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.
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:
54 * keyboard ----->| core |-----> display
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
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"
83 # ifndef WIN32_LEAN_AND_MEAN
84 # define WIN32_LEAN_AND_MEAN
88 # undef WIN32 /* Some windows.h define WIN32, we don't want that here. */
92 #if (defined(WIN3264) || defined(WIN32UNIX)) && !defined(__MINGW32__)
97 # include <X11/Intrinsic.h>
100 #include <X11/Xlocale.h>
103 #if defined(FEAT_GUI_GTK) && defined(FEAT_XIM) && defined(HAVE_GTK2)
104 # include <gdk/gdkkeysyms.h>
106 # include <gdk/gdkwin32.h>
108 # include <gdk/gdkx.h>
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
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 */
176 xim_log(char *s
, ...)
179 static FILE *fd
= NULL
;
181 if (fd
== (FILE *)-1)
185 fd
= mch_fopen("xim.log", "w");
188 EMSG("Cannot open xim.log");
194 va_start(arglist
, s
);
195 vfprintf(fd
, s
, arglist
);
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.
208 { char *name
; int prop
; int codepage
;}
211 #define IDX_LATIN_1 0
212 {"latin1", ENC_8BIT
+ ENC_LATIN1
, 1252},
214 {"iso-8859-2", ENC_8BIT
, 0},
216 {"iso-8859-3", ENC_8BIT
, 0},
218 {"iso-8859-4", ENC_8BIT
, 0},
220 {"iso-8859-5", ENC_8BIT
, 0},
222 {"iso-8859-6", ENC_8BIT
, 0},
224 {"iso-8859-7", ENC_8BIT
, 0},
226 {"iso-8859-8", ENC_8BIT
, 0},
228 {"iso-8859-9", ENC_8BIT
, 0},
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},
244 {"utf-8", ENC_UNICODE
, 0},
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},
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},
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. */
260 {"debug", ENC_DBCS
, DBCS_DEBUG
},
261 #define IDX_EUC_JP 24
262 {"euc-jp", ENC_DBCS
, DBCS_JPNU
},
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
},
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. */
278 {"cp437", ENC_8BIT
, 437}, /* like iso-8859-1 */
280 {"cp737", ENC_8BIT
, 737}, /* like iso-8859-7 */
282 {"cp775", ENC_8BIT
, 775}, /* Baltic */
284 {"cp850", ENC_8BIT
, 850}, /* like iso-8859-4 */
286 {"cp852", ENC_8BIT
, 852}, /* like iso-8859-1 */
288 {"cp855", ENC_8BIT
, 855}, /* like iso-8859-2 */
290 {"cp857", ENC_8BIT
, 857}, /* like iso-8859-5 */
292 {"cp860", ENC_8BIT
, 860}, /* like iso-8859-9 */
294 {"cp861", ENC_8BIT
, 861}, /* like iso-8859-1 */
296 {"cp862", ENC_8BIT
, 862}, /* like iso-8859-1 */
298 {"cp863", ENC_8BIT
, 863}, /* like iso-8859-8 */
300 {"cp865", ENC_8BIT
, 865}, /* like iso-8859-1 */
302 {"cp866", ENC_8BIT
, 866}, /* like iso-8859-5 */
304 {"cp869", ENC_8BIT
, 869}, /* like iso-8859-7 */
306 {"cp874", ENC_8BIT
, 874}, /* Thai */
308 {"cp932", ENC_DBCS
, DBCS_JPN
},
310 {"cp936", ENC_DBCS
, DBCS_CHS
},
312 {"cp949", ENC_DBCS
, DBCS_KOR
},
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 */
343 * Aliases for encoding names.
346 { char *name
; int canon
;}
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
},
358 {"hebrew", IDX_CP1255
},
360 {"hebrew", IDX_ISO_8
},
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
},
371 {"unicode", 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
},
381 {"ucs4be", IDX_UCS4
},
382 {"ucs-4be", IDX_UCS4
},
383 {"ucs4le", IDX_UCS4LE
},
385 {"utf-32", IDX_UCS4
},
386 {"utf32be", IDX_UCS4
},
387 {"utf-32be", IDX_UCS4
},
388 {"utf32le", IDX_UCS4LE
},
389 {"utf-32le", IDX_UCS4LE
},
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
},
408 {"chinese", IDX_CP936
},
409 {"taiwan", IDX_CP950
},
412 {"japan", IDX_EUC_JP
},
413 {"korea", IDX_EUC_KR
},
415 {"chinese", IDX_EUC_CN
},
416 {"taiwan", IDX_EUC_TW
},
420 {"mac", IDX_MACROMAN
},
421 {"mac-roman", IDX_MACROMAN
},
426 # define CP_UTF8 65001 /* magic number from winnls.h */
430 * Find encoding "name" in the list of canonical encoding names.
431 * Returns -1 if not found.
434 enc_canon_search(name
)
439 for (i
= 0; i
< IDX_COUNT
; ++i
)
440 if (STRCMP(name
, enc_canon_table
[i
].name
) == 0)
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
)
459 i
= enc_canon_search(name
);
461 return enc_canon_table
[i
].prop
;
463 if (name
[0] == 'c' && name
[1] == 'p' && VIM_ISDIGIT(name
[2]))
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 */
472 if (cpinfo
.MaxCharSize
== 2
473 && (cpinfo
.LeadByte
[0] != 0 || cpinfo
.LeadByte
[1] != 0))
474 /* must be a DBCS encoding */
480 if (STRNCMP(name
, "2byte-", 6) == 0)
482 if (STRNCMP(name
, "8bit-", 5) == 0 || STRNCMP(name
, "iso-8859-", 9) == 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
505 int enc_dbcs_new
= 0;
506 #if defined(USE_ICONV) && !defined(WIN3264) && !defined(WIN32UNIX) \
508 # define LEN_FROM_CONV
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
;
525 if (p_enc
[0] == 'c' && p_enc
[1] == 'p' && VIM_ISDIGIT(p_enc
[2]))
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 */
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);
545 goto codepage_invalid
;
547 else if (GetLastError() == ERROR_INVALID_PARAMETER
)
550 return (char_u
*)N_("E543: Not a valid codepage");
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. */
561 else if (STRNCMP(p_enc
, "2byte-", 6) == 0)
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)
569 /* Unix: accept any "2byte-" name, assume current locale. */
570 enc_dbcs_new
= DBCS_2BYTE
;
573 else if ((idx
= enc_canon_search(p_enc
)) >= 0)
575 i
= enc_canon_table
[idx
].prop
;
580 if (i
& (ENC_2BYTE
| ENC_2WORD
))
582 else if (i
& ENC_4BYTE
)
587 else if (i
& ENC_DBCS
)
589 /* 2byte, handle below */
590 enc_dbcs_new
= enc_canon_table
[idx
].codepage
;
599 else /* Don't know what encoding this is, reject it. */
602 if (enc_dbcs_new
!= 0)
605 /* Check if the DBCS code page is OK. */
606 if (!IsValidCodePage(enc_dbcs_new
))
607 goto codepage_invalid
;
612 enc_dbcs
= enc_dbcs_new
;
613 has_mbyte
= (enc_dbcs
!= 0 || enc_utf8
);
616 enc_codepage
= encname2codepage(p_enc
);
617 enc_latin9
= (STRCMP(p_enc
, "iso-8859-15") == 0);
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.
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
;
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().
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
;
677 if (p
== NULL
|| STRCMP(p
, p_enc
) != 0)
679 convert_setup(&vimconv
, p_enc
, (char_u
*)"utf-8");
680 vimconv
.vc_fail
= TRUE
;
686 for (i
= 0; i
< 256; ++i
)
688 /* Our own function to reliably check the length of UTF-8 characters,
689 * independent of mblen(). */
692 else if (enc_dbcs
== 0)
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
700 n
= IsDBCSLeadByteEx(enc_dbcs
, (BYTE
)i
) ? 2 : 1;
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;
710 char buf
[MB_MAXBYTES
];
713 # define mblen _Xmblen
716 if (i
== NUL
) /* just in case mblen() can't handle "" */
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
);
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)
758 mb_bytelen_tab
[i
] = n
;
762 convert_setup(&vimconv
, NULL
, NULL
);
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[] */
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
);
784 /* When changing 'encoding' while starting up, then convert the command
785 * line arguments from the active codepage to 'encoding'. */
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
);
797 /* Need to reload spell dictionaries */
805 * Return the size of the BOM for the current buffer:
807 * 2 - UCS-2 or UTF-16 BOM
816 if (curbuf
->b_p_bomb
&& !curbuf
->b_p_bin
)
818 if (*curbuf
->b_p_fenc
== NUL
)
822 if (enc_unicode
!= 0)
828 else if (STRCMP(curbuf
->b_p_fenc
, "utf-8") == 0)
830 else if (STRNCMP(curbuf
->b_p_fenc
, "ucs-2", 5) == 0
831 || STRNCMP(curbuf
->b_p_fenc
, "utf-16", 6) == 0)
833 else if (STRNCMP(curbuf
->b_p_fenc
, "ucs-4", 5) == 0)
840 * Get class of pointer:
843 * 2 for an (ASCII) word character
844 * >2 for other word characters
850 if (MB_BYTE2LEN(p
[0]) == 1)
852 if (p
[0] == NUL
|| vim_iswhite(p
[0]))
854 if (vim_iswordc(p
[0]))
858 if (enc_dbcs
!= 0 && p
[0] != NUL
&& p
[1] != NUL
)
859 return dbcs_class(p
[0], p
[1]);
861 return utf_class(utf_ptr2char(p
));
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
)
876 /* please add classfy routine for your language in here */
878 case DBCS_JPNU
: /* ? */
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 */
889 lb
= (lb
- 0x81) * 2 + 0x21;
891 lb
= (lb
- 0xc1) * 2 + 0x21;
903 * XXX: Code page identification can not use with all
904 * system! So, some other encoding information
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 */
916 switch (lb
<< 8 | tb
)
918 case 0x2121: /* ZENKAKU space */
920 case 0x2122: /* KU-TEN (Japanese comma) */
921 case 0x2123: /* TOU-TEN (Japanese period) */
922 case 0x2124: /* ZENKAKU comma */
923 case 0x2125: /* ZENKAKU period */
925 case 0x213c: /* prolongedsound handled as KATAKANA */
928 /* sieved by KU code */
933 /* special symbols */
959 case DBCS_KORU
: /* ? */
962 /* KS code classification */
963 unsigned char c1
= lead
;
964 unsigned char c2
= trail
;
970 * 23 : Alpha-numeric/Roman Letter (Full width)
971 * 24 : Hangul Letter(Alphabet)
972 * 25 : Roman Numeral/Greek Letter
975 * 28 : Circled/Parenthesized Letter
976 * 29 : Hirigana/Katakana
977 * 30 : Cyrillic Letter
980 if (c1
>= 0xB0 && c1
<= 0xC8)
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
992 else if (c1
>= 0xCA && c1
<= 0xFD)
1005 /* Hangul Letter(Alphabet) */
1008 /* Roman Numeral/Greek Letter */
1019 return 25; /* Roman Letter */
1020 else if (c2
>= 0xF6)
1021 return 22; /* Symbols */
1023 /* Circled/Parenthesized Letter */
1027 /* Hirigana/Katakana */
1030 /* Cyrillic Letter */
1041 * mb_char2len() function pointer.
1042 * Return length in bytes of character "c".
1043 * Returns 1 for a single-byte character.
1062 * mb_char2bytes() function pointer.
1063 * Convert a character to its bytes.
1064 * Returns the length in bytes.
1067 latin_char2bytes(c
, buf
)
1076 dbcs_char2bytes(c
, buf
)
1082 buf
[0] = (unsigned)c
>> 8;
1084 /* Never use a NUL byte, it causes lots of trouble. It's an invalid
1085 * character anyway. */
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.
1104 return MB_BYTE2LEN(*p
);
1113 /* Check if second byte is not missing. */
1114 len
= MB_BYTE2LEN(*p
);
1115 if (len
== 2 && p
[1] == NUL
)
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
)
1131 if (size
< 1 || *p
== NUL
)
1137 dbcs_ptr2len_len(p
, size
)
1143 if (size
< 1 || *p
== NUL
)
1147 /* Check that second byte is not missing. */
1148 len
= MB_BYTE2LEN(*p
);
1149 if (len
== 2 && p
[1] == NUL
)
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)]".
1165 intable(table
, size
, c
)
1166 struct interval
*table
;
1172 /* first quick check for Latin1 etc. characters */
1173 if (c
< table
[0].first
)
1176 /* binary search in table */
1178 top
= (int)(size
/ sizeof(struct interval
) - 1);
1181 mid
= (bot
+ top
) / 2;
1182 if (table
[mid
].last
< c
)
1184 else if (table
[mid
].first
> c
)
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).
1203 /* Sorted list of non-overlapping intervals of East Asian Ambiguous
1204 * characters, generated with ../runtime/tools/unicode.vim. */
1205 static struct interval ambiguous
[] =
1387 {0x100000, 0x10fffd}
1392 #ifdef USE_WCHAR_FUNCTIONS
1394 * Assume the library function wcwidth() works better than our own
1395 * stuff. It should return 1 for ambiguous width chars!
1400 return 6; /* unprintable, displays <xxxx> */
1404 if (!utf_printable(c
))
1405 return 6; /* unprintable, displays <xxxx> */
1407 && (c
<= 0x115f /* Hangul Jamo */
1410 || (c
>= 0x2e80 && c
<= 0xa4cf
1411 && c
!= 0x303f) /* CJK ... Yi */
1412 || (c
>= 0xac00 && c
<= 0xd7a3) /* Hangul Syllables */
1413 || (c
>= 0xf900 && c
<= 0xfaff) /* CJK Compatibility
1415 || (c
>= 0xfe30 && c
<= 0xfe6f) /* CJK Compatibility Forms */
1416 || (c
>= 0xff00 && c
<= 0xff60) /* Fullwidth Forms */
1417 || (c
>= 0xffe0 && c
<= 0xffe6)
1418 || (c
>= 0x20000 && c
<= 0x2fffd)
1419 || (c
>= 0x30000 && c
<= 0x3fffd)))
1424 /* Characters below 0x100 are influenced by 'isprint' option */
1425 else if (c
>= 0x80 && !vim_isprintc(c
))
1426 return 4; /* unprintable, displays <xx> */
1428 if (c
>= 0x80 && *p_ambw
== 'd' && intable(ambiguous
, sizeof(ambiguous
), c
))
1435 * mb_ptr2cells() function pointer.
1436 * Return the number of display cells character at "*p" occupies.
1437 * This doesn't take care of unprintable characters, use ptr2cells() for that.
1452 /* Need to convert to a wide character. */
1455 c
= utf_ptr2char(p
);
1456 /* An illegal byte is displayed as <xx>. */
1457 if (utf_ptr2len(p
) == 1 || c
== NUL
)
1459 /* If the char is ASCII it must be an overlong sequence. */
1461 return char2cells(c
);
1462 return utf_char2cells(c
);
1471 /* Number of cells is equal to number of bytes, except for euc-jp when
1472 * the first byte is 0x8e. */
1473 if (enc_dbcs
== DBCS_JPNU
&& *p
== 0x8e)
1475 return MB_BYTE2LEN(*p
);
1479 * mb_ptr2cells_len() function pointer.
1480 * Like mb_ptr2cells(), but limit string length to "size".
1481 * For an empty string or truncated character returns 1.
1484 latin_ptr2cells_len(p
, size
)
1492 utf_ptr2cells_len(p
, size
)
1498 /* Need to convert to a wide character. */
1499 if (size
> 0 && *p
>= 0x80)
1501 if (utf_ptr2len_len(p
, size
) < utf8len_tab
[*p
])
1502 return 1; /* truncated */
1503 c
= utf_ptr2char(p
);
1504 /* An illegal byte is displayed as <xx>. */
1505 if (utf_ptr2len(p
) == 1 || c
== NUL
)
1507 /* If the char is ASCII it must be an overlong sequence. */
1509 return char2cells(c
);
1510 return utf_char2cells(c
);
1516 dbcs_ptr2cells_len(p
, size
)
1520 /* Number of cells is equal to number of bytes, except for euc-jp when
1521 * the first byte is 0x8e. */
1522 if (size
<= 1 || (enc_dbcs
== DBCS_JPNU
&& *p
== 0x8e))
1524 return MB_BYTE2LEN(*p
);
1528 * mb_char2cells() function pointer.
1529 * Return the number of display cells character "c" occupies.
1530 * Only takes care of multi-byte chars, not "^C" and such.
1543 /* Number of cells is equal to number of bytes, except for euc-jp when
1544 * the first byte is 0x8e. */
1545 if (enc_dbcs
== DBCS_JPNU
&& ((unsigned)c
>> 8) == 0x8e)
1547 /* use the first byte */
1548 return MB_BYTE2LEN((unsigned)c
>> 8);
1552 * mb_off2cells() function pointer.
1553 * Return number of display cells for char at ScreenLines[off].
1554 * We make sure that the offset used is less than "max_off".
1557 latin_off2cells(off
, max_off
)
1558 unsigned off UNUSED
;
1559 unsigned max_off UNUSED
;
1565 dbcs_off2cells(off
, max_off
)
1569 /* never check beyond end of the line */
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
&& ScreenLines
[off
] == 0x8e)
1577 return MB_BYTE2LEN(ScreenLines
[off
]);
1581 utf_off2cells(off
, max_off
)
1585 return (off
+ 1 < max_off
&& ScreenLines
[off
+ 1] == 0) ? 2 : 1;
1589 * mb_ptr2char() function pointer.
1590 * Convert a byte sequence into a character.
1603 if (MB_BYTE2LEN(*p
) > 1 && p
[1] != NUL
)
1604 return (p
[0] << 8) + p
[1];
1609 * Convert a UTF-8 byte sequence to a wide character.
1610 * If the sequence is illegal or truncated by a NUL the first byte is
1612 * Does not include composing characters, of course.
1620 if (p
[0] < 0x80) /* be quick for ASCII */
1623 len
= utf8len_tab_zero
[p
[0]];
1624 if (len
> 1 && (p
[1] & 0xc0) == 0x80)
1627 return ((p
[0] & 0x1f) << 6) + (p
[1] & 0x3f);
1628 if ((p
[2] & 0xc0) == 0x80)
1631 return ((p
[0] & 0x0f) << 12) + ((p
[1] & 0x3f) << 6)
1633 if ((p
[3] & 0xc0) == 0x80)
1636 return ((p
[0] & 0x07) << 18) + ((p
[1] & 0x3f) << 12)
1637 + ((p
[2] & 0x3f) << 6) + (p
[3] & 0x3f);
1638 if ((p
[4] & 0xc0) == 0x80)
1641 return ((p
[0] & 0x03) << 24) + ((p
[1] & 0x3f) << 18)
1642 + ((p
[2] & 0x3f) << 12) + ((p
[3] & 0x3f) << 6)
1644 if ((p
[5] & 0xc0) == 0x80 && len
== 6)
1645 return ((p
[0] & 0x01) << 30) + ((p
[1] & 0x3f) << 24)
1646 + ((p
[2] & 0x3f) << 18) + ((p
[3] & 0x3f) << 12)
1647 + ((p
[4] & 0x3f) << 6) + (p
[5] & 0x3f);
1652 /* Illegal value, just return the first byte */
1657 * Get character at **pp and advance *pp to the next character.
1658 * Note: composing characters are skipped!
1666 c
= (*mb_ptr2char
)(*pp
);
1667 *pp
+= (*mb_ptr2len
)(*pp
);
1672 * Get character at **pp and advance *pp to the next character.
1673 * Note: composing characters are returned as separate characters.
1676 mb_cptr2char_adv(pp
)
1681 c
= (*mb_ptr2char
)(*pp
);
1683 *pp
+= utf_ptr2len(*pp
);
1685 *pp
+= (*mb_ptr2len
)(*pp
);
1689 #if defined(FEAT_ARABIC) || defined(PROTO)
1691 * Check whether we are dealing with Arabic combining characters.
