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 two composing characters
30 * are allowed. These are drawn on top of the first char.
31 * For most editing the sequence of bytes with composing
32 * characters included is considered to be one character.
33 * "enc_unicode" When 2 use 16-bit Unicode characters (or UTF-16).
34 * When 4 use 32-but Unicode characters.
35 * Internally characters are stored in UTF-8 encoding to
36 * avoid NUL bytes. Conversion happens when doing I/O.
37 * "enc_utf8" will also be TRUE.
39 * "has_mbyte" is set when "enc_dbcs" or "enc_utf8" is non-zero.
41 * If none of these is TRUE, 8-bit bytes are used for a character. The
42 * encoding isn't currently specified (TODO).
44 * 'encoding' specifies the encoding used in the core. This is in registers,
45 * text manipulation, buffers, etc. Conversion has to be done when characters
46 * in another encoding are received or send:
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_char2cells
__ARGS((int c
));
131 static int dbcs_ptr2char
__ARGS((char_u
*p
));
133 /* Lookup table to quickly get the length in bytes of a UTF-8 character from
134 * the first byte of a UTF-8 string. Bytes which are illegal when used as the
135 * first byte have a one, because these will be used separately. */
136 static char utf8len_tab
[256] =
138 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
139 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
140 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
141 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
142 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/
143 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /*bogus*/
144 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,
145 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,6,6,1,1,
149 * XIM often causes trouble. Define XIM_DEBUG to get a log of XIM callbacks
150 * in the "xim.log" file.
152 /* #define XIM_DEBUG */
155 xim_log(char *s
, ...)
158 static FILE *fd
= NULL
;
160 if (fd
== (FILE *)-1)
164 fd
= mch_fopen("xim.log", "w");
167 EMSG("Cannot open xim.log");
173 va_start(arglist
, s
);
174 vfprintf(fd
, s
, arglist
);
181 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO)
183 * Canonical encoding names and their properties.
184 * "iso-8859-n" is handled by enc_canonize() directly.
187 { char *name
; int prop
; int codepage
;}
190 #define IDX_LATIN_1 0
191 {"latin1", ENC_8BIT
+ ENC_LATIN1
, 1252},
193 {"iso-8859-2", ENC_8BIT
, 0},
195 {"iso-8859-3", ENC_8BIT
, 0},
197 {"iso-8859-4", ENC_8BIT
, 0},
199 {"iso-8859-5", ENC_8BIT
, 0},
201 {"iso-8859-6", ENC_8BIT
, 0},
203 {"iso-8859-7", ENC_8BIT
, 0},
205 {"iso-8859-8", ENC_8BIT
, 0},
207 {"iso-8859-9", ENC_8BIT
, 0},
209 {"iso-8859-10", ENC_8BIT
, 0},
210 #define IDX_ISO_11 10
211 {"iso-8859-11", ENC_8BIT
, 0},
212 #define IDX_ISO_13 11
213 {"iso-8859-13", ENC_8BIT
, 0},
214 #define IDX_ISO_14 12
215 {"iso-8859-14", ENC_8BIT
, 0},
216 #define IDX_ISO_15 13
217 {"iso-8859-15", ENC_8BIT
+ ENC_LATIN9
, 0},
218 #define IDX_KOI8_R 14
219 {"koi8-r", ENC_8BIT
, 0},
220 #define IDX_KOI8_U 15
221 {"koi8-u", ENC_8BIT
, 0},
223 {"utf-8", ENC_UNICODE
, 0},
225 {"ucs-2", ENC_UNICODE
+ ENC_ENDIAN_B
+ ENC_2BYTE
, 0},
226 #define IDX_UCS2LE 18
227 {"ucs-2le", ENC_UNICODE
+ ENC_ENDIAN_L
+ ENC_2BYTE
, 0},
229 {"utf-16", ENC_UNICODE
+ ENC_ENDIAN_B
+ ENC_2WORD
, 0},
230 #define IDX_UTF16LE 20
231 {"utf-16le", ENC_UNICODE
+ ENC_ENDIAN_L
+ ENC_2WORD
, 0},
233 {"ucs-4", ENC_UNICODE
+ ENC_ENDIAN_B
+ ENC_4BYTE
, 0},
234 #define IDX_UCS4LE 22
235 {"ucs-4le", ENC_UNICODE
+ ENC_ENDIAN_L
+ ENC_4BYTE
, 0},
237 /* For debugging DBCS encoding on Unix. */
239 {"debug", ENC_DBCS
, DBCS_DEBUG
},
240 #define IDX_EUC_JP 24
241 {"euc-jp", ENC_DBCS
, DBCS_JPNU
},
243 {"sjis", ENC_DBCS
, DBCS_JPN
},
244 #define IDX_EUC_KR 26
245 {"euc-kr", ENC_DBCS
, DBCS_KORU
},
246 #define IDX_EUC_CN 27
247 {"euc-cn", ENC_DBCS
, DBCS_CHSU
},
248 #define IDX_EUC_TW 28
249 {"euc-tw", ENC_DBCS
, DBCS_CHTU
},
251 {"big5", ENC_DBCS
, DBCS_CHT
},
253 /* MS-DOS and MS-Windows codepages are included here, so that they can be
254 * used on Unix too. Most of them are similar to ISO-8859 encodings, but
255 * not exactly the same. */
257 {"cp437", ENC_8BIT
, 437}, /* like iso-8859-1 */
259 {"cp737", ENC_8BIT
, 737}, /* like iso-8859-7 */
261 {"cp775", ENC_8BIT
, 775}, /* Baltic */
263 {"cp850", ENC_8BIT
, 850}, /* like iso-8859-4 */
265 {"cp852", ENC_8BIT
, 852}, /* like iso-8859-1 */
267 {"cp855", ENC_8BIT
, 855}, /* like iso-8859-2 */
269 {"cp857", ENC_8BIT
, 857}, /* like iso-8859-5 */
271 {"cp860", ENC_8BIT
, 860}, /* like iso-8859-9 */
273 {"cp861", ENC_8BIT
, 861}, /* like iso-8859-1 */
275 {"cp862", ENC_8BIT
, 862}, /* like iso-8859-1 */
277 {"cp863", ENC_8BIT
, 863}, /* like iso-8859-8 */
279 {"cp865", ENC_8BIT
, 865}, /* like iso-8859-1 */
281 {"cp866", ENC_8BIT
, 866}, /* like iso-8859-5 */
283 {"cp869", ENC_8BIT
, 869}, /* like iso-8859-7 */
285 {"cp874", ENC_8BIT
, 874}, /* Thai */
287 {"cp932", ENC_DBCS
, DBCS_JPN
},
289 {"cp936", ENC_DBCS
, DBCS_CHS
},
291 {"cp949", ENC_DBCS
, DBCS_KOR
},
293 {"cp950", ENC_DBCS
, DBCS_CHT
},
294 #define IDX_CP1250 49
295 {"cp1250", ENC_8BIT
, 1250}, /* Czech, Polish, etc. */
296 #define IDX_CP1251 50
297 {"cp1251", ENC_8BIT
, 1251}, /* Cyrillic */
298 /* cp1252 is considered to be equal to latin1 */
299 #define IDX_CP1253 51
300 {"cp1253", ENC_8BIT
, 1253}, /* Greek */
301 #define IDX_CP1254 52
302 {"cp1254", ENC_8BIT
, 1254}, /* Turkish */
303 #define IDX_CP1255 53
304 {"cp1255", ENC_8BIT
, 1255}, /* Hebrew */
305 #define IDX_CP1256 54
306 {"cp1256", ENC_8BIT
, 1256}, /* Arabic */
307 #define IDX_CP1257 55
308 {"cp1257", ENC_8BIT
, 1257}, /* Baltic */
309 #define IDX_CP1258 56
310 {"cp1258", ENC_8BIT
, 1258}, /* Vietnamese */
312 #define IDX_MACROMAN 57
313 {"macroman", ENC_8BIT
+ ENC_MACROMAN
, 0}, /* Mac OS */
314 #define IDX_DECMCS 58
315 {"dec-mcs", ENC_8BIT
, 0}, /* DEC MCS */
316 #define IDX_HPROMAN8 59
317 {"hp-roman8", ENC_8BIT
, 0}, /* HP Roman8 */
322 * Aliases for encoding names.
325 { char *name
; int canon
;}
328 {"ansi", IDX_LATIN_1
},
329 {"iso-8859-1", IDX_LATIN_1
},
330 {"latin2", IDX_ISO_2
},
331 {"latin3", IDX_ISO_3
},
332 {"latin4", IDX_ISO_4
},
333 {"cyrillic", IDX_ISO_5
},
334 {"arabic", IDX_ISO_6
},
335 {"greek", IDX_ISO_7
},
337 {"hebrew", IDX_CP1255
},
339 {"hebrew", IDX_ISO_8
},
341 {"latin5", IDX_ISO_9
},
342 {"turkish", IDX_ISO_9
}, /* ? */
343 {"latin6", IDX_ISO_10
},
344 {"nordic", IDX_ISO_10
}, /* ? */
345 {"thai", IDX_ISO_11
}, /* ? */
346 {"latin7", IDX_ISO_13
},
347 {"latin8", IDX_ISO_14
},
348 {"latin9", IDX_ISO_15
},
350 {"unicode", IDX_UCS2
},
352 {"ucs2be", IDX_UCS2
},
353 {"ucs-2be", IDX_UCS2
},
354 {"ucs2le", IDX_UCS2LE
},
355 {"utf16", IDX_UTF16
},
356 {"utf16be", IDX_UTF16
},
357 {"utf-16be", IDX_UTF16
},
358 {"utf16le", IDX_UTF16LE
},
360 {"ucs4be", IDX_UCS4
},
361 {"ucs-4be", IDX_UCS4
},
362 {"ucs4le", IDX_UCS4LE
},
368 {"eucjp", IDX_EUC_JP
},
369 {"unix-jis", IDX_EUC_JP
},
370 {"ujis", IDX_EUC_JP
},
371 {"shift-jis", IDX_SJIS
},
372 {"euckr", IDX_EUC_KR
},
373 {"5601", IDX_EUC_KR
}, /* Sun: KS C 5601 */
374 {"euccn", IDX_EUC_CN
},
375 {"gb2312", IDX_EUC_CN
},
376 {"euctw", IDX_EUC_TW
},
377 #if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS)
378 {"japan", IDX_CP932
},
379 {"korea", IDX_CP949
},
381 {"chinese", IDX_CP936
},
382 {"taiwan", IDX_CP950
},
385 {"japan", IDX_EUC_JP
},
386 {"korea", IDX_EUC_KR
},
388 {"chinese", IDX_EUC_CN
},
389 {"taiwan", IDX_EUC_TW
},
393 {"mac", IDX_MACROMAN
},
394 {"mac-roman", IDX_MACROMAN
},
399 # define CP_UTF8 65001 /* magic number from winnls.h */
403 * Find encoding "name" in the list of canonical encoding names.
404 * Returns -1 if not found.
407 enc_canon_search(name
)
412 for (i
= 0; i
< IDX_COUNT
; ++i
)
413 if (STRCMP(name
, enc_canon_table
[i
].name
) == 0)
420 #if defined(FEAT_MBYTE) || defined(PROTO)
423 * Find canonical encoding "name" in the list and return its properties.
424 * Returns 0 if not found.
427 enc_canon_props(name
)
432 i
= enc_canon_search(name
);
434 return enc_canon_table
[i
].prop
;
436 if (name
[0] == 'c' && name
[1] == 'p' && VIM_ISDIGIT(name
[2]))
440 /* Get info on this codepage to find out what it is. */
441 if (GetCPInfo(atoi(name
+ 2), &cpinfo
) != 0)
443 if (cpinfo
.MaxCharSize
== 1) /* some single-byte encoding */
445 if (cpinfo
.MaxCharSize
== 2
446 && (cpinfo
.LeadByte
[0] != 0 || cpinfo
.LeadByte
[1] != 0))
447 /* must be a DBCS encoding */
453 if (STRNCMP(name
, "2byte-", 6) == 0)
455 if (STRNCMP(name
, "8bit-", 5) == 0 || STRNCMP(name
, "iso-8859-", 9) == 0)
461 * Set up for using multi-byte characters.
462 * Called in three cases:
463 * - by main() to initialize (p_enc == NULL)
464 * - by set_init_1() after 'encoding' was set to its default.
465 * - by do_set() when 'encoding' has been set.
466 * p_enc must have been passed through enc_canonize() already.
467 * Sets the "enc_unicode", "enc_utf8", "enc_dbcs" and "has_mbyte" flags.
468 * Fills mb_bytelen_tab[] and returns NULL when there are no problems.
469 * When there is something wrong: Returns an error message and doesn't change
478 int enc_dbcs_new
= 0;
479 #if defined(USE_ICONV) && !defined(WIN3264) && !defined(WIN32UNIX) \
481 # define LEN_FROM_CONV
488 /* Just starting up: set the whole table to one's. */
489 for (i
= 0; i
< 256; ++i
)
490 mb_bytelen_tab
[i
] = 1;
491 input_conv
.vc_type
= CONV_NONE
;
492 input_conv
.vc_factor
= 1;
493 output_conv
.vc_type
= CONV_NONE
;
498 if (p_enc
[0] == 'c' && p_enc
[1] == 'p' && VIM_ISDIGIT(p_enc
[2]))
502 /* Get info on this codepage to find out what it is. */
503 if (GetCPInfo(atoi(p_enc
+ 2), &cpinfo
) != 0)
505 if (cpinfo
.MaxCharSize
== 1)
507 /* some single-byte encoding */
511 else if (cpinfo
.MaxCharSize
== 2
512 && (cpinfo
.LeadByte
[0] != 0 || cpinfo
.LeadByte
[1] != 0))
514 /* must be a DBCS encoding, check below */
515 enc_dbcs_new
= atoi(p_enc
+ 2);
518 goto codepage_invalid
;
520 else if (GetLastError() == ERROR_INVALID_PARAMETER
)
523 return (char_u
*)N_("E543: Not a valid codepage");
527 else if (STRNCMP(p_enc
, "8bit-", 5) == 0
528 || STRNCMP(p_enc
, "iso-8859-", 9) == 0)
530 /* Accept any "8bit-" or "iso-8859-" name. */
534 else if (STRNCMP(p_enc
, "2byte-", 6) == 0)
537 /* Windows: accept only valid codepage numbers, check below. */
538 if (p_enc
[6] != 'c' || p_enc
[7] != 'p'
539 || (enc_dbcs_new
= atoi(p_enc
+ 8)) == 0)
542 /* Unix: accept any "2byte-" name, assume current locale. */
543 enc_dbcs_new
= DBCS_2BYTE
;
546 else if ((idx
= enc_canon_search(p_enc
)) >= 0)
548 i
= enc_canon_table
[idx
].prop
;
553 if (i
& (ENC_2BYTE
| ENC_2WORD
))
555 else if (i
& ENC_4BYTE
)
560 else if (i
& ENC_DBCS
)
562 /* 2byte, handle below */
563 enc_dbcs_new
= enc_canon_table
[idx
].codepage
;
572 else /* Don't know what encoding this is, reject it. */
575 if (enc_dbcs_new
!= 0)
578 /* Check if the DBCS code page is OK. */
579 if (!IsValidCodePage(enc_dbcs_new
))
580 goto codepage_invalid
;
585 enc_dbcs
= enc_dbcs_new
;
586 has_mbyte
= (enc_dbcs
!= 0 || enc_utf8
);
589 enc_codepage
= encname2codepage(p_enc
);
590 enc_latin9
= (STRCMP(p_enc
, "iso-8859-15") == 0);
593 /* Detect an encoding that uses latin1 characters. */
594 enc_latin1like
= (enc_utf8
|| STRCMP(p_enc
, "latin1") == 0
595 || STRCMP(p_enc
, "iso-8859-15") == 0);
598 * Set the function pointers.
602 mb_ptr2len
= utfc_ptr2len
;
603 mb_char2len
= utf_char2len
;
604 mb_char2bytes
= utf_char2bytes
;
605 mb_ptr2cells
= utf_ptr2cells
;
606 mb_char2cells
= utf_char2cells
;
607 mb_off2cells
= utf_off2cells
;
608 mb_ptr2char
= utf_ptr2char
;
609 mb_head_off
= utf_head_off
;
611 else if (enc_dbcs
!= 0)
613 mb_ptr2len
= dbcs_ptr2len
;
614 mb_char2len
= dbcs_char2len
;
615 mb_char2bytes
= dbcs_char2bytes
;
616 mb_ptr2cells
= dbcs_ptr2cells
;
617 mb_char2cells
= dbcs_char2cells
;
618 mb_off2cells
= dbcs_off2cells
;
619 mb_ptr2char
= dbcs_ptr2char
;
620 mb_head_off
= dbcs_head_off
;
624 mb_ptr2len
= latin_ptr2len
;
625 mb_char2len
= latin_char2len
;
626 mb_char2bytes
= latin_char2bytes
;
627 mb_ptr2cells
= latin_ptr2cells
;
628 mb_char2cells
= latin_char2cells
;
629 mb_off2cells
= latin_off2cells
;
630 mb_ptr2char
= latin_ptr2char
;
631 mb_head_off
= latin_head_off
;
635 * Fill the mb_bytelen_tab[] for MB_BYTE2LEN().
638 /* When 'encoding' is different from the current locale mblen() won't
639 * work. Use conversion to "utf-8" instead. */
640 vimconv
.vc_type
= CONV_NONE
;
644 if (p
== NULL
|| STRCMP(p
, p_enc
) != 0)
646 convert_setup(&vimconv
, p_enc
, (char_u
*)"utf-8");
647 vimconv
.vc_fail
= TRUE
;
653 for (i
= 0; i
< 256; ++i
)
655 /* Our own function to reliably check the length of UTF-8 characters,
656 * independent of mblen(). */
659 else if (enc_dbcs
== 0)
663 #if defined(WIN3264) || defined(WIN32UNIX)
664 /* enc_dbcs is set by setting 'fileencoding'. It becomes a Windows
665 * CodePage identifier, which we can pass directly in to Windows
667 n
= IsDBCSLeadByteEx(enc_dbcs
, (BYTE
)i
) ? 2 : 1;
669 # if defined(MACOS) || defined(__amigaos4__)
671 * if mblen() is not available, character which MSB is turned on
672 * are treated as leading byte character. (note : This assumption
673 * is not always true.)
675 n
= (i
& 0x80) ? 2 : 1;
677 char buf
[MB_MAXBYTES
];
680 # define mblen _Xmblen
683 if (i
== NUL
) /* just in case mblen() can't handle "" */
690 if (vimconv
.vc_type
!= CONV_NONE
)
693 * string_convert() should fail when converting the first
694 * byte of a double-byte character.
696 p
= string_convert(&vimconv
, (char_u
*)buf
, NULL
);
709 * mblen() should return -1 for invalid (means the leading
710 * multibyte) character. However there are some platforms
711 * where mblen() returns 0 for invalid character.
712 * Therefore, following condition includes 0.
