin9 to utf-8 conversion */
6773 retval = alloc(len * 3 + 1);
6774 if (retval == NULL)
6775 break;
6776 d = retval;
6777 for (i = 0; i < len; ++i)
6779 c = ptr[i];
6780 switch (c)
6782 case 0xa4: c = 0x20ac; break; /* euro */
6783 case 0xa6: c = 0x0160; break; /* S hat */
6784 case 0xa8: c = 0x0161; break; /* S -hat */
6785 case 0xb4: c = 0x017d; break; /* Z hat */
6786 case 0xb8: c = 0x017e; break; /* Z -hat */
6787 case 0xbc: c = 0x0152; break; /* OE */
6788 case 0xbd: c = 0x0153; break; /* oe */
6789 case 0xbe: c = 0x0178; break; /* Y */
6791 d += utf_char2bytes(c, d);
6793 *d = NUL;
6794 if (lenp != NULL)
6795 *lenp = (int)(d - retval);
6796 break;
6798 case CONV_TO_LATIN1: /* utf-8 to latin1 conversion */
6799 case CONV_TO_LATIN9: /* utf-8 to latin9 conversion */
6800 retval = alloc(len + 1);
6801 if (retval == NULL)
6802 break;
6803 d = retval;
6804 for (i = 0; i < len; ++i)
6806 l = utf_ptr2len_len(ptr + i, len - i);
6807 if (l == 0)
6808 *d++ = NUL;
6809 else if (l == 1)
6811 int l_w = utf8len_tab_zero[ptr[i]];
6813 if (l_w == 0)
6815 /* Illegal utf-8 byte cannot be converted */
6816 vim_free(retval);
6817 return NULL;
6819 if (unconvlenp != NULL && l_w > len - i)
6821 /* Incomplete sequence at the end. */
6822 *unconvlenp = len - i;
6823 break;
6825 *d++ = ptr[i];
6827 else
6829 c = utf_ptr2char(ptr + i);
6830 if (vcp->vc_type == CONV_TO_LATIN9)
6831 switch (c)
6833 case 0x20ac: c = 0xa4; break; /* euro */
6834 case 0x0160: c = 0xa6; break; /* S hat */
6835 case 0x0161: c = 0xa8; break; /* S -hat */
6836 case 0x017d: c = 0xb4; break; /* Z hat */
6837 case 0x017e: c = 0xb8; break; /* Z -hat */
6838 case 0x0152: c = 0xbc; break; /* OE */
6839 case 0x0153: c = 0xbd; break; /* oe */
6840 case 0x0178: c = 0xbe; break; /* Y */
6841 case 0xa4:
6842 case 0xa6:
6843 case 0xa8:
6844 case 0xb4:
6845 case 0xb8:
6846 case 0xbc:
6847 case 0xbd:
6848 case 0xbe: c = 0x100; break; /* not in latin9 */
6850 if (!utf_iscomposing(c)) /* skip composing chars */
6852 if (c < 0x100)
6853 *d++ = c;
6854 else if (vcp->vc_fail)
6856 vim_free(retval);
6857 return NULL;
6859 else
6861 *d++ = 0xbf;
6862 if (utf_char2cells(c) > 1)
6863 *d++ = '?';
6866 i += l - 1;
6869 *d = NUL;
6870 if (lenp != NULL)
6871 *lenp = (int)(d - retval);
6872 break;
6874 # ifdef MACOS_CONVERT
6875 case CONV_MAC_LATIN1:
6876 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6877 'm', 'l', unconvlenp);
6878 break;
6880 case CONV_LATIN1_MAC:
6881 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6882 'l', 'm', unconvlenp);
6883 break;
6885 case CONV_MAC_UTF8:
6886 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6887 'm', 'u', unconvlenp);
6888 break;
6890 case CONV_UTF8_MAC:
6891 retval = mac_string_convert(ptr, len, lenp, vcp->vc_fail,
6892 'u', 'm', unconvlenp);
6893 break;
6894 # endif
6896 # ifdef USE_ICONV
6897 case CONV_ICONV: /* conversion with output_conv.vc_fd */
6898 retval = iconv_string(vcp, ptr, len, unconvlenp, lenp);
6899 break;
6900 # endif
6901 # ifdef WIN3264
6902 case CONV_CODEPAGE: /* codepage -> codepage */
6904 int retlen;
6905 int tmp_len;
6906 short_u *tmp;
6908 /* 1. codepage/UTF-8 -> ucs-2. */
6909 if (vcp->vc_cpfrom == 0)
6910 tmp_len = utf8_to_utf16(ptr, len, NULL, NULL);
6911 else
6912 tmp_len = MultiByteToWideChar(vcp->vc_cpfrom, 0,
6913 ptr, len, 0, 0);
6914 tmp = (short_u *)alloc(sizeof(short_u) * tmp_len);
6915 if (tmp == NULL)
6916 break;
6917 if (vcp->vc_cpfrom == 0)
6918 utf8_to_utf16(ptr, len, tmp, unconvlenp);
6919 else
6920 MultiByteToWideChar(vcp->vc_cpfrom, 0, ptr, len, tmp, tmp_len);
6922 /* 2. ucs-2 -> codepage/UTF-8. */
6923 if (vcp->vc_cpto == 0)
6924 retlen = utf16_to_utf8(tmp, tmp_len, NULL);
6925 else
6926 retlen = WideCharToMultiByte(vcp->vc_cpto, 0,
6927 tmp, tmp_len, 0, 0, 0, 0);
6928 retval = alloc(retlen + 1);
6929 if (retval != NULL)
6931 if (vcp->vc_cpto == 0)
6932 utf16_to_utf8(tmp, tmp_len, retval);
6933 else
6934 WideCharToMultiByte(vcp->vc_cpto, 0,
6935 tmp, tmp_len, retval, retlen, 0, 0);
6936 retval[retlen] = NUL;
6937 if (lenp != NULL)
6938 *lenp = retlen;
6940 vim_free(tmp);
6941 break;
6943 # endif
6946 return retval;
6948 #endif