1692 * Note: these are NOT really composing characters!
1695 arabic_combine(one
, two
)
1696 int one
; /* first character */
1697 int two
; /* character just after "one" */
1700 return arabic_maycombine(two
);
1705 * Check whether we are dealing with a character that could be regarded as an
1706 * Arabic combining character, need to check the character before this.
1709 arabic_maycombine(two
)
1712 if (p_arshape
&& !p_tbidi
)
1713 return (two
== a_ALEF_MADDA
1714 || two
== a_ALEF_HAMZA_ABOVE
1715 || two
== a_ALEF_HAMZA_BELOW
1721 * Check if the character pointed to by "p2" is a composing character when it
1722 * comes after "p1". For Arabic sometimes "ab" is replaced with "c", which
1723 * behaves like a composing character.
1726 utf_composinglike(p1
, p2
)
1732 c2
= utf_ptr2char(p2
);
1733 if (utf_iscomposing(c2
))
1735 if (!arabic_maycombine(c2
))
1737 return arabic_combine(utf_ptr2char(p1
), c2
);
1742 * Convert a UTF-8 byte string to a wide character. Also get up to MAX_MCO
1743 * composing characters.
1746 utfc_ptr2char(p
, pcc
)
1748 int *pcc
; /* return: composing chars, last one is 0 */
1755 c
= utf_ptr2char(p
);
1756 len
= utf_ptr2len(p
);
1758 /* Only accept a composing char when the first char isn't illegal. */
1759 if ((len
> 1 || *p
< 0x80)
1761 && UTF_COMPOSINGLIKE(p
, p
+ len
))
1763 cc
= utf_ptr2char(p
+ len
);
1769 len
+= utf_ptr2len(p
+ len
);
1770 if (p
[len
] < 0x80 || !utf_iscomposing(cc
= utf_ptr2char(p
+ len
)))
1775 if (i
< MAX_MCO
) /* last composing char must be 0 */
1782 * Convert a UTF-8 byte string to a wide character. Also get up to MAX_MCO
1783 * composing characters. Use no more than p[maxlen].
1786 utfc_ptr2char_len(p
, pcc
, maxlen
)
1788 int *pcc
; /* return: composing chars, last one is 0 */
1796 c
= utf_ptr2char(p
);
1797 len
= utf_ptr2len_len(p
, maxlen
);
1798 /* Only accept a composing char when the first char isn't illegal. */
1799 if ((len
> 1 || *p
< 0x80)
1802 && UTF_COMPOSINGLIKE(p
, p
+ len
))
1804 cc
= utf_ptr2char(p
+ len
);
1810 len
+= utf_ptr2len_len(p
+ len
, maxlen
- len
);
1813 || !utf_iscomposing(cc
= utf_ptr2char(p
+ len
)))
1818 if (i
< MAX_MCO
) /* last composing char must be 0 */
1825 * Convert the character at screen position "off" to a sequence of bytes.
1826 * Includes the composing characters.
1827 * "buf" must at least have the length MB_MAXBYTES.
1828 * Returns the produced number of bytes.
1831 utfc_char2bytes(off
, buf
)
1838 len
= utf_char2bytes(ScreenLinesUC
[off
], buf
);
1839 for (i
= 0; i
< Screen_mco
; ++i
)
1841 if (ScreenLinesC
[i
][off
] == 0)
1843 len
+= utf_char2bytes(ScreenLinesC
[i
][off
], buf
+ len
);
1849 * Get the length of a UTF-8 byte sequence, not including any following
1850 * composing characters.
1852 * Returns 1 for an illegal byte sequence.
1863 len
= utf8len_tab
[*p
];
1864 for (i
= 1; i
< len
; ++i
)
1865 if ((p
[i
] & 0xc0) != 0x80)
1871 * Return length of UTF-8 character, obtained from the first byte.
1872 * "b" must be between 0 and 255!
1873 * Returns 1 for an invalid first byte value.
1879 return utf8len_tab
[b
];
1883 * Get the length of UTF-8 byte sequence "p[size]". Does not include any
1884 * following composing characters.
1886 * Returns 1 for an illegal byte sequence (also in incomplete byte seq.).
1887 * Returns number > "size" for an incomplete byte sequence.
1888 * Never returns zero.
1891 utf_ptr2len_len(p
, size
)
1899 len
= utf8len_tab
[*p
];
1901 return 1; /* NUL, ascii or illegal lead byte */
1903 m
= size
; /* incomplete byte sequence. */
1906 for (i
= 1; i
< m
; ++i
)
1907 if ((p
[i
] & 0xc0) != 0x80)
1913 * Return the number of bytes the UTF-8 encoding of the character at "p" takes.
1914 * This includes following composing characters.
1928 if (b0
< 0x80 && p
[1] < 0x80) /* be quick for ASCII */
1931 /* Skip over first UTF-8 char, stopping at a NUL byte. */
1932 len
= utf_ptr2len(p
);
1934 /* Check for illegal byte. */
1935 if (len
== 1 && b0
>= 0x80)
1939 * Check for composing characters. We can handle only the first six, but
1940 * skip all of them (otherwise the cursor would get stuck).
1947 if (p
[len
] < 0x80 || !UTF_COMPOSINGLIKE(p
+ prevlen
, p
+ len
))
1950 /* Skip over composing char */
1954 len
+= utf_ptr2len(p
+ len
);
1959 * Return the number of bytes the UTF-8 encoding of the character at "p[size]"
1960 * takes. This includes following composing characters.
1961 * Returns 0 for an empty string.
1962 * Returns 1 for an illegal char or an incomplete byte sequence.
1965 utfc_ptr2len_len(p
, size
)
1974 if (size
< 1 || *p
== NUL
)
1976 if (p
[0] < 0x80 && (size
== 1 || p
[1] < 0x80)) /* be quick for ASCII */
1979 /* Skip over first UTF-8 char, stopping at a NUL byte. */
1980 len
= utf_ptr2len_len(p
, size
);
1982 /* Check for illegal byte and incomplete byte sequence. */
1983 if ((len
== 1 && p
[0] >= 0x80) || len
> size
)
1987 * Check for composing characters. We can handle only the first six, but
1988 * skip all of them (otherwise the cursor would get stuck).
2001 * Next character length should not go beyond size to ensure that
2002 * UTF_COMPOSINGLIKE(...) does not read beyond size.
2004 len_next_char
= utf_ptr2len_len(p
+ len
, size
- len
);
2005 if (len_next_char
> size
- len
)
2008 if (!UTF_COMPOSINGLIKE(p
+ prevlen
, p
+ len
))
2011 /* Skip over composing char */
2015 len
+= len_next_char
;
2021 * Return the number of bytes the UTF-8 encoding of character "c" takes.
2022 * This does not include composing characters.
2042 * Convert Unicode character "c" to UTF-8 string in "buf[]".
2043 * Returns the number of bytes.
2044 * This does not include composing characters.
2047 utf_char2bytes(c
, buf
)
2051 if (c
< 0x80) /* 7 bits */
2056 if (c
< 0x800) /* 11 bits */
2058 buf
[0] = 0xc0 + ((unsigned)c
>> 6);
2059 buf
[1] = 0x80 + (c
& 0x3f);
2062 if (c
< 0x10000) /* 16 bits */
2064 buf
[0] = 0xe0 + ((unsigned)c
>> 12);
2065 buf
[1] = 0x80 + (((unsigned)c
>> 6) & 0x3f);
2066 buf
[2] = 0x80 + (c
& 0x3f);
2069 if (c
< 0x200000) /* 21 bits */
2071 buf
[0] = 0xf0 + ((unsigned)c
>> 18);
2072 buf
[1] = 0x80 + (((unsigned)c
>> 12) & 0x3f);
2073 buf
[2] = 0x80 + (((unsigned)c
>> 6) & 0x3f);
2074 buf
[3] = 0x80 + (c
& 0x3f);
2077 if (c
< 0x4000000) /* 26 bits */
2079 buf
[0] = 0xf8 + ((unsigned)c
>> 24);
2080 buf
[1] = 0x80 + (((unsigned)c
>> 18) & 0x3f);
2081 buf
[2] = 0x80 + (((unsigned)c
>> 12) & 0x3f);
2082 buf
[3] = 0x80 + (((unsigned)c
>> 6) & 0x3f);
2083 buf
[4] = 0x80 + (c
& 0x3f);
2087 buf
[0] = 0xfc + ((unsigned)c
>> 30);
2088 buf
[1] = 0x80 + (((unsigned)c
>> 24) & 0x3f);
2089 buf
[2] = 0x80 + (((unsigned)c
>> 18) & 0x3f);
2090 buf
[3] = 0x80 + (((unsigned)c
>> 12) & 0x3f);
2091 buf
[4] = 0x80 + (((unsigned)c
>> 6) & 0x3f);
2092 buf
[5] = 0x80 + (c
& 0x3f);
2097 * Return TRUE if "c" is a composing UTF-8 character. This means it will be
2098 * drawn on top of the preceding character.
2099 * Based on code from Markus Kuhn.
2105 /* Sorted list of non-overlapping intervals.
2106 * Generated by ../runtime/tools/unicode.vim. */
2107 static struct interval combining
[] =
2299 return intable(combining
, sizeof(combining
), c
);
2303 * Return TRUE for characters that can be displayed in a normal way.
2304 * Only for characters of 0x100 and above!
2310 #ifdef USE_WCHAR_FUNCTIONS
2312 * Assume the iswprint() library function works better than our own stuff.
2316 /* Sorted list of non-overlapping intervals.
2317 * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */
2318 static struct interval nonprint
[] =
2320 {0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e},
2321 {0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb},
2325 return !intable(nonprint
, sizeof(nonprint
), c
);
2330 * Get class of a Unicode character.
2333 * 2 or bigger: some class of word character.
2339 /* sorted list of non-overlapping intervals */
2340 static struct clinterval
2342 unsigned short first
;
2343 unsigned short last
;
2344 unsigned short class;
2347 {0x037e, 0x037e, 1}, /* Greek question mark */
2348 {0x0387, 0x0387, 1}, /* Greek ano teleia */
2349 {0x055a, 0x055f, 1}, /* Armenian punctuation */
2350 {0x0589, 0x0589, 1}, /* Armenian full stop */
2351 {0x05be, 0x05be, 1},
2352 {0x05c0, 0x05c0, 1},
2353 {0x05c3, 0x05c3, 1},
2354 {0x05f3, 0x05f4, 1},
2355 {0x060c, 0x060c, 1},
2356 {0x061b, 0x061b, 1},
2357 {0x061f, 0x061f, 1},
2358 {0x066a, 0x066d, 1},
2359 {0x06d4, 0x06d4, 1},
2360 {0x0700, 0x070d, 1}, /* Syriac punctuation */
2361 {0x0964, 0x0965, 1},
2362 {0x0970, 0x0970, 1},
2363 {0x0df4, 0x0df4, 1},
2364 {0x0e4f, 0x0e4f, 1},
2365 {0x0e5a, 0x0e5b, 1},
2366 {0x0f04, 0x0f12, 1},
2367 {0x0f3a, 0x0f3d, 1},
2368 {0x0f85, 0x0f85, 1},
2369 {0x104a, 0x104f, 1}, /* Myanmar punctuation */
2370 {0x10fb, 0x10fb, 1}, /* Georgian punctuation */
2371 {0x1361, 0x1368, 1}, /* Ethiopic punctuation */
2372 {0x166d, 0x166e, 1}, /* Canadian Syl. punctuation */
2373 {0x1680, 0x1680, 0},
2374 {0x169b, 0x169c, 1},
2375 {0x16eb, 0x16ed, 1},
2376 {0x1735, 0x1736, 1},
2377 {0x17d4, 0x17dc, 1}, /* Khmer punctuation */
2378 {0x1800, 0x180a, 1}, /* Mongolian punctuation */
2379 {0x2000, 0x200b, 0}, /* spaces */
2380 {0x200c, 0x2027, 1}, /* punctuation and symbols */
2381 {0x2028, 0x2029, 0},
2382 {0x202a, 0x202e, 1}, /* punctuation and symbols */
2383 {0x202f, 0x202f, 0},
2384 {0x2030, 0x205e, 1}, /* punctuation and symbols */
2385 {0x205f, 0x205f, 0},
2386 {0x2060, 0x27ff, 1}, /* punctuation and symbols */
2387 {0x2070, 0x207f, 0x2070}, /* superscript */
2388 {0x2080, 0x2094, 0x2080}, /* subscript */
2389 {0x20a0, 0x27ff, 1}, /* all kinds of symbols */
2390 {0x2800, 0x28ff, 0x2800}, /* braille */
2391 {0x2900, 0x2998, 1}, /* arrows, brackets, etc. */
2392 {0x29d8, 0x29db, 1},
2393 {0x29fc, 0x29fd, 1},
2394 {0x3000, 0x3000, 0}, /* ideographic space */
2395 {0x3001, 0x3020, 1}, /* ideographic punctuation */
2396 {0x3030, 0x3030, 1},
2397 {0x303d, 0x303d, 1},
2398 {0x3040, 0x309f, 0x3040}, /* Hiragana */
2399 {0x30a0, 0x30ff, 0x30a0}, /* Katakana */
2400 {0x3300, 0x9fff, 0x4e00}, /* CJK Ideographs */
2401 {0xac00, 0xd7a3, 0xac00}, /* Hangul Syllables */
2402 {0xf900, 0xfaff, 0x4e00}, /* CJK Ideographs */
2403 {0xfd3e, 0xfd3f, 1},
2404 {0xfe30, 0xfe6b, 1}, /* punctuation forms */
2405 {0xff00, 0xff0f, 1}, /* half/fullwidth ASCII */
2406 {0xff1a, 0xff20, 1}, /* half/fullwidth ASCII */
2407 {0xff3b, 0xff40, 1}, /* half/fullwidth ASCII */
2408 {0xff5b, 0xff65, 1}, /* half/fullwidth ASCII */
2411 int top
= sizeof(classes
) / sizeof(struct clinterval
) - 1;
2414 /* First quick check for Latin1 characters, use 'iskeyword'. */
2417 if (c
== ' ' || c
== '\t' || c
== NUL
|| c
== 0xa0)
2418 return 0; /* blank */
2420 return 2; /* word character */
2421 return 1; /* punctuation */
2424 /* binary search in table */
2427 mid
= (bot
+ top
) / 2;
2428 if (classes
[mid
].last
< c
)
2430 else if (classes
[mid
].first
> c
)
2433 return (int)classes
[mid
].class;
2436 /* most other characters are "word" characters */
2441 * Code for Unicode case-dependent operations. Based on notes in
2442 * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt
2443 * This code uses simple case folding, not full case folding.
2444 * Last updated for Unicode 5.2.
2448 * The following tables are built by ../runtime/tools/unicode.vim.
2449 * They must be in numeric order, because we use binary search.
2450 * An entry such as {0x41,0x5a,1,32} means that Unicode characters in the
2451 * range from 0x41 to 0x5a inclusive, stepping by 1, are changed to
2452 * folded/upper/lower by adding 32.
2462 static convertStruct foldCase
[] =
2472 {0x178,0x178,-1,-121},
2474 {0x17f,0x17f,-1,-268},
2475 {0x181,0x181,-1,210},
2477 {0x186,0x186,-1,206},
2479 {0x189,0x18a,1,205},
2481 {0x18e,0x18e,-1,79},
2482 {0x18f,0x18f,-1,202},
2483 {0x190,0x190,-1,203},
2485 {0x193,0x193,-1,205},
2486 {0x194,0x194,-1,207},
2487 {0x196,0x196,-1,211},
2488 {0x197,0x197,-1,209},
2490 {0x19c,0x19c,-1,211},
2491 {0x19d,0x19d,-1,213},
2492 {0x19f,0x19f,-1,214},
2494 {0x1a6,0x1a6,-1,218},
2496 {0x1a9,0x1a9,-1,218},
2498 {0x1ae,0x1ae,-1,218},
2500 {0x1b1,0x1b2,1,217},
2502 {0x1b7,0x1b7,-1,219},
2513 {0x1f6,0x1f6,-1,-97},
2514 {0x1f7,0x1f7,-1,-56},
2516 {0x220,0x220,-1,-130},
2518 {0x23a,0x23a,-1,10795},
2520 {0x23d,0x23d,-1,-163},
2521 {0x23e,0x23e,-1,10792},
2523 {0x243,0x243,-1,-195},
2524 {0x244,0x244,-1,69},
2525 {0x245,0x245,-1,71},
2527 {0x345,0x345,-1,116},
2530 {0x386,0x386,-1,38},
2532 {0x38c,0x38c,-1,64},
2538 {0x3d0,0x3d0,-1,-30},
2539 {0x3d1,0x3d1,-1,-25},
2540 {0x3d5,0x3d5,-1,-15},
2541 {0x3d6,0x3d6,-1,-22},
2543 {0x3f0,0x3f0,-1,-54},
2544 {0x3f1,0x3f1,-1,-48},
2545 {0x3f4,0x3f4,-1,-60},
2546 {0x3f5,0x3f5,-1,-64},
2548 {0x3f9,0x3f9,-1,-7},
2550 {0x3fd,0x3ff,1,-130},
2555 {0x4c0,0x4c0,-1,15},
2559 {0x10a0,0x10c5,1,7264},
2560 {0x1e00,0x1e94,2,1},
2561 {0x1e9b,0x1e9b,-1,-58},
2562 {0x1e9e,0x1e9e,-1,-7615},
2563 {0x1ea0,0x1efe,2,1},
2564 {0x1f08,0x1f0f,1,-8},
2565 {0x1f18,0x1f1d,1,-8},
2566 {0x1f28,0x1f2f,1,-8},
2567 {0x1f38,0x1f3f,1,-8},
2568 {0x1f48,0x1f4d,1,-8},
2569 {0x1f59,0x1f5f,2,-8},
2570 {0x1f68,0x1f6f,1,-8},
2571 {0x1f88,0x1f8f,1,-8},
2572 {0x1f98,0x1f9f,1,-8},
2573 {0x1fa8,0x1faf,1,-8},
2574 {0x1fb8,0x1fb9,1,-8},
2575 {0x1fba,0x1fbb,1,-74},
2576 {0x1fbc,0x1fbc,-1,-9},
2577 {0x1fbe,0x1fbe,-1,-7173},
2578 {0x1fc8,0x1fcb,1,-86},
2579 {0x1fcc,0x1fcc,-1,-9},
2580 {0x1fd8,0x1fd9,1,-8},
2581 {0x1fda,0x1fdb,1,-100},
2582 {0x1fe8,0x1fe9,1,-8},
2583 {0x1fea,0x1feb,1,-112},
2584 {0x1fec,0x1fec,-1,-7},
2585 {0x1ff8,0x1ff9,1,-128},
2586 {0x1ffa,0x1ffb,1,-126},
2587 {0x1ffc,0x1ffc,-1,-9},
2588 {0x2126,0x2126,-1,-7517},
2589 {0x212a,0x212a,-1,-8383},
2590 {0x212b,0x212b,-1,-8262},
2591 {0x2132,0x2132,-1,28},
2592 {0x2160,0x216f,1,16},
2593 {0x2183,0x2183,-1,1},
2594 {0x24b6,0x24cf,1,26},
2595 {0x2c00,0x2c2e,1,48},
2596 {0x2c60,0x2c60,-1,1},
2597 {0x2c62,0x2c62,-1,-10743},
2598 {0x2c63,0x2c63,-1,-3814},
2599 {0x2c64,0x2c64,-1,-10727},
2600 {0x2c67,0x2c6b,2,1},
2601 {0x2c6d,0x2c6d,-1,-10780},
2602 {0x2c6e,0x2c6e,-1,-10749},
2603 {0x2c6f,0x2c6f,-1,-10783},
2604 {0x2c70,0x2c70,-1,-10782},
2605 {0x2c72,0x2c75,3,1},
2606 {0x2c7e,0x2c7f,1,-10815},
2607 {0x2c80,0x2ce2,2,1},
2608 {0x2ceb,0x2ced,2,1},
2609 {0xa640,0xa65e,2,1},
2610 {0xa662,0xa66c,2,1},
2611 {0xa680,0xa696,2,1},
2612 {0xa722,0xa72e,2,1},
2613 {0xa732,0xa76e,2,1},
2614 {0xa779,0xa77b,2,1},
2615 {0xa77d,0xa77d,-1,-35332},
2616 {0xa77e,0xa786,2,1},
2617 {0xa78b,0xa78b,-1,1},
2618 {0xff21,0xff3a,1,32},
2619 {0x10400,0x10427,1,40}
2622 static int utf_convert(int a
, convertStruct table
[], int tableSize
);
2625 * Generic conversion function for case operations.