714 (void)mblen(NULL
, 0); /* First reset the state. */
715 if (mblen(buf
, (size_t)1) <= 0)
725 mb_bytelen_tab
[i
] = n
;
729 convert_setup(&vimconv
, NULL
, NULL
);
732 /* The cell width depends on the type of multi-byte characters. */
733 (void)init_chartab();
735 /* When enc_utf8 is set or reset, (de)allocate ScreenLinesUC[] */
738 /* When using Unicode, set default for 'fileencodings'. */
739 if (enc_utf8
&& !option_was_set((char_u
*)"fencs"))
740 set_string_option_direct((char_u
*)"fencs", -1,
741 (char_u
*)"ucs-bom,utf-8,default,latin1", OPT_FREE
, 0);
743 #if defined(HAVE_BIND_TEXTDOMAIN_CODESET) && defined(FEAT_GETTEXT)
744 /* GNU gettext 0.10.37 supports this feature: set the codeset used for
745 * translated messages independently from the current locale. */
746 (void)bind_textdomain_codeset(VIMPACKAGE
,
747 enc_utf8
? "utf-8" : (char *)p_enc
);
751 /* When changing 'encoding' while starting up, then convert the command
752 * line arguments from the active codepage to 'encoding'. */
758 /* Fire an autocommand to let people do custom font setup. This must be
759 * after Vim has been setup for the new encoding. */
760 apply_autocmds(EVENT_ENCODINGCHANGED
, NULL
, (char_u
*)"", FALSE
, curbuf
);
764 /* Need to reload spell dictionaries */
772 * Return the size of the BOM for the current buffer:
774 * 2 - UCS-2 or UTF-16 BOM
783 if (curbuf
->b_p_bomb
&& !curbuf
->b_p_bin
)
785 if (*curbuf
->b_p_fenc
== NUL
)
789 if (enc_unicode
!= 0)
795 else if (STRCMP(curbuf
->b_p_fenc
, "utf-8") == 0)
797 else if (STRNCMP(curbuf
->b_p_fenc
, "ucs-2", 5) == 0
798 || STRNCMP(curbuf
->b_p_fenc
, "utf-16", 6) == 0)
800 else if (STRNCMP(curbuf
->b_p_fenc
, "ucs-4", 5) == 0)
807 * Get class of pointer:
810 * 2 for an (ASCII) word character
811 * >2 for other word characters
817 if (MB_BYTE2LEN(p
[0]) == 1)
819 if (p
[0] == NUL
|| vim_iswhite(p
[0]))
821 if (vim_iswordc(p
[0]))
825 if (enc_dbcs
!= 0 && p
[0] != NUL
&& p
[1] != NUL
)
826 return dbcs_class(p
[0], p
[1]);
828 return utf_class(utf_ptr2char(p
));
833 * Get class of a double-byte character. This always returns 3 or bigger.
834 * TODO: Should return 1 for punctuation.
837 dbcs_class(lead
, trail
)
843 /* please add classfy routine for your language in here */
845 case DBCS_JPNU
: /* ? */
848 /* JIS code classification */
849 unsigned char lb
= lead
;
850 unsigned char tb
= trail
;
852 /* convert process code to JIS */
853 # if defined(WIN3264) || defined(WIN32UNIX) || defined(MACOS)
854 /* process code is SJIS */
856 lb
= (lb
- 0x81) * 2 + 0x21;
858 lb
= (lb
- 0xc1) * 2 + 0x21;
870 * XXX: Code page identification can not use with all
871 * system! So, some other encoding information
873 * In japanese: SJIS,EUC,UNICODE,(JIS)
874 * Note that JIS-code system don't use as
875 * process code in most system because it uses
876 * escape sequences(JIS is context depend encoding).
878 /* assume process code is JAPANESE-EUC */
883 switch (lb
<< 8 | tb
)
885 case 0x2121: /* ZENKAKU space */
887 case 0x2122: /* KU-TEN (Japanese comma) */
888 case 0x2123: /* TOU-TEN (Japanese period) */
889 case 0x2124: /* ZENKAKU comma */
890 case 0x2125: /* ZENKAKU period */
892 case 0x213c: /* prolongedsound handled as KATAKANA */
895 /* sieved by KU code */
900 /* special symbols */
926 case DBCS_KORU
: /* ? */
929 /* KS code classification */
930 unsigned char c1
= lead
;
931 unsigned char c2
= trail
;
937 * 23 : Alpha-numeric/Roman Letter (Full width)
938 * 24 : Hangul Letter(Alphabet)
939 * 25 : Roman Numeral/Greek Letter
942 * 28 : Circled/Parenthesized Letter
943 * 29 : Hirigana/Katakana
944 * 30 : Cyrillic Letter
947 if (c1
>= 0xB0 && c1
<= 0xC8)
950 #if defined(WIN3264) || defined(WIN32UNIX)
951 else if (c1
<= 0xA0 || c2
<= 0xA0)
952 /* Extended Hangul Region : MS UHC(Unified Hangul Code) */
953 /* c1: 0x81-0xA0 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xFE
954 * c1: 0xA1-0xC6 with c2: 0x41-0x5A, 0x61-0x7A, 0x81-0xA0
959 else if (c1
>= 0xCA && c1
<= 0xFD)
972 /* Hangul Letter(Alphabet) */
975 /* Roman Numeral/Greek Letter */
986 return 25; /* Roman Letter */
988 return 22; /* Symbols */
990 /* Circled/Parenthesized Letter */
994 /* Hirigana/Katakana */
997 /* Cyrillic Letter */
1008 * mb_char2len() function pointer.
1009 * Return length in bytes of character "c".
1010 * Returns 1 for a single-byte character.
1030 * mb_char2bytes() function pointer.
1031 * Convert a character to its bytes.
1032 * Returns the length in bytes.
1035 latin_char2bytes(c
, buf
)
1044 dbcs_char2bytes(c
, buf
)
1050 buf
[0] = (unsigned)c
>> 8;
1052 /* Never use a NUL byte, it causes lots of trouble. It's an invalid
1053 * character anyway. */
1063 * mb_ptr2len() function pointer.
1064 * Get byte length of character at "*p" but stop at a NUL.
1065 * For UTF-8 this includes following composing characters.
1066 * Returns 0 when *p is NUL.
1073 return MB_BYTE2LEN(*p
);
1082 /* Check if second byte is not missing. */
1083 len
= MB_BYTE2LEN(*p
);
1084 if (len
== 2 && p
[1] == NUL
)
1091 unsigned short first
;
1092 unsigned short last
;
1094 static int intable
__ARGS((struct interval
*table
, size_t size
, int c
));
1097 * Return TRUE if "c" is in "table[size / sizeof(struct interval)]".
1100 intable(table
, size
, c
)
1101 struct interval
*table
;
1107 /* first quick check for Latin1 etc. characters */
1108 if (c
< table
[0].first
)
1111 /* binary search in table */
1113 top
= (int)(size
/ sizeof(struct interval
) - 1);
1116 mid
= (bot
+ top
) / 2;
1117 if (table
[mid
].last
< c
)
1119 else if (table
[mid
].first
> c
)
1128 * For UTF-8 character "c" return 2 for a double-width character, 1 for others.
1129 * Returns 4 or 6 for an unprintable character.
1130 * Is only correct for characters >= 0x80.
1131 * When p_ambw is "double", return 2 for a character with East Asian Width
1132 * class 'A'(mbiguous).
1138 /* sorted list of non-overlapping intervals of East Asian Ambiguous
1139 * characters, generated with:
1140 * "uniset +WIDTH-A -cat=Me -cat=Mn -cat=Cf c" */
1141 static struct interval ambiguous
[] = {
1142 {0x00A1, 0x00A1}, {0x00A4, 0x00A4}, {0x00A7, 0x00A8},
1143 {0x00AA, 0x00AA}, {0x00AE, 0x00AE}, {0x00B0, 0x00B4},
1144 {0x00B6, 0x00BA}, {0x00BC, 0x00BF}, {0x00C6, 0x00C6},
1145 {0x00D0, 0x00D0}, {0x00D7, 0x00D8}, {0x00DE, 0x00E1},
1146 {0x00E6, 0x00E6}, {0x00E8, 0x00EA}, {0x00EC, 0x00ED},
1147 {0x00F0, 0x00F0}, {0x00F2, 0x00F3}, {0x00F7, 0x00FA},
1148 {0x00FC, 0x00FC}, {0x00FE, 0x00FE}, {0x0101, 0x0101},
1149 {0x0111, 0x0111}, {0x0113, 0x0113}, {0x011B, 0x011B},
1150 {0x0126, 0x0127}, {0x012B, 0x012B}, {0x0131, 0x0133},
1151 {0x0138, 0x0138}, {0x013F, 0x0142}, {0x0144, 0x0144},
1152 {0x0148, 0x014B}, {0x014D, 0x014D}, {0x0152, 0x0153},
1153 {0x0166, 0x0167}, {0x016B, 0x016B}, {0x01CE, 0x01CE},
1154 {0x01D0, 0x01D0}, {0x01D2, 0x01D2}, {0x01D4, 0x01D4},
1155 {0x01D6, 0x01D6}, {0x01D8, 0x01D8}, {0x01DA, 0x01DA},
1156 {0x01DC, 0x01DC}, {0x0251, 0x0251}, {0x0261, 0x0261},
1157 {0x02C4, 0x02C4}, {0x02C7, 0x02C7}, {0x02C9, 0x02CB},
1158 {0x02CD, 0x02CD}, {0x02D0, 0x02D0}, {0x02D8, 0x02DB},
1159 {0x02DD, 0x02DD}, {0x02DF, 0x02DF}, {0x0391, 0x03A1},
1160 {0x03A3, 0x03A9}, {0x03B1, 0x03C1}, {0x03C3, 0x03C9},
1161 {0x0401, 0x0401}, {0x0410, 0x044F}, {0x0451, 0x0451},
1162 {0x2010, 0x2010}, {0x2013, 0x2016}, {0x2018, 0x2019},
1163 {0x201C, 0x201D}, {0x2020, 0x2022}, {0x2024, 0x2027},
1164 {0x2030, 0x2030}, {0x2032, 0x2033}, {0x2035, 0x2035},
1165 {0x203B, 0x203B}, {0x203E, 0x203E}, {0x2074, 0x2074},
1166 {0x207F, 0x207F}, {0x2081, 0x2084}, {0x20AC, 0x20AC},
1167 {0x2103, 0x2103}, {0x2105, 0x2105}, {0x2109, 0x2109},
1168 {0x2113, 0x2113}, {0x2116, 0x2116}, {0x2121, 0x2122},
1169 {0x2126, 0x2126}, {0x212B, 0x212B}, {0x2153, 0x2154},
1170 {0x215B, 0x215E}, {0x2160, 0x216B}, {0x2170, 0x2179},
1171 {0x2190, 0x2199}, {0x21B8, 0x21B9}, {0x21D2, 0x21D2},
1172 {0x21D4, 0x21D4}, {0x21E7, 0x21E7}, {0x2200, 0x2200},
1173 {0x2202, 0x2203}, {0x2207, 0x2208}, {0x220B, 0x220B},
1174 {0x220F, 0x220F}, {0x2211, 0x2211}, {0x2215, 0x2215},
1175 {0x221A, 0x221A}, {0x221D, 0x2220}, {0x2223, 0x2223},
1176 {0x2225, 0x2225}, {0x2227, 0x222C}, {0x222E, 0x222E},
1177 {0x2234, 0x2237}, {0x223C, 0x223D}, {0x2248, 0x2248},
1178 {0x224C, 0x224C}, {0x2252, 0x2252}, {0x2260, 0x2261},
1179 {0x2264, 0x2267}, {0x226A, 0x226B}, {0x226E, 0x226F},
1180 {0x2282, 0x2283}, {0x2286, 0x2287}, {0x2295, 0x2295},
1181 {0x2299, 0x2299}, {0x22A5, 0x22A5}, {0x22BF, 0x22BF},
1182 {0x2312, 0x2312}, {0x2460, 0x24E9}, {0x24EB, 0x254B},
1183 {0x2550, 0x2573}, {0x2580, 0x258F}, {0x2592, 0x2595},
1184 {0x25A0, 0x25A1}, {0x25A3, 0x25A9}, {0x25B2, 0x25B3},
1185 {0x25B6, 0x25B7}, {0x25BC, 0x25BD}, {0x25C0, 0x25C1},
1186 {0x25C6, 0x25C8}, {0x25CB, 0x25CB}, {0x25CE, 0x25D1},
1187 {0x25E2, 0x25E5}, {0x25EF, 0x25EF}, {0x2605, 0x2606},
1188 {0x2609, 0x2609}, {0x260E, 0x260F}, {0x2614, 0x2615},
1189 {0x261C, 0x261C}, {0x261E, 0x261E}, {0x2640, 0x2640},
1190 {0x2642, 0x2642}, {0x2660, 0x2661}, {0x2663, 0x2665},
1191 {0x2667, 0x266A}, {0x266C, 0x266D}, {0x266F, 0x266F},
1192 {0x273D, 0x273D}, {0x2776, 0x277F}, {0xE000, 0xF8FF},
1193 {0xFFFD, 0xFFFD}, /* {0xF0000, 0xFFFFD}, {0x100000, 0x10FFFD} */
1198 #ifdef USE_WCHAR_FUNCTIONS
1200 * Assume the library function wcwidth() works better than our own
1201 * stuff. It should return 1 for ambiguous width chars!
1206 return 6; /* unprintable, displays <xxxx> */
1210 if (!utf_printable(c
))
1211 return 6; /* unprintable, displays <xxxx> */
1213 && (c
<= 0x115f /* Hangul Jamo */
1216 || (c
>= 0x2e80 && c
<= 0xa4cf
1217 && c
!= 0x303f) /* CJK ... Yi */
1218 || (c
>= 0xac00 && c
<= 0xd7a3) /* Hangul Syllables */
1219 || (c
>= 0xf900 && c
<= 0xfaff) /* CJK Compatibility
1221 || (c
>= 0xfe30 && c
<= 0xfe6f) /* CJK Compatibility Forms */
1222 || (c
>= 0xff00 && c
<= 0xff60) /* Fullwidth Forms */
1223 || (c
>= 0xffe0 && c
<= 0xffe6)
1224 || (c
>= 0x20000 && c
<= 0x2fffd)
1225 || (c
>= 0x30000 && c
<= 0x3fffd)))
1230 /* Characters below 0x100 are influenced by 'isprint' option */
1231 else if (c
>= 0x80 && !vim_isprintc(c
))
1232 return 4; /* unprintable, displays <xx> */
1234 if (c
>= 0x80 && *p_ambw
== 'd' && intable(ambiguous
, sizeof(ambiguous
), c
))
1241 * mb_ptr2cells() function pointer.
1242 * Return the number of display cells character at "*p" occupies.
1243 * This doesn't take care of unprintable characters, use ptr2cells() for that.
1259 /* Need to convert to a wide character. */
1262 c
= utf_ptr2char(p
);
1263 /* An illegal byte is displayed as <xx>. */
1264 if (utf_ptr2len(p
) == 1 || c
== NUL
)
1266 /* If the char is ASCII it must be an overlong sequence. */
1268 return char2cells(c
);
1269 return utf_char2cells(c
);
1278 /* Number of cells is equal to number of bytes, except for euc-jp when
1279 * the first byte is 0x8e. */
1280 if (enc_dbcs
== DBCS_JPNU
&& *p
== 0x8e)
1282 return MB_BYTE2LEN(*p
);
1286 * mb_char2cells() function pointer.
1287 * Return the number of display cells character "c" occupies.
1288 * Only takes care of multi-byte chars, not "^C" and such.
1302 /* Number of cells is equal to number of bytes, except for euc-jp when
1303 * the first byte is 0x8e. */
1304 if (enc_dbcs
== DBCS_JPNU
&& ((unsigned)c
>> 8) == 0x8e)
1306 /* use the first byte */
1307 return MB_BYTE2LEN((unsigned)c
>> 8);
1311 * mb_off2cells() function pointer.
1312 * Return number of display cells for char at ScreenLines[off].
1313 * Caller must make sure "off" and "off + 1" are valid!
1317 latin_off2cells(off
)
1327 /* Number of cells is equal to number of bytes, except for euc-jp when
1328 * the first byte is 0x8e. */
1329 if (enc_dbcs
== DBCS_JPNU
&& ScreenLines
[off
] == 0x8e)
1331 return MB_BYTE2LEN(ScreenLines
[off
]);
1338 return ScreenLines
[off
+ 1] == 0 ? 2 : 1;
1342 * mb_ptr2char() function pointer.
1343 * Convert a byte sequence into a character.
1356 if (MB_BYTE2LEN(*p
) > 1 && p
[1] != NUL
)
1357 return (p
[0] << 8) + p
[1];
1362 * Convert a UTF-8 byte sequence to a wide character.
1363 * If the sequence is illegal or truncated by a NUL the first byte is
1365 * Does not include composing characters, of course.
1373 if (p
[0] < 0x80) /* be quick for ASCII */
1376 len
= utf8len_tab
[p
[0]];
1377 if ((p
[1] & 0xc0) == 0x80)
1380 return ((p
[0] & 0x1f) << 6) + (p
[1] & 0x3f);
1381 if ((p
[2] & 0xc0) == 0x80)
1384 return ((p
[0] & 0x0f) << 12) + ((p
[1] & 0x3f) << 6)
1386 if ((p
[3] & 0xc0) == 0x80)
1389 return ((p
[0] & 0x07) << 18) + ((p
[1] & 0x3f) << 12)
1390 + ((p
[2] & 0x3f) << 6) + (p
[3] & 0x3f);
1391 if ((p
[4] & 0xc0) == 0x80)
1394 return ((p
[0] & 0x03) << 24) + ((p
[1] & 0x3f) << 18)
1395 + ((p
[2] & 0x3f) << 12) + ((p
[3] & 0x3f) << 6)
1397 if ((p
[5] & 0xc0) == 0x80 && len
== 6)
1398 return ((p
[0] & 0x01) << 30) + ((p
[1] & 0x3f) << 24)
1399 + ((p
[2] & 0x3f) << 18) + ((p
[3] & 0x3f) << 12)
1400 + ((p
[4] & 0x3f) << 6) + (p
[5] & 0x3f);
1405 /* Illegal value, just return the first byte */
1410 * Get character at **pp and advance *pp to the next character.
1411 * Note: composing characters are skipped!
1419 c
= (*mb_ptr2char
)(*pp
);
1420 *pp
+= (*mb_ptr2len
)(*pp
);
1425 * Get character at **pp and advance *pp to the next character.
1426 * Note: composing characters are returned as separate characters.
1429 mb_cptr2char_adv(pp
)
1434 c
= (*mb_ptr2char
)(*pp
);
1436 *pp
+= utf_ptr2len(*pp
);
1438 *pp
+= (*mb_ptr2len
)(*pp
);
1442 #if defined(FEAT_ARABIC) || defined(PROTO)
1444 * Check whether we are dealing with Arabic combining characters.
1445 * Note: these are NOT really composing characters!
1448 arabic_combine(one
, two
)
1449 int one
; /* first character */
1450 int two
; /* character just after "one" */
1453 return arabic_maycombine(two
);
1458 * Check whether we are dealing with a character that could be regarded as an
1459 * Arabic combining character, need to check the character before this.
1462 arabic_maycombine(two
)
1465 if (p_arshape
&& !p_tbidi
)
1466 return (two
== a_ALEF_MADDA
1467 || two
== a_ALEF_HAMZA_ABOVE
1468 || two
== a_ALEF_HAMZA_BELOW
1474 * Check if the character pointed to by "p2" is a composing character when it
1475 * comes after "p1". For Arabic sometimes "ab" is replaced with "c", which
1476 * behaves like a composing character.
1479 utf_composinglike(p1
, p2
)
1485 c2
= utf_ptr2char(p2
);
1486 if (utf_iscomposing(c2
))
1488 if (!arabic_maycombine(c2
))
1490 return arabic_combine(utf_ptr2char(p1
), c2
);
1495 * Convert a UTF-8 byte string to a wide character. Also get up to MAX_MCO
1496 * composing characters.
1499 utfc_ptr2char(p
, pcc
)
1501 int *pcc
; /* return: composing chars, last one is 0 */
1508 c
= utf_ptr2char(p
);
1509 len
= utf_ptr2len(p
);
1511 /* Only accept a composing char when the first char isn't illegal. */
1512 if ((len
> 1 || *p
< 0x80)
1514 && UTF_COMPOSINGLIKE(p
, p
+ len
))
1516 cc
= utf_ptr2char(p
+ len
);
1522 len
+= utf_ptr2len(p
+ len
);
1523 if (p
[len
] < 0x80 || !utf_iscomposing(cc
= utf_ptr2char(p
+ len
)))
1528 if (i
< MAX_MCO
) /* last composing char must be 0 */
1535 * Convert a UTF-8 byte string to a wide character. Also get up to MAX_MCO
1536 * composing characters. Use no more than p[maxlen].
1539 utfc_ptr2char_len(p
, pcc
, maxlen
)
1541 int *pcc
; /* return: composing chars, last one is 0 */
1549 c
= utf_ptr2char(p
);
1550 len
= utf_ptr2len_len(p
, maxlen
);
1551 /* Only accept a composing char when the first char isn't illegal. */
1552 if ((len
> 1 || *p
< 0x80)
1555 && UTF_COMPOSINGLIKE(p
, p
+ len
))
1557 cc
= utf_ptr2char(p
+ len
);
1563 len
+= utf_ptr2len_len(p
+ len
, maxlen
- len
);
1566 || !utf_iscomposing(cc
= utf_ptr2char(p
+ len
)))
1571 if (i
< MAX_MCO
) /* last composing char must be 0 */
1578 * Convert the character at screen position "off" to a sequence of bytes.
1579 * Includes the composing characters.
1580 * "buf" must at least have the length MB_MAXBYTES.
1581 * Returns the produced number of bytes.
1584 utfc_char2bytes(off
, buf
)
1591 len
= utf_char2bytes(ScreenLinesUC
[off
], buf
);
1592 for (i
= 0; i
< Screen_mco
; ++i
)
1594 if (ScreenLinesC
[i
][off
] == 0)
1596 len
+= utf_char2bytes(ScreenLinesC
[i
][off
], buf
+ len
);
1602 * Get the length of a UTF-8 byte sequence, not including any following
1603 * composing characters.
1605 * Returns 1 for an illegal byte sequence.
1616 len
= utf8len_tab
[*p
];
1617 for (i
= 1; i
< len
; ++i
)
1618 if ((p
[i
] & 0xc0) != 0x80)
1624 * Return length of UTF-8 character, obtained from the first byte.
1625 * "b" must be between 0 and 255!
1631 return utf8len_tab
[b
];
1635 * Get the length of UTF-8 byte sequence "p[size]". Does not include any
1636 * following composing characters.
1638 * Returns 1 for an illegal byte sequence.
1639 * Returns number > "size" for an incomplete byte sequence.
1642 utf_ptr2len_len(p
, size
)
1651 len
= utf8len_tab
[*p
];
1653 return len
; /* incomplete byte sequence. */
1654 for (i
= 1; i
< len
; ++i
)
1655 if ((p
[i
] & 0xc0) != 0x80)
1661 * Return the number of bytes the UTF-8 encoding of the character at "p" takes.
1662 * This includes following composing characters.
1676 if (b0
< 0x80 && p
[1] < 0x80) /* be quick for ASCII */
1679 /* Skip over first UTF-8 char, stopping at a NUL byte. */
1680 len
= utf_ptr2len(p
);
1682 /* Check for illegal byte. */
1683 if (len
== 1 && b0
>= 0x80)
1687 * Check for composing characters. We can handle only the first two, but
1688 * skip all of them (otherwise the cursor would get stuck).
1695 if (p
[len
] < 0x80 || !UTF_COMPOSINGLIKE(p
+ prevlen
, p
+ len
))
1698 /* Skip over composing char */
1702 len
+= utf_ptr2len(p
+ len
);
1707 * Return the number of bytes the UTF-8 encoding of the character at "p[size]"
1708 * takes. This includes following composing characters.
1709 * Returns 1 for an illegal char or an incomplete byte sequence.
1712 utfc_ptr2len_len(p
, size
)
1723 if (p
[0] < 0x80 && (size
== 1 || p
[1] < 0x80)) /* be quick for ASCII */
1726 /* Skip over first UTF-8 char, stopping at a NUL byte. */
1727 len
= utf_ptr2len_len(p
, size
);
1729 /* Check for illegal byte and incomplete byte sequence. */
1730 if ((len
== 1 && p
[0] >= 0x80) || len
> size
)
1734 * Check for composing characters. We can handle only the first two, but
1735 * skip all of them (otherwise the cursor would get stuck).
1742 if (p
[len
] < 0x80 || !UTF_COMPOSINGLIKE(p
+ prevlen
, p
+ len
))
1745 /* Skip over composing char */
1749 len
+= utf_ptr2len_len(p
+ len
, size
- len
);
1755 * Return the number of bytes the UTF-8 encoding of character "c" takes.
1756 * This does not include composing characters.
1776 * Convert Unicode character "c" to UTF-8 string in "buf[]".
1777 * Returns the number of bytes.
1778 * This does not include composing characters.
1781 utf_char2bytes(c
, buf
)
1785 if (c
< 0x80) /* 7 bits */
1790 if (c
< 0x800) /* 11 bits */
1792 buf
[0] = 0xc0 + ((unsigned)c
>> 6);
1793 buf
[1] = 0x80 + (c
& 0x3f);
1796 if (c
< 0x10000) /* 16 bits */
1798 buf
[0] = 0xe0 + ((unsigned)c
>> 12);
1799 buf
[1] = 0x80 + (((unsigned)c
>> 6) & 0x3f);
1800 buf
[2] = 0x80 + (c
& 0x3f);
1803 if (c
< 0x200000) /* 21 bits */
1805 buf
[0] = 0xf0 + ((unsigned)c
>> 18);
1806 buf
[1] = 0x80 + (((unsigned)c
>> 12) & 0x3f);
1807 buf
[2] = 0x80 + (((unsigned)c
>> 6) & 0x3f);
1808 buf
[3] = 0x80 + (c
& 0x3f);
1811 if (c
< 0x4000000) /* 26 bits */
1813 buf
[0] = 0xf8 + ((unsigned)c
>> 24);
1814 buf
[1] = 0x80 + (((unsigned)c
>> 18) & 0x3f);
1815 buf
[2] = 0x80 + (((unsigned)c
>> 12) & 0x3f);
1816 buf
[3] = 0x80 + (((unsigned)c
>> 6) & 0x3f);
1817 buf
[4] = 0x80 + (c
& 0x3f);
1821 buf
[0] = 0xfc + ((unsigned)c
>> 30);
1822 buf
[1] = 0x80 + (((unsigned)c
>> 24) & 0x3f);
1823 buf
[2] = 0x80 + (((unsigned)c
>> 18) & 0x3f);
1824 buf
[3] = 0x80 + (((unsigned)c
>> 12) & 0x3f);
1825 buf
[4] = 0x80 + (((unsigned)c
>> 6) & 0x3f);
1826 buf
[5] = 0x80 + (c
& 0x3f);
1831 * Return TRUE if "c" is a composing UTF-8 character. This means it will be
1832 * drawn on top of the preceding character.
1833 * Based on code from Markus Kuhn.
1839 /* sorted list of non-overlapping intervals */
1840 static struct interval combining
[] =
1842 {0x0300, 0x034f}, {0x0360, 0x036f}, {0x0483, 0x0486}, {0x0488, 0x0489},
1843 {0x0591, 0x05a1}, {0x05a3, 0x05b9}, {0x05bb, 0x05bd}, {0x05bf, 0x05bf},
1844 {0x05c1, 0x05c2}, {0x05c4, 0x05c4}, {0x0610, 0x0615}, {0x064b, 0x0658},
1845 {0x0670, 0x0670}, {0x06d6, 0x06dc}, {0x06de, 0x06e4}, {0x06e7, 0x06e8},
1846 {0x06ea, 0x06ed}, {0x0711, 0x0711}, {0x0730, 0x074a}, {0x07a6, 0x07b0},
1847 {0x0901, 0x0903}, {0x093c, 0x093c}, {0x093e, 0x094d}, {0x0951, 0x0954},
1848 {0x0962, 0x0963}, {0x0981, 0x0983}, {0x09bc, 0x09bc}, {0x09be, 0x09c4},
1849 {0x09c7, 0x09c8}, {0x09cb, 0x09cd}, {0x09d7, 0x09d7}, {0x09e2, 0x09e3},
1850 {0x0a01, 0x0a03}, {0x0a3c, 0x0a3c}, {0x0a3e, 0x0a42}, {0x0a47, 0x0a48},
1851 {0x0a4b, 0x0a4d}, {0x0a70, 0x0a71}, {0x0a81, 0x0a83}, {0x0abc, 0x0abc},
1852 {0x0abe, 0x0ac5}, {0x0ac7, 0x0ac9}, {0x0acb, 0x0acd}, {0x0ae2, 0x0ae3},
1853 {0x0b01, 0x0b03}, {0x0b3c, 0x0b3c}, {0x0b3e, 0x0b43}, {0x0b47, 0x0b48},
1854 {0x0b4b, 0x0b4d}, {0x0b56, 0x0b57}, {0x0b82, 0x0b82}, {0x0bbe, 0x0bc2},
1855 {0x0bc6, 0x0bc8}, {0x0bca, 0x0bcd}, {0x0bd7, 0x0bd7}, {0x0c01, 0x0c03},
1856 {0x0c3e, 0x0c44}, {0x0c46, 0x0c48}, {0x0c4a, 0x0c4d}, {0x0c55, 0x0c56},
1857 {0x0c82, 0x0c83}, {0x0cbc, 0x0cbc}, {0x0cbe, 0x0cc4}, {0x0cc6, 0x0cc8},
1858 {0x0cca, 0x0ccd}, {0x0cd5, 0x0cd6}, {0x0d02, 0x0d03}, {0x0d3e, 0x0d43},
1859 {0x0d46, 0x0d48}, {0x0d4a, 0x0d4d}, {0x0d57, 0x0d57}, {0x0d82, 0x0d83},
1860 {0x0dca, 0x0dca}, {0x0dcf, 0x0dd4}, {0x0dd6, 0x0dd6}, {0x0dd8, 0x0ddf},
1861 {0x0df2, 0x0df3}, {0x0e31, 0x0e31}, {0x0e34, 0x0e3a}, {0x0e47, 0x0e4e},
1862 {0x0eb1, 0x0eb1}, {0x0eb4, 0x0eb9}, {0x0ebb, 0x0ebc}, {0x0ec8, 0x0ecd},
1863 {0x0f18, 0x0f19}, {0x0f35, 0x0f35}, {0x0f37, 0x0f37}, {0x0f39, 0x0f39},
1864 {0x0f3e, 0x0f3f}, {0x0f71, 0x0f84}, {0x0f86, 0x0f87}, {0x0f90, 0x0f97},
1865 {0x0f99, 0x0fbc}, {0x0fc6, 0x0fc6}, {0x102c, 0x1032}, {0x1036, 0x1039},
1866 {0x1056, 0x1059}, {0x1712, 0x1714}, {0x1732, 0x1734}, {0x1752, 0x1753},
1867 {0x1772, 0x1773}, {0x17b6, 0x17d3}, {0x17dd, 0x17dd}, {0x180b, 0x180d},
1868 {0x18a9, 0x18a9}, {0x1920, 0x192b}, {0x1930, 0x193b}, {0x20d0, 0x20ea},
1869 {0x302a, 0x302f}, {0x3099, 0x309a}, {0xfb1e, 0xfb1e}, {0xfe00, 0xfe0f},
1873 return intable(combining
, sizeof(combining
), c
);
1877 * Return TRUE for characters that can be displayed in a normal way.
1878 * Only for characters of 0x100 and above!
1884 #ifdef USE_WCHAR_FUNCTIONS
1886 * Assume the iswprint() library function works better than our own stuff.
1890 /* Sorted list of non-overlapping intervals.
1891 * 0xd800-0xdfff is reserved for UTF-16, actually illegal. */
1892 static struct interval nonprint
[] =
1894 {0x070f, 0x070f}, {0x180b, 0x180e}, {0x200b, 0x200f}, {0x202a, 0x202e},
1895 {0x206a, 0x206f}, {0xd800, 0xdfff}, {0xfeff, 0xfeff}, {0xfff9, 0xfffb},
1899 return !intable(nonprint
, sizeof(nonprint
), c
);
1904 * Get class of a Unicode character.
1907 * 2 or bigger: some class of word character.
1913 /* sorted list of non-overlapping intervals */
1914 static struct clinterval
1916 unsigned short first
;
1917 unsigned short last
;
1918 unsigned short class;
1921 {0x037e, 0x037e, 1}, /* Greek question mark */
1922 {0x0387, 0x0387, 1}, /* Greek ano teleia */
1923 {0x055a, 0x055f, 1}, /* Armenian punctuation */
1924 {0x0589, 0x0589, 1}, /* Armenian full stop */
1925 {0x05be, 0x05be, 1},
1926 {0x05c0, 0x05c0, 1},
1927 {0x05c3, 0x05c3, 1},
1928 {0x05f3, 0x05f4, 1},
1929 {0x060c, 0x060c, 1},
1930 {0x061b, 0x061b, 1},
1931 {0x061f, 0x061f, 1},
1932 {0x066a, 0x066d, 1},
1933 {0x06d4, 0x06d4, 1},
1934 {0x0700, 0x070d, 1}, /* Syriac punctuation */
1935 {0x0964, 0x0965, 1},
1936 {0x0970, 0x0970, 1},
1937 {0x0df4, 0x0df4, 1},
1938 {0x0e4f, 0x0e4f, 1},
1939 {0x0e5a, 0x0e5b, 1},
1940 {0x0f04, 0x0f12, 1},
1941 {0x0f3a, 0x0f3d, 1},
1942 {0x0f85, 0x0f85, 1},
1943 {0x104a, 0x104f, 1}, /* Myanmar punctuation */
1944 {0x10fb, 0x10fb, 1}, /* Georgian punctuation */
1945 {0x1361, 0x1368, 1}, /* Ethiopic punctuation */
1946 {0x166d, 0x166e, 1}, /* Canadian Syl. punctuation */
1947 {0x1680, 0x1680, 0},
1948 {0x169b, 0x169c, 1},
1949 {0x16eb, 0x16ed, 1},
1950 {0x1735, 0x1736, 1},
1951 {0x17d4, 0x17dc, 1}, /* Khmer punctuation */
1952 {0x1800, 0x180a, 1}, /* Mongolian punctuation */
1953 {0x2000, 0x200b, 0}, /* spaces */
1954 {0x200c, 0x2027, 1}, /* punctuation and symbols */
1955 {0x2028, 0x2029, 0},
1956 {0x202a, 0x202e, 1}, /* punctuation and symbols */
1957 {0x202f, 0x202f, 0},
1958 {0x2030, 0x205e, 1}, /* punctuation and symbols */
1959 {0x205f, 0x205f, 0},
1960 {0x2060, 0x27ff, 1}, /* punctuation and symbols */
1961 {0x2070, 0x207f, 0x2070}, /* superscript */
1962 {0x2080, 0x208f, 0x2080}, /* subscript */
1963 {0x2983, 0x2998, 1},
1964 {0x29d8, 0x29db, 1},
1965 {0x29fc, 0x29fd, 1},
1966 {0x3000, 0x3000, 0}, /* ideographic space */
1967 {0x3001, 0x3020, 1}, /* ideographic punctuation */
1968 {0x3030, 0x3030, 1},
1969 {0x303d, 0x303d, 1},
1970 {0x3040, 0x309f, 0x3040}, /* Hiragana */
1971 {0x30a0, 0x30ff, 0x30a0}, /* Katakana */
1972 {0x3300, 0x9fff, 0x4e00}, /* CJK Ideographs */
1973 {0xac00, 0xd7a3, 0xac00}, /* Hangul Syllables */
1974 {0xf900, 0xfaff, 0x4e00}, /* CJK Ideographs */
1975 {0xfd3e, 0xfd3f, 1},
1976 {0xfe30, 0xfe6b, 1}, /* punctuation forms */
1977 {0xff00, 0xff0f, 1}, /* half/fullwidth ASCII */
1978 {0xff1a, 0xff20, 1}, /* half/fullwidth ASCII */
1979 {0xff3b, 0xff40, 1}, /* half/fullwidth ASCII */
1980 {0xff5b, 0xff65, 1}, /* half/fullwidth ASCII */
1983 int top
= sizeof(classes
) / sizeof(struct clinterval
) - 1;
1986 /* First quick check for Latin1 characters, use 'iskeyword'. */
1989 if (c
== ' ' || c
== '\t' || c
== NUL
|| c
== 0xa0)
1990 return 0; /* blank */
1992 return 2; /* word character */
1993 return 1; /* punctuation */
1996 /* binary search in table */
1999 mid
= (bot
+ top
) / 2;
2000 if (classes
[mid
].last
< c
)
2002 else if (classes
[mid
].first
> c
)
2005 return (int)classes
[mid
].class;
2008 /* most other characters are "word" characters */
2013 * Code for Unicode case-dependent operations. Based on notes in
2014 * http://www.unicode.org/Public/UNIDATA/CaseFolding.txt
2015 * This code uses simple case folding, not full case folding.
2019 * The following table is built by foldExtract.pl < CaseFolding.txt .
2020 * It must be in numeric order, because we use binary search on it.
2021 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range
2022 * from 0x41 to 0x5a inclusive, stepping by 1, are folded by adding 32.