2626 * Return the converted equivalent of "a", which is a UCS-4 character. Use
2627 * the given conversion "table". Uses binary search on "table".
2630 utf_convert(a
, table
, tableSize
)
2632 convertStruct table
[];
2635 int start
, mid
, end
; /* indices into table */
2638 end
= tableSize
/ sizeof(convertStruct
);
2641 /* need to search further */
2642 mid
= (end
+ start
) /2;
2643 if (table
[mid
].rangeEnd
< a
)
2648 if (table
[start
].rangeStart
<= a
&& a
<= table
[start
].rangeEnd
2649 && (a
- table
[start
].rangeStart
) % table
[start
].step
== 0)
2650 return (a
+ table
[start
].offset
);
2656 * Return the folded-case equivalent of "a", which is a UCS-4 character. Uses
2657 * simple case folding.
2663 return utf_convert(a
, foldCase
, sizeof(foldCase
));
2666 static convertStruct toLower
[] =
2672 {0x130,0x130,-1,-199},
2676 {0x178,0x178,-1,-121},
2678 {0x181,0x181,-1,210},
2680 {0x186,0x186,-1,206},
2682 {0x189,0x18a,1,205},
2684 {0x18e,0x18e,-1,79},
2685 {0x18f,0x18f,-1,202},
2686 {0x190,0x190,-1,203},
2688 {0x193,0x193,-1,205},
2689 {0x194,0x194,-1,207},
2690 {0x196,0x196,-1,211},
2691 {0x197,0x197,-1,209},
2693 {0x19c,0x19c,-1,211},
2694 {0x19d,0x19d,-1,213},
2695 {0x19f,0x19f,-1,214},
2697 {0x1a6,0x1a6,-1,218},
2699 {0x1a9,0x1a9,-1,218},
2701 {0x1ae,0x1ae,-1,218},
2703 {0x1b1,0x1b2,1,217},
2705 {0x1b7,0x1b7,-1,219},
2716 {0x1f6,0x1f6,-1,-97},
2717 {0x1f7,0x1f7,-1,-56},
2719 {0x220,0x220,-1,-130},
2721 {0x23a,0x23a,-1,10795},
2723 {0x23d,0x23d,-1,-163},
2724 {0x23e,0x23e,-1,10792},
2726 {0x243,0x243,-1,-195},
2727 {0x244,0x244,-1,69},
2728 {0x245,0x245,-1,71},
2732 {0x386,0x386,-1,38},
2734 {0x38c,0x38c,-1,64},
2740 {0x3f4,0x3f4,-1,-60},
2742 {0x3f9,0x3f9,-1,-7},
2744 {0x3fd,0x3ff,1,-130},
2749 {0x4c0,0x4c0,-1,15},
2753 {0x10a0,0x10c5,1,7264},
2754 {0x1e00,0x1e94,2,1},
2755 {0x1e9e,0x1e9e,-1,-7615},
2756 {0x1ea0,0x1efe,2,1},
2757 {0x1f08,0x1f0f,1,-8},
2758 {0x1f18,0x1f1d,1,-8},
2759 {0x1f28,0x1f2f,1,-8},
2760 {0x1f38,0x1f3f,1,-8},
2761 {0x1f48,0x1f4d,1,-8},
2762 {0x1f59,0x1f5f,2,-8},
2763 {0x1f68,0x1f6f,1,-8},
2764 {0x1f88,0x1f8f,1,-8},
2765 {0x1f98,0x1f9f,1,-8},
2766 {0x1fa8,0x1faf,1,-8},
2767 {0x1fb8,0x1fb9,1,-8},
2768 {0x1fba,0x1fbb,1,-74},
2769 {0x1fbc,0x1fbc,-1,-9},
2770 {0x1fc8,0x1fcb,1,-86},
2771 {0x1fcc,0x1fcc,-1,-9},
2772 {0x1fd8,0x1fd9,1,-8},
2773 {0x1fda,0x1fdb,1,-100},
2774 {0x1fe8,0x1fe9,1,-8},
2775 {0x1fea,0x1feb,1,-112},
2776 {0x1fec,0x1fec,-1,-7},
2777 {0x1ff8,0x1ff9,1,-128},
2778 {0x1ffa,0x1ffb,1,-126},
2779 {0x1ffc,0x1ffc,-1,-9},
2780 {0x2126,0x2126,-1,-7517},
2781 {0x212a,0x212a,-1,-8383},
2782 {0x212b,0x212b,-1,-8262},
2783 {0x2132,0x2132,-1,28},
2784 {0x2160,0x216f,1,16},
2785 {0x2183,0x2183,-1,1},
2786 {0x24b6,0x24cf,1,26},
2787 {0x2c00,0x2c2e,1,48},
2788 {0x2c60,0x2c60,-1,1},
2789 {0x2c62,0x2c62,-1,-10743},
2790 {0x2c63,0x2c63,-1,-3814},
2791 {0x2c64,0x2c64,-1,-10727},
2792 {0x2c67,0x2c6b,2,1},
2793 {0x2c6d,0x2c6d,-1,-10780},
2794 {0x2c6e,0x2c6e,-1,-10749},
2795 {0x2c6f,0x2c6f,-1,-10783},
2796 {0x2c70,0x2c70,-1,-10782},
2797 {0x2c72,0x2c75,3,1},
2798 {0x2c7e,0x2c7f,1,-10815},
2799 {0x2c80,0x2ce2,2,1},
2800 {0x2ceb,0x2ced,2,1},
2801 {0xa640,0xa65e,2,1},
2802 {0xa662,0xa66c,2,1},
2803 {0xa680,0xa696,2,1},
2804 {0xa722,0xa72e,2,1},
2805 {0xa732,0xa76e,2,1},
2806 {0xa779,0xa77b,2,1},
2807 {0xa77d,0xa77d,-1,-35332},
2808 {0xa77e,0xa786,2,1},
2809 {0xa78b,0xa78b,-1,1},
2810 {0xff21,0xff3a,1,32},
2811 {0x10400,0x10427,1,40}
2814 static convertStruct toUpper
[] =
2822 {0x131,0x131,-1,-232},
2827 {0x17f,0x17f,-1,-300},
2828 {0x180,0x180,-1,195},
2831 {0x192,0x192,-1,-1},
2832 {0x195,0x195,-1,97},
2833 {0x199,0x199,-1,-1},
2834 {0x19a,0x19a,-1,163},
2835 {0x19e,0x19e,-1,130},
2840 {0x1bd,0x1bd,-1,-1},
2841 {0x1bf,0x1bf,-1,56},
2842 {0x1c5,0x1c5,-1,-1},
2843 {0x1c6,0x1c6,-1,-2},
2844 {0x1c8,0x1c8,-1,-1},
2845 {0x1c9,0x1c9,-1,-2},
2846 {0x1cb,0x1cb,-1,-1},
2847 {0x1cc,0x1cc,-1,-2},
2849 {0x1dd,0x1dd,-1,-79},
2851 {0x1f2,0x1f2,-1,-1},
2852 {0x1f3,0x1f3,-1,-2},
2856 {0x23c,0x23c,-1,-1},
2857 {0x23f,0x240,1,10815},
2860 {0x250,0x250,-1,10783},
2861 {0x251,0x251,-1,10780},
2862 {0x252,0x252,-1,10782},
2863 {0x253,0x253,-1,-210},
2864 {0x254,0x254,-1,-206},
2865 {0x256,0x257,1,-205},
2866 {0x259,0x259,-1,-202},
2867 {0x25b,0x25b,-1,-203},
2868 {0x260,0x260,-1,-205},
2869 {0x263,0x263,-1,-207},
2870 {0x268,0x268,-1,-209},
2871 {0x269,0x269,-1,-211},
2872 {0x26b,0x26b,-1,10743},
2873 {0x26f,0x26f,-1,-211},
2874 {0x271,0x271,-1,10749},
2875 {0x272,0x272,-1,-213},
2876 {0x275,0x275,-1,-214},
2877 {0x27d,0x27d,-1,10727},
2878 {0x280,0x283,3,-218},
2879 {0x288,0x288,-1,-218},
2880 {0x289,0x289,-1,-69},
2881 {0x28a,0x28b,1,-217},
2882 {0x28c,0x28c,-1,-71},
2883 {0x292,0x292,-1,-219},
2884 {0x345,0x345,-1,84},
2886 {0x377,0x377,-1,-1},
2887 {0x37b,0x37d,1,130},
2888 {0x3ac,0x3ac,-1,-38},
2889 {0x3ad,0x3af,1,-37},
2890 {0x3b1,0x3c1,1,-32},
2891 {0x3c2,0x3c2,-1,-31},
2892 {0x3c3,0x3cb,1,-32},
2893 {0x3cc,0x3cc,-1,-64},
2894 {0x3cd,0x3ce,1,-63},
2895 {0x3d0,0x3d0,-1,-62},
2896 {0x3d1,0x3d1,-1,-57},
2897 {0x3d5,0x3d5,-1,-47},
2898 {0x3d6,0x3d6,-1,-54},
2899 {0x3d7,0x3d7,-1,-8},
2901 {0x3f0,0x3f0,-1,-86},
2902 {0x3f1,0x3f1,-1,-80},
2904 {0x3f5,0x3f5,-1,-96},
2906 {0x430,0x44f,1,-32},
2907 {0x450,0x45f,1,-80},
2911 {0x4cf,0x4cf,-1,-15},
2913 {0x561,0x586,1,-48},
2914 {0x1d79,0x1d79,-1,35332},
2915 {0x1d7d,0x1d7d,-1,3814},
2916 {0x1e01,0x1e95,2,-1},
2917 {0x1e9b,0x1e9b,-1,-59},
2918 {0x1ea1,0x1eff,2,-1},
2919 {0x1f00,0x1f07,1,8},
2920 {0x1f10,0x1f15,1,8},
2921 {0x1f20,0x1f27,1,8},
2922 {0x1f30,0x1f37,1,8},
2923 {0x1f40,0x1f45,1,8},
2924 {0x1f51,0x1f57,2,8},
2925 {0x1f60,0x1f67,1,8},
2926 {0x1f70,0x1f71,1,74},
2927 {0x1f72,0x1f75,1,86},
2928 {0x1f76,0x1f77,1,100},
2929 {0x1f78,0x1f79,1,128},
2930 {0x1f7a,0x1f7b,1,112},
2931 {0x1f7c,0x1f7d,1,126},
2932 {0x1f80,0x1f87,1,8},
2933 {0x1f90,0x1f97,1,8},
2934 {0x1fa0,0x1fa7,1,8},
2935 {0x1fb0,0x1fb1,1,8},
2936 {0x1fb3,0x1fb3,-1,9},
2937 {0x1fbe,0x1fbe,-1,-7205},
2938 {0x1fc3,0x1fc3,-1,9},
2939 {0x1fd0,0x1fd1,1,8},
2940 {0x1fe0,0x1fe1,1,8},
2941 {0x1fe5,0x1fe5,-1,7},
2942 {0x1ff3,0x1ff3,-1,9},
2943 {0x214e,0x214e,-1,-28},
2944 {0x2170,0x217f,1,-16},
2945 {0x2184,0x2184,-1,-1},
2946 {0x24d0,0x24e9,1,-26},
2947 {0x2c30,0x2c5e,1,-48},
2948 {0x2c61,0x2c61,-1,-1},
2949 {0x2c65,0x2c65,-1,-10795},
2950 {0x2c66,0x2c66,-1,-10792},
2951 {0x2c68,0x2c6c,2,-1},
2952 {0x2c73,0x2c76,3,-1},
2953 {0x2c81,0x2ce3,2,-1},
2954 {0x2cec,0x2cee,2,-1},
2955 {0x2d00,0x2d25,1,-7264},
2956 {0xa641,0xa65f,2,-1},
2957 {0xa663,0xa66d,2,-1},
2958 {0xa681,0xa697,2,-1},
2959 {0xa723,0xa72f,2,-1},
2960 {0xa733,0xa76f,2,-1},
2961 {0xa77a,0xa77c,2,-1},
2962 {0xa77f,0xa787,2,-1},
2963 {0xa78c,0xa78c,-1,-1},
2964 {0xff41,0xff5a,1,-32},
2965 {0x10428,0x1044f,1,-40}
2969 * Return the upper-case equivalent of "a", which is a UCS-4 character. Use
2970 * simple case folding.
2976 /* If 'casemap' contains "keepascii" use ASCII style toupper(). */
2977 if (a
< 128 && (cmp_flags
& CMP_KEEPASCII
))
2978 return TOUPPER_ASC(a
);
2980 #if defined(HAVE_TOWUPPER) && defined(__STDC_ISO_10646__)
2981 /* If towupper() is available and handles Unicode, use it. */
2982 if (!(cmp_flags
& CMP_INTERNAL
))
2986 /* For characters below 128 use locale sensitive toupper(). */
2988 return TOUPPER_LOC(a
);
2990 /* For any other characters use the above mapping table. */
2991 return utf_convert(a
, toUpper
, sizeof(toUpper
));
2998 return (utf_toupper(a
) != a
);
3002 * Return the lower-case equivalent of "a", which is a UCS-4 character. Use
3003 * simple case folding.
3009 /* If 'casemap' contains "keepascii" use ASCII style tolower(). */
3010 if (a
< 128 && (cmp_flags
& CMP_KEEPASCII
))
3011 return TOLOWER_ASC(a
);
3013 #if defined(HAVE_TOWLOWER) && defined(__STDC_ISO_10646__)
3014 /* If towlower() is available and handles Unicode, use it. */
3015 if (!(cmp_flags
& CMP_INTERNAL
))
3019 /* For characters below 128 use locale sensitive tolower(). */
3021 return TOLOWER_LOC(a
);
3023 /* For any other characters use the above mapping table. */
3024 return utf_convert(a
, toLower
, sizeof(toLower
));
3031 return (utf_tolower(a
) != a
);
3035 * Version of strnicmp() that handles multi-byte characters.
3036 * Needed for Big5, Sjift-JIS and UTF-8 encoding. Other DBCS encodings can
3037 * probably use strnicmp(), because there are no ASCII characters in the
3039 * Returns zero if s1 and s2 are equal (ignoring case), the difference between
3040 * two characters otherwise.
3043 mb_strnicmp(s1
, s2
, nn
)
3049 int incomplete
= FALSE
;
3052 for (i
= 0; i
< n
; i
+= l
)
3054 if (s1
[i
] == NUL
&& s2
[i
] == NUL
) /* both strings end */
3058 l
= utf_byte2len(s1
[i
]);
3061 l
= n
- i
; /* incomplete character */
3064 /* Check directly first, it's faster. */
3065 for (j
= 0; j
< l
; ++j
)
3067 if (s1
[i
+ j
] != s2
[i
+ j
])
3070 /* Both stings have the same bytes but are incomplete or
3071 * have illegal bytes, accept them as equal. */
3076 /* If one of the two characters is incomplete return -1. */
3077 if (incomplete
|| i
+ utf_byte2len(s2
[i
]) > n
)
3079 cdiff
= utf_fold(utf_ptr2char(s1
+ i
))
3080 - utf_fold(utf_ptr2char(s2
+ i
));
3087 l
= (*mb_ptr2len
)(s1
+ i
);
3090 /* Single byte: first check normally, then with ignore case. */
3093 cdiff
= MB_TOLOWER(s1
[i
]) - MB_TOLOWER(s2
[i
]);
3100 /* For non-Unicode multi-byte don't ignore case. */
3103 cdiff
= STRNCMP(s1
+ i
, s2
+ i
, l
);
3113 * "g8": show bytes of the UTF-8 char under the cursor. Doesn't matter what
3114 * 'encoding' has been set to.
3125 /* Get the byte length of the char under the cursor, including composing
3127 line
= ml_get_cursor();
3128 len
= utfc_ptr2len(line
);
3136 for (i
= 0; i
< len
; ++i
)
3140 /* start of (composing) character, get its length */
3143 STRCPY(IObuff
+ rlen
, "+ ");
3146 clen
= utf_ptr2len(line
+ i
);
3148 sprintf((char *)IObuff
+ rlen
, "%02x ", line
[i
]);
3150 rlen
+= (int)STRLEN(IObuff
+ rlen
);
3151 if (rlen
> IOSIZE
- 20)
3159 * mb_head_off() function pointer.
3160 * Return offset from "p" to the first byte of the character it points into.
3161 * If "p" points to the NUL at the end of the string return 0.
3162 * Returns 0 when already at the first byte of a character.
3165 latin_head_off(base
, p
)
3166 char_u
*base UNUSED
;
3173 dbcs_head_off(base
, p
)
3179 /* It can't be a trailing byte when not using DBCS, at the start of the
3180 * string or the previous byte can't start a double-byte. */
3181 if (p
<= base
|| MB_BYTE2LEN(p
[-1]) == 1 || *p
== NUL
)
3184 /* This is slow: need to start at the base and go forward until the
3185 * byte we are looking for. Return 1 when we went past it, 0 otherwise. */
3188 q
+= dbcs_ptr2len(q
);
3189 return (q
== p
) ? 0 : 1;
3193 * Special version of dbcs_head_off() that works for ScreenLines[], where
3194 * single-width DBCS_JPNU characters are stored separately.
3197 dbcs_screen_head_off(base
, p
)
3203 /* It can't be a trailing byte when not using DBCS, at the start of the
3204 * string or the previous byte can't start a double-byte.
3205 * For euc-jp an 0x8e byte in the previous cell always means we have a
3206 * lead byte in the current cell. */
3208 || (enc_dbcs
== DBCS_JPNU
&& p
[-1] == 0x8e)
3209 || MB_BYTE2LEN(p
[-1]) == 1
3213 /* This is slow: need to start at the base and go forward until the
3214 * byte we are looking for. Return 1 when we went past it, 0 otherwise.
3215 * For DBCS_JPNU look out for 0x8e, which means the second byte is not
3216 * stored as the next byte. */
3220 if (enc_dbcs
== DBCS_JPNU
&& *q
== 0x8e)
3223 q
+= dbcs_ptr2len(q
);
3225 return (q
== p
) ? 0 : 1;
3229 utf_head_off(base
, p
)
3241 if (*p
< 0x80) /* be quick for ASCII */
3244 /* Skip backwards over trailing bytes: 10xx.xxxx
3245 * Skip backwards again if on a composing char. */
3248 /* Move s to the last byte of this char. */
3249 for (s
= q
; (s
[1] & 0xc0) == 0x80; ++s
)
3251 /* Move q to the first byte of this char. */
3252 while (q
> base
&& (*q
& 0xc0) == 0x80)
3254 /* Check for illegal sequence. Do allow an illegal byte after where we
3256 len
= utf8len_tab
[*q
];
3257 if (len
!= (int)(s
- q
+ 1) && len
!= (int)(p
- q
+ 1))
3263 c
= utf_ptr2char(q
);
3264 if (utf_iscomposing(c
))
3268 if (arabic_maycombine(c
))
3270 /* Advance to get a sneak-peak at the next char */
3273 /* Move j to the first byte of this char. */
3274 while (j
> base
&& (*j
& 0xc0) == 0x80)
3276 if (arabic_combine(utf_ptr2char(j
), c
))
3283 return (int)(p
- q
);
3287 * Copy a character from "*fp" to "*tp" and advance the pointers.
3290 mb_copy_char(fp
, tp
)
3294 int l
= (*mb_ptr2len
)(*fp
);
3296 mch_memmove(*tp
, *fp
, (size_t)l
);
3302 * Return the offset from "p" to the first byte of a character. When "p" is
3303 * at the start of a character 0 is returned, otherwise the offset to the next
3304 * character. Can start anywhere in a stream of bytes.
3307 mb_off_next(base
, p
)
3316 if (*p
< 0x80) /* be quick for ASCII */
3319 /* Find the next character that isn't 10xx.xxxx */
3320 for (i
= 0; (p
[i
] & 0xc0) == 0x80; ++i
)
3324 /* Check for illegal sequence. */
3325 for (j
= 0; p
- j
> base
; ++j
)
3326 if ((p
[-j
] & 0xc0) != 0x80)
3328 if (utf8len_tab
[p
[-j
]] != i
+ j
)
3334 /* Only need to check if we're on a trail byte, it doesn't matter if we
3335 * want the offset to the next or current character. */
3336 return (*mb_head_off
)(base
, p
);
3340 * Return the offset from "p" to the last byte of the character it points
3341 * into. Can start anywhere in a stream of bytes.
3344 mb_tail_off(base
, p
)
3356 /* Find the last character that is 10xx.xxxx */
3357 for (i
= 0; (p
[i
+ 1] & 0xc0) == 0x80; ++i
)
3359 /* Check for illegal sequence. */
3360 for (j
= 0; p
- j
> base
; ++j
)
3361 if ((p
[-j
] & 0xc0) != 0x80)
3363 if (utf8len_tab
[p
[-j
]] != i
+ j
+ 1)
3368 /* It can't be the first byte if a double-byte when not using DBCS, at the
3369 * end of the string or the byte can't start a double-byte. */
3370 if (enc_dbcs
== 0 || p
[1] == NUL
|| MB_BYTE2LEN(*p
) == 1)
3373 /* Return 1 when on the lead byte, 0 when on the tail byte. */
3374 return 1 - dbcs_head_off(base
, p
);
3378 * Find the next illegal byte sequence.
3383 pos_T pos
= curwin
->w_cursor
;
3387 char_u
*tofree
= NULL
;
3389 vimconv
.vc_type
= CONV_NONE
;
3390 if (enc_utf8
&& (enc_canon_props(curbuf
->b_p_fenc
) & ENC_8BIT
))
3392 /* 'encoding' is "utf-8" but we are editing a 8-bit encoded file,
3393 * possibly a utf-8 file with illegal bytes. Setup for conversion
3394 * from utf-8 to 'fileencoding'. */
3395 convert_setup(&vimconv
, p_enc
, curbuf
->b_p_fenc
);
3398 #ifdef FEAT_VIRTUALEDIT
3399 curwin
->w_cursor
.coladd
= 0;
3403 p
= ml_get_cursor();
3404 if (vimconv
.vc_type
!= CONV_NONE
)
3407 tofree
= string_convert(&vimconv
, p
, NULL
);
3415 /* Illegal means that there are not enough trail bytes (checked by
3416 * utf_ptr2len()) or too many of them (overlong sequence). */
3417 len
= utf_ptr2len(p
);
3418 if (*p
>= 0x80 && (len
== 1
3419 || utf_char2len(utf_ptr2char(p
)) != len
))
3421 if (vimconv
.vc_type
== CONV_NONE
)
3422 curwin
->w_cursor
.col
+= (colnr_T
)(p
- ml_get_cursor());
3427 len
= (int)(p
- tofree
);
3428 for (p
= ml_get_cursor(); *p
!= NUL
&& len
-- > 0; p
+= l
)
3431 curwin
->w_cursor
.col
+= l
;
3438 if (curwin
->w_cursor
.lnum
== curbuf
->b_ml
.ml_line_count
)
3440 ++curwin
->w_cursor
.lnum
;
3441 curwin
->w_cursor
.col
= 0;
3444 /* didn't find it: don't move and beep */
3445 curwin
->w_cursor
= pos
;
3450 convert_setup(&vimconv
, NULL
, NULL
);
3453 #if defined(HAVE_GTK2) || defined(PROTO)
3455 * Return TRUE if string "s" is a valid utf-8 string.
3456 * When "end" is NULL stop at the first NUL.
3457 * When "end" is positive stop there.
3460 utf_valid_string(s
, end
)
3467 while (end
== NULL
? *p
!= NUL
: p
< end
)
3469 l
= utf8len_tab_zero
[*p
];
3471 return FALSE
; /* invalid lead byte */
3472 if (end
!= NULL
&& p
+ l
> end
)
3473 return FALSE
; /* incomplete byte sequence */
3476 if ((*p
++ & 0xc0) != 0x80)
3477 return FALSE
; /* invalid trail byte */
3483 #if defined(FEAT_GUI) || defined(PROTO)
3485 * Special version of mb_tail_off() for use in ScreenLines[].
3488 dbcs_screen_tail_off(base
, p
)
3492 /* It can't be the first byte if a double-byte when not using DBCS, at the
3493 * end of the string or the byte can't start a double-byte.
3494 * For euc-jp an 0x8e byte always means we have a lead byte in the current
3496 if (*p
== NUL
|| p
[1] == NUL
3497 || (enc_dbcs
== DBCS_JPNU
&& *p
== 0x8e)
3498 || MB_BYTE2LEN(*p
) == 1)
3501 /* Return 1 when on the lead byte, 0 when on the tail byte. */
3502 return 1 - dbcs_screen_head_off(base
, p
);
3507 * If the cursor moves on an trail byte, set the cursor on the lead byte.
3508 * Thus it moves left if necessary.
3509 * Return TRUE when the cursor was adjusted.
3514 mb_adjustpos(&curwin
->w_cursor
);
3518 * Adjust position "*lp" to point to the first byte of a multi-byte character.
3519 * If it points to a tail byte it's moved backwards to the head byte.
3528 #ifdef FEAT_VIRTUALEDIT
3533 p
= ml_get(lp
->lnum
);
3534 lp
->col
-= (*mb_head_off
)(p
, p
+ lp
->col
);
3535 #ifdef FEAT_VIRTUALEDIT
3536 /* Reset "coladd" when the cursor would be on the right half of a
3537 * double-wide character. */
3539 && p
[lp
->col
] != TAB
3540 && vim_isprintc((*mb_ptr2char
)(p
+ lp
->col
))
3541 && ptr2cells(p
+ lp
->col
) > 1)
3548 * Return a pointer to the character before "*p", if there is one.
3552 char_u
*line
; /* start of the string */
3556 mb_ptr_back(line
, p
);
3561 * Return the character length of "str". Each multi-byte character (with
3562 * following composing characters) counts as one.
3574 for (count
= 0; *p
!= NUL
; count
++)
3575 p
+= (*mb_ptr2len
)(p
);
3580 #if defined(FEAT_SPELL) || defined(PROTO)
3582 * Like mb_charlen() but for a string with specified length.
3585 mb_charlen_len(str
, len
)
3592 for (count
= 0; *p
!= NUL
&& p
< str
+ len
; count
++)
3593 p
+= (*mb_ptr2len
)(p
);
3600 * Try to un-escape a multi-byte character.
3601 * Used for the "to" and "from" part of a mapping.
3602 * Return the un-escaped string if it is a multi-byte character, and advance
3603 * "pp" to just after the bytes that formed it.
3604 * Return NULL if no multi-byte char was found.
3610 static char_u buf
[MB_MAXBYTES
+ 1];
3614 /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI
3615 * KS_EXTRA KE_CSI to CSI. */
3616 for (n
= 0; str
[n
] != NUL
&& m
<= MB_MAXBYTES
; ++n
)
3618 if (str
[n
] == K_SPECIAL
3619 && str
[n
+ 1] == KS_SPECIAL
3620 && str
[n
+ 2] == KE_FILLER
)
3622 buf
[m
++] = K_SPECIAL
;
3625 else if ((str
[n
] == K_SPECIAL
3630 && str
[n
+ 1] == KS_EXTRA
3631 && str
[n
+ 2] == (int)KE_CSI
)
3636 else if (str
[n
] == K_SPECIAL
3641 break; /* a special key can't be a multibyte char */
3646 /* Return a multi-byte character if it's found. An illegal sequence
3647 * will result in a 1 here. */
3648 if ((*mb_ptr2len
)(buf
) > 1)
3658 * Return TRUE if the character at "row"/"col" on the screen is the left side
3659 * of a double-width character.
3660 * Caller must make sure "row" and "col" are not invalid!
3663 mb_lefthalve(row
, col
)
3667 #ifdef FEAT_HANGULIN
3668 if (composing_hangul
)
3671 return (*mb_off2cells
)(LineOffset
[row
] + col
,
3672 LineOffset
[row
] + screen_Columns
) > 1;
3676 * Correct a position on the screen, if it's the right half of a double-wide
3677 * char move it to the left half. Returns the corrected column.
3680 mb_fix_col(col
, row
)
3684 col
= check_col(col
);
3685 row
= check_row(row
);
3686 if (has_mbyte
&& ScreenLines
!= NULL
&& col
> 0
3688 && ScreenLines
[LineOffset
[row
] + col
] != NUL
3689 && dbcs_screen_head_off(ScreenLines
+ LineOffset
[row
],
3690 ScreenLines
+ LineOffset
[row
] + col
))
3691 || (enc_utf8
&& ScreenLines
[LineOffset
[row
] + col
] == 0)))
3697 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO)
3698 static int enc_alias_search
__ARGS((char_u
*name
));
3701 * Skip the Vim specific head of a 'encoding' name.
3707 if (STRNCMP(p
, "2byte-", 6) == 0)
3709 if (STRNCMP(p
, "8bit-", 5) == 0)
3715 * Find the canonical name for encoding "enc".
3716 * When the name isn't recognized, returns "enc" itself, but with all lower
3717 * case characters and '_' replaced with '-'.
3718 * Returns an allocated string. NULL for out-of-memory.
3729 if (STRCMP(enc
, "default") == 0)
3731 /* Use the default encoding as it's found by set_init_1(). */
3732 r
= get_encoding_default();
3734 r
= (char_u
*)"latin1";
3735 return vim_strsave(r
);
3739 /* copy "enc" to allocated memory, with room for two '-' */
3740 r
= alloc((unsigned)(STRLEN(enc
) + 3));
3743 /* Make it all lower case and replace '_' with '-'. */
3745 for (s
= enc
; *s
!= NUL
; ++s
)
3750 *p
++ = TOLOWER_ASC(*s
);
3754 /* Skip "2byte-" and "8bit-". */
3757 /* Change "microsoft-cp" to "cp". Used in some spell files. */
3758 if (STRNCMP(p
, "microsoft-cp", 12) == 0)
3761 /* "iso8859" -> "iso-8859" */
3762 if (STRNCMP(p
, "iso8859", 7) == 0)
3764 STRMOVE(p
+ 4, p
+ 3);
3768 /* "iso-8859n" -> "iso-8859-n" */
3769 if (STRNCMP(p
, "iso-8859", 8) == 0 && p
[8] != '-')
3771 STRMOVE(p
+ 9, p
+ 8);
3775 /* "latin-N" -> "latinN" */
3776 if (STRNCMP(p
, "latin-", 6) == 0)
3777 STRMOVE(p
+ 5, p
+ 6);
3779 if (enc_canon_search(p
) >= 0)
3781 /* canonical name can be used unmodified */
3785 else if ((i
= enc_alias_search(p
)) >= 0)
3787 /* alias recognized, get canonical name */
3789 r
= vim_strsave((char_u
*)enc_canon_table
[i
].name
);
3796 * Search for an encoding alias of "name".
3797 * Returns -1 when not found.
3800 enc_alias_search(name
)
3805 for (i
= 0; enc_alias_table
[i
].name
!= NULL
; ++i
)
3806 if (STRCMP(name
, enc_alias_table
[i
].name
) == 0)
3807 return enc_alias_table
[i
].canon
;
3812 #if defined(FEAT_MBYTE) || defined(PROTO)
3814 #ifdef HAVE_LANGINFO_H
3815 # include <langinfo.h>
3819 * Get the canonicalized encoding of the current locale.
3820 * Returns an allocated string when successful, NULL when not.
3832 long acp
= GetACP();
3835 STRCPY(buf
, "ucs-2le");
3836 else if (acp
== 1252) /* cp1252 is used as latin1 */
3837 STRCPY(buf
, "latin1");
3839 sprintf(buf
, "cp%ld", acp
);
3841 # ifdef HAVE_NL_LANGINFO_CODESET
3842 if ((s
= nl_langinfo(CODESET
)) == NULL
|| *s
== NUL
)
3844 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE)
3845 if ((s
= setlocale(LC_CTYPE
, NULL
)) == NULL
|| *s
== NUL
)
3847 if ((s
= getenv("LC_ALL")) == NULL
|| *s
== NUL
)
3848 if ((s
= getenv("LC_CTYPE")) == NULL
|| *s
== NUL
)
3851 if (s
== NULL
|| *s
== NUL
)
3854 /* The most generic locale format is:
3855 * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]]
3856 * If there is a '.' remove the part before it.
3857 * if there is something after the codeset, remove it.
3858 * Make the name lowercase and replace '_' with '-'.
3859 * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn",
3860 * "ko_KR.EUC" == "euc-kr"
3862 if ((p
= (char *)vim_strchr((char_u
*)s
, '.')) != NULL
)
3864 if (p
> s
+ 2 && STRNICMP(p
+ 1, "EUC", 3) == 0
3865 && !isalnum((int)p
[4]) && p
[4] != '-' && p
[-3] == '_')
3867 /* copy "XY.EUC" to "euc-XY" to buf[10] */
3868 STRCPY(buf
+ 10, "euc-");
3877 for (i
= 0; s
[i
] != NUL
&& i
< (int)sizeof(buf
) - 1; ++i
)
3879 if (s
[i
] == '_' || s
[i
] == '-')
3881 else if (isalnum((int)s
[i
]))
3882 buf
[i
] = TOLOWER_ASC(s
[i
]);
3889 return enc_canonize((char_u
*)buf
);
3892 #if defined(WIN3264) || defined(PROTO)
3894 * Convert an encoding name to an MS-Windows codepage.
3895 * Returns zero if no codepage can be figured out.
3898 encname2codepage(name
)
3905 if (STRNCMP(p
, "8bit-", 5) == 0)
3907 else if (STRNCMP(p_enc
, "2byte-", 6) == 0)
3910 if (p
[0] == 'c' && p
[1] == 'p')
3912 else if ((idx
= enc_canon_search(p
)) >= 0)
3913 cp
= enc_canon_table
[idx
].codepage
;
3916 if (IsValidCodePage(cp
))
3922 # if defined(USE_ICONV) || defined(PROTO)
3924 static char_u
*iconv_string
__ARGS((vimconv_T
*vcp
, char_u
*str
, int slen
, int *unconvlenp
, int *resultlenp
));
3927 * Call iconv_open() with a check if iconv() works properly (there are broken
3929 * Returns (void *)-1 if failed.
3930 * (should return iconv_t, but that causes problems with prototypes).
3933 my_iconv_open(to
, from
)
3938 #define ICONV_TESTLEN 400
3939 char_u tobuf
[ICONV_TESTLEN
];
3942 static int iconv_ok
= -1;
3944 if (iconv_ok
== FALSE
)
3945 return (void *)-1; /* detected a broken iconv() previously */
3947 #ifdef DYNAMIC_ICONV
3948 /* Check if the iconv.dll can be found. */
3949 if (!iconv_enabled(TRUE
))
3953 fd
= iconv_open((char *)enc_skip(to
), (char *)enc_skip(from
));
3955 if (fd
!= (iconv_t
)-1 && iconv_ok
== -1)
3958 * Do a dummy iconv() call to check if it actually works. There is a
3959 * version of iconv() on Linux that is broken. We can't ignore it,
3960 * because it's wide-spread. The symptoms are that after outputting
3961 * the initial shift state the "to" pointer is NULL and conversion
3962 * stops for no apparent reason after about 8160 characters.
3965 tolen
= ICONV_TESTLEN
;
3966 (void)iconv(fd
, NULL
, NULL
, &p
, &tolen
);
3981 * Convert the string "str[slen]" with iconv().
3982 * If "unconvlenp" is not NULL handle the string ending in an incomplete
3983 * sequence and set "*unconvlenp" to the length of it.
3984 * Returns the converted string in allocated memory. NULL for an error.
3985 * If resultlenp is not NULL, sets it to the result length in bytes.
3988 iconv_string(vcp
, str
, slen
, unconvlenp
, resultlenp
)
4001 char_u
*result
= NULL
;
4009 if (len
== 0 || ICONV_ERRNO
== ICONV_E2BIG
)
4011 /* Allocate enough room for most conversions. When re-allocating
4012 * increase the buffer size. */
4013 len
= len
+ fromlen
* 2 + 40;
4014 p
= alloc((unsigned)len
);
4015 if (p
!= NULL
&& done
> 0)
4016 mch_memmove(p
, result
, done
);
4019 if (result
== NULL
) /* out of memory */
4023 to
= (char *)result
+ done
;
4024 tolen
= len
- done
- 2;
4025 /* Avoid a warning for systems with a wrong iconv() prototype by
4026 * casting the second argument to void *. */
4027 if (iconv(vcp
->vc_fd
, (void *)&from
, &fromlen
, &to
, &tolen
)
4030 /* Finished, append a NUL. */
4035 /* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded
4036 * iconv library may use one of them. */
4037 if (!vcp
->vc_fail
&& unconvlenp
!= NULL
4038 && (ICONV_ERRNO
== ICONV_EINVAL
|| ICONV_ERRNO
== EINVAL
))
4040 /* Handle an incomplete sequence at the end. */
4042 *unconvlenp
= (int)fromlen
;
4046 /* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded
4047 * iconv library may use one of them. */
4048 else if (!vcp
->vc_fail
4049 && (ICONV_ERRNO
== ICONV_EILSEQ
|| ICONV_ERRNO
== EILSEQ
4050 || ICONV_ERRNO
== ICONV_EINVAL
|| ICONV_ERRNO
== EINVAL
))
4052 /* Can't convert: insert a '?' and skip a character. This assumes
4053 * conversion from 'encoding' to something else. In other
4054 * situations we don't know what to skip anyway. */
4056 if ((*mb_ptr2cells
)((char_u
*)from
) > 1)
4059 l
= utfc_ptr2len_len((char_u
*)from
, (int)fromlen
);
4062 l
= (*mb_ptr2len
)((char_u
*)from
);
4063 if (l
> (int)fromlen
)
4069 else if (ICONV_ERRNO
!= ICONV_E2BIG
)
4071 /* conversion failed */
4076 /* Not enough room or skipping illegal sequence. */
4077 done
= to
- (char *)result
;
4080 if (resultlenp
!= NULL
)
4081 *resultlenp
= (int)(to
- (char *)result
);
4085 # if defined(DYNAMIC_ICONV) || defined(PROTO)
4087 * Dynamically load the "iconv.dll" on Win32.
4090 #ifndef DYNAMIC_ICONV /* just generating prototypes */
4091 # define HINSTANCE int
4093 static HINSTANCE hIconvDLL
= 0;
4094 static HINSTANCE hMsvcrtDLL
= 0;
4096 # ifndef DYNAMIC_ICONV_DLL
4097 # define DYNAMIC_ICONV_DLL "iconv.dll"
4098 # define DYNAMIC_ICONV_DLL_ALT "libiconv.dll"
4100 # ifndef DYNAMIC_MSVCRT_DLL
4101 # define DYNAMIC_MSVCRT_DLL "msvcrt.dll"
4105 * Try opening the iconv.dll and return TRUE if iconv() can be used.