2033 static convertStruct foldCase
[] =
2035 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32},
2036 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1},
2037 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121},
2038 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1},
2039 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205},
2040 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202},
2041 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205},
2042 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209},
2043 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213},
2044 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218},
2045 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1},
2046 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217},
2047 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1},
2048 {0x1c4,0x1c4,-1,2}, {0x1c5,0x1c5,-1,1}, {0x1c7,0x1c7,-1,2},
2049 {0x1c8,0x1c8,-1,1}, {0x1ca,0x1ca,-1,2}, {0x1cb,0x1db,2,1},
2050 {0x1de,0x1ee,2,1}, {0x1f1,0x1f1,-1,2}, {0x1f2,0x1f4,2,1},
2051 {0x1f6,0x1f6,-1,-97}, {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1},
2052 {0x220,0x220,-1,-130}, {0x222,0x232,2,1}, {0x386,0x386,-1,38},
2053 {0x388,0x38a,1,37}, {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63},
2054 {0x391,0x3a1,1,32}, {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1},
2055 {0x3f4,0x3f4,-1,-60}, {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7},
2056 {0x3fa,0x3fa,-1,1}, {0x400,0x40f,1,80}, {0x410,0x42f,1,32},
2057 {0x460,0x480,2,1}, {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1},
2058 {0x4d0,0x4f4,2,1}, {0x4f8,0x500,8,1}, {0x502,0x50e,2,1},
2059 {0x531,0x556,1,48}, {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1},
2060 {0x1f08,0x1f0f,1,-8}, {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8},
2061 {0x1f38,0x1f3f,1,-8}, {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8},
2062 {0x1f68,0x1f6f,1,-8}, {0x1f88,0x1f8f,1,-8}, {0x1f98,0x1f9f,1,-8},
2063 {0x1fa8,0x1faf,1,-8}, {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74},
2064 {0x1fbc,0x1fbc,-1,-9}, {0x1fc8,0x1fcb,1,-86}, {0x1fcc,0x1fcc,-1,-9},
2065 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8},
2066 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128},
2067 {0x1ffa,0x1ffb,1,-126}, {0x1ffc,0x1ffc,-1,-9}, {0x2126,0x2126,-1,-7517},
2068 {0x212a,0x212a,-1,-8383}, {0x212b,0x212b,-1,-8262},
2069 {0x2160,0x216f,1,16}, {0x24b6,0x24cf,1,26}, {0xff21,0xff3a,1,32},
2070 {0x10400,0x10427,1,40}
2073 static int utf_convert(int a
, convertStruct table
[], int tableSize
);
2076 * Generic conversion function for case operations.
2077 * Return the converted equivalent of "a", which is a UCS-4 character. Use
2078 * the given conversion "table". Uses binary search on "table".
2081 utf_convert(a
, table
, tableSize
)
2083 convertStruct table
[];
2086 int start
, mid
, end
; /* indices into table */
2089 end
= tableSize
/ sizeof(convertStruct
);
2092 /* need to search further */
2093 mid
= (end
+ start
) /2;
2094 if (table
[mid
].rangeEnd
< a
)
2099 if (table
[start
].rangeStart
<= a
&& a
<= table
[start
].rangeEnd
2100 && (a
- table
[start
].rangeStart
) % table
[start
].step
== 0)
2101 return (a
+ table
[start
].offset
);
2107 * Return the folded-case equivalent of "a", which is a UCS-4 character. Uses
2108 * simple case folding.
2114 return utf_convert(a
, foldCase
, sizeof(foldCase
));
2118 * The following tables are built by upperLowerExtract.pl < UnicodeData.txt .
2119 * They must be in numeric order, because we use binary search on them.
2120 * An entry such as {0x41,0x5a,1,32} means that UCS-4 characters in the range
2121 * from 0x41 to 0x5a inclusive, stepping by 1, are switched to lower (for
2122 * example) by adding 32.
2124 static convertStruct toLower
[] =
2126 {0x41,0x5a,1,32}, {0xc0,0xd6,1,32}, {0xd8,0xde,1,32},
2127 {0x100,0x12e,2,1}, {0x130,0x130,-1,-199}, {0x132,0x136,2,1},
2128 {0x139,0x147,2,1}, {0x14a,0x176,2,1}, {0x178,0x178,-1,-121},
2129 {0x179,0x17d,2,1}, {0x181,0x181,-1,210}, {0x182,0x184,2,1},
2130 {0x186,0x186,-1,206}, {0x187,0x187,-1,1}, {0x189,0x18a,1,205},
2131 {0x18b,0x18b,-1,1}, {0x18e,0x18e,-1,79}, {0x18f,0x18f,-1,202},
2132 {0x190,0x190,-1,203}, {0x191,0x191,-1,1}, {0x193,0x193,-1,205},
2133 {0x194,0x194,-1,207}, {0x196,0x196,-1,211}, {0x197,0x197,-1,209},
2134 {0x198,0x198,-1,1}, {0x19c,0x19c,-1,211}, {0x19d,0x19d,-1,213},
2135 {0x19f,0x19f,-1,214}, {0x1a0,0x1a4,2,1}, {0x1a6,0x1a6,-1,218},
2136 {0x1a7,0x1a7,-1,1}, {0x1a9,0x1a9,-1,218}, {0x1ac,0x1ac,-1,1},
2137 {0x1ae,0x1ae,-1,218}, {0x1af,0x1af,-1,1}, {0x1b1,0x1b2,1,217},
2138 {0x1b3,0x1b5,2,1}, {0x1b7,0x1b7,-1,219}, {0x1b8,0x1bc,4,1},
2139 {0x1c4,0x1ca,3,2}, {0x1cd,0x1db,2,1}, {0x1de,0x1ee,2,1},
2140 {0x1f1,0x1f1,-1,2}, {0x1f4,0x1f4,-1,1}, {0x1f6,0x1f6,-1,-97},
2141 {0x1f7,0x1f7,-1,-56}, {0x1f8,0x21e,2,1}, {0x220,0x220,-1,-130},
2142 {0x222,0x232,2,1}, {0x386,0x386,-1,38}, {0x388,0x38a,1,37},
2143 {0x38c,0x38c,-1,64}, {0x38e,0x38f,1,63}, {0x391,0x3a1,1,32},
2144 {0x3a3,0x3ab,1,32}, {0x3d8,0x3ee,2,1}, {0x3f4,0x3f4,-1,-60},
2145 {0x3f7,0x3f7,-1,1}, {0x3f9,0x3f9,-1,-7}, {0x3fa,0x3fa,-1,1},
2146 {0x400,0x40f,1,80}, {0x410,0x42f,1,32}, {0x460,0x480,2,1},
2147 {0x48a,0x4be,2,1}, {0x4c1,0x4cd,2,1}, {0x4d0,0x4f4,2,1},
2148 {0x4f8,0x500,8,1}, {0x502,0x50e,2,1}, {0x531,0x556,1,48},
2149 {0x1e00,0x1e94,2,1}, {0x1ea0,0x1ef8,2,1}, {0x1f08,0x1f0f,1,-8},
2150 {0x1f18,0x1f1d,1,-8}, {0x1f28,0x1f2f,1,-8}, {0x1f38,0x1f3f,1,-8},
2151 {0x1f48,0x1f4d,1,-8}, {0x1f59,0x1f5f,2,-8}, {0x1f68,0x1f6f,1,-8},
2152 {0x1fb8,0x1fb9,1,-8}, {0x1fba,0x1fbb,1,-74}, {0x1fc8,0x1fcb,1,-86},
2153 {0x1fd8,0x1fd9,1,-8}, {0x1fda,0x1fdb,1,-100}, {0x1fe8,0x1fe9,1,-8},
2154 {0x1fea,0x1feb,1,-112}, {0x1fec,0x1fec,-1,-7}, {0x1ff8,0x1ff9,1,-128},
2155 {0x1ffa,0x1ffb,1,-126}, {0x2126,0x2126,-1,-7517}, {0x212a,0x212a,-1,-8383},
2156 {0x212b,0x212b,-1,-8262}, {0xff21,0xff3a,1,32}, {0x10400,0x10427,1,40}
2159 static convertStruct toUpper
[] =
2161 {0x61,0x7a,1,-32}, {0xb5,0xb5,-1,743}, {0xe0,0xf6,1,-32},
2162 {0xf8,0xfe,1,-32}, {0xff,0xff,-1,121}, {0x101,0x12f,2,-1},
2163 {0x131,0x131,-1,-232}, {0x133,0x137,2,-1}, {0x13a,0x148,2,-1},
2164 {0x14b,0x177,2,-1}, {0x17a,0x17e,2,-1}, {0x17f,0x17f,-1,-300},
2165 {0x183,0x185,2,-1}, {0x188,0x18c,4,-1}, {0x192,0x192,-1,-1},
2166 {0x195,0x195,-1,97}, {0x199,0x199,-1,-1}, {0x19e,0x19e,-1,130},
2167 {0x1a1,0x1a5,2,-1}, {0x1a8,0x1ad,5,-1}, {0x1b0,0x1b4,4,-1},
2168 {0x1b6,0x1b9,3,-1}, {0x1bd,0x1bd,-1,-1}, {0x1bf,0x1bf,-1,56},
2169 {0x1c5,0x1c6,1,-1}, {0x1c8,0x1c9,1,-1}, {0x1cb,0x1cc,1,-1},
2170 {0x1ce,0x1dc,2,-1}, {0x1dd,0x1dd,-1,-79}, {0x1df,0x1ef,2,-1},
2171 {0x1f2,0x1f3,1,-1}, {0x1f5,0x1f9,4,-1}, {0x1fb,0x21f,2,-1},
2172 {0x223,0x233,2,-1}, {0x253,0x253,-1,-210}, {0x254,0x254,-1,-206},
2173 {0x256,0x257,1,-205}, {0x259,0x259,-1,-202}, {0x25b,0x25b,-1,-203},
2174 {0x260,0x260,-1,-205}, {0x263,0x263,-1,-207}, {0x268,0x268,-1,-209},
2175 {0x269,0x26f,6,-211}, {0x272,0x272,-1,-213}, {0x275,0x275,-1,-214},
2176 {0x280,0x283,3,-218}, {0x288,0x288,-1,-218}, {0x28a,0x28b,1,-217},
2177 {0x292,0x292,-1,-219}, {0x3ac,0x3ac,-1,-38}, {0x3ad,0x3af,1,-37},
2178 {0x3b1,0x3c1,1,-32}, {0x3c2,0x3c2,-1,-31}, {0x3c3,0x3cb,1,-32},
2179 {0x3cc,0x3cc,-1,-64}, {0x3cd,0x3ce,1,-63}, {0x3d0,0x3d0,-1,-62},
2180 {0x3d1,0x3d1,-1,-57}, {0x3d5,0x3d5,-1,-47}, {0x3d6,0x3d6,-1,-54},
2181 {0x3d9,0x3ef,2,-1}, {0x3f0,0x3f0,-1,-86}, {0x3f1,0x3f1,-1,-80},
2182 {0x3f2,0x3f2,-1,7}, {0x3f5,0x3f5,-1,-96}, {0x3f8,0x3fb,3,-1},
2183 {0x430,0x44f,1,-32}, {0x450,0x45f,1,-80}, {0x461,0x481,2,-1},
2184 {0x48b,0x4bf,2,-1}, {0x4c2,0x4ce,2,-1}, {0x4d1,0x4f5,2,-1},
2185 {0x4f9,0x501,8,-1}, {0x503,0x50f,2,-1}, {0x561,0x586,1,-48},
2186 {0x1e01,0x1e95,2,-1}, {0x1e9b,0x1e9b,-1,-59}, {0x1ea1,0x1ef9,2,-1},
2187 {0x1f00,0x1f07,1,8}, {0x1f10,0x1f15,1,8}, {0x1f20,0x1f27,1,8},
2188 {0x1f30,0x1f37,1,8}, {0x1f40,0x1f45,1,8}, {0x1f51,0x1f57,2,8},
2189 {0x1f60,0x1f67,1,8}, {0x1f70,0x1f71,1,74}, {0x1f72,0x1f75,1,86},
2190 {0x1f76,0x1f77,1,100}, {0x1f78,0x1f79,1,128}, {0x1f7a,0x1f7b,1,112},
2191 {0x1f7c,0x1f7d,1,126}, {0x1f80,0x1f87,1,8}, {0x1f90,0x1f97,1,8},
2192 {0x1fa0,0x1fa7,1,8}, {0x1fb0,0x1fb1,1,8}, {0x1fb3,0x1fb3,-1,9},
2193 {0x1fbe,0x1fbe,-1,-7205}, {0x1fc3,0x1fc3,-1,9}, {0x1fd0,0x1fd1,1,8},
2194 {0x1fe0,0x1fe1,1,8}, {0x1fe5,0x1fe5,-1,7}, {0x1ff3,0x1ff3,-1,9},
2195 {0xff41,0xff5a,1,-32}, {0x10428,0x1044f,1,-40}
2199 * Return the upper-case equivalent of "a", which is a UCS-4 character. Use
2200 * simple case folding.
2206 /* If 'casemap' contains "keepascii" use ASCII style toupper(). */
2207 if (a
< 128 && (cmp_flags
& CMP_KEEPASCII
))
2208 return TOUPPER_ASC(a
);
2210 #if defined(HAVE_TOWUPPER) && defined(__STDC_ISO_10646__)
2211 /* If towupper() is available and handles Unicode, use it. */
2212 if (!(cmp_flags
& CMP_INTERNAL
))
2216 /* For characters below 128 use locale sensitive toupper(). */
2218 return TOUPPER_LOC(a
);
2220 /* For any other characters use the above mapping table. */
2221 return utf_convert(a
, toUpper
, sizeof(toUpper
));
2228 return (utf_toupper(a
) != a
);
2232 * Return the lower-case equivalent of "a", which is a UCS-4 character. Use
2233 * simple case folding.
2239 /* If 'casemap' contains "keepascii" use ASCII style tolower(). */
2240 if (a
< 128 && (cmp_flags
& CMP_KEEPASCII
))
2241 return TOLOWER_ASC(a
);
2243 #if defined(HAVE_TOWLOWER) && defined(__STDC_ISO_10646__)
2244 /* If towlower() is available and handles Unicode, use it. */
2245 if (!(cmp_flags
& CMP_INTERNAL
))
2249 /* For characters below 128 use locale sensitive tolower(). */
2251 return TOLOWER_LOC(a
);
2253 /* For any other characters use the above mapping table. */
2254 return utf_convert(a
, toLower
, sizeof(toLower
));
2261 return (utf_tolower(a
) != a
);
2265 * Version of strnicmp() that handles multi-byte characters.
2266 * Needed for Big5, Sjift-JIS and UTF-8 encoding. Other DBCS encodings can
2267 * probably use strnicmp(), because there are no ASCII characters in the
2269 * Returns zero if s1 and s2 are equal (ignoring case), the difference between
2270 * two characters otherwise.
2273 mb_strnicmp(s1
, s2
, nn
)
2279 int incomplete
= FALSE
;
2282 for (i
= 0; i
< n
; i
+= l
)
2284 if (s1
[i
] == NUL
&& s2
[i
] == NUL
) /* both strings end */
2288 l
= utf_byte2len(s1
[i
]);
2291 l
= n
- i
; /* incomplete character */
2294 /* Check directly first, it's faster. */
2295 for (j
= 0; j
< l
; ++j
)
2297 if (s1
[i
+ j
] != s2
[i
+ j
])
2300 /* Both stings have the same bytes but are incomplete or
2301 * have illegal bytes, accept them as equal. */
2306 /* If one of the two characters is incomplete return -1. */
2307 if (incomplete
|| i
+ utf_byte2len(s2
[i
]) > n
)
2309 cdiff
= utf_fold(utf_ptr2char(s1
+ i
))
2310 - utf_fold(utf_ptr2char(s2
+ i
));
2317 l
= (*mb_ptr2len
)(s1
+ i
);
2320 /* Single byte: first check normally, then with ignore case. */
2323 cdiff
= MB_TOLOWER(s1
[i
]) - MB_TOLOWER(s2
[i
]);
2330 /* For non-Unicode multi-byte don't ignore case. */
2333 cdiff
= STRNCMP(s1
+ i
, s2
+ i
, l
);
2343 * "g8": show bytes of the UTF-8 char under the cursor. Doesn't matter what
2344 * 'encoding' has been set to.
2355 /* Get the byte length of the char under the cursor, including composing
2357 line
= ml_get_cursor();
2358 len
= utfc_ptr2len(line
);
2366 for (i
= 0; i
< len
; ++i
)
2370 /* start of (composing) character, get its length */
2373 STRCPY(IObuff
+ rlen
, "+ ");
2376 clen
= utf_ptr2len(line
+ i
);
2378 sprintf((char *)IObuff
+ rlen
, "%02x ", line
[i
]);
2380 rlen
+= (int)STRLEN(IObuff
+ rlen
);
2381 if (rlen
> IOSIZE
- 20)
2389 * mb_head_off() function pointer.
2390 * Return offset from "p" to the first byte of the character it points into.
2391 * Returns 0 when already at the first byte of a character.
2395 latin_head_off(base
, p
)
2403 dbcs_head_off(base
, p
)
2409 /* It can't be a trailing byte when not using DBCS, at the start of the
2410 * string or the previous byte can't start a double-byte. */
2411 if (p
<= base
|| MB_BYTE2LEN(p
[-1]) == 1)
2414 /* This is slow: need to start at the base and go forward until the
2415 * byte we are looking for. Return 1 when we went past it, 0 otherwise. */
2418 q
+= dbcs_ptr2len(q
);
2419 return (q
== p
) ? 0 : 1;
2422 #if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \
2425 * Special version of dbcs_head_off() that works for ScreenLines[], where
2426 * single-width DBCS_JPNU characters are stored separately.
2429 dbcs_screen_head_off(base
, p
)
2435 /* It can't be a trailing byte when not using DBCS, at the start of the
2436 * string or the previous byte can't start a double-byte.
2437 * For euc-jp an 0x8e byte in the previous cell always means we have a
2438 * lead byte in the current cell. */
2440 || (enc_dbcs
== DBCS_JPNU
&& p
[-1] == 0x8e)
2441 || MB_BYTE2LEN(p
[-1]) == 1)
2444 /* This is slow: need to start at the base and go forward until the
2445 * byte we are looking for. Return 1 when we went past it, 0 otherwise.
2446 * For DBCS_JPNU look out for 0x8e, which means the second byte is not
2447 * stored as the next byte. */
2451 if (enc_dbcs
== DBCS_JPNU
&& *q
== 0x8e)
2454 q
+= dbcs_ptr2len(q
);
2456 return (q
== p
) ? 0 : 1;
2461 utf_head_off(base
, p
)
2472 if (*p
< 0x80) /* be quick for ASCII */
2475 /* Skip backwards over trailing bytes: 10xx.xxxx
2476 * Skip backwards again if on a composing char. */
2479 /* Move s to the last byte of this char. */
2480 for (s
= q
; (s
[1] & 0xc0) == 0x80; ++s
)
2482 /* Move q to the first byte of this char. */
2483 while (q
> base
&& (*q
& 0xc0) == 0x80)
2485 /* Check for illegal sequence. Do allow an illegal byte after where we
2487 if (utf8len_tab
[*q
] != (int)(s
- q
+ 1)
2488 && utf8len_tab
[*q
] != (int)(p
- q
+ 1))
2494 c
= utf_ptr2char(q
);
2495 if (utf_iscomposing(c
))
2499 if (arabic_maycombine(c
))
2501 /* Advance to get a sneak-peak at the next char */
2504 /* Move j to the first byte of this char. */
2505 while (j
> base
&& (*j
& 0xc0) == 0x80)
2507 if (arabic_combine(utf_ptr2char(j
), c
))
2514 return (int)(p
- q
);
2517 #if defined(FEAT_EVAL) || defined(PROTO)
2519 * Copy a character from "*fp" to "*tp" and advance the pointers.
2522 mb_copy_char(fp
, tp
)
2526 int l
= (*mb_ptr2len
)(*fp
);
2528 mch_memmove(*tp
, *fp
, (size_t)l
);
2535 * Return the offset from "p" to the first byte of a character. When "p" is
2536 * at the start of a character 0 is returned, otherwise the offset to the next
2537 * character. Can start anywhere in a stream of bytes.