4108 iconv_enabled(verbose
)
4111 if (hIconvDLL
!= 0 && hMsvcrtDLL
!= 0)
4113 hIconvDLL
= LoadLibrary(DYNAMIC_ICONV_DLL
);
4114 if (hIconvDLL
== 0) /* sometimes it's called libiconv.dll */
4115 hIconvDLL
= LoadLibrary(DYNAMIC_ICONV_DLL_ALT
);
4117 hMsvcrtDLL
= LoadLibrary(DYNAMIC_MSVCRT_DLL
);
4118 if (hIconvDLL
== 0 || hMsvcrtDLL
== 0)
4120 /* Only give the message when 'verbose' is set, otherwise it might be
4121 * done whenever a conversion is attempted. */
4122 if (verbose
&& p_verbose
> 0)
4126 hIconvDLL
== 0 ? DYNAMIC_ICONV_DLL
: DYNAMIC_MSVCRT_DLL
);
4133 iconv
= (void *)GetProcAddress(hIconvDLL
, "libiconv");
4134 iconv_open
= (void *)GetProcAddress(hIconvDLL
, "libiconv_open");
4135 iconv_close
= (void *)GetProcAddress(hIconvDLL
, "libiconv_close");
4136 iconvctl
= (void *)GetProcAddress(hIconvDLL
, "libiconvctl");
4137 iconv_errno
= (void *)GetProcAddress(hMsvcrtDLL
, "_errno");
4138 if (iconv
== NULL
|| iconv_open
== NULL
|| iconv_close
== NULL
4139 || iconvctl
== NULL
|| iconv_errno
== NULL
)
4142 if (verbose
&& p_verbose
> 0)
4145 EMSG2(_(e_loadfunc
), "for libiconv");
4156 /* Don't use iconv() when inputting or outputting characters. */
4157 if (input_conv
.vc_type
== CONV_ICONV
)
4158 convert_setup(&input_conv
, NULL
, NULL
);
4159 if (output_conv
.vc_type
== CONV_ICONV
)
4160 convert_setup(&output_conv
, NULL
, NULL
);
4163 FreeLibrary(hIconvDLL
);
4164 if (hMsvcrtDLL
!= 0)
4165 FreeLibrary(hMsvcrtDLL
);
4169 # endif /* DYNAMIC_ICONV */
4170 # endif /* USE_ICONV */
4172 #endif /* FEAT_MBYTE */
4174 #if defined(FEAT_XIM) || defined(PROTO)
4176 # ifdef FEAT_GUI_GTK
4177 static int xim_has_preediting
INIT(= FALSE
); /* IM current status */
4180 * Set preedit_start_col to the current cursor position.
4183 init_preedit_start_col(void)
4185 if (State
& CMDLINE
)
4186 preedit_start_col
= cmdline_getvcol_cursor();
4187 else if (curwin
!= NULL
)
4188 getvcol(curwin
, &curwin
->w_cursor
, &preedit_start_col
, NULL
, NULL
);
4189 /* Prevent that preediting marks the buffer as changed. */
4190 xim_changed_while_preediting
= curbuf
->b_changed
;
4194 # if defined(HAVE_GTK2) && !defined(PROTO)
4196 static int im_is_active
= FALSE
; /* IM is enabled for current mode */
4197 static int preedit_is_active
= FALSE
;
4198 static int im_preedit_cursor
= 0; /* cursor offset in characters */
4199 static int im_preedit_trailing
= 0; /* number of characters after cursor */
4201 static unsigned long im_commit_handler_id
= 0;
4202 static unsigned int im_activatekey_keyval
= GDK_VoidSymbol
;
4203 static unsigned int im_activatekey_state
= 0;
4206 im_set_active(int active
)
4210 was_active
= !!im_is_active
;
4211 im_is_active
= (active
&& !p_imdisable
);
4213 if (im_is_active
!= was_active
)
4218 xim_set_focus(int focus
)
4223 gtk_im_context_focus_in(xic
);
4225 gtk_im_context_focus_out(xic
);
4230 im_set_position(int row
, int col
)
4236 area
.x
= FILL_X(col
);
4237 area
.y
= FILL_Y(row
);
4238 area
.width
= gui
.char_width
* (mb_lefthalve(row
, col
) ? 2 : 1);
4239 area
.height
= gui
.char_height
;
4241 gtk_im_context_set_cursor_location(xic
, &area
);
4245 # if 0 || defined(PROTO) /* apparently only used in gui_x11.c */
4247 xim_set_preedit(void)
4249 im_set_position(gui
.row
, gui
.col
);
4254 im_add_to_input(char_u
*str
, int len
)
4256 /* Convert from 'termencoding' (always "utf-8") to 'encoding' */
4257 if (input_conv
.vc_type
!= CONV_NONE
)
4259 str
= string_convert(&input_conv
, str
, &len
);
4260 g_return_if_fail(str
!= NULL
);
4263 add_to_input_buf_csi(str
, len
);
4265 if (input_conv
.vc_type
!= CONV_NONE
)
4268 if (p_mh
) /* blank out the pointer if necessary */
4269 gui_mch_mousehide(TRUE
);
4273 im_delete_preedit(void)
4275 char_u bskey
[] = {CSI
, 'k', 'b'};
4276 char_u delkey
[] = {CSI
, 'k', 'D'};
4280 im_preedit_cursor
= 0;
4283 for (; im_preedit_cursor
> 0; --im_preedit_cursor
)
4284 add_to_input_buf(bskey
, (int)sizeof(bskey
));
4286 for (; im_preedit_trailing
> 0; --im_preedit_trailing
)
4287 add_to_input_buf(delkey
, (int)sizeof(delkey
));
4291 * Move the cursor left by "num_move_back" characters.
4292 * Note that ins_left() checks im_is_preediting() to avoid breaking undo for
4293 * these K_LEFT keys.
4296 im_correct_cursor(int num_move_back
)
4298 char_u backkey
[] = {CSI
, 'k', 'l'};
4302 # ifdef FEAT_RIGHTLEFT
4303 if ((State
& CMDLINE
) == 0 && curwin
!= NULL
&& curwin
->w_p_rl
)
4306 for (; num_move_back
> 0; --num_move_back
)
4307 add_to_input_buf(backkey
, (int)sizeof(backkey
));
4310 static int xim_expected_char
= NUL
;
4311 static int xim_ignored_char
= FALSE
;
4314 * Update the mode and cursor while in an IM callback.
4321 old_vgetc_busy
= vgetc_busy
;
4324 vgetc_busy
= old_vgetc_busy
;
4330 * Callback invoked when the user finished preediting.
4331 * Put the final string into the input buffer.
4334 im_commit_cb(GtkIMContext
*context UNUSED
,
4336 gpointer data UNUSED
)
4338 int slen
= (int)STRLEN(str
);
4339 int add_to_input
= TRUE
;
4342 int commit_with_preedit
= TRUE
;
4346 xim_log("im_commit_cb(): %s\n", str
);
4349 /* The imhangul module doesn't reset the preedit string before
4350 * committing. Call im_delete_preedit() to work around that. */
4351 im_delete_preedit();
4353 /* Indicate that preediting has finished. */
4354 if (preedit_start_col
== MAXCOL
)
4356 init_preedit_start_col();
4357 commit_with_preedit
= FALSE
;
4360 /* The thing which setting "preedit_start_col" to MAXCOL means that
4361 * "preedit_start_col" will be set forcely when calling
4362 * preedit_changed_cb() next time.
4363 * "preedit_start_col" should not reset with MAXCOL on this part. Vim
4364 * is simulating the preediting by using add_to_input_str(). when
4365 * preedit begin immediately before committed, the typebuf is not
4366 * flushed to screen, then it can't get correct "preedit_start_col".
4367 * Thus, it should calculate the cells by adding cells of the committed
4369 if (input_conv
.vc_type
!= CONV_NONE
)
4371 im_str
= string_convert(&input_conv
, (char_u
*)str
, &len
);
4372 g_return_if_fail(im_str
!= NULL
);
4375 im_str
= (char_u
*)str
;
4377 for (p
= im_str
; p
< im_str
+ len
; p
+= (*mb_ptr2len
)(p
))
4378 clen
+= (*mb_ptr2cells
)(p
);
4379 if (input_conv
.vc_type
!= CONV_NONE
)
4381 preedit_start_col
+= clen
;
4383 /* Is this a single character that matches a keypad key that's just
4384 * been pressed? If so, we don't want it to be entered as such - let
4385 * us carry on processing the raw keycode so that it may be used in
4386 * mappings as <kSomething>. */
4387 if (xim_expected_char
!= NUL
)
4389 /* We're currently processing a keypad or other special key */
4390 if (slen
== 1 && str
[0] == xim_expected_char
)
4392 /* It's a match - don't do it here */
4393 xim_ignored_char
= TRUE
;
4394 add_to_input
= FALSE
;
4399 xim_ignored_char
= FALSE
;
4404 im_add_to_input((char_u
*)str
, slen
);
4406 /* Inserting chars while "im_is_active" is set does not cause a change of
4407 * buffer. When the chars are committed the buffer must be marked as
4409 if (!commit_with_preedit
)
4410 preedit_start_col
= MAXCOL
;
4412 /* This flag is used in changed() at next call. */
4413 xim_changed_while_preediting
= TRUE
;
4415 if (gtk_main_level() > 0)
4420 * Callback invoked after start to the preedit.
4423 im_preedit_start_cb(GtkIMContext
*context UNUSED
, gpointer data UNUSED
)
4426 xim_log("im_preedit_start_cb()\n");
4429 im_is_active
= TRUE
;
4430 preedit_is_active
= TRUE
;
4431 gui_update_cursor(TRUE
, FALSE
);
4436 * Callback invoked after end to the preedit.
4439 im_preedit_end_cb(GtkIMContext
*context UNUSED
, gpointer data UNUSED
)
4442 xim_log("im_preedit_end_cb()\n");
4444 im_delete_preedit();
4446 /* Indicate that preediting has finished */
4447 preedit_start_col
= MAXCOL
;
4448 xim_has_preediting
= FALSE
;
4451 /* Removal of this line suggested by Takuhiro Nishioka. Fixes that IM was
4452 * switched off unintentionally. We now use preedit_is_active (added by
4454 im_is_active
= FALSE
;
4456 preedit_is_active
= FALSE
;
4457 gui_update_cursor(TRUE
, FALSE
);
4462 * Callback invoked after changes to the preedit string. If the preedit
4463 * string was empty before, remember the preedit start column so we know
4464 * where to apply feedback attributes. Delete the previous preedit string
4465 * if there was one, save the new preedit cursor offset, and put the new
4466 * string into the input buffer.
4468 * TODO: The pragmatic "put into input buffer" approach used here has
4469 * several fundamental problems:
4471 * - The characters in the preedit string are subject to remapping.
4472 * That's broken, only the finally committed string should be remapped.
4474 * - There is a race condition involved: The retrieved value for the
4475 * current cursor position will be wrong if any unprocessed characters
4476 * are still queued in the input buffer.
4478 * - Due to the lack of synchronization between the file buffer in memory
4479 * and any typed characters, it's practically impossible to implement the
4480 * "retrieve_surrounding" and "delete_surrounding" signals reliably. IM
4481 * modules for languages such as Thai are likely to rely on this feature
4482 * for proper operation.
4484 * Conclusions: I think support for preediting needs to be moved to the
4485 * core parts of Vim. Ideally, until it has been committed, the preediting
4486 * string should only be displayed and not affect the buffer content at all.
4487 * The question how to deal with the synchronization issue still remains.
4488 * Circumventing the input buffer is probably not desirable. Anyway, I think
4489 * implementing "retrieve_surrounding" is the only hard problem.
4491 * One way to solve all of this in a clean manner would be to queue all key
4492 * press/release events "as is" in the input buffer, and apply the IM filtering
4493 * at the receiving end of the queue. This, however, would have a rather large
4494 * impact on the code base. If there is an easy way to force processing of all
4495 * remaining input from within the "retrieve_surrounding" signal handler, this
4496 * might not be necessary. Gotta ask on vim-dev for opinions.
4499 im_preedit_changed_cb(GtkIMContext
*context
, gpointer data UNUSED
)
4501 char *preedit_string
= NULL
;
4502 int cursor_index
= 0;
4503 int num_move_back
= 0;
4508 gtk_im_context_get_preedit_string(context
,
4509 &preedit_string
, NULL
,
4513 xim_log("im_preedit_changed_cb(): %s\n", preedit_string
);
4516 g_return_if_fail(preedit_string
!= NULL
); /* just in case */
4518 /* If preedit_start_col is MAXCOL set it to the current cursor position. */
4519 if (preedit_start_col
== MAXCOL
&& preedit_string
[0] != '\0')
4521 xim_has_preediting
= TRUE
;
4523 /* Urgh, this breaks if the input buffer isn't empty now */
4524 init_preedit_start_col();
4526 else if (cursor_index
== 0 && preedit_string
[0] == '\0')
4528 xim_has_preediting
= FALSE
;
4530 /* If at the start position (after typing backspace)
4531 * preedit_start_col must be reset. */
4532 preedit_start_col
= MAXCOL
;
4535 im_delete_preedit();
4538 * Compute the end of the preediting area: "preedit_end_col".
4539 * According to the documentation of gtk_im_context_get_preedit_string(),
4540 * the cursor_pos output argument returns the offset in bytes. This is
4541 * unfortunately not true -- real life shows the offset is in characters,
4542 * and the GTK+ source code agrees with me. Will file a bug later.
4544 if (preedit_start_col
!= MAXCOL
)
4545 preedit_end_col
= preedit_start_col
;
4546 str
= (char_u
*)preedit_string
;
4547 for (p
= str
, i
= 0; *p
!= NUL
; p
+= utf_byte2len(*p
), ++i
)
4551 is_composing
= ((*p
& 0x80) != 0 && utf_iscomposing(utf_ptr2char(p
)));
4553 * These offsets are used as counters when generating <BS> and <Del>
4554 * to delete the preedit string. So don't count composing characters
4555 * unless 'delcombine' is enabled.
4557 if (!is_composing
|| p_deco
)
4559 if (i
< cursor_index
)
4560 ++im_preedit_cursor
;
4562 ++im_preedit_trailing
;
4564 if (!is_composing
&& i
>= cursor_index
)
4566 /* This is essentially the same as im_preedit_trailing, except
4567 * composing characters are not counted even if p_deco is set. */
4570 if (preedit_start_col
!= MAXCOL
)
4571 preedit_end_col
+= utf_ptr2cells(p
);
4576 im_add_to_input(str
, (int)(p
- str
));
4577 im_correct_cursor(num_move_back
);
4580 g_free(preedit_string
);
4582 if (gtk_main_level() > 0)
4587 * Translate the Pango attributes at iter to Vim highlighting attributes.
4588 * Ignore attributes not supported by Vim highlighting. This shouldn't have
4589 * too much impact -- right now we handle even more attributes than necessary
4590 * for the IM modules I tested with.
4593 translate_pango_attributes(PangoAttrIterator
*iter
)
4595 PangoAttribute
*attr
;
4596 int char_attr
= HL_NORMAL
;
4598 attr
= pango_attr_iterator_get(iter
, PANGO_ATTR_UNDERLINE
);
4599 if (attr
!= NULL
&& ((PangoAttrInt
*)attr
)->value
4600 != (int)PANGO_UNDERLINE_NONE
)
4601 char_attr
|= HL_UNDERLINE
;
4603 attr
= pango_attr_iterator_get(iter
, PANGO_ATTR_WEIGHT
);
4604 if (attr
!= NULL
&& ((PangoAttrInt
*)attr
)->value
>= (int)PANGO_WEIGHT_BOLD
)
4605 char_attr
|= HL_BOLD
;
4607 attr
= pango_attr_iterator_get(iter
, PANGO_ATTR_STYLE
);
4608 if (attr
!= NULL
&& ((PangoAttrInt
*)attr
)->value
4609 != (int)PANGO_STYLE_NORMAL
)
4610 char_attr
|= HL_ITALIC
;
4612 attr
= pango_attr_iterator_get(iter
, PANGO_ATTR_BACKGROUND
);
4615 const PangoColor
*color
= &((PangoAttrColor
*)attr
)->color
;
4617 /* Assume inverse if black background is requested */
4618 if ((color
->red
| color
->green
| color
->blue
) == 0)
4619 char_attr
|= HL_INVERSE
;
4626 * Retrieve the highlighting attributes at column col in the preedit string.
4627 * Return -1 if not in preediting mode or if col is out of range.
4630 im_get_feedback_attr(int col
)
4632 char *preedit_string
= NULL
;
4633 PangoAttrList
*attr_list
= NULL
;
4639 gtk_im_context_get_preedit_string(xic
, &preedit_string
, &attr_list
, NULL
);
4641 if (preedit_string
!= NULL
&& attr_list
!= NULL
)
4645 /* Get the byte index as used by PangoAttrIterator */
4646 for (idx
= 0; col
> 0 && preedit_string
[idx
] != '\0'; --col
)
4647 idx
+= utfc_ptr2len((char_u
*)preedit_string
+ idx
);
4649 if (preedit_string
[idx
] != '\0')
4651 PangoAttrIterator
*iter
;
4654 char_attr
= HL_NORMAL
;
4655 iter
= pango_attr_list_get_iterator(attr_list
);
4657 /* Extract all relevant attributes from the list. */
4660 pango_attr_iterator_range(iter
, &start
, &end
);
4662 if (idx
>= start
&& idx
< end
)
4663 char_attr
|= translate_pango_attributes(iter
);
4665 while (pango_attr_iterator_next(iter
));
4667 pango_attr_iterator_destroy(iter
);
4671 if (attr_list
!= NULL
)
4672 pango_attr_list_unref(attr_list
);
4673 g_free(preedit_string
);
4682 xim_log("xim_init()\n");
4685 g_return_if_fail(gui
.drawarea
!= NULL
);
4686 g_return_if_fail(gui
.drawarea
->window
!= NULL
);
4688 xic
= gtk_im_multicontext_new();
4691 im_commit_handler_id
= g_signal_connect(G_OBJECT(xic
), "commit",
4692 G_CALLBACK(&im_commit_cb
), NULL
);
4693 g_signal_connect(G_OBJECT(xic
), "preedit_changed",
4694 G_CALLBACK(&im_preedit_changed_cb
), NULL
);
4695 g_signal_connect(G_OBJECT(xic
), "preedit_start",
4696 G_CALLBACK(&im_preedit_start_cb
), NULL
);
4697 g_signal_connect(G_OBJECT(xic
), "preedit_end",
4698 G_CALLBACK(&im_preedit_end_cb
), NULL
);
4700 gtk_im_context_set_client_window(xic
, gui
.drawarea
->window
);
4707 xim_log("im_shutdown()\n");
4712 gtk_im_context_focus_out(xic
);
4713 g_object_unref(xic
);
4716 im_is_active
= FALSE
;
4717 im_commit_handler_id
= 0;
4718 preedit_start_col
= MAXCOL
;
4719 xim_has_preediting
= FALSE
;
4723 * Convert the string argument to keyval and state for GdkEventKey.
4724 * If str is valid return TRUE, otherwise FALSE.
4726 * See 'imactivatekey' for documentation of the format.
4729 im_string_to_keyval(const char *str
, unsigned int *keyval
, unsigned int *state
)
4731 const char *mods_end
;
4732 unsigned tmp_keyval
;
4733 unsigned tmp_state
= 0;
4735 mods_end
= strrchr(str
, '-');
4736 mods_end
= (mods_end
!= NULL
) ? mods_end
+ 1 : str
;
4738 /* Parse modifier keys */
4739 while (str
< mods_end
)
4743 case 'S': case 's': tmp_state
|= (unsigned)GDK_SHIFT_MASK
; break;
4744 case 'L': case 'l': tmp_state
|= (unsigned)GDK_LOCK_MASK
; break;
4745 case 'C': case 'c': tmp_state
|= (unsigned)GDK_CONTROL_MASK
;break;
4746 case '1': tmp_state
|= (unsigned)GDK_MOD1_MASK
; break;
4747 case '2': tmp_state
|= (unsigned)GDK_MOD2_MASK
; break;
4748 case '3': tmp_state
|= (unsigned)GDK_MOD3_MASK
; break;
4749 case '4': tmp_state
|= (unsigned)GDK_MOD4_MASK
; break;
4750 case '5': tmp_state
|= (unsigned)GDK_MOD5_MASK
; break;
4755 tmp_keyval
= gdk_keyval_from_name(str
);
4757 if (tmp_keyval
== 0 || tmp_keyval
== GDK_VoidSymbol
)
4761 *keyval
= tmp_keyval
;
4769 * Return TRUE if p_imak is valid, otherwise FALSE. As a special case, an
4770 * empty string is also regarded as valid.