2540 mb_off_next(base
, p
)
2549 if (*p
< 0x80) /* be quick for ASCII */
2552 /* Find the next character that isn't 10xx.xxxx */
2553 for (i
= 0; (p
[i
] & 0xc0) == 0x80; ++i
)
2557 /* Check for illegal sequence. */
2558 for (j
= 0; p
- j
> base
; ++j
)
2559 if ((p
[-j
] & 0xc0) != 0x80)
2561 if (utf8len_tab
[p
[-j
]] != i
+ j
)
2567 /* Only need to check if we're on a trail byte, it doesn't matter if we
2568 * want the offset to the next or current character. */
2569 return (*mb_head_off
)(base
, p
);
2573 * Return the offset from "p" to the last byte of the character it points
2574 * into. Can start anywhere in a stream of bytes.
2577 mb_tail_off(base
, p
)
2589 /* Find the last character that is 10xx.xxxx */
2590 for (i
= 0; (p
[i
+ 1] & 0xc0) == 0x80; ++i
)
2592 /* Check for illegal sequence. */
2593 for (j
= 0; p
- j
> base
; ++j
)
2594 if ((p
[-j
] & 0xc0) != 0x80)
2596 if (utf8len_tab
[p
[-j
]] != i
+ j
+ 1)
2601 /* It can't be the first byte if a double-byte when not using DBCS, at the
2602 * end of the string or the byte can't start a double-byte. */
2603 if (enc_dbcs
== 0 || p
[1] == NUL
|| MB_BYTE2LEN(*p
) == 1)
2606 /* Return 1 when on the lead byte, 0 when on the tail byte. */
2607 return 1 - dbcs_head_off(base
, p
);
2611 * Find the next illegal byte sequence.
2616 pos_T pos
= curwin
->w_cursor
;
2620 char_u
*tofree
= NULL
;
2622 vimconv
.vc_type
= CONV_NONE
;
2623 if (enc_utf8
&& (enc_canon_props(curbuf
->b_p_fenc
) & ENC_8BIT
))
2625 /* 'encoding' is "utf-8" but we are editing a 8-bit encoded file,
2626 * possibly a utf-8 file with illegal bytes. Setup for conversion
2627 * from utf-8 to 'fileencoding'. */
2628 convert_setup(&vimconv
, p_enc
, curbuf
->b_p_fenc
);
2631 #ifdef FEAT_VIRTUALEDIT
2632 curwin
->w_cursor
.coladd
= 0;
2636 p
= ml_get_cursor();
2637 if (vimconv
.vc_type
!= CONV_NONE
)
2640 tofree
= string_convert(&vimconv
, p
, NULL
);
2648 /* Illegal means that there are not enough trail bytes (checked by
2649 * utf_ptr2len()) or too many of them (overlong sequence). */
2650 len
= utf_ptr2len(p
);
2651 if (*p
>= 0x80 && (len
== 1
2652 || utf_char2len(utf_ptr2char(p
)) != len
))
2654 if (vimconv
.vc_type
== CONV_NONE
)
2655 curwin
->w_cursor
.col
+= (colnr_T
)(p
- ml_get_cursor());
2660 len
= (int)(p
- tofree
);
2661 for (p
= ml_get_cursor(); *p
!= NUL
&& len
-- > 0; p
+= l
)
2664 curwin
->w_cursor
.col
+= l
;
2671 if (curwin
->w_cursor
.lnum
== curbuf
->b_ml
.ml_line_count
)
2673 ++curwin
->w_cursor
.lnum
;
2674 curwin
->w_cursor
.col
= 0;
2677 /* didn't find it: don't move and beep */
2678 curwin
->w_cursor
= pos
;
2683 convert_setup(&vimconv
, NULL
, NULL
);
2686 #if defined(HAVE_GTK2) || defined(PROTO)
2688 * Return TRUE if string "s" is a valid utf-8 string.
2689 * When "end" is NULL stop at the first NUL.
2690 * When "end" is positive stop there.
2693 utf_valid_string(s
, end
)
2700 while (end
== NULL
? *p
!= NUL
: p
< end
)
2702 if ((*p
& 0xc0) == 0x80)
2703 return FALSE
; /* invalid lead byte */
2704 l
= utf8len_tab
[*p
];
2705 if (end
!= NULL
&& p
+ l
> end
)
2706 return FALSE
; /* incomplete byte sequence */
2709 if ((*p
++ & 0xc0) != 0x80)
2710 return FALSE
; /* invalid trail byte */
2716 #if defined(FEAT_GUI) || defined(PROTO)
2718 * Special version of mb_tail_off() for use in ScreenLines[].
2721 dbcs_screen_tail_off(base
, p
)
2725 /* It can't be the first byte if a double-byte when not using DBCS, at the
2726 * end of the string or the byte can't start a double-byte.
2727 * For euc-jp an 0x8e byte always means we have a lead byte in the current
2729 if (*p
== NUL
|| p
[1] == NUL
2730 || (enc_dbcs
== DBCS_JPNU
&& *p
== 0x8e)
2731 || MB_BYTE2LEN(*p
) == 1)
2734 /* Return 1 when on the lead byte, 0 when on the tail byte. */
2735 return 1 - dbcs_screen_head_off(base
, p
);
2740 * If the cursor moves on an trail byte, set the cursor on the lead byte.
2741 * Thus it moves left if necessary.
2742 * Return TRUE when the cursor was adjusted.
2747 mb_adjustpos(&curwin
->w_cursor
);
2751 * Adjust position "*lp" to point to the first byte of a multi-byte character.
2752 * If it points to a tail byte it's moved backwards to the head byte.
2761 #ifdef FEAT_VIRTUALEDIT
2766 p
= ml_get(lp
->lnum
);
2767 lp
->col
-= (*mb_head_off
)(p
, p
+ lp
->col
);
2768 #ifdef FEAT_VIRTUALEDIT
2769 /* Reset "coladd" when the cursor would be on the right half of a
2770 * double-wide character. */
2772 && p
[lp
->col
] != TAB
2773 && vim_isprintc((*mb_ptr2char
)(p
+ lp
->col
))
2774 && ptr2cells(p
+ lp
->col
) > 1)
2781 * Return a pointer to the character before "*p", if there is one.
2785 char_u
*line
; /* start of the string */
2789 mb_ptr_back(line
, p
);
2794 * Return the character length of "str". Each multi-byte character (with
2795 * following composing characters) counts as one.
2807 for (count
= 0; *p
!= NUL
; count
++)
2808 p
+= (*mb_ptr2len
)(p
);
2813 #if defined(FEAT_SPELL) || defined(PROTO)
2815 * Like mb_charlen() but for a string with specified length.
2818 mb_charlen_len(str
, len
)
2825 for (count
= 0; *p
!= NUL
&& p
< str
+ len
; count
++)
2826 p
+= (*mb_ptr2len
)(p
);
2833 * Try to un-escape a multi-byte character.
2834 * Used for the "to" and "from" part of a mapping.
2835 * Return the un-escaped string if it is a multi-byte character, and advance
2836 * "pp" to just after the bytes that formed it.
2837 * Return NULL if no multi-byte char was found.
2843 static char_u buf
[MB_MAXBYTES
+ 1];
2847 /* Must translate K_SPECIAL KS_SPECIAL KE_FILLER to K_SPECIAL and CSI
2848 * KS_EXTRA KE_CSI to CSI. */
2849 for (n
= 0; str
[n
] != NUL
&& m
<= MB_MAXBYTES
; ++n
)
2851 if (str
[n
] == K_SPECIAL
2852 && str
[n
+ 1] == KS_SPECIAL
2853 && str
[n
+ 2] == KE_FILLER
)
2855 buf
[m
++] = K_SPECIAL
;
2859 else if (str
[n
] == CSI
2860 && str
[n
+ 1] == KS_EXTRA
2861 && str
[n
+ 2] == (int)KE_CSI
)
2867 else if (str
[n
] == K_SPECIAL
2872 break; /* a special key can't be a multibyte char */
2877 /* Return a multi-byte character if it's found. An illegal sequence
2878 * will result in a 1 here. */
2879 if ((*mb_ptr2len
)(buf
) > 1)
2889 * Return TRUE if the character at "row"/"col" on the screen is the left side
2890 * of a double-width character.
2891 * Caller must make sure "row" and "col" are not invalid!
2894 mb_lefthalve(row
, col
)
2898 #ifdef FEAT_HANGULIN
2899 if (composing_hangul
)
2903 return dbcs_off2cells(LineOffset
[row
] + col
) > 1;
2905 return (col
+ 1 < Columns
2906 && ScreenLines
[LineOffset
[row
] + col
+ 1] == 0);
2910 # if defined(FEAT_CLIPBOARD) || defined(FEAT_GUI) || defined(FEAT_RIGHTLEFT) \
2913 * Correct a position on the screen, if it's the right halve of a double-wide
2914 * char move it to the left halve. Returns the corrected column.
2917 mb_fix_col(col
, row
)
2921 col
= check_col(col
);
2922 row
= check_row(row
);
2923 if (has_mbyte
&& ScreenLines
!= NULL
&& col
> 0
2925 && ScreenLines
[LineOffset
[row
] + col
] != NUL
2926 && dbcs_screen_head_off(ScreenLines
+ LineOffset
[row
],
2927 ScreenLines
+ LineOffset
[row
] + col
))
2928 || (enc_utf8
&& ScreenLines
[LineOffset
[row
] + col
] == 0)))
2935 #if defined(FEAT_MBYTE) || defined(FEAT_POSTSCRIPT) || defined(PROTO)
2936 static int enc_alias_search
__ARGS((char_u
*name
));
2939 * Skip the Vim specific head of a 'encoding' name.
2945 if (STRNCMP(p
, "2byte-", 6) == 0)
2947 if (STRNCMP(p
, "8bit-", 5) == 0)
2953 * Find the canonical name for encoding "enc".
2954 * When the name isn't recognized, returns "enc" itself, but with all lower
2955 * case characters and '_' replaced with '-'.
2956 * Returns an allocated string. NULL for out-of-memory.
2967 if (STRCMP(enc
, "default") == 0)
2969 /* Use the default encoding as it's found by set_init_1(). */
2970 r
= get_encoding_default();
2972 r
= (char_u
*)"latin1";
2973 return vim_strsave(r
);
2977 /* copy "enc" to allocated memory, with room for two '-' */
2978 r
= alloc((unsigned)(STRLEN(enc
) + 3));
2981 /* Make it all lower case and replace '_' with '-'. */
2983 for (s
= enc
; *s
!= NUL
; ++s
)
2988 *p
++ = TOLOWER_ASC(*s
);
2992 /* Skip "2byte-" and "8bit-". */
2995 /* Change "microsoft-cp" to "cp". Used in some spell files. */
2996 if (STRNCMP(p
, "microsoft-cp", 12) == 0)
2997 mch_memmove(p
, p
+ 10, STRLEN(p
+ 10) + 1);
2999 /* "iso8859" -> "iso-8859" */
3000 if (STRNCMP(p
, "iso8859", 7) == 0)
3002 mch_memmove(p
+ 4, p
+ 3, STRLEN(p
+ 2));
3006 /* "iso-8859n" -> "iso-8859-n" */
3007 if (STRNCMP(p
, "iso-8859", 8) == 0 && p
[8] != '-')
3009 mch_memmove(p
+ 9, p
+ 8, STRLEN(p
+ 7));
3013 /* "latin-N" -> "latinN" */
3014 if (STRNCMP(p
, "latin-", 6) == 0)
3015 mch_memmove(p
+ 5, p
+ 6, STRLEN(p
+ 5));
3017 if (enc_canon_search(p
) >= 0)
3019 /* canonical name can be used unmodified */
3021 mch_memmove(r
, p
, STRLEN(p
) + 1);
3023 else if ((i
= enc_alias_search(p
)) >= 0)
3025 /* alias recognized, get canonical name */
3027 r
= vim_strsave((char_u
*)enc_canon_table
[i
].name
);
3034 * Search for an encoding alias of "name".
3035 * Returns -1 when not found.
3038 enc_alias_search(name
)
3043 for (i
= 0; enc_alias_table
[i
].name
!= NULL
; ++i
)
3044 if (STRCMP(name
, enc_alias_table
[i
].name
) == 0)
3045 return enc_alias_table
[i
].canon
;
3050 #if defined(FEAT_MBYTE) || defined(PROTO)
3052 #ifdef HAVE_LANGINFO_H
3053 # include <langinfo.h>
3057 * Get the canonicalized encoding of the current locale.
3058 * Returns an allocated string when successful, NULL when not.
3070 long acp
= GetACP();
3073 STRCPY(buf
, "ucs-2le");
3074 else if (acp
== 1252) /* cp1252 is used as latin1 */
3075 STRCPY(buf
, "latin1");
3077 sprintf(buf
, "cp%ld", acp
);
3079 # ifdef HAVE_NL_LANGINFO_CODESET
3080 if ((s
= nl_langinfo(CODESET
)) == NULL
|| *s
== NUL
)
3082 # if defined(HAVE_LOCALE_H) || defined(X_LOCALE)
3083 if ((s
= setlocale(LC_CTYPE
, NULL
)) == NULL
|| *s
== NUL
)
3085 if ((s
= getenv("LC_ALL")) == NULL
|| *s
== NUL
)
3086 if ((s
= getenv("LC_CTYPE")) == NULL
|| *s
== NUL
)
3089 if (s
== NULL
|| *s
== NUL
)
3092 /* The most generic locale format is:
3093 * language[_territory][.codeset][@modifier][+special][,[sponsor][_revision]]
3094 * If there is a '.' remove the part before it.
3095 * if there is something after the codeset, remove it.
3096 * Make the name lowercase and replace '_' with '-'.
3097 * Exception: "ja_JP.EUC" == "euc-jp", "zh_CN.EUC" = "euc-cn",
3098 * "ko_KR.EUC" == "euc-kr"
3100 if ((p
= (char *)vim_strchr((char_u
*)s
, '.')) != NULL
)
3102 if (p
> s
+ 2 && STRNICMP(p
+ 1, "EUC", 3) == 0
3103 && !isalnum((int)p
[4]) && p
[4] != '-' && p
[-3] == '_')
3105 /* copy "XY.EUC" to "euc-XY" to buf[10] */
3106 STRCPY(buf
+ 10, "euc-");
3115 for (i
= 0; s
[i
] != NUL
&& s
+ i
< buf
+ sizeof(buf
) - 1; ++i
)
3117 if (s
[i
] == '_' || s
[i
] == '-')
3119 else if (isalnum((int)s
[i
]))
3120 buf
[i
] = TOLOWER_ASC(s
[i
]);
3127 return enc_canonize((char_u
*)buf
);
3130 #if defined(WIN3264) || defined(PROTO)
3132 * Convert an encoding name to an MS-Windows codepage.
3133 * Returns zero if no codepage can be figured out.
3136 encname2codepage(name
)
3143 if (STRNCMP(p
, "8bit-", 5) == 0)
3145 else if (STRNCMP(p_enc
, "2byte-", 6) == 0)
3148 if (p
[0] == 'c' && p
[1] == 'p')
3150 else if ((idx
= enc_canon_search(p
)) >= 0)
3151 cp
= enc_canon_table
[idx
].codepage
;
3154 if (IsValidCodePage(cp
))
3160 # if defined(USE_ICONV) || defined(PROTO)
3162 static char_u
*iconv_string
__ARGS((vimconv_T
*vcp
, char_u
*str
, int slen
, int *unconvlenp
));
3165 * Call iconv_open() with a check if iconv() works properly (there are broken
3167 * Returns (void *)-1 if failed.
3168 * (should return iconv_t, but that causes problems with prototypes).
3171 my_iconv_open(to
, from
)
3176 #define ICONV_TESTLEN 400
3177 char_u tobuf
[ICONV_TESTLEN
];
3180 static int iconv_ok
= -1;
3182 if (iconv_ok
== FALSE
)
3183 return (void *)-1; /* detected a broken iconv() previously */
3185 #ifdef DYNAMIC_ICONV
3186 /* Check if the iconv.dll can be found. */
3187 if (!iconv_enabled(TRUE
))
3191 fd
= iconv_open((char *)enc_skip(to
), (char *)enc_skip(from
));
3193 if (fd
!= (iconv_t
)-1 && iconv_ok
== -1)
3196 * Do a dummy iconv() call to check if it actually works. There is a
3197 * version of iconv() on Linux that is broken. We can't ignore it,
3198 * because it's wide-spread. The symptoms are that after outputting
3199 * the initial shift state the "to" pointer is NULL and conversion
3200 * stops for no apparent reason after about 8160 characters.
3203 tolen
= ICONV_TESTLEN
;
3204 (void)iconv(fd
, NULL
, NULL
, &p
, &tolen
);
3219 * Convert the string "str[slen]" with iconv().
3220 * If "unconvlenp" is not NULL handle the string ending in an incomplete
3221 * sequence and set "*unconvlenp" to the length of it.
3222 * Returns the converted string in allocated memory. NULL for an error.
3225 iconv_string(vcp
, str
, slen
, unconvlenp
)
3237 char_u
*result
= NULL
;
3245 if (len
== 0 || ICONV_ERRNO
== ICONV_E2BIG
)
3247 /* Allocate enough room for most conversions. When re-allocating
3248 * increase the buffer size. */
3249 len
= len
+ fromlen
* 2 + 40;
3250 p
= alloc((unsigned)len
);
3251 if (p
!= NULL
&& done
> 0)
3252 mch_memmove(p
, result
, done
);
3255 if (result
== NULL
) /* out of memory */
3259 to
= (char *)result
+ done
;
3260 tolen
= len
- done
- 2;
3261 /* Avoid a warning for systems with a wrong iconv() prototype by
3262 * casting the second argument to void *. */
3263 if (iconv(vcp
->vc_fd
, (void *)&from
, &fromlen
, &to
, &tolen
)
3266 /* Finished, append a NUL. */
3271 /* Check both ICONV_EINVAL and EINVAL, because the dynamically loaded
3272 * iconv library may use one of them. */
3273 if (!vcp
->vc_fail
&& unconvlenp
!= NULL
3274 && (ICONV_ERRNO
== ICONV_EINVAL
|| ICONV_ERRNO
== EINVAL
))
3276 /* Handle an incomplete sequence at the end. */
3278 *unconvlenp
= (int)fromlen
;
3282 /* Check both ICONV_EILSEQ and EILSEQ, because the dynamically loaded
3283 * iconv library may use one of them. */
3284 else if (!vcp
->vc_fail
3285 && (ICONV_ERRNO
== ICONV_EILSEQ
|| ICONV_ERRNO
== EILSEQ
3286 || ICONV_ERRNO
== ICONV_EINVAL
|| ICONV_ERRNO
== EINVAL
))
3288 /* Can't convert: insert a '?' and skip a character. This assumes
3289 * conversion from 'encoding' to something else. In other
3290 * situations we don't know what to skip anyway. */
3292 if ((*mb_ptr2cells
)((char_u
*)from
) > 1)
3295 l
= utfc_ptr2len_len((char_u
*)from
, (int)fromlen
);
3298 l
= (*mb_ptr2len
)((char_u
*)from
);
3299 if (l
> (int)fromlen
)
3305 else if (ICONV_ERRNO
!= ICONV_E2BIG
)
3307 /* conversion failed */
3312 /* Not enough room or skipping illegal sequence. */
3313 done
= to
- (char *)result
;
3318 # if defined(DYNAMIC_ICONV) || defined(PROTO)
3320 * Dynamically load the "iconv.dll" on Win32.
3323 #ifndef DYNAMIC_ICONV /* just generating prototypes */
3324 # define HINSTANCE int
3326 static HINSTANCE hIconvDLL
= 0;
3327 static HINSTANCE hMsvcrtDLL
= 0;
3329 # ifndef DYNAMIC_ICONV_DLL
3330 # define DYNAMIC_ICONV_DLL "iconv.dll"
3331 # define DYNAMIC_ICONV_DLL_ALT "libiconv.dll"
3333 # ifndef DYNAMIC_MSVCRT_DLL
3334 # define DYNAMIC_MSVCRT_DLL "msvcrt.dll"
3338 * Try opening the iconv.dll and return TRUE if iconv() can be used.
3341 iconv_enabled(verbose
)
3344 if (hIconvDLL
!= 0 && hMsvcrtDLL
!= 0)
3346 hIconvDLL
= LoadLibrary(DYNAMIC_ICONV_DLL
);
3347 if (hIconvDLL
== 0) /* sometimes it's called libiconv.dll */
3348 hIconvDLL
= LoadLibrary(DYNAMIC_ICONV_DLL_ALT
);
3350 hMsvcrtDLL
= LoadLibrary(DYNAMIC_MSVCRT_DLL
);
3351 if (hIconvDLL
== 0 || hMsvcrtDLL
== 0)
3353 /* Only give the message when 'verbose' is set, otherwise it might be
3354 * done whenever a conversion is attempted. */
3355 if (verbose
&& p_verbose
> 0)
3359 hIconvDLL
== 0 ? DYNAMIC_ICONV_DLL
: DYNAMIC_MSVCRT_DLL
);
3366 iconv
= (void *)GetProcAddress(hIconvDLL
, "libiconv");
3367 iconv_open
= (void *)GetProcAddress(hIconvDLL
, "libiconv_open");
3368 iconv_close
= (void *)GetProcAddress(hIconvDLL
, "libiconv_close");
3369 iconvctl
= (void *)GetProcAddress(hIconvDLL
, "libiconvctl");
3370 iconv_errno
= (void *)GetProcAddress(hMsvcrtDLL
, "_errno");
3371 if (iconv
== NULL
|| iconv_open
== NULL
|| iconv_close
== NULL
3372 || iconvctl
== NULL
|| iconv_errno
== NULL
)
3375 if (verbose
&& p_verbose
> 0)
3378 EMSG2(_(e_loadfunc
), "for libiconv");
3389 /* Don't use iconv() when inputting or outputting characters. */
3390 if (input_conv
.vc_type
== CONV_ICONV
)
3391 convert_setup(&input_conv
, NULL
, NULL
);
3392 if (output_conv
.vc_type
== CONV_ICONV
)
3393 convert_setup(&output_conv
, NULL
, NULL
);
3396 FreeLibrary(hIconvDLL
);
3397 if (hMsvcrtDLL
!= 0)
3398 FreeLibrary(hMsvcrtDLL
);
3402 # endif /* DYNAMIC_ICONV */
3403 # endif /* USE_ICONV */
3405 #endif /* FEAT_MBYTE */
3407 #if defined(FEAT_XIM) || defined(PROTO)
3409 # ifdef FEAT_GUI_GTK
3410 static int xim_has_preediting
INIT(= FALSE
); /* IM current status */
3413 * Set preedit_start_col to the current cursor position.
3416 init_preedit_start_col(void)
3418 if (State
& CMDLINE
)
3419 preedit_start_col
= cmdline_getvcol_cursor();
3420 else if (curwin
!= NULL
)
3421 getvcol(curwin
, &curwin
->w_cursor
, &preedit_start_col
, NULL
, NULL
);
3422 /* Prevent that preediting marks the buffer as changed. */
3423 xim_changed_while_preediting
= curbuf
->b_changed
;
3427 # if defined(HAVE_GTK2) && !defined(PROTO)
3429 static int im_is_active
= FALSE
; /* IM is enabled for current mode */
3430 static int im_preedit_cursor
= 0; /* cursor offset in characters */
3431 static int im_preedit_trailing
= 0; /* number of characters after cursor */
3433 static unsigned long im_commit_handler_id
= 0;
3434 static unsigned int im_activatekey_keyval
= GDK_VoidSymbol
;
3435 static unsigned int im_activatekey_state
= 0;
3438 im_set_active(int active
)
3442 was_active
= !!im_is_active
;
3443 im_is_active
= (active
&& !p_imdisable
);
3445 if (im_is_active
!= was_active
)
3450 xim_set_focus(int focus
)
3455 gtk_im_context_focus_in(xic
);
3457 gtk_im_context_focus_out(xic
);
3462 im_set_position(int row
, int col
)
3468 area
.x
= FILL_X(col
);
3469 area
.y
= FILL_Y(row
);
3470 area
.width
= gui
.char_width
* (mb_lefthalve(row
, col
) ? 2 : 1);
3471 area
.height
= gui
.char_height
;
3473 gtk_im_context_set_cursor_location(xic
, &area
);
3477 # if 0 || defined(PROTO) /* apparently only used in gui_x11.c */
3479 xim_set_preedit(void)
3481 im_set_position(gui
.row
, gui
.col
);
3486 im_add_to_input(char_u
*str
, int len
)
3488 /* Convert from 'termencoding' (always "utf-8") to 'encoding' */
3489 if (input_conv
.vc_type
!= CONV_NONE
)
3491 str
= string_convert(&input_conv
, str
, &len
);
3492 g_return_if_fail(str
!= NULL
);
3495 add_to_input_buf_csi(str
, len
);
3497 if (input_conv
.vc_type
!= CONV_NONE
)
3500 if (p_mh
) /* blank out the pointer if necessary */
3501 gui_mch_mousehide(TRUE
);
3505 im_delete_preedit(void)
3507 char_u bskey
[] = {CSI
, 'k', 'b'};
3508 char_u delkey
[] = {CSI
, 'k', 'D'};
3512 im_preedit_cursor
= 0;
3515 for (; im_preedit_cursor
> 0; --im_preedit_cursor
)
3516 add_to_input_buf(bskey
, (int)sizeof(bskey
));
3518 for (; im_preedit_trailing
> 0; --im_preedit_trailing
)
3519 add_to_input_buf(delkey
, (int)sizeof(delkey
));
3523 * Move the cursor left by "num_move_back" characters.
3524 * Note that ins_left() checks im_is_preediting() to avoid breaking undo for
3525 * these K_LEFT keys.
3528 im_correct_cursor(int num_move_back
)
3530 char_u backkey
[] = {CSI
, 'k', 'l'};
3534 # ifdef FEAT_RIGHTLEFT
3535 if ((State
& CMDLINE
) == 0 && curwin
!= NULL
&& curwin
->w_p_rl
)
3538 for (; num_move_back
> 0; --num_move_back
)
3539 add_to_input_buf(backkey
, (int)sizeof(backkey
));
3542 static int xim_expected_char
= NUL
;
3543 static int xim_ignored_char
= FALSE
;
3546 * Update the mode and cursor while in an IM callback.
3553 old_vgetc_busy
= vgetc_busy
;
3556 vgetc_busy
= old_vgetc_busy
;
3562 * Callback invoked when the user finished preediting.
3563 * Put the final string into the input buffer.
3567 im_commit_cb(GtkIMContext
*context
, const gchar
*str
, gpointer data
)
3569 int slen
= (int)STRLEN(str
);
3570 int add_to_input
= TRUE
;
3573 int commit_with_preedit
= TRUE
;
3577 xim_log("im_commit_cb(): %s\n", str
);
3580 /* The imhangul module doesn't reset the preedit string before
3581 * committing. Call im_delete_preedit() to work around that. */
3582 im_delete_preedit();
3584 /* Indicate that preediting has finished. */
3585 if (preedit_start_col
== MAXCOL
)
3587 init_preedit_start_col();
3588 commit_with_preedit
= FALSE
;
3591 /* The thing which setting "preedit_start_col" to MAXCOL means that
3592 * "preedit_start_col" will be set forcely when calling
3593 * preedit_changed_cb() next time.
3594 * "preedit_start_col" should not reset with MAXCOL on this part. Vim
3595 * is simulating the preediting by using add_to_input_str(). when
3596 * preedit begin immediately before committed, the typebuf is not
3597 * flushed to screen, then it can't get correct "preedit_start_col".
3598 * Thus, it should calculate the cells by adding cells of the committed
3600 if (input_conv
.vc_type
!= CONV_NONE
)
3602 im_str
= string_convert(&input_conv
, (char_u
*)str
, &len
);
3603 g_return_if_fail(im_str
!= NULL
);
3606 im_str
= (char_u
*)str
;
3608 for (p
= im_str
; p
< im_str
+ len
; p
+= (*mb_ptr2len
)(p
))
3609 clen
+= (*mb_ptr2cells
)(p
);
3610 if (input_conv
.vc_type
!= CONV_NONE
)
3612 preedit_start_col
+= clen
;
3614 /* Is this a single character that matches a keypad key that's just
3615 * been pressed? If so, we don't want it to be entered as such - let
3616 * us carry on processing the raw keycode so that it may be used in
3617 * mappings as <kSomething>. */
3618 if (xim_expected_char
!= NUL
)
3620 /* We're currently processing a keypad or other special key */
3621 if (slen
== 1 && str
[0] == xim_expected_char
)
3623 /* It's a match - don't do it here */
3624 xim_ignored_char
= TRUE
;
3625 add_to_input
= FALSE
;
3630 xim_ignored_char
= FALSE
;
3635 im_add_to_input((char_u
*)str
, slen
);
3637 /* Inserting chars while "im_is_active" is set does not cause a change of
3638 * buffer. When the chars are committed the buffer must be marked as
3640 if (!commit_with_preedit
)
3641 preedit_start_col
= MAXCOL
;
3643 /* This flag is used in changed() at next call. */
3644 xim_changed_while_preediting
= TRUE
;
3646 if (gtk_main_level() > 0)
3651 * Callback invoked after start to the preedit.
3655 im_preedit_start_cb(GtkIMContext
*context
, gpointer data
)
3658 xim_log("im_preedit_start_cb()\n");
3661 im_is_active
= TRUE
;
3662 gui_update_cursor(TRUE
, FALSE
);
3666 * Callback invoked after end to the preedit.
3670 im_preedit_end_cb(GtkIMContext
*context
, gpointer data
)
3673 xim_log("im_preedit_end_cb()\n");
3675 im_delete_preedit();
3677 /* Indicate that preediting has finished */
3678 preedit_start_col
= MAXCOL
;
3679 xim_has_preediting
= FALSE
;
3681 im_is_active
= FALSE
;
3682 gui_update_cursor(TRUE
, FALSE
);
3687 * Callback invoked after changes to the preedit string. If the preedit
3688 * string was empty before, remember the preedit start column so we know
3689 * where to apply feedback attributes. Delete the previous preedit string
3690 * if there was one, save the new preedit cursor offset, and put the new
3691 * string into the input buffer.
3693 * TODO: The pragmatic "put into input buffer" approach used here has
3694 * several fundamental problems:
3696 * - The characters in the preedit string are subject to remapping.
3697 * That's broken, only the finally committed string should be remapped.
3699 * - There is a race condition involved: The retrieved value for the
3700 * current cursor position will be wrong if any unprocessed characters
3701 * are still queued in the input buffer.
3703 * - Due to the lack of synchronization between the file buffer in memory
3704 * and any typed characters, it's practically impossible to implement the
3705 * "retrieve_surrounding" and "delete_surrounding" signals reliably. IM
3706 * modules for languages such as Thai are likely to rely on this feature
3707 * for proper operation.
3709 * Conclusions: I think support for preediting needs to be moved to the
3710 * core parts of Vim. Ideally, until it has been committed, the preediting
3711 * string should only be displayed and not affect the buffer content at all.
3712 * The question how to deal with the synchronization issue still remains.
3713 * Circumventing the input buffer is probably not desirable. Anyway, I think
3714 * implementing "retrieve_surrounding" is the only hard problem.
3716 * One way to solve all of this in a clean manner would be to queue all key
3717 * press/release events "as is" in the input buffer, and apply the IM filtering
3718 * at the receiving end of the queue. This, however, would have a rather large
3719 * impact on the code base. If there is an easy way to force processing of all
3720 * remaining input from within the "retrieve_surrounding" signal handler, this
3721 * might not be necessary. Gotta ask on vim-dev for opinions.
3725 im_preedit_changed_cb(GtkIMContext
*context
, gpointer data
)
3727 char *preedit_string
= NULL
;
3728 int cursor_index
= 0;
3729 int num_move_back
= 0;
3734 gtk_im_context_get_preedit_string(context
,
3735 &preedit_string
, NULL
,
3739 xim_log("im_preedit_changed_cb(): %s\n", preedit_string
);
3742 g_return_if_fail(preedit_string
!= NULL
); /* just in case */
3744 /* If preedit_start_col is MAXCOL set it to the current cursor position. */
3745 if (preedit_start_col
== MAXCOL
&& preedit_string
[0] != '\0')
3747 xim_has_preediting
= TRUE
;
3749 /* Urgh, this breaks if the input buffer isn't empty now */
3750 init_preedit_start_col();
3752 else if (cursor_index
== 0 && preedit_string
[0] == '\0')
3754 xim_has_preediting
= FALSE
;
3756 /* If at the start position (after typing backspace)
3757 * preedit_start_col must be reset. */
3758 preedit_start_col
= MAXCOL
;
3761 im_delete_preedit();
3764 * Compute the end of the preediting area: "preedit_end_col".
3765 * According to the documentation of gtk_im_context_get_preedit_string(),
3766 * the cursor_pos output argument returns the offset in bytes. This is
3767 * unfortunately not true -- real life shows the offset is in characters,
3768 * and the GTK+ source code agrees with me. Will file a bug later.
3770 if (preedit_start_col
!= MAXCOL
)
3771 preedit_end_col
= preedit_start_col
;
3772 str
= (char_u
*)preedit_string
;
3773 for (p
= str
, i
= 0; *p
!= NUL
; p
+= utf_byte2len(*p
), ++i
)
3777 is_composing
= ((*p
& 0x80) != 0 && utf_iscomposing(utf_ptr2char(p
)));
3779 * These offsets are used as counters when generating <BS> and <Del>
3780 * to delete the preedit string. So don't count composing characters
3781 * unless 'delcombine' is enabled.
3783 if (!is_composing
|| p_deco
)
3785 if (i
< cursor_index
)
3786 ++im_preedit_cursor
;
3788 ++im_preedit_trailing
;
3790 if (!is_composing
&& i
>= cursor_index
)
3792 /* This is essentially the same as im_preedit_trailing, except
3793 * composing characters are not counted even if p_deco is set. */
3796 if (preedit_start_col
!= MAXCOL
)
3797 preedit_end_col
+= utf_ptr2cells(p
);
3802 im_add_to_input(str
, (int)(p
- str
));
3803 im_correct_cursor(num_move_back
);
3806 g_free(preedit_string
);
3808 if (gtk_main_level() > 0)
3813 * Translate the Pango attributes at iter to Vim highlighting attributes.
3814 * Ignore attributes not supported by Vim highlighting. This shouldn't have
3815 * too much impact -- right now we handle even more attributes than necessary
3816 * for the IM modules I tested with.
3819 translate_pango_attributes(PangoAttrIterator
*iter
)
3821 PangoAttribute
*attr
;
3822 int char_attr
= HL_NORMAL
;
3824 attr
= pango_attr_iterator_get(iter
, PANGO_ATTR_UNDERLINE
);
3825 if (attr
!= NULL
&& ((PangoAttrInt
*)attr
)->value
3826 != (int)PANGO_UNDERLINE_NONE
)
3827 char_attr
|= HL_UNDERLINE
;
3829 attr
= pango_attr_iterator_get(iter
, PANGO_ATTR_WEIGHT
);
3830 if (attr
!= NULL
&& ((PangoAttrInt
*)attr
)->value
>= (int)PANGO_WEIGHT_BOLD
)
3831 char_attr
|= HL_BOLD
;
3833 attr
= pango_attr_iterator_get(iter
, PANGO_ATTR_STYLE
);
3834 if (attr
!= NULL
&& ((PangoAttrInt
*)attr
)->value
3835 != (int)PANGO_STYLE_NORMAL
)
3836 char_attr
|= HL_ITALIC
;
3838 attr
= pango_attr_iterator_get(iter
, PANGO_ATTR_BACKGROUND
);
3841 const PangoColor
*color
= &((PangoAttrColor
*)attr
)->color
;
3843 /* Assume inverse if black background is requested */
3844 if ((color
->red
| color
->green
| color
->blue
) == 0)
3845 char_attr
|= HL_INVERSE
;
3852 * Retrieve the highlighting attributes at column col in the preedit string.
3853 * Return -1 if not in preediting mode or if col is out of range.
3856 im_get_feedback_attr(int col
)
3858 char *preedit_string
= NULL
;
3859 PangoAttrList
*attr_list
= NULL
;
3865 gtk_im_context_get_preedit_string(xic
, &preedit_string
, &attr_list
, NULL
);
3867 if (preedit_string
!= NULL
&& attr_list
!= NULL
)
3871 /* Get the byte index as used by PangoAttrIterator */
3872 for (idx
= 0; col
> 0 && preedit_string
[idx
] != '\0'; --col
)
3873 idx
+= utfc_ptr2len((char_u
*)preedit_string
+ idx
);
3875 if (preedit_string
[idx
] != '\0')
3877 PangoAttrIterator
*iter
;
3880 char_attr
= HL_NORMAL
;
3881 iter
= pango_attr_list_get_iterator(attr_list
);
3883 /* Extract all relevant attributes from the list. */
3886 pango_attr_iterator_range(iter
, &start
, &end
);
3888 if (idx
>= start
&& idx
< end
)
3889 char_attr
|= translate_pango_attributes(iter
);
3891 while (pango_attr_iterator_next(iter
));
3893 pango_attr_iterator_destroy(iter
);
3897 if (attr_list
!= NULL
)
3898 pango_attr_list_unref(attr_list
);
3899 g_free(preedit_string
);
3908 xim_log("xim_init()\n");
3911 g_return_if_fail(gui
.drawarea
!= NULL
);
3912 g_return_if_fail(gui
.drawarea
->window
!= NULL
);
3914 xic
= gtk_im_multicontext_new();
3917 im_commit_handler_id
= g_signal_connect(G_OBJECT(xic
), "commit",
3918 G_CALLBACK(&im_commit_cb
), NULL
);
3919 g_signal_connect(G_OBJECT(xic
), "preedit_changed",
3920 G_CALLBACK(&im_preedit_changed_cb
), NULL
);
3921 g_signal_connect(G_OBJECT(xic
), "preedit_start",
3922 G_CALLBACK(&im_preedit_start_cb
), NULL
);
3923 g_signal_connect(G_OBJECT(xic
), "preedit_end",
3924 G_CALLBACK(&im_preedit_end_cb
), NULL
);
3926 gtk_im_context_set_client_window(xic
, gui
.drawarea
->window
);
3933 xim_log("im_shutdown()\n");
3938 gtk_im_context_focus_out(xic
);
3939 g_object_unref(xic
);
3942 im_is_active
= FALSE
;
3943 im_commit_handler_id
= 0;
3944 preedit_start_col
= MAXCOL
;
3945 xim_has_preediting
= FALSE
;
3949 * Convert the string argument to keyval and state for GdkEventKey.