4772 * Note: The numerical key value of p_imak is cached if it was valid; thus
4773 * boldly assuming im_xim_isvalid_imactivate() will always be called whenever
4774 * 'imak' changes. This is currently the case but not obvious -- should
4775 * probably rename the function for clarity.
4778 im_xim_isvalid_imactivate(void)
4780 if (p_imak
[0] == NUL
)
4782 im_activatekey_keyval
= GDK_VoidSymbol
;
4783 im_activatekey_state
= 0;
4787 return im_string_to_keyval((const char *)p_imak
,
4788 &im_activatekey_keyval
,
4789 &im_activatekey_state
);
4793 im_synthesize_keypress(unsigned int keyval
, unsigned int state
)
4797 # ifdef HAVE_GTK_MULTIHEAD
4798 event
= (GdkEventKey
*)gdk_event_new(GDK_KEY_PRESS
);
4799 g_object_ref(gui
.drawarea
->window
); /* unreffed by gdk_event_free() */
4801 event
= (GdkEventKey
*)g_malloc0((gulong
)sizeof(GdkEvent
));
4802 event
->type
= GDK_KEY_PRESS
;
4804 event
->window
= gui
.drawarea
->window
;
4805 event
->send_event
= TRUE
;
4806 event
->time
= GDK_CURRENT_TIME
;
4807 event
->state
= state
;
4808 event
->keyval
= keyval
;
4809 event
->hardware_keycode
= /* needed for XIM */
4810 XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event
->window
), (KeySym
)keyval
);
4812 event
->string
= NULL
;
4814 gtk_im_context_filter_keypress(xic
, event
);
4816 /* For consistency, also send the corresponding release event. */
4817 event
->type
= GDK_KEY_RELEASE
;
4818 event
->send_event
= FALSE
;
4819 gtk_im_context_filter_keypress(xic
, event
);
4821 # ifdef HAVE_GTK_MULTIHEAD
4822 gdk_event_free((GdkEvent
*)event
);
4834 * The third-party imhangul module (and maybe others too) ignores
4835 * gtk_im_context_reset() or at least doesn't reset the active state.
4836 * Thus sending imactivatekey would turn it off if it was on before,
4837 * which is clearly not what we want. Fortunately we can work around
4838 * that for imhangul by sending GDK_Escape, but I don't know if it
4839 * works with all IM modules that support an activation key :/
4841 * An alternative approach would be to destroy the IM context and
4842 * recreate it. But that means loading/unloading the IM module on
4843 * every mode switch, which causes a quite noticable delay even on
4844 * my rather fast box...
4846 * Moreover, there are some XIM which cannot respond to
4847 * im_synthesize_keypress(). we hope that they reset by
4850 if (im_activatekey_keyval
!= GDK_VoidSymbol
&& im_is_active
)
4851 im_synthesize_keypress(GDK_Escape
, 0U);
4853 gtk_im_context_reset(xic
);
4856 * HACK for Ami: This sequence of function calls makes Ami handle
4857 * the IM reset graciously, without breaking loads of other stuff.
4858 * It seems to force English mode as well, which is exactly what we
4859 * want because it makes the Ami status display work reliably.
4861 gtk_im_context_set_use_preedit(xic
, FALSE
);
4867 gtk_im_context_set_use_preedit(xic
, TRUE
);
4868 xim_set_focus(gui
.in_focus
);
4870 if (im_activatekey_keyval
!= GDK_VoidSymbol
)
4874 g_signal_handler_block(xic
, im_commit_handler_id
);
4875 im_synthesize_keypress(im_activatekey_keyval
,
4876 im_activatekey_state
);
4877 g_signal_handler_unblock(xic
, im_commit_handler_id
);
4884 xim_set_focus(gui
.in_focus
);
4889 preedit_start_col
= MAXCOL
;
4890 xim_has_preediting
= FALSE
;
4894 xim_queue_key_press_event(GdkEventKey
*event
, int down
)
4899 * Workaround GTK2 XIM 'feature' that always converts keypad keys to
4900 * chars., even when not part of an IM sequence (ref. feature of
4902 * Flag any keypad keys that might represent a single char.
4903 * If this (on its own - i.e., not part of an IM sequence) is
4904 * committed while we're processing one of these keys, we can ignore
4905 * that commit and go ahead & process it ourselves. That way we can
4906 * still distinguish keypad keys for use in mappings.
4907 * Also add GDK_space to make <S-Space> work.
4909 switch (event
->keyval
)
4911 case GDK_KP_Add
: xim_expected_char
= '+'; break;
4912 case GDK_KP_Subtract
: xim_expected_char
= '-'; break;
4913 case GDK_KP_Divide
: xim_expected_char
= '/'; break;
4914 case GDK_KP_Multiply
: xim_expected_char
= '*'; break;
4915 case GDK_KP_Decimal
: xim_expected_char
= '.'; break;
4916 case GDK_KP_Equal
: xim_expected_char
= '='; break;
4917 case GDK_KP_0
: xim_expected_char
= '0'; break;
4918 case GDK_KP_1
: xim_expected_char
= '1'; break;
4919 case GDK_KP_2
: xim_expected_char
= '2'; break;
4920 case GDK_KP_3
: xim_expected_char
= '3'; break;
4921 case GDK_KP_4
: xim_expected_char
= '4'; break;
4922 case GDK_KP_5
: xim_expected_char
= '5'; break;
4923 case GDK_KP_6
: xim_expected_char
= '6'; break;
4924 case GDK_KP_7
: xim_expected_char
= '7'; break;
4925 case GDK_KP_8
: xim_expected_char
= '8'; break;
4926 case GDK_KP_9
: xim_expected_char
= '9'; break;
4927 case GDK_space
: xim_expected_char
= ' '; break;
4928 default: xim_expected_char
= NUL
;
4930 xim_ignored_char
= FALSE
;
4934 * When typing fFtT, XIM may be activated. Thus it must pass
4935 * gtk_im_context_filter_keypress() in Normal mode.
4936 * And while doing :sh too.
4938 if (xic
!= NULL
&& !p_imdisable
4939 && (State
& (INSERT
| CMDLINE
| NORMAL
| EXTERNCMD
)) != 0)
4942 * Filter 'imactivatekey' and map it to CTRL-^. This way, Vim is
4943 * always aware of the current status of IM, and can even emulate
4944 * the activation key for modules that don't support one.
4946 if (event
->keyval
== im_activatekey_keyval
4947 && (event
->state
& im_activatekey_state
) == im_activatekey_state
)
4949 unsigned int state_mask
;
4951 /* Require the state of the 3 most used modifiers to match exactly.
4952 * Otherwise e.g. <S-C-space> would be unusable for other purposes
4953 * if the IM activate key is <S-space>. */
4954 state_mask
= im_activatekey_state
;
4955 state_mask
|= ((int)GDK_SHIFT_MASK
| (int)GDK_CONTROL_MASK
4956 | (int)GDK_MOD1_MASK
);
4958 if ((event
->state
& state_mask
) != im_activatekey_state
)
4961 /* Don't send it a second time on GDK_KEY_RELEASE. */
4962 if (event
->type
!= GDK_KEY_PRESS
)
4965 if (map_to_exists_mode((char_u
*)"", LANGMAP
, FALSE
))
4967 im_set_active(FALSE
);
4969 /* ":lmap" mappings exists, toggle use of mappings. */
4971 if (State
& LANGMAP
)
4973 curbuf
->b_p_iminsert
= B_IMODE_NONE
;
4978 curbuf
->b_p_iminsert
= B_IMODE_LMAP
;
4984 return gtk_im_context_filter_keypress(xic
, event
);
4987 /* Don't filter events through the IM context if IM isn't active
4988 * right now. Unlike with GTK+ 1.2 we cannot rely on the IM module
4989 * not doing anything before the activation key was sent. */
4990 if (im_activatekey_keyval
== GDK_VoidSymbol
|| im_is_active
)
4992 int imresult
= gtk_im_context_filter_keypress(xic
, event
);
4994 /* Some XIM send following sequence:
4995 * 1. preedited string.
4996 * 2. committed string.
4997 * 3. line changed key.
4998 * 4. preedited string.
4999 * 5. remove preedited string.
5000 * if 3, Vim can't move back the above line for 5.
5001 * thus, this part should not parse the key. */
5002 if (!imresult
&& preedit_start_col
!= MAXCOL
5003 && event
->keyval
== GDK_Return
)
5005 im_synthesize_keypress(GDK_Return
, 0U);
5009 /* If XIM tried to commit a keypad key as a single char.,
5010 * ignore it so we can use the keypad key 'raw', for mappings. */
5011 if (xim_expected_char
!= NUL
&& xim_ignored_char
)
5012 /* We had a keypad key, and XIM tried to thieve it */
5015 /* Normal processing */
5026 return im_is_active
;
5029 # else /* !HAVE_GTK2 */
5031 static int xim_is_active
= FALSE
; /* XIM should be active in the current
5033 static int xim_has_focus
= FALSE
; /* XIM is really being used for Vim */
5035 static XIMStyle input_style
;
5036 static int status_area_enabled
= TRUE
;
5041 # include <gdk/gdkwin32.h>
5043 # include <gdk/gdkx.h>
5047 /* Define a few things to be able to generate prototypes while not configured
5050 # define gboolean int
5051 typedef int GdkEvent
;
5052 typedef int GdkEventKey
;
5057 #if defined(FEAT_GUI_GTK) || defined(PROTO)
5058 static int preedit_buf_len
= 0;
5059 static int xim_can_preediting
INIT(= FALSE
); /* XIM in showmode() */
5060 static int xim_input_style
;
5061 #ifndef FEAT_GUI_GTK
5062 # define gboolean int
5064 static gboolean use_status_area
= 0;
5066 static int im_xim_str2keycode
__ARGS((unsigned int *code
, unsigned int *state
));
5067 static void im_xim_send_event_imactivate
__ARGS((void));
5070 * Convert string to keycode and state for XKeyEvent.
5071 * When string is valid return OK, when invalid return FAIL.
5073 * See 'imactivatekey' documentation for the format.
5076 im_xim_str2keycode(code
, state
)
5078 unsigned int *state
;
5082 unsigned keycode
= 0, keystate
= 0;
5090 len
= STRLEN(p_imak
);
5091 for (flag_end
= p_imak
+ len
- 1;
5092 flag_end
> p_imak
&& *flag_end
!= '-'; --flag_end
)
5095 /* Parse modifier keys */
5096 for (str
= p_imak
; str
< flag_end
; ++str
)
5101 keystate
|= ShiftMask
;
5104 keystate
|= LockMask
;
5107 keystate
|= ControlMask
;
5110 keystate
|= Mod1Mask
;
5113 keystate
|= Mod2Mask
;
5116 keystate
|= Mod3Mask
;
5119 keystate
|= Mod4Mask
;
5122 keystate
|= Mod5Mask
;
5133 /* Get keycode from string. */
5134 gui_get_x11_windis(&window
, &display
);
5136 keycode
= XKeysymToKeycode(display
, XStringToKeysym((char *)str
));
5149 im_xim_send_event_imactivate()
5151 /* Force turn on preedit state by symulate keypress event.
5152 * Keycode and state is specified by 'imactivatekey'.
5156 gui_get_x11_windis(&ev
.window
, &ev
.display
);
5157 ev
.root
= RootWindow(ev
.display
, DefaultScreen(ev
.display
));
5158 ev
.subwindow
= None
;
5159 ev
.time
= CurrentTime
;
5166 if (im_xim_str2keycode(&ev
.keycode
, &ev
.state
) == OK
)
5167 XSendEvent(ev
.display
, ev
.window
, 1, KeyPressMask
, (XEvent
*)&ev
);
5171 * Return TRUE if 'imactivatekey' has a valid value.
5174 im_xim_isvalid_imactivate()
5176 return im_xim_str2keycode(NULL
, NULL
) == OK
;
5178 #endif /* FEAT_GUI_GTK */
5181 * Switch using XIM on/off. This is used by the code that changes "State".
5184 im_set_active(active
)
5190 /* If 'imdisable' is set, XIM is never active. */
5193 #if !defined (FEAT_GUI_GTK)
5194 else if (input_style
& XIMPreeditPosition
)
5195 /* There is a problem in switching XIM off when preediting is used,
5196 * and it is not clear how this can be solved. For now, keep XIM on
5197 * all the time, like it was done in Vim 5.8. */
5201 /* Remember the active state, it is needed when Vim gets keyboard focus. */
5202 xim_is_active
= active
;
5205 /* When 'imactivatekey' has valid key-string, try to control XIM preedit
5206 * state. When 'imactivatekey' has no or invalid string, try old XIM
5212 * Destroy old Input Context (XIC), and create new one. New XIC
5213 * would have a state of preedit that is off. When argument:active
5214 * is false, that's all. Else argument:active is true, send a key
5215 * event specified by 'imactivatekey' to activate XIM preedit state.
5218 xim_is_active
= TRUE
; /* Disable old XIM focus control */
5219 /* If we can monitor preedit state with preedit callback functions,
5220 * try least creation of new XIC.
5222 if (xim_input_style
& (int)GDK_IM_PREEDIT_CALLBACKS
)
5224 if (xim_can_preediting
&& !active
)
5226 /* Force turn off preedit state. With some IM
5227 * implementations, we cannot turn off preedit state by
5228 * symulate keypress event. It is why using such a method
5229 * that destroy old IC (input context), and create new one.
5230 * When create new IC, its preedit state is usually off.
5233 xim_set_focus(FALSE
);
5234 gdk_ic_destroy(xic
);
5236 xim_can_preediting
= FALSE
;
5238 else if (!xim_can_preediting
&& active
)
5239 im_xim_send_event_imactivate();
5243 /* First, force destroy old IC, and create new one. It
5244 * symulates "turning off preedit state".
5246 xim_set_focus(FALSE
);
5247 gdk_ic_destroy(xic
);
5249 xim_can_preediting
= FALSE
;
5251 /* 2nd, when requested to activate IM, symulate this by sending
5256 im_xim_send_event_imactivate();
5257 xim_can_preediting
= TRUE
;
5263 # ifndef XIMPreeditUnKnown
5264 /* X11R5 doesn't have these, it looks safe enough to define here. */
5265 typedef unsigned long XIMPreeditState
;
5266 # define XIMPreeditUnKnown 0L
5267 # define XIMPreeditEnable 1L
5268 # define XIMPreeditDisable (1L<<1)
5269 # define XNPreeditState "preeditState"
5271 XIMPreeditState preedit_state
= XIMPreeditUnKnown
;
5272 XVaNestedList preedit_attr
;
5275 preedit_attr
= XVaCreateNestedList(0,
5276 XNPreeditState
, &preedit_state
,
5278 pxic
= ((GdkICPrivate
*)xic
)->xic
;
5280 if (!XGetICValues(pxic
, XNPreeditAttributes
, preedit_attr
, NULL
))
5282 XFree(preedit_attr
);
5283 preedit_attr
= XVaCreateNestedList(0,
5285 active
? XIMPreeditEnable
: XIMPreeditDisable
,
5287 XSetICValues(pxic
, XNPreeditAttributes
, preedit_attr
, NULL
);
5288 xim_can_preediting
= active
;
5289 xim_is_active
= active
;
5291 XFree(preedit_attr
);
5293 if (xim_input_style
& XIMPreeditCallbacks
)
5295 preedit_buf_len
= 0;
5296 init_preedit_start_col();
5300 /* When had tested kinput2 + canna + Athena GUI version with
5301 * 'imactivatekey' is "s-space", im_xim_send_event_imactivate() did not
5302 * work correctly. It just inserted one space. I don't know why we
5303 * couldn't switch state of XIM preediting. This is reason why these
5304 * codes are commented out.
5306 /* First, force destroy old IC, and create new one. It symulates
5307 * "turning off preedit state".
5309 xim_set_focus(FALSE
);
5314 /* 2nd, when requested to activate IM, symulate this by sending the
5318 im_xim_send_event_imactivate();
5325 * Adjust using XIM for gaining or losing keyboard focus. Also called when
5326 * "xim_is_active" changes.
5329 xim_set_focus(focus
)
5336 * XIM only gets focus when the Vim window has keyboard focus and XIM has
5337 * been set active for the current mode.
5339 if (focus
&& xim_is_active
)
5343 xim_has_focus
= TRUE
;
5345 gdk_im_begin(xic
, gui
.drawarea
->window
);
5355 xim_has_focus
= FALSE
;
5366 im_set_position(row
, col
)
5374 * Set the XIM to the current cursor position.
5382 xim_set_focus(TRUE
);
5396 # ifdef FEAT_XFONTSET
5397 if ((xim_input_style
& (int)GDK_IM_PREEDIT_POSITION
)
5398 && gui
.fontset
!= NOFONTSET
5399 && gui
.fontset
->type
== GDK_FONT_FONTSET
)
5403 if (attr
->spot_location
.y
>= 0)
5405 attr
->spot_location
.x
= 0;
5406 attr
->spot_location
.y
= -100;
5407 attrmask
|= (int)GDK_IC_SPOT_LOCATION
;
5414 if (attr
->spot_location
.x
!= TEXT_X(gui
.col
)
5415 || attr
->spot_location
.y
!= TEXT_Y(gui
.row
))
5417 attr
->spot_location
.x
= TEXT_X(gui
.col
);
5418 attr
->spot_location
.y
= TEXT_Y(gui
.row
);
5419 attrmask
|= (int)GDK_IC_SPOT_LOCATION
;
5422 gdk_window_get_size(gui
.drawarea
->window
, &width
, &height
);
5423 width
-= 2 * gui
.border_offset
;
5424 height
-= 2 * gui
.border_offset
;
5425 if (xim_input_style
& (int)GDK_IM_STATUS_AREA
)
5426 height
-= gui
.char_height
;
5427 if (attr
->preedit_area
.width
!= width
5428 || attr
->preedit_area
.height
!= height
)
5430 attr
->preedit_area
.x
= gui
.border_offset
;
5431 attr
->preedit_area
.y
= gui
.border_offset
;
5432 attr
->preedit_area
.width
= width
;
5433 attr
->preedit_area
.height
= height
;
5434 attrmask
|= (int)GDK_IC_PREEDIT_AREA
;
5437 if (attr
->preedit_fontset
!= gui
.current_font
)
5439 attr
->preedit_fontset
= gui
.current_font
;
5440 attrmask
|= (int)GDK_IC_PREEDIT_FONTSET
;
5444 # endif /* FEAT_XFONTSET */
5446 if (xim_fg_color
== INVALCOLOR
)
5448 xim_fg_color
= gui
.def_norm_pixel
;
5449 xim_bg_color
= gui
.def_back_pixel
;
5451 if (attr
->preedit_foreground
.pixel
!= xim_fg_color
)
5453 attr
->preedit_foreground
.pixel
= xim_fg_color
;
5454 attrmask
|= (int)GDK_IC_PREEDIT_FOREGROUND
;
5456 if (attr
->preedit_background
.pixel
!= xim_bg_color
)
5458 attr
->preedit_background
.pixel
= xim_bg_color
;
5459 attrmask
|= (int)GDK_IC_PREEDIT_BACKGROUND
;
5463 gdk_ic_set_attr(xic
, attr
, (GdkICAttributesType
)attrmask
);
5465 #else /* FEAT_GUI_GTK */
5467 XVaNestedList attr_list
;
5468 XRectangle spot_area
;
5474 /* hide XIM cursor */
5476 over_spot
.y
= -100; /* arbitrary invisible position */
5477 attr_list
= (XVaNestedList
) XVaCreateNestedList(0,
5481 XSetICValues(xic
, XNPreeditAttributes
, attr_list
, NULL
);
5486 if (input_style
& XIMPreeditPosition
)
5488 if (xim_fg_color
== INVALCOLOR
)
5490 xim_fg_color
= gui
.def_norm_pixel
;
5491 xim_bg_color
= gui
.def_back_pixel
;
5493 over_spot
.x
= TEXT_X(gui
.col
);
5494 over_spot
.y
= TEXT_Y(gui
.row
);
5497 spot_area
.height
= gui
.char_height
* Rows
;
5498 spot_area
.width
= gui
.char_width
* Columns
;
5499 line_space
= gui
.char_height
;
5500 attr_list
= (XVaNestedList
) XVaCreateNestedList(0,
5501 XNSpotLocation
, &over_spot
,
5502 XNForeground
, (Pixel
) xim_fg_color
,
5503 XNBackground
, (Pixel
) xim_bg_color
,
5505 XNLineSpace
, line_space
,
5507 if (XSetICValues(xic
, XNPreeditAttributes
, attr_list
, NULL
))
5508 EMSG(_("E284: Cannot set IC values"));
5512 #endif /* FEAT_GUI_GTK */
5516 * Set up the status area.