3950 * If str is valid return TRUE, otherwise FALSE.
3952 * See 'imactivatekey' for documentation of the format.
3955 im_string_to_keyval(const char *str
, unsigned int *keyval
, unsigned int *state
)
3957 const char *mods_end
;
3958 unsigned tmp_keyval
;
3959 unsigned tmp_state
= 0;
3961 mods_end
= strrchr(str
, '-');
3962 mods_end
= (mods_end
!= NULL
) ? mods_end
+ 1 : str
;
3964 /* Parse modifier keys */
3965 while (str
< mods_end
)
3969 case 'S': case 's': tmp_state
|= (unsigned)GDK_SHIFT_MASK
; break;
3970 case 'L': case 'l': tmp_state
|= (unsigned)GDK_LOCK_MASK
; break;
3971 case 'C': case 'c': tmp_state
|= (unsigned)GDK_CONTROL_MASK
;break;
3972 case '1': tmp_state
|= (unsigned)GDK_MOD1_MASK
; break;
3973 case '2': tmp_state
|= (unsigned)GDK_MOD2_MASK
; break;
3974 case '3': tmp_state
|= (unsigned)GDK_MOD3_MASK
; break;
3975 case '4': tmp_state
|= (unsigned)GDK_MOD4_MASK
; break;
3976 case '5': tmp_state
|= (unsigned)GDK_MOD5_MASK
; break;
3981 tmp_keyval
= gdk_keyval_from_name(str
);
3983 if (tmp_keyval
== 0 || tmp_keyval
== GDK_VoidSymbol
)
3987 *keyval
= tmp_keyval
;
3995 * Return TRUE if p_imak is valid, otherwise FALSE. As a special case, an
3996 * empty string is also regarded as valid.
3998 * Note: The numerical key value of p_imak is cached if it was valid; thus
3999 * boldly assuming im_xim_isvalid_imactivate() will always be called whenever
4000 * 'imak' changes. This is currently the case but not obvious -- should
4001 * probably rename the function for clarity.
4004 im_xim_isvalid_imactivate(void)
4006 if (p_imak
[0] == NUL
)
4008 im_activatekey_keyval
= GDK_VoidSymbol
;
4009 im_activatekey_state
= 0;
4013 return im_string_to_keyval((const char *)p_imak
,
4014 &im_activatekey_keyval
,
4015 &im_activatekey_state
);
4019 im_synthesize_keypress(unsigned int keyval
, unsigned int state
)
4023 # ifdef HAVE_GTK_MULTIHEAD
4024 event
= (GdkEventKey
*)gdk_event_new(GDK_KEY_PRESS
);
4025 g_object_ref(gui
.drawarea
->window
); /* unreffed by gdk_event_free() */
4027 event
= (GdkEventKey
*)g_malloc0((gulong
)sizeof(GdkEvent
));
4028 event
->type
= GDK_KEY_PRESS
;
4030 event
->window
= gui
.drawarea
->window
;
4031 event
->send_event
= TRUE
;
4032 event
->time
= GDK_CURRENT_TIME
;
4033 event
->state
= state
;
4034 event
->keyval
= keyval
;
4035 event
->hardware_keycode
= /* needed for XIM */
4036 XKeysymToKeycode(GDK_WINDOW_XDISPLAY(event
->window
), (KeySym
)keyval
);
4038 event
->string
= NULL
;
4040 gtk_im_context_filter_keypress(xic
, event
);
4042 /* For consistency, also send the corresponding release event. */
4043 event
->type
= GDK_KEY_RELEASE
;
4044 event
->send_event
= FALSE
;
4045 gtk_im_context_filter_keypress(xic
, event
);
4047 # ifdef HAVE_GTK_MULTIHEAD
4048 gdk_event_free((GdkEvent
*)event
);
4060 * The third-party imhangul module (and maybe others too) ignores
4061 * gtk_im_context_reset() or at least doesn't reset the active state.
4062 * Thus sending imactivatekey would turn it off if it was on before,
4063 * which is clearly not what we want. Fortunately we can work around
4064 * that for imhangul by sending GDK_Escape, but I don't know if it
4065 * works with all IM modules that support an activation key :/
4067 * An alternative approach would be to destroy the IM context and
4068 * recreate it. But that means loading/unloading the IM module on
4069 * every mode switch, which causes a quite noticable delay even on
4070 * my rather fast box...
4072 * Moreover, there are some XIM which cannot respond to
4073 * im_synthesize_keypress(). we hope that they reset by
4076 if (im_activatekey_keyval
!= GDK_VoidSymbol
&& im_is_active
)
4077 im_synthesize_keypress(GDK_Escape
, 0U);
4079 gtk_im_context_reset(xic
);
4082 * HACK for Ami: This sequence of function calls makes Ami handle
4083 * the IM reset graciously, without breaking loads of other stuff.
4084 * It seems to force English mode as well, which is exactly what we
4085 * want because it makes the Ami status display work reliably.
4087 gtk_im_context_set_use_preedit(xic
, FALSE
);
4093 gtk_im_context_set_use_preedit(xic
, TRUE
);
4094 xim_set_focus(gui
.in_focus
);
4096 if (im_activatekey_keyval
!= GDK_VoidSymbol
)
4100 g_signal_handler_block(xic
, im_commit_handler_id
);
4101 im_synthesize_keypress(im_activatekey_keyval
,
4102 im_activatekey_state
);
4103 g_signal_handler_unblock(xic
, im_commit_handler_id
);
4110 xim_set_focus(gui
.in_focus
);
4115 preedit_start_col
= MAXCOL
;
4116 xim_has_preediting
= FALSE
;
4120 xim_queue_key_press_event(GdkEventKey
*event
, int down
)
4125 * Workaround GTK2 XIM 'feature' that always converts keypad keys to
4126 * chars., even when not part of an IM sequence (ref. feature of
4128 * Flag any keypad keys that might represent a single char.
4129 * If this (on its own - i.e., not part of an IM sequence) is
4130 * committed while we're processing one of these keys, we can ignore
4131 * that commit and go ahead & process it ourselves. That way we can
4132 * still distinguish keypad keys for use in mappings.
4134 switch (event
->keyval
)
4136 case GDK_KP_Add
: xim_expected_char
= '+'; break;
4137 case GDK_KP_Subtract
: xim_expected_char
= '-'; break;
4138 case GDK_KP_Divide
: xim_expected_char
= '/'; break;
4139 case GDK_KP_Multiply
: xim_expected_char
= '*'; break;
4140 case GDK_KP_Decimal
: xim_expected_char
= '.'; break;
4141 case GDK_KP_Equal
: xim_expected_char
= '='; break;
4142 case GDK_KP_0
: xim_expected_char
= '0'; break;
4143 case GDK_KP_1
: xim_expected_char
= '1'; break;
4144 case GDK_KP_2
: xim_expected_char
= '2'; break;
4145 case GDK_KP_3
: xim_expected_char
= '3'; break;
4146 case GDK_KP_4
: xim_expected_char
= '4'; break;
4147 case GDK_KP_5
: xim_expected_char
= '5'; break;
4148 case GDK_KP_6
: xim_expected_char
= '6'; break;
4149 case GDK_KP_7
: xim_expected_char
= '7'; break;
4150 case GDK_KP_8
: xim_expected_char
= '8'; break;
4151 case GDK_KP_9
: xim_expected_char
= '9'; break;
4152 default: xim_expected_char
= NUL
;
4154 xim_ignored_char
= FALSE
;
4158 * When typing fFtT, XIM may be activated. Thus it must pass
4159 * gtk_im_context_filter_keypress() in Normal mode.
4160 * And while doing :sh too.
4162 if (xic
!= NULL
&& !p_imdisable
4163 && (State
& (INSERT
| CMDLINE
| NORMAL
| EXTERNCMD
)) != 0)
4166 * Filter 'imactivatekey' and map it to CTRL-^. This way, Vim is
4167 * always aware of the current status of IM, and can even emulate
4168 * the activation key for modules that don't support one.
4170 if (event
->keyval
== im_activatekey_keyval
4171 && (event
->state
& im_activatekey_state
) == im_activatekey_state
)
4173 unsigned int state_mask
;
4175 /* Require the state of the 3 most used modifiers to match exactly.
4176 * Otherwise e.g. <S-C-space> would be unusable for other purposes
4177 * if the IM activate key is <S-space>. */
4178 state_mask
= im_activatekey_state
;
4179 state_mask
|= ((int)GDK_SHIFT_MASK
| (int)GDK_CONTROL_MASK
4180 | (int)GDK_MOD1_MASK
);
4182 if ((event
->state
& state_mask
) != im_activatekey_state
)
4185 /* Don't send it a second time on GDK_KEY_RELEASE. */
4186 if (event
->type
!= GDK_KEY_PRESS
)
4189 if (map_to_exists_mode((char_u
*)"", LANGMAP
, FALSE
))
4191 im_set_active(FALSE
);
4193 /* ":lmap" mappings exists, toggle use of mappings. */
4195 if (State
& LANGMAP
)
4197 curbuf
->b_p_iminsert
= B_IMODE_NONE
;
4202 curbuf
->b_p_iminsert
= B_IMODE_LMAP
;
4208 return gtk_im_context_filter_keypress(xic
, event
);
4211 /* Don't filter events through the IM context if IM isn't active
4212 * right now. Unlike with GTK+ 1.2 we cannot rely on the IM module
4213 * not doing anything before the activation key was sent. */
4214 if (im_activatekey_keyval
== GDK_VoidSymbol
|| im_is_active
)
4216 int imresult
= gtk_im_context_filter_keypress(xic
, event
);
4218 /* Some XIM send following sequence:
4219 * 1. preedited string.
4220 * 2. committed string.
4221 * 3. line changed key.
4222 * 4. preedited string.
4223 * 5. remove preedited string.
4224 * if 3, Vim can't move back the above line for 5.
4225 * thus, this part should not parse the key. */
4226 if (!imresult
&& preedit_start_col
!= MAXCOL
4227 && event
->keyval
== GDK_Return
)
4229 im_synthesize_keypress(GDK_Return
, 0U);
4233 /* If XIM tried to commit a keypad key as a single char.,
4234 * ignore it so we can use the keypad key 'raw', for mappings. */
4235 if (xim_expected_char
!= NUL
&& xim_ignored_char
)
4236 /* We had a keypad key, and XIM tried to thieve it */
4239 /* Normal processing */
4250 return im_is_active
;
4253 # else /* !HAVE_GTK2 */
4255 static int xim_is_active
= FALSE
; /* XIM should be active in the current
4257 static int xim_has_focus
= FALSE
; /* XIM is really being used for Vim */
4259 static XIMStyle input_style
;
4260 static int status_area_enabled
= TRUE
;
4265 # include <gdk/gdkwin32.h>
4267 # include <gdk/gdkx.h>
4271 /* Define a few things to be able to generate prototypes while not configured
4274 # define gboolean int
4275 typedef int GdkEvent
;
4276 typedef int GdkEventKey
;
4281 #if defined(FEAT_GUI_GTK) || defined(PROTO)
4282 static int preedit_buf_len
= 0;
4283 static int xim_can_preediting
INIT(= FALSE
); /* XIM in showmode() */
4284 static int xim_input_style
;
4285 #ifndef FEAT_GUI_GTK
4286 # define gboolean int
4288 static gboolean use_status_area
= 0;
4290 static int im_xim_str2keycode
__ARGS((unsigned int *code
, unsigned int *state
));
4291 static void im_xim_send_event_imactivate
__ARGS((void));
4294 * Convert string to keycode and state for XKeyEvent.
4295 * When string is valid return OK, when invalid return FAIL.
4297 * See 'imactivatekey' documentation for the format.
4300 im_xim_str2keycode(code
, state
)
4302 unsigned int *state
;
4306 unsigned keycode
= 0, keystate
= 0;
4314 len
= STRLEN(p_imak
);
4315 for (flag_end
= p_imak
+ len
- 1;
4316 flag_end
> p_imak
&& *flag_end
!= '-'; --flag_end
)
4319 /* Parse modifier keys */
4320 for (str
= p_imak
; str
< flag_end
; ++str
)
4325 keystate
|= ShiftMask
;
4328 keystate
|= LockMask
;
4331 keystate
|= ControlMask
;
4334 keystate
|= Mod1Mask
;
4337 keystate
|= Mod2Mask
;
4340 keystate
|= Mod3Mask
;
4343 keystate
|= Mod4Mask
;
4346 keystate
|= Mod5Mask
;
4357 /* Get keycode from string. */
4358 gui_get_x11_windis(&window
, &display
);
4360 keycode
= XKeysymToKeycode(display
, XStringToKeysym((char *)str
));
4373 im_xim_send_event_imactivate()
4375 /* Force turn on preedit state by symulate keypress event.
4376 * Keycode and state is specified by 'imactivatekey'.
4380 gui_get_x11_windis(&ev
.window
, &ev
.display
);
4381 ev
.root
= RootWindow(ev
.display
, DefaultScreen(ev
.display
));
4382 ev
.subwindow
= None
;
4383 ev
.time
= CurrentTime
;
4390 if (im_xim_str2keycode(&ev
.keycode
, &ev
.state
) == OK
)
4391 XSendEvent(ev
.display
, ev
.window
, 1, KeyPressMask
, (XEvent
*)&ev
);
4395 * Return TRUE if 'imactivatekey' has a valid value.
4398 im_xim_isvalid_imactivate()
4400 return im_xim_str2keycode(NULL
, NULL
) == OK
;
4402 #endif /* FEAT_GUI_GTK */
4405 * Switch using XIM on/off. This is used by the code that changes "State".
4408 im_set_active(active
)
4414 /* If 'imdisable' is set, XIM is never active. */
4417 #if !defined (FEAT_GUI_GTK)
4418 else if (input_style
& XIMPreeditPosition
)
4419 /* There is a problem in switching XIM off when preediting is used,
4420 * and it is not clear how this can be solved. For now, keep XIM on
4421 * all the time, like it was done in Vim 5.8. */
4425 /* Remember the active state, it is needed when Vim gets keyboard focus. */
4426 xim_is_active
= active
;
4429 /* When 'imactivatekey' has valid key-string, try to control XIM preedit
4430 * state. When 'imactivatekey' has no or invalid string, try old XIM
4436 * Destroy old Input Context (XIC), and create new one. New XIC
4437 * would have a state of preedit that is off. When argument:active
4438 * is false, that's all. Else argument:active is true, send a key
4439 * event specified by 'imactivatekey' to activate XIM preedit state.
4442 xim_is_active
= TRUE
; /* Disable old XIM focus control */
4443 /* If we can monitor preedit state with preedit callback functions,
4444 * try least creation of new XIC.
4446 if (xim_input_style
& (int)GDK_IM_PREEDIT_CALLBACKS
)
4448 if (xim_can_preediting
&& !active
)
4450 /* Force turn off preedit state. With some IM
4451 * implementations, we cannot turn off preedit state by
4452 * symulate keypress event. It is why using such a method
4453 * that destroy old IC (input context), and create new one.
4454 * When create new IC, its preedit state is usually off.
4457 xim_set_focus(FALSE
);
4458 gdk_ic_destroy(xic
);
4460 xim_can_preediting
= FALSE
;
4462 else if (!xim_can_preediting
&& active
)
4463 im_xim_send_event_imactivate();
4467 /* First, force destroy old IC, and create new one. It
4468 * symulates "turning off preedit state".
4470 xim_set_focus(FALSE
);
4471 gdk_ic_destroy(xic
);
4473 xim_can_preediting
= FALSE
;
4475 /* 2nd, when requested to activate IM, symulate this by sending
4480 im_xim_send_event_imactivate();
4481 xim_can_preediting
= TRUE
;
4487 # ifndef XIMPreeditUnKnown
4488 /* X11R5 doesn't have these, it looks safe enough to define here. */
4489 typedef unsigned long XIMPreeditState
;
4490 # define XIMPreeditUnKnown 0L
4491 # define XIMPreeditEnable 1L
4492 # define XIMPreeditDisable (1L<<1)
4493 # define XNPreeditState "preeditState"
4495 XIMPreeditState preedit_state
= XIMPreeditUnKnown
;
4496 XVaNestedList preedit_attr
;
4499 preedit_attr
= XVaCreateNestedList(0,
4500 XNPreeditState
, &preedit_state
,
4502 pxic
= ((GdkICPrivate
*)xic
)->xic
;
4504 if (!XGetICValues(pxic
, XNPreeditAttributes
, preedit_attr
, NULL
))
4506 XFree(preedit_attr
);
4507 preedit_attr
= XVaCreateNestedList(0,
4509 active
? XIMPreeditEnable
: XIMPreeditDisable
,
4511 XSetICValues(pxic
, XNPreeditAttributes
, preedit_attr
, NULL
);
4512 xim_can_preediting
= active
;
4513 xim_is_active
= active
;
4515 XFree(preedit_attr
);
4517 if (xim_input_style
& XIMPreeditCallbacks
)
4519 preedit_buf_len
= 0;
4520 init_preedit_start_col();
4524 /* When had tested kinput2 + canna + Athena GUI version with
4525 * 'imactivatekey' is "s-space", im_xim_send_event_imactivate() did not
4526 * work correctly. It just inserted one space. I don't know why we
4527 * couldn't switch state of XIM preediting. This is reason why these
4528 * codes are commented out.
4530 /* First, force destroy old IC, and create new one. It symulates
4531 * "turning off preedit state".
4533 xim_set_focus(FALSE
);
4538 /* 2nd, when requested to activate IM, symulate this by sending the
4542 im_xim_send_event_imactivate();
4549 * Adjust using XIM for gaining or losing keyboard focus. Also called when
4550 * "xim_is_active" changes.
4553 xim_set_focus(focus
)
4560 * XIM only gets focus when the Vim window has keyboard focus and XIM has
4561 * been set active for the current mode.
4563 if (focus
&& xim_is_active
)
4567 xim_has_focus
= TRUE
;
4569 gdk_im_begin(xic
, gui
.drawarea
->window
);
4579 xim_has_focus
= FALSE
;
4591 im_set_position(row
, col
)
4599 * Set the XIM to the current cursor position.