5518 * This should use a separate Widget, but that seems not possible, because
5519 * preedit_area and status_area should be set to the same window as for the
5520 * text input. Unfortunately this means the status area pollutes the text
5524 xim_set_status_area()
5530 # if defined(FEAT_XFONTSET)
5531 if (use_status_area
)
5544 style
= (int)gdk_ic_get_style(xic
);
5545 if ((style
& (int)GDK_IM_STATUS_MASK
) == (int)GDK_IM_STATUS_AREA
)
5547 if (gui
.fontset
!= NOFONTSET
5548 && gui
.fontset
->type
== GDK_FONT_FONTSET
)
5550 widget
= gui
.mainwin
;
5551 gdk_window_get_size(widget
->window
, &width
, &height
);
5553 attrmask
|= (int)GDK_IC_STATUS_AREA
;
5554 attr
->status_area
.x
= 0;
5555 attr
->status_area
.y
= height
- gui
.char_height
- 1;
5556 attr
->status_area
.width
= width
;
5557 attr
->status_area
.height
= gui
.char_height
;
5561 gdk_ic_set_attr(xic
, attr
, (GdkICAttributesType
)attrmask
);
5566 XVaNestedList preedit_list
= 0, status_list
= 0, list
= 0;
5567 XRectangle pre_area
, status_area
;
5569 if (input_style
& XIMStatusArea
)
5571 if (input_style
& XIMPreeditArea
)
5573 XRectangle
*needed_rect
;
5575 /* to get status_area width */
5576 status_list
= XVaCreateNestedList(0, XNAreaNeeded
,
5577 &needed_rect
, NULL
);
5578 XGetICValues(xic
, XNStatusAttributes
, status_list
, NULL
);
5581 status_area
.width
= needed_rect
->width
;
5584 status_area
.width
= gui
.char_width
* Columns
;
5587 status_area
.y
= gui
.char_height
* Rows
+ gui
.border_offset
;
5588 if (gui
.which_scrollbars
[SBAR_BOTTOM
])
5589 status_area
.y
+= gui
.scrollbar_height
;
5591 if (gui
.menu_is_active
)
5592 status_area
.y
+= gui
.menu_height
;
5594 status_area
.height
= gui
.char_height
;
5595 status_list
= XVaCreateNestedList(0, XNArea
, &status_area
, NULL
);
5600 status_area
.y
= gui
.char_height
* Rows
+ gui
.border_offset
;
5601 if (gui
.which_scrollbars
[SBAR_BOTTOM
])
5602 status_area
.y
+= gui
.scrollbar_height
;
5604 if (gui
.menu_is_active
)
5605 status_area
.y
+= gui
.menu_height
;
5607 status_area
.width
= 0;
5608 status_area
.height
= gui
.char_height
;
5611 if (input_style
& XIMPreeditArea
) /* off-the-spot */
5613 pre_area
.x
= status_area
.x
+ status_area
.width
;
5614 pre_area
.y
= gui
.char_height
* Rows
+ gui
.border_offset
;
5615 pre_area
.width
= gui
.char_width
* Columns
- pre_area
.x
;
5616 if (gui
.which_scrollbars
[SBAR_BOTTOM
])
5617 pre_area
.y
+= gui
.scrollbar_height
;
5619 if (gui
.menu_is_active
)
5620 pre_area
.y
+= gui
.menu_height
;
5622 pre_area
.height
= gui
.char_height
;
5623 preedit_list
= XVaCreateNestedList(0, XNArea
, &pre_area
, NULL
);
5625 else if (input_style
& XIMPreeditPosition
) /* over-the-spot */
5629 pre_area
.height
= gui
.char_height
* Rows
;
5630 pre_area
.width
= gui
.char_width
* Columns
;
5631 preedit_list
= XVaCreateNestedList(0, XNArea
, &pre_area
, NULL
);
5634 if (preedit_list
&& status_list
)
5635 list
= XVaCreateNestedList(0, XNPreeditAttributes
, preedit_list
,
5636 XNStatusAttributes
, status_list
, NULL
);
5637 else if (preedit_list
)
5638 list
= XVaCreateNestedList(0, XNPreeditAttributes
, preedit_list
,
5640 else if (status_list
)
5641 list
= XVaCreateNestedList(0, XNStatusAttributes
, status_list
,
5648 XSetICValues(xic
, XNVaNestedList
, list
, NULL
);
5654 XFree(preedit_list
);
5659 #if defined(FEAT_GUI_X11) || defined(FEAT_GUI_GTK)
5660 static char e_xim
[] = N_("E285: Failed to create input context");
5663 #if defined(FEAT_GUI_X11) || defined(PROTO)
5664 # if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(sun)
5665 # define USE_X11R6_XIM
5668 static int xim_real_init
__ARGS((Window x11_window
, Display
*x11_display
));
5671 #ifdef USE_X11R6_XIM
5672 static void xim_instantiate_cb
__ARGS((Display
*display
, XPointer client_data
, XPointer call_data
));
5673 static void xim_destroy_cb
__ARGS((XIM im
, XPointer client_data
, XPointer call_data
));
5676 xim_instantiate_cb(display
, client_data
, call_data
)
5678 XPointer client_data UNUSED
;
5679 XPointer call_data UNUSED
;
5682 Display
*x11_display
;
5685 xim_log("xim_instantiate_cb()\n");
5688 gui_get_x11_windis(&x11_window
, &x11_display
);
5689 if (display
!= x11_display
)
5692 xim_real_init(x11_window
, x11_display
);
5693 gui_set_shellsize(FALSE
, FALSE
, RESIZE_BOTH
);
5695 XUnregisterIMInstantiateCallback(x11_display
, NULL
, NULL
, NULL
,
5696 xim_instantiate_cb
, NULL
);
5700 xim_destroy_cb(im
, client_data
, call_data
)
5702 XPointer client_data UNUSED
;
5703 XPointer call_data UNUSED
;
5706 Display
*x11_display
;
5709 xim_log("xim_destroy_cb()\n");
5711 gui_get_x11_windis(&x11_window
, &x11_display
);
5714 status_area_enabled
= FALSE
;
5716 gui_set_shellsize(FALSE
, FALSE
, RESIZE_BOTH
);
5718 XRegisterIMInstantiateCallback(x11_display
, NULL
, NULL
, NULL
,
5719 xim_instantiate_cb
, NULL
);
5727 Display
*x11_display
;
5730 xim_log("xim_init()\n");
5733 gui_get_x11_windis(&x11_window
, &x11_display
);
5737 if (xim_real_init(x11_window
, x11_display
))
5740 gui_set_shellsize(FALSE
, FALSE
, RESIZE_BOTH
);
5742 #ifdef USE_X11R6_XIM
5743 XRegisterIMInstantiateCallback(x11_display
, NULL
, NULL
, NULL
,
5744 xim_instantiate_cb
, NULL
);
5749 xim_real_init(x11_window
, x11_display
)
5751 Display
*x11_display
;
5759 #define IMLEN_MAX 40
5760 char buf
[IMLEN_MAX
+ 7];
5762 XIMStyles
*xim_styles
;
5763 XIMStyle this_input_style
= 0;
5766 XVaNestedList preedit_list
, status_list
;
5769 status_area_enabled
= FALSE
;
5774 if (gui
.rsrc_input_method
!= NULL
&& *gui
.rsrc_input_method
!= NUL
)
5776 strcpy(tmp
, gui
.rsrc_input_method
);
5777 for (ns
= s
= tmp
; ns
!= NULL
&& *s
!= NUL
;)
5779 s
= (char *)skipwhite((char_u
*)s
);
5782 if ((ns
= end
= strchr(s
, ',')) == NULL
)
5783 end
= s
+ strlen(s
);
5784 while (isspace(((char_u
*)end
)[-1]))
5788 if (strlen(s
) <= IMLEN_MAX
)
5790 strcpy(buf
, "@im=");
5792 if ((p
= XSetLocaleModifiers(buf
)) != NULL
&& *p
!= NUL
5793 && (xim
= XOpenIM(x11_display
, NULL
, NULL
, NULL
))
5802 if (xim
== NULL
&& (p
= XSetLocaleModifiers("")) != NULL
&& *p
!= NUL
)
5803 xim
= XOpenIM(x11_display
, NULL
, NULL
, NULL
);
5805 /* This is supposed to be useful to obtain characters through
5806 * XmbLookupString() without really using a XIM. */
5807 if (xim
== NULL
&& (p
= XSetLocaleModifiers("@im=none")) != NULL
5809 xim
= XOpenIM(x11_display
, NULL
, NULL
, NULL
);
5813 /* Only give this message when verbose is set, because too many people
5814 * got this message when they didn't want to use a XIM. */
5818 EMSG(_("E286: Failed to open input method"));
5824 #ifdef USE_X11R6_XIM
5826 XIMCallback destroy_cb
;
5828 destroy_cb
.callback
= xim_destroy_cb
;
5829 destroy_cb
.client_data
= NULL
;
5830 if (XSetIMValues(xim
, XNDestroyCallback
, &destroy_cb
, NULL
))
5831 EMSG(_("E287: Warning: Could not set destroy callback to IM"));
5835 if (XGetIMValues(xim
, XNQueryInputStyle
, &xim_styles
, NULL
) || !xim_styles
)
5837 EMSG(_("E288: input method doesn't support any style"));
5843 strcpy(tmp
, gui
.rsrc_preedit_type_name
);
5844 for (s
= tmp
; s
&& !found
; )
5846 while (*s
&& isspace((unsigned char)*s
))
5850 if ((ns
= end
= strchr(s
, ',')) != 0)
5853 end
= s
+ strlen(s
);
5854 while (isspace((unsigned char)*end
))
5858 if (!strcmp(s
, "OverTheSpot"))
5859 this_input_style
= (XIMPreeditPosition
| XIMStatusArea
);
5860 else if (!strcmp(s
, "OffTheSpot"))
5861 this_input_style
= (XIMPreeditArea
| XIMStatusArea
);
5862 else if (!strcmp(s
, "Root"))
5863 this_input_style
= (XIMPreeditNothing
| XIMStatusNothing
);
5865 for (i
= 0; (unsigned short)i
< xim_styles
->count_styles
; i
++)
5867 if (this_input_style
== xim_styles
->supported_styles
[i
])
5874 for (i
= 0; (unsigned short)i
< xim_styles
->count_styles
; i
++)
5876 if ((xim_styles
->supported_styles
[i
] & this_input_style
)
5877 == (this_input_style
& ~XIMStatusArea
))
5879 this_input_style
&= ~XIMStatusArea
;
5891 /* Only give this message when verbose is set, because too many people
5892 * got this message when they didn't want to use a XIM. */
5896 EMSG(_("E289: input method doesn't support my preedit type"));
5903 over_spot
.x
= TEXT_X(gui
.col
);
5904 over_spot
.y
= TEXT_Y(gui
.row
);
5905 input_style
= this_input_style
;
5907 /* A crash was reported when trying to pass gui.norm_font as XNFontSet,
5908 * thus that has been removed. Hopefully the default works... */
5909 #ifdef FEAT_XFONTSET
5910 if (gui
.fontset
!= NOFONTSET
)
5912 preedit_list
= XVaCreateNestedList(0,
5913 XNSpotLocation
, &over_spot
,
5914 XNForeground
, (Pixel
)gui
.def_norm_pixel
,
5915 XNBackground
, (Pixel
)gui
.def_back_pixel
,
5916 XNFontSet
, (XFontSet
)gui
.fontset
,
5918 status_list
= XVaCreateNestedList(0,
5919 XNForeground
, (Pixel
)gui
.def_norm_pixel
,
5920 XNBackground
, (Pixel
)gui
.def_back_pixel
,
5921 XNFontSet
, (XFontSet
)gui
.fontset
,
5927 preedit_list
= XVaCreateNestedList(0,
5928 XNSpotLocation
, &over_spot
,
5929 XNForeground
, (Pixel
)gui
.def_norm_pixel
,
5930 XNBackground
, (Pixel
)gui
.def_back_pixel
,
5932 status_list
= XVaCreateNestedList(0,
5933 XNForeground
, (Pixel
)gui
.def_norm_pixel
,
5934 XNBackground
, (Pixel
)gui
.def_back_pixel
,
5938 xic
= XCreateIC(xim
,
5939 XNInputStyle
, input_style
,
5940 XNClientWindow
, x11_window
,
5941 XNFocusWindow
, gui
.wid
,
5942 XNPreeditAttributes
, preedit_list
,
5943 XNStatusAttributes
, status_list
,
5946 XFree(preedit_list
);
5949 if (input_style
& XIMStatusArea
)
5951 xim_set_status_area();
5952 status_area_enabled
= TRUE
;
5955 gui_set_shellsize(FALSE
, FALSE
, RESIZE_BOTH
);
5967 #endif /* FEAT_GUI_X11 */
5969 #if defined(FEAT_GUI_GTK) || defined(PROTO)
5971 # ifdef FEAT_XFONTSET
5972 static char e_overthespot
[] = N_("E290: over-the-spot style requires fontset");
5980 xim_decide_input_style()
5982 /* GDK_IM_STATUS_CALLBACKS was disabled, enabled it to allow Japanese
5984 int supported_style
= (int)GDK_IM_PREEDIT_NONE
|
5985 (int)GDK_IM_PREEDIT_NOTHING
|
5986 (int)GDK_IM_PREEDIT_POSITION
|
5987 (int)GDK_IM_PREEDIT_CALLBACKS
|
5988 (int)GDK_IM_STATUS_CALLBACKS
|
5989 (int)GDK_IM_STATUS_AREA
|
5990 (int)GDK_IM_STATUS_NONE
|
5991 (int)GDK_IM_STATUS_NOTHING
;
5994 xim_log("xim_decide_input_style()\n");
5997 if (!gdk_im_ready())
5998 xim_input_style
= 0;
6001 if (gtk_major_version
> 1
6002 || (gtk_major_version
== 1
6003 && (gtk_minor_version
> 2
6004 || (gtk_minor_version
== 2 && gtk_micro_version
>= 3))))
6005 use_status_area
= TRUE
;
6008 EMSG(_("E291: Your GTK+ is older than 1.2.3. Status area disabled"));
6009 use_status_area
= FALSE
;
6011 #ifdef FEAT_XFONTSET
6012 if (gui
.fontset
== NOFONTSET
|| gui
.fontset
->type
!= GDK_FONT_FONTSET
)
6014 supported_style
&= ~((int)GDK_IM_PREEDIT_POSITION
6015 | (int)GDK_IM_STATUS_AREA
);
6016 if (!use_status_area
)
6017 supported_style
&= ~(int)GDK_IM_STATUS_AREA
;
6018 xim_input_style
= (int)gdk_im_decide_style((GdkIMStyle
)supported_style
);
6023 preedit_start_cbproc(XIC thexic UNUSED
,
6024 XPointer client_data UNUSED
,
6025 XPointer call_data UNUSED
)
6028 xim_log("xim_decide_input_style()\n");
6031 draw_feedback
= NULL
;
6032 xim_can_preediting
= TRUE
;
6033 xim_has_preediting
= TRUE
;
6034 gui_update_cursor(TRUE
, FALSE
);
6043 xim_back_delete(int n
)
6051 add_to_input_buf(str
, 3);
6054 static GSList
*key_press_event_queue
= NULL
;
6055 static gboolean processing_queued_event
= FALSE
;
6058 preedit_draw_cbproc(XIC thexic UNUSED
,
6059 XPointer client_data UNUSED
,
6062 XIMPreeditDrawCallbackStruct
*draw_data
;
6065 GSList
*event_queue
;
6068 xim_log("preedit_draw_cbproc()\n");
6071 draw_data
= (XIMPreeditDrawCallbackStruct
*) call_data
;
6072 text
= (XIMText
*) draw_data
->text
;
6074 if ((text
== NULL
&& draw_data
->chg_length
== preedit_buf_len
)
6075 || preedit_buf_len
== 0)
6077 init_preedit_start_col();
6078 vim_free(draw_feedback
);
6079 draw_feedback
= NULL
;
6081 if (draw_data
->chg_length
> 0)
6085 if (draw_data
->chg_length
> preedit_buf_len
)
6086 bs_cnt
= preedit_buf_len
;
6088 bs_cnt
= draw_data
->chg_length
;
6089 xim_back_delete(bs_cnt
);
6090 preedit_buf_len
-= bs_cnt
;
6097 unsigned int nfeedback
= 0;
6101 src
= text
->string
.multi_byte
;
6102 if (src
!= NULL
&& !text
->encoding_is_wchar
)
6105 ptr
= (char_u
*)src
;
6106 /* Avoid the enter for decision */
6111 if (input_conv
.vc_type
!= CONV_NONE
6112 && (buf
= string_convert(&input_conv
,
6113 (char_u
*)src
, &len
)) != NULL
)
6115 /* Converted from 'termencoding' to 'encoding'. */
6116 add_to_input_buf_csi(buf
, len
);
6121 add_to_input_buf_csi((char_u
*)src
, len
);
6122 /* Add count of character to preedit_buf_len */
6126 if (draw_data
->text
->feedback
!= NULL
)
6128 if (draw_feedback
== NULL
)
6129 draw_feedback
= (char *)alloc(draw_data
->chg_first
6132 draw_feedback
= vim_realloc(draw_feedback
,
6133 draw_data
->chg_first
+ text
->length
);
6134 if (draw_feedback
!= NULL
)
6136 draw_feedback
[nfeedback
+ draw_data
->chg_first
]
6137 = draw_data
->text
->feedback
[nfeedback
];
6142 ptr
+= (*mb_ptr2len
)(ptr
);
6151 preedit_end_col
= MAXCOL
;
6154 if (text
!= NULL
|| draw_data
->chg_length
> 0)
6156 event_queue
= key_press_event_queue
;
6157 processing_queued_event
= TRUE
;
6158 while (event_queue
!= NULL
&& processing_queued_event
)
6160 GdkEvent
*ev
= event_queue
->data
;
6163 gtk_signal_emit_by_name((GtkObject
*)gui
.mainwin
, "key_press_event",
6166 event_queue
= event_queue
->next
;
6168 processing_queued_event
= FALSE
;
6169 if (key_press_event_queue
)
6171 g_slist_free(key_press_event_queue
);
6172 key_press_event_queue
= NULL
;
6175 if (gtk_main_level() > 0)
6180 * Retrieve the highlighting attributes at column col in the preedit string.
6181 * Return -1 if not in preediting mode or if col is out of range.
6184 im_get_feedback_attr(int col
)
6186 if (draw_feedback
!= NULL
&& col
< preedit_buf_len
)
6188 if (draw_feedback
[col
] & XIMReverse
)
6190 else if (draw_feedback
[col
] & XIMUnderline
)
6191 return HL_UNDERLINE
;
6193 return hl_attr(HLF_V
);
6200 preedit_caret_cbproc(XIC thexic UNUSED
,
6201 XPointer client_data UNUSED
,
6202 XPointer call_data UNUSED
)
6205 xim_log("preedit_caret_cbproc()\n");
6210 preedit_done_cbproc(XIC thexic UNUSED
,
6211 XPointer client_data UNUSED
,
6212 XPointer call_data UNUSED
)
6215 xim_log("preedit_done_cbproc()\n");
6218 vim_free(draw_feedback
);
6219 draw_feedback
= NULL
;
6220 xim_can_preediting
= FALSE
;
6221 xim_has_preediting
= FALSE
;
6222 gui_update_cursor(TRUE
, FALSE
);
6236 xim_log("xim_reset()\n");
6241 text
= XmbResetIC(((GdkICPrivate
*)xic
)->xic
);
6242 if (text
!= NULL
&& !(xim_input_style
& (int)GDK_IM_PREEDIT_CALLBACKS
))
6243 add_to_input_buf_csi((char_u
*)text
, strlen(text
));
6245 preedit_buf_len
= 0;
6252 xim_queue_key_press_event(GdkEventKey
*event
, int down UNUSED
)
6255 xim_log("xim_queue_key_press_event()\n");
6258 if (preedit_buf_len
<= 0)
6260 if (processing_queued_event
)
6261 processing_queued_event
= FALSE
;
6263 key_press_event_queue
= g_slist_append(key_press_event_queue
,
6264 gdk_event_copy((GdkEvent
*)event
));
6269 preedit_callback_setup(GdkIC
*ic UNUSED
)
6272 XVaNestedList preedit_attr
;
6273 XIMCallback preedit_start_cb
;
6274 XIMCallback preedit_draw_cb
;
6275 XIMCallback preedit_caret_cb
;
6276 XIMCallback preedit_done_cb
;
6278 xxic
= ((GdkICPrivate
*)xic
)->xic
;
6279 preedit_start_cb
.callback
= (XIMProc
)preedit_start_cbproc
;
6280 preedit_draw_cb
.callback
= (XIMProc
)preedit_draw_cbproc
;
6281 preedit_caret_cb
.callback
= (XIMProc
)preedit_caret_cbproc
;
6282 preedit_done_cb
.callback
= (XIMProc
)preedit_done_cbproc
;
6284 = XVaCreateNestedList(0,
6285 XNPreeditStartCallback
, &preedit_start_cb
,
6286 XNPreeditDrawCallback
, &preedit_draw_cb
,
6287 XNPreeditCaretCallback
, &preedit_caret_cb
,
6288 XNPreeditDoneCallback
, &preedit_done_cb
,
6290 XSetICValues(xxic
, XNPreeditAttributes
, preedit_attr
, NULL
);
6291 XFree(preedit_attr
);
6295 reset_state_setup(GdkIC
*ic UNUSED
)
6297 #ifdef USE_X11R6_XIM
6298 /* don't change the input context when we call reset */
6299 XSetICValues(((GdkICPrivate
*)ic
)->xic
, XNResetState
, XIMPreserveState
,
6308 xim_log("xim_init()\n");
6314 if (!gdk_im_ready())
6319 EMSG(_("E292: Input Method Server is not running"));
6324 if ((xic_attr
= gdk_ic_attr_new()) != NULL
)
6326 #ifdef FEAT_XFONTSET
6330 GdkColormap
*colormap
;
6331 GdkICAttr
*attr
= xic_attr
;
6332 int attrmask
= (int)GDK_IC_ALL_REQ
;
6333 GtkWidget
*widget
= gui
.drawarea
;
6335 attr
->style
= (GdkIMStyle
)xim_input_style
;
6336 attr
->client_window
= gui
.mainwin
->window
;
6338 if ((colormap
= gtk_widget_get_colormap(widget
)) !=
6339 gtk_widget_get_default_colormap())
6341 attrmask
|= (int)GDK_IC_PREEDIT_COLORMAP
;
6342 attr
->preedit_colormap
= colormap
;
6344 attrmask
|= (int)GDK_IC_PREEDIT_FOREGROUND
;
6345 attrmask
|= (int)GDK_IC_PREEDIT_BACKGROUND
;
6346 attr
->preedit_foreground
= widget
->style
->fg
[GTK_STATE_NORMAL
];
6347 attr
->preedit_background
= widget
->style
->base
[GTK_STATE_NORMAL
];
6349 #ifdef FEAT_XFONTSET
6350 if ((xim_input_style
& (int)GDK_IM_PREEDIT_MASK
)
6351 == (int)GDK_IM_PREEDIT_POSITION
)
6353 if (gui
.fontset
== NOFONTSET
6354 || gui
.fontset
->type
!= GDK_FONT_FONTSET
)
6356 EMSG(_(e_overthespot
));
6360 gdk_window_get_size(widget
->window
, &width
, &height
);
6362 attrmask
|= (int)GDK_IC_PREEDIT_POSITION_REQ
;
6363 attr
->spot_location
.x
= TEXT_X(0);
6364 attr
->spot_location
.y
= TEXT_Y(0);
6365 attr
->preedit_area
.x
= gui
.border_offset
;
6366 attr
->preedit_area
.y
= gui
.border_offset
;
6367 attr
->preedit_area
.width
= width
- 2*gui
.border_offset
;
6368 attr
->preedit_area
.height
= height
- 2*gui
.border_offset
;
6369 attr
->preedit_fontset
= gui
.fontset
;
6373 if ((xim_input_style
& (int)GDK_IM_STATUS_MASK
)
6374 == (int)GDK_IM_STATUS_AREA
)
6376 if (gui
.fontset
== NOFONTSET
6377 || gui
.fontset
->type
!= GDK_FONT_FONTSET
)
6379 EMSG(_(e_overthespot
));
6383 gdk_window_get_size(gui
.mainwin
->window
, &width
, &height
);
6384 attrmask
|= (int)GDK_IC_STATUS_AREA_REQ
;
6385 attr
->status_area
.x
= 0;
6386 attr
->status_area
.y
= height
- gui
.char_height
- 1;
6387 attr
->status_area
.width
= width
;
6388 attr
->status_area
.height
= gui
.char_height
;
6389 attr
->status_fontset
= gui
.fontset
;
6392 else if ((xim_input_style
& (int)GDK_IM_STATUS_MASK
)
6393 == (int)GDK_IM_STATUS_CALLBACKS
)
6399 xic
= gdk_ic_new(attr
, (GdkICAttributesType
)attrmask
);
6405 mask
= (int)gdk_window_get_events(widget
->window
);
6406 mask
|= (int)gdk_ic_get_events(xic
);
6407 gdk_window_set_events(widget
->window
, (GdkEventMask
)mask
);
6408 if (xim_input_style
& (int)GDK_IM_PREEDIT_CALLBACKS
)
6409 preedit_callback_setup(xic
);
6410 reset_state_setup(xic
);
6419 xim_log("im_shutdown()\n");
6425 gdk_ic_destroy(xic
);
6428 xim_is_active
= FALSE
;
6429 xim_can_preediting
= FALSE
;
6430 preedit_start_col
= MAXCOL
;
6431 xim_has_preediting
= FALSE
;
6434 #endif /* FEAT_GUI_GTK */
6437 xim_get_status_area_height()
6440 if (xim_input_style
& (int)GDK_IM_STATUS_AREA
)
6441 return gui
.char_height
;
6443 if (status_area_enabled
)
6444 return gui
.char_height
;
6450 * Get IM status. When IM is on, return TRUE. Else return FALSE.
6451 * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is
6452 * active, when not having focus XIM may still be active (e.g., when using a
6453 * tear-off menu item).
6458 # ifdef FEAT_GUI_GTK
6459 if (xim_input_style
& (int)GDK_IM_PREEDIT_CALLBACKS
)
6460 return xim_can_preediting
;
6462 return xim_has_focus
;
6465 # endif /* !HAVE_GTK2 */
6467 # if defined(HAVE_GTK2) || defined(PROTO)
6469 preedit_get_status(void)
6471 return preedit_is_active
;
6475 # if defined(FEAT_GUI_GTK) || defined(PROTO)
6479 return xim_has_preediting
;
6482 #endif /* FEAT_XIM */
6484 #if defined(FEAT_MBYTE) || defined(PROTO)
6487 * Setup "vcp" for conversion from "from" to "to".
6488 * The names must have been made canonical with enc_canonize().
6489 * vcp->vc_type must have been initialized to CONV_NONE.
6490 * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8
6492 * Afterwards invoke with "from" and "to" equal to NULL to cleanup.
6493 * Return FAIL when conversion is not supported, OK otherwise.
6496 convert_setup(vcp
, from
, to
)
6501 return convert_setup_ext(vcp
, from
, TRUE
, to
, TRUE
);
6505 * As convert_setup(), but only when from_unicode_is_utf8 is TRUE will all
6506 * "from" unicode charsets be considered utf-8. Same for "to".
6509 convert_setup_ext(vcp
, from
, from_unicode_is_utf8
, to
, to_unicode_is_utf8
)
6512 int from_unicode_is_utf8
;
6514 int to_unicode_is_utf8
;
6521 /* Reset to no conversion. */
6523 if (vcp
->vc_type
== CONV_ICONV
&& vcp
->vc_fd
!= (iconv_t
)-1)
6524 iconv_close(vcp
->vc_fd
);
6526 vcp
->vc_type
= CONV_NONE
;
6528 vcp
->vc_fail
= FALSE
;
6530 /* No conversion when one of the names is empty or they are equal. */
6531 if (from
== NULL
|| *from
== NUL
|| to
== NULL
|| *to
== NUL
6532 || STRCMP(from
, to
) == 0)
6535 from_prop
= enc_canon_props(from
);
6536 to_prop
= enc_canon_props(to
);
6537 if (from_unicode_is_utf8
)
6538 from_is_utf8
= from_prop
& ENC_UNICODE
;
6540 from_is_utf8
= from_prop
== ENC_UNICODE
;
6541 if (to_unicode_is_utf8
)
6542 to_is_utf8
= to_prop
& ENC_UNICODE
;
6544 to_is_utf8
= to_prop
== ENC_UNICODE
;
6546 if ((from_prop
& ENC_LATIN1
) && to_is_utf8
)
6548 /* Internal latin1 -> utf-8 conversion. */
6549 vcp
->vc_type
= CONV_TO_UTF8
;
6550 vcp
->vc_factor
= 2; /* up to twice as long */
6552 else if ((from_prop
& ENC_LATIN9
) && to_is_utf8
)
6554 /* Internal latin9 -> utf-8 conversion. */
6555 vcp
->vc_type
= CONV_9_TO_UTF8
;
6556 vcp
->vc_factor
= 3; /* up to three as long (euro sign) */
6558 else if (from_is_utf8
&& (to_prop
& ENC_LATIN1
))
6560 /* Internal utf-8 -> latin1 conversion. */
6561 vcp
->vc_type
= CONV_TO_LATIN1
;
6563 else if (from_is_utf8
&& (to_prop
& ENC_LATIN9
))
6565 /* Internal utf-8 -> latin9 conversion. */
6566 vcp
->vc_type
= CONV_TO_LATIN9
;
6569 /* Win32-specific codepage <-> codepage conversion without iconv. */
6570 else if ((from_is_utf8
|| encname2codepage(from
) > 0)
6571 && (to_is_utf8
|| encname2codepage(to
) > 0))
6573 vcp
->vc_type
= CONV_CODEPAGE
;
6574 vcp
->vc_factor
= 2; /* up to twice as long */
6575 vcp
->vc_cpfrom
= from_is_utf8
? 0 : encname2codepage(from
);
6576 vcp
->vc_cpto
= to_is_utf8
? 0 : encname2codepage(to
);
6580 else if ((from_prop
& ENC_MACROMAN
) && (to_prop
& ENC_LATIN1
))
6582 vcp
->vc_type
= CONV_MAC_LATIN1
;
6584 else if ((from_prop
& ENC_MACROMAN
) && to_is_utf8
)
6586 vcp
->vc_type
= CONV_MAC_UTF8
;
6587 vcp
->vc_factor
= 2; /* up to twice as long */
6589 else if ((from_prop
& ENC_LATIN1
) && (to_prop
& ENC_MACROMAN
))
6591 vcp
->vc_type
= CONV_LATIN1_MAC
;
6593 else if (from_is_utf8
&& (to_prop
& ENC_MACROMAN
))
6595 vcp
->vc_type
= CONV_UTF8_MAC
;
6601 /* Use iconv() for conversion. */
6602 vcp
->vc_fd
= (iconv_t
)my_iconv_open(
6603 to_is_utf8
? (char_u
*)"utf-8" : to
,
6604 from_is_utf8
? (char_u
*)"utf-8" : from
);
6605 if (vcp
->vc_fd
!= (iconv_t
)-1)
6607 vcp
->vc_type
= CONV_ICONV
;
6608 vcp
->vc_factor
= 4; /* could be longer too... */
6612 if (vcp
->vc_type
== CONV_NONE
)
6618 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \
6619 || defined(MSDOS) || defined(PROTO)
6621 * Do conversion on typed input characters in-place.
6622 * The input and output are not NUL terminated!
6623 * Returns the length after conversion.
6626 convert_input(ptr
, len
, maxlen
)
6631 return convert_input_safe(ptr
, len
, maxlen
, NULL
, NULL
);
6636 * Like convert_input(), but when there is an incomplete byte sequence at the
6637 * end return that as an allocated string in "restp" and set "*restlenp" to
6638 * the length. If "restp" is NULL it is not used.
6641 convert_input_safe(ptr
, len
, maxlen
, restp
, restlenp
)
6650 int unconvertlen
= 0;
6652 d
= string_convert_ext(&input_conv
, ptr
, &dlen
,
6653 restp
== NULL
? NULL
: &unconvertlen
);
6658 if (unconvertlen
> 0)
6660 /* Move the unconverted characters to allocated memory. */
6661 *restp
= alloc(unconvertlen
);
6663 mch_memmove(*restp
, ptr
+ len
- unconvertlen
, unconvertlen
);
6664 *restlenp
= unconvertlen
;
6666 mch_memmove(ptr
, d
, dlen
);
6669 /* result is too long, keep the unconverted text (the caller must
6670 * have done something wrong!) */
6678 * Convert text "ptr[*lenp]" according to "vcp".
6679 * Returns the result in allocated memory and sets "*lenp".
6680 * When "lenp" is NULL, use NUL terminated strings.
6681 * Illegal chars are often changed to "?", unless vcp->vc_fail is set.
6682 * When something goes wrong, NULL is returned and "*lenp" is unchanged.
6685 string_convert(vcp
, ptr
, lenp
)
6690 return string_convert_ext(vcp
, ptr
, lenp
, NULL
);
6694 * Like string_convert(), but when "unconvlenp" is not NULL and there are is
6695 * an incomplete sequence at the end it is not converted and "*unconvlenp" is
6696 * set to the number of remaining bytes.
6699 string_convert_ext(vcp
, ptr
, lenp
, unconvlenp
)
6705 char_u
*retval
= NULL
;
6713 len
= (int)STRLEN(ptr
);
6717 return vim_strsave((char_u
*)"");
6719 switch (vcp
->vc_type
)
6721 case CONV_TO_UTF8
: /* latin1 to utf-8 conversion */
6722 retval
= alloc(len
* 2 + 1);
6726 for (i
= 0; i
< len
; ++i
)
6733 *d
++ = 0xc0 + ((unsigned)c
>> 6);
6734 *d
++ = 0x80 + (c
& 0x3f);
6739 *lenp
= (int)(d
- retval
);
6742 case CONV_9_TO_UTF8
: /* latin9 to utf-8 conversion */
6743 retval
= alloc(len
* 3 + 1);
6747 for (i
= 0; i
< len
; ++i
)
6752 case 0xa4: c
= 0x20ac; break; /* euro */
6753 case 0xa6: c
= 0x0160; break; /* S hat */
6754 case 0xa8: c
= 0x0161; break; /* S -hat */
6755 case 0xb4: c
= 0x017d; break; /* Z hat */
6756 case 0xb8: c
= 0x017e; break; /* Z -hat */
6757 case 0xbc: c
= 0x0152; break; /* OE */
6758 case 0xbd: c
= 0x0153; break; /* oe */
6759 case 0xbe: c
= 0x0178; break; /* Y */
6761 d
+= utf_char2bytes(c
, d
);
6765 *lenp
= (int)(d
- retval
);
6768 case CONV_TO_LATIN1
: /* utf-8 to latin1 conversion */
6769 case CONV_TO_LATIN9
: /* utf-8 to latin9 conversion */
6770 retval
= alloc(len
+ 1);
6774 for (i
= 0; i
< len
; ++i
)
6776 l
= utf_ptr2len_len(ptr
+ i
, len
- i
);
6781 int l_w
= utf8len_tab_zero
[ptr
[i
]];
6785 /* Illegal utf-8 byte cannot be converted */
6789 if (unconvlenp
!= NULL
&& l_w
> len
- i
)
6791 /* Incomplete sequence at the end. */
6792 *unconvlenp
= len
- i
;
6799 c
= utf_ptr2char(ptr
+ i
);
6800 if (vcp
->vc_type
== CONV_TO_LATIN9
)
6803 case 0x20ac: c
= 0xa4; break; /* euro */
6804 case 0x0160: c
= 0xa6; break; /* S hat */
6805 case 0x0161: c
= 0xa8; break; /* S -hat */
6806 case 0x017d: c
= 0xb4; break; /* Z hat */
6807 case 0x017e: c
= 0xb8; break; /* Z -hat */
6808 case 0x0152: c
= 0xbc; break; /* OE */
6809 case 0x0153: c
= 0xbd; break; /* oe */
6810 case 0x0178: c
= 0xbe; break; /* Y */
6818 case 0xbe: c
= 0x100; break; /* not in latin9 */
6820 if (!utf_iscomposing(c
)) /* skip composing chars */
6824 else if (vcp
->vc_fail
)
6832 if (utf_char2cells(c
) > 1)
6841 *lenp
= (int)(d
- retval
);
6844 # ifdef MACOS_CONVERT
6845 case CONV_MAC_LATIN1
:
6846 retval
= mac_string_convert(ptr
, len
, lenp
, vcp
->vc_fail
,
6847 'm', 'l', unconvlenp
);
6850 case CONV_LATIN1_MAC
:
6851 retval
= mac_string_convert(ptr
, len
, lenp
, vcp
->vc_fail
,
6852 'l', 'm', unconvlenp
);
6856 retval
= mac_string_convert(ptr
, len
, lenp
, vcp
->vc_fail
,
6857 'm', 'u', unconvlenp
);
6861 retval
= mac_string_convert(ptr
, len
, lenp
, vcp
->vc_fail
,
6862 'u', 'm', unconvlenp
);
6867 case CONV_ICONV
: /* conversion with output_conv.vc_fd */
6868 retval
= iconv_string(vcp
, ptr
, len
, unconvlenp
, lenp
);
6872 case CONV_CODEPAGE
: /* codepage -> codepage */
6878 /* 1. codepage/UTF-8 -> ucs-2. */
6879 if (vcp
->vc_cpfrom
== 0)
6880 tmp_len
= utf8_to_utf16(ptr
, len
, NULL
, NULL
);
6882 tmp_len
= MultiByteToWideChar(vcp
->vc_cpfrom
, 0,
6884 tmp
= (short_u
*)alloc(sizeof(short_u
) * tmp_len
);
6887 if (vcp
->vc_cpfrom
== 0)
6888 utf8_to_utf16(ptr
, len
, tmp
, unconvlenp
);
6890 MultiByteToWideChar(vcp
->vc_cpfrom
, 0, ptr
, len
, tmp
, tmp_len
);
6892 /* 2. ucs-2 -> codepage/UTF-8. */
6893 if (vcp
->vc_cpto
== 0)
6894 retlen
= utf16_to_utf8(tmp
, tmp_len
, NULL
);
6896 retlen
= WideCharToMultiByte(vcp
->vc_cpto
, 0,
6897 tmp
, tmp_len
, 0, 0, 0, 0);
6898 retval
= alloc(retlen
+ 1);
6901 if (vcp
->vc_cpto
== 0)
6902 utf16_to_utf8(tmp
, tmp_len
, retval
);
6904 WideCharToMultiByte(vcp
->vc_cpto
, 0,
6905 tmp
, tmp_len
, retval
, retlen
, 0, 0);
6906 retval
[retlen
] = NUL
;