4607 xim_set_focus(TRUE
);
4621 # ifdef FEAT_XFONTSET
4622 if ((xim_input_style
& (int)GDK_IM_PREEDIT_POSITION
)
4623 && gui
.fontset
!= NOFONTSET
4624 && gui
.fontset
->type
== GDK_FONT_FONTSET
)
4628 if (attr
->spot_location
.y
>= 0)
4630 attr
->spot_location
.x
= 0;
4631 attr
->spot_location
.y
= -100;
4632 attrmask
|= (int)GDK_IC_SPOT_LOCATION
;
4639 if (attr
->spot_location
.x
!= TEXT_X(gui
.col
)
4640 || attr
->spot_location
.y
!= TEXT_Y(gui
.row
))
4642 attr
->spot_location
.x
= TEXT_X(gui
.col
);
4643 attr
->spot_location
.y
= TEXT_Y(gui
.row
);
4644 attrmask
|= (int)GDK_IC_SPOT_LOCATION
;
4647 gdk_window_get_size(gui
.drawarea
->window
, &width
, &height
);
4648 width
-= 2 * gui
.border_offset
;
4649 height
-= 2 * gui
.border_offset
;
4650 if (xim_input_style
& (int)GDK_IM_STATUS_AREA
)
4651 height
-= gui
.char_height
;
4652 if (attr
->preedit_area
.width
!= width
4653 || attr
->preedit_area
.height
!= height
)
4655 attr
->preedit_area
.x
= gui
.border_offset
;
4656 attr
->preedit_area
.y
= gui
.border_offset
;
4657 attr
->preedit_area
.width
= width
;
4658 attr
->preedit_area
.height
= height
;
4659 attrmask
|= (int)GDK_IC_PREEDIT_AREA
;
4662 if (attr
->preedit_fontset
!= gui
.current_font
)
4664 attr
->preedit_fontset
= gui
.current_font
;
4665 attrmask
|= (int)GDK_IC_PREEDIT_FONTSET
;
4669 # endif /* FEAT_XFONTSET */
4671 if (xim_fg_color
== INVALCOLOR
)
4673 xim_fg_color
= gui
.def_norm_pixel
;
4674 xim_bg_color
= gui
.def_back_pixel
;
4676 if (attr
->preedit_foreground
.pixel
!= xim_fg_color
)
4678 attr
->preedit_foreground
.pixel
= xim_fg_color
;
4679 attrmask
|= (int)GDK_IC_PREEDIT_FOREGROUND
;
4681 if (attr
->preedit_background
.pixel
!= xim_bg_color
)
4683 attr
->preedit_background
.pixel
= xim_bg_color
;
4684 attrmask
|= (int)GDK_IC_PREEDIT_BACKGROUND
;
4688 gdk_ic_set_attr(xic
, attr
, (GdkICAttributesType
)attrmask
);
4690 #else /* FEAT_GUI_GTK */
4692 XVaNestedList attr_list
;
4693 XRectangle spot_area
;
4699 /* hide XIM cursor */
4701 over_spot
.y
= -100; /* arbitrary invisible position */
4702 attr_list
= (XVaNestedList
) XVaCreateNestedList(0,
4706 XSetICValues(xic
, XNPreeditAttributes
, attr_list
, NULL
);
4711 if (input_style
& XIMPreeditPosition
)
4713 if (xim_fg_color
== INVALCOLOR
)
4715 xim_fg_color
= gui
.def_norm_pixel
;
4716 xim_bg_color
= gui
.def_back_pixel
;
4718 over_spot
.x
= TEXT_X(gui
.col
);
4719 over_spot
.y
= TEXT_Y(gui
.row
);
4722 spot_area
.height
= gui
.char_height
* Rows
;
4723 spot_area
.width
= gui
.char_width
* Columns
;
4724 line_space
= gui
.char_height
;
4725 attr_list
= (XVaNestedList
) XVaCreateNestedList(0,
4726 XNSpotLocation
, &over_spot
,
4727 XNForeground
, (Pixel
) xim_fg_color
,
4728 XNBackground
, (Pixel
) xim_bg_color
,
4730 XNLineSpace
, line_space
,
4732 if (XSetICValues(xic
, XNPreeditAttributes
, attr_list
, NULL
))
4733 EMSG(_("E284: Cannot set IC values"));
4737 #endif /* FEAT_GUI_GTK */
4741 * Set up the status area.
4743 * This should use a separate Widget, but that seems not possible, because
4744 * preedit_area and status_area should be set to the same window as for the
4745 * text input. Unfortunately this means the status area pollutes the text
4749 xim_set_status_area()
4755 # if defined(FEAT_XFONTSET)
4756 if (use_status_area
)
4769 style
= (int)gdk_ic_get_style(xic
);
4770 if ((style
& (int)GDK_IM_STATUS_MASK
) == (int)GDK_IM_STATUS_AREA
)
4772 if (gui
.fontset
!= NOFONTSET
4773 && gui
.fontset
->type
== GDK_FONT_FONTSET
)
4775 widget
= gui
.mainwin
;
4776 gdk_window_get_size(widget
->window
, &width
, &height
);
4778 attrmask
|= (int)GDK_IC_STATUS_AREA
;
4779 attr
->status_area
.x
= 0;
4780 attr
->status_area
.y
= height
- gui
.char_height
- 1;
4781 attr
->status_area
.width
= width
;
4782 attr
->status_area
.height
= gui
.char_height
;
4786 gdk_ic_set_attr(xic
, attr
, (GdkICAttributesType
)attrmask
);
4791 XVaNestedList preedit_list
= 0, status_list
= 0, list
= 0;
4792 XRectangle pre_area
, status_area
;
4794 if (input_style
& XIMStatusArea
)
4796 if (input_style
& XIMPreeditArea
)
4798 XRectangle
*needed_rect
;
4800 /* to get status_area width */
4801 status_list
= XVaCreateNestedList(0, XNAreaNeeded
,
4802 &needed_rect
, NULL
);
4803 XGetICValues(xic
, XNStatusAttributes
, status_list
, NULL
);
4806 status_area
.width
= needed_rect
->width
;
4809 status_area
.width
= gui
.char_width
* Columns
;
4812 status_area
.y
= gui
.char_height
* Rows
+ gui
.border_offset
;
4813 if (gui
.which_scrollbars
[SBAR_BOTTOM
])
4814 status_area
.y
+= gui
.scrollbar_height
;
4816 if (gui
.menu_is_active
)
4817 status_area
.y
+= gui
.menu_height
;
4819 status_area
.height
= gui
.char_height
;
4820 status_list
= XVaCreateNestedList(0, XNArea
, &status_area
, NULL
);
4825 status_area
.y
= gui
.char_height
* Rows
+ gui
.border_offset
;
4826 if (gui
.which_scrollbars
[SBAR_BOTTOM
])
4827 status_area
.y
+= gui
.scrollbar_height
;
4829 if (gui
.menu_is_active
)
4830 status_area
.y
+= gui
.menu_height
;
4832 status_area
.width
= 0;
4833 status_area
.height
= gui
.char_height
;
4836 if (input_style
& XIMPreeditArea
) /* off-the-spot */
4838 pre_area
.x
= status_area
.x
+ status_area
.width
;
4839 pre_area
.y
= gui
.char_height
* Rows
+ gui
.border_offset
;
4840 pre_area
.width
= gui
.char_width
* Columns
- pre_area
.x
;
4841 if (gui
.which_scrollbars
[SBAR_BOTTOM
])
4842 pre_area
.y
+= gui
.scrollbar_height
;
4844 if (gui
.menu_is_active
)
4845 pre_area
.y
+= gui
.menu_height
;
4847 pre_area
.height
= gui
.char_height
;
4848 preedit_list
= XVaCreateNestedList(0, XNArea
, &pre_area
, NULL
);
4850 else if (input_style
& XIMPreeditPosition
) /* over-the-spot */
4854 pre_area
.height
= gui
.char_height
* Rows
;
4855 pre_area
.width
= gui
.char_width
* Columns
;
4856 preedit_list
= XVaCreateNestedList(0, XNArea
, &pre_area
, NULL
);
4859 if (preedit_list
&& status_list
)
4860 list
= XVaCreateNestedList(0, XNPreeditAttributes
, preedit_list
,
4861 XNStatusAttributes
, status_list
, NULL
);
4862 else if (preedit_list
)
4863 list
= XVaCreateNestedList(0, XNPreeditAttributes
, preedit_list
,
4865 else if (status_list
)
4866 list
= XVaCreateNestedList(0, XNStatusAttributes
, status_list
,
4873 XSetICValues(xic
, XNVaNestedList
, list
, NULL
);
4879 XFree(preedit_list
);
4884 #if defined(FEAT_GUI_X11) || defined(FEAT_GUI_GTK)
4885 static char e_xim
[] = N_("E285: Failed to create input context");
4888 #if defined(FEAT_GUI_X11) || defined(PROTO)
4889 # if defined(XtSpecificationRelease) && XtSpecificationRelease >= 6 && !defined(sun)
4890 # define USE_X11R6_XIM
4893 static int xim_real_init
__ARGS((Window x11_window
, Display
*x11_display
));
4896 #ifdef USE_X11R6_XIM
4897 static void xim_instantiate_cb
__ARGS((Display
*display
, XPointer client_data
, XPointer call_data
));
4898 static void xim_destroy_cb
__ARGS((XIM im
, XPointer client_data
, XPointer call_data
));
4902 xim_instantiate_cb(display
, client_data
, call_data
)
4904 XPointer client_data
;
4908 Display
*x11_display
;
4911 xim_log("xim_instantiate_cb()\n");
4914 gui_get_x11_windis(&x11_window
, &x11_display
);
4915 if (display
!= x11_display
)
4918 xim_real_init(x11_window
, x11_display
);
4919 gui_set_shellsize(FALSE
, FALSE
, RESIZE_BOTH
);
4921 XUnregisterIMInstantiateCallback(x11_display
, NULL
, NULL
, NULL
,
4922 xim_instantiate_cb
, NULL
);
4927 xim_destroy_cb(im
, client_data
, call_data
)
4929 XPointer client_data
;
4933 Display
*x11_display
;
4936 xim_log("xim_destroy_cb()\n");
4938 gui_get_x11_windis(&x11_window
, &x11_display
);
4941 status_area_enabled
= FALSE
;
4943 gui_set_shellsize(FALSE
, FALSE
, RESIZE_BOTH
);
4945 XRegisterIMInstantiateCallback(x11_display
, NULL
, NULL
, NULL
,
4946 xim_instantiate_cb
, NULL
);
4954 Display
*x11_display
;
4957 xim_log("xim_init()\n");
4960 gui_get_x11_windis(&x11_window
, &x11_display
);
4964 if (xim_real_init(x11_window
, x11_display
))
4967 gui_set_shellsize(FALSE
, FALSE
, RESIZE_BOTH
);
4969 #ifdef USE_X11R6_XIM
4970 XRegisterIMInstantiateCallback(x11_display
, NULL
, NULL
, NULL
,
4971 xim_instantiate_cb
, NULL
);
4976 xim_real_init(x11_window
, x11_display
)
4978 Display
*x11_display
;
4986 #define IMLEN_MAX 40
4987 char buf
[IMLEN_MAX
+ 7];
4989 XIMStyles
*xim_styles
;
4990 XIMStyle this_input_style
= 0;
4993 XVaNestedList preedit_list
, status_list
;
4996 status_area_enabled
= FALSE
;
5001 if (gui
.rsrc_input_method
!= NULL
&& *gui
.rsrc_input_method
!= NUL
)
5003 strcpy(tmp
, gui
.rsrc_input_method
);
5004 for (ns
= s
= tmp
; ns
!= NULL
&& *s
!= NUL
;)
5006 s
= (char *)skipwhite((char_u
*)s
);
5009 if ((ns
= end
= strchr(s
, ',')) == NULL
)
5010 end
= s
+ strlen(s
);
5011 while (isspace(((char_u
*)end
)[-1]))
5015 if (strlen(s
) <= IMLEN_MAX
)
5017 strcpy(buf
, "@im=");
5019 if ((p
= XSetLocaleModifiers(buf
)) != NULL
&& *p
!= NUL
5020 && (xim
= XOpenIM(x11_display
, NULL
, NULL
, NULL
))
5029 if (xim
== NULL
&& (p
= XSetLocaleModifiers("")) != NULL
&& *p
!= NUL
)
5030 xim
= XOpenIM(x11_display
, NULL
, NULL
, NULL
);
5032 /* This is supposed to be useful to obtain characters through
5033 * XmbLookupString() without really using a XIM. */
5034 if (xim
== NULL
&& (p
= XSetLocaleModifiers("@im=none")) != NULL
5036 xim
= XOpenIM(x11_display
, NULL
, NULL
, NULL
);
5040 /* Only give this message when verbose is set, because too many people
5041 * got this message when they didn't want to use a XIM. */
5045 EMSG(_("E286: Failed to open input method"));
5051 #ifdef USE_X11R6_XIM
5053 XIMCallback destroy_cb
;
5055 destroy_cb
.callback
= xim_destroy_cb
;
5056 destroy_cb
.client_data
= NULL
;
5057 if (XSetIMValues(xim
, XNDestroyCallback
, &destroy_cb
, NULL
))
5058 EMSG(_("E287: Warning: Could not set destroy callback to IM"));
5062 if (XGetIMValues(xim
, XNQueryInputStyle
, &xim_styles
, NULL
) || !xim_styles
)
5064 EMSG(_("E288: input method doesn't support any style"));
5070 strcpy(tmp
, gui
.rsrc_preedit_type_name
);
5071 for (s
= tmp
; s
&& !found
; )
5073 while (*s
&& isspace((unsigned char)*s
))
5077 if ((ns
= end
= strchr(s
, ',')) != 0)
5080 end
= s
+ strlen(s
);
5081 while (isspace((unsigned char)*end
))
5085 if (!strcmp(s
, "OverTheSpot"))
5086 this_input_style
= (XIMPreeditPosition
| XIMStatusArea
);
5087 else if (!strcmp(s
, "OffTheSpot"))
5088 this_input_style
= (XIMPreeditArea
| XIMStatusArea
);
5089 else if (!strcmp(s
, "Root"))
5090 this_input_style
= (XIMPreeditNothing
| XIMStatusNothing
);
5092 for (i
= 0; (unsigned short)i
< xim_styles
->count_styles
; i
++)
5094 if (this_input_style
== xim_styles
->supported_styles
[i
])
5101 for (i
= 0; (unsigned short)i
< xim_styles
->count_styles
; i
++)
5103 if ((xim_styles
->supported_styles
[i
] & this_input_style
)
5104 == (this_input_style
& ~XIMStatusArea
))
5106 this_input_style
&= ~XIMStatusArea
;
5118 /* Only give this message when verbose is set, because too many people
5119 * got this message when they didn't want to use a XIM. */
5123 EMSG(_("E289: input method doesn't support my preedit type"));
5130 over_spot
.x
= TEXT_X(gui
.col
);
5131 over_spot
.y
= TEXT_Y(gui
.row
);
5132 input_style
= this_input_style
;
5134 /* A crash was reported when trying to pass gui.norm_font as XNFontSet,
5135 * thus that has been removed. Hopefully the default works... */
5136 #ifdef FEAT_XFONTSET
5137 if (gui
.fontset
!= NOFONTSET
)
5139 preedit_list
= XVaCreateNestedList(0,
5140 XNSpotLocation
, &over_spot
,
5141 XNForeground
, (Pixel
)gui
.def_norm_pixel
,
5142 XNBackground
, (Pixel
)gui
.def_back_pixel
,
5143 XNFontSet
, (XFontSet
)gui
.fontset
,
5145 status_list
= XVaCreateNestedList(0,
5146 XNForeground
, (Pixel
)gui
.def_norm_pixel
,
5147 XNBackground
, (Pixel
)gui
.def_back_pixel
,
5148 XNFontSet
, (XFontSet
)gui
.fontset
,
5154 preedit_list
= XVaCreateNestedList(0,
5155 XNSpotLocation
, &over_spot
,
5156 XNForeground
, (Pixel
)gui
.def_norm_pixel
,
5157 XNBackground
, (Pixel
)gui
.def_back_pixel
,
5159 status_list
= XVaCreateNestedList(0,
5160 XNForeground
, (Pixel
)gui
.def_norm_pixel
,
5161 XNBackground
, (Pixel
)gui
.def_back_pixel
,
5165 xic
= XCreateIC(xim
,
5166 XNInputStyle
, input_style
,
5167 XNClientWindow
, x11_window
,
5168 XNFocusWindow
, gui
.wid
,
5169 XNPreeditAttributes
, preedit_list
,
5170 XNStatusAttributes
, status_list
,
5173 XFree(preedit_list
);
5176 if (input_style
& XIMStatusArea
)
5178 xim_set_status_area();
5179 status_area_enabled
= TRUE
;
5182 gui_set_shellsize(FALSE
, FALSE
, RESIZE_BOTH
);
5194 #endif /* FEAT_GUI_X11 */
5196 #if defined(FEAT_GUI_GTK) || defined(PROTO)
5198 # ifdef FEAT_XFONTSET
5199 static char e_overthespot
[] = N_("E290: over-the-spot style requires fontset");
5207 xim_decide_input_style()
5209 /* GDK_IM_STATUS_CALLBACKS was disabled, enabled it to allow Japanese
5211 int supported_style
= (int)GDK_IM_PREEDIT_NONE
|
5212 (int)GDK_IM_PREEDIT_NOTHING
|
5213 (int)GDK_IM_PREEDIT_POSITION
|
5214 (int)GDK_IM_PREEDIT_CALLBACKS
|
5215 (int)GDK_IM_STATUS_CALLBACKS
|
5216 (int)GDK_IM_STATUS_AREA
|
5217 (int)GDK_IM_STATUS_NONE
|
5218 (int)GDK_IM_STATUS_NOTHING
;
5221 xim_log("xim_decide_input_style()\n");
5224 if (!gdk_im_ready())
5225 xim_input_style
= 0;
5228 if (gtk_major_version
> 1
5229 || (gtk_major_version
== 1
5230 && (gtk_minor_version
> 2
5231 || (gtk_minor_version
== 2 && gtk_micro_version
>= 3))))
5232 use_status_area
= TRUE
;
5235 EMSG(_("E291: Your GTK+ is older than 1.2.3. Status area disabled"));
5236 use_status_area
= FALSE
;
5238 #ifdef FEAT_XFONTSET
5239 if (gui
.fontset
== NOFONTSET
|| gui
.fontset
->type
!= GDK_FONT_FONTSET
)
5241 supported_style
&= ~((int)GDK_IM_PREEDIT_POSITION
5242 | (int)GDK_IM_STATUS_AREA
);
5243 if (!use_status_area
)
5244 supported_style
&= ~(int)GDK_IM_STATUS_AREA
;
5245 xim_input_style
= (int)gdk_im_decide_style((GdkIMStyle
)supported_style
);
5251 preedit_start_cbproc(XIC xic
, XPointer client_data
, XPointer call_data
)
5254 xim_log("xim_decide_input_style()\n");
5257 draw_feedback
= NULL
;
5258 xim_can_preediting
= TRUE
;
5259 xim_has_preediting
= TRUE
;
5260 gui_update_cursor(TRUE
, FALSE
);
5269 xim_back_delete(int n
)
5277 add_to_input_buf(str
, 3);
5280 static GSList
*key_press_event_queue
= NULL
;
5281 static gboolean processing_queued_event
= FALSE
;
5285 preedit_draw_cbproc(XIC xic
, XPointer client_data
, XPointer call_data
)
5287 XIMPreeditDrawCallbackStruct
*draw_data
;
5290 GSList
*event_queue
;
5293 xim_log("preedit_draw_cbproc()\n");
5296 draw_data
= (XIMPreeditDrawCallbackStruct
*) call_data
;
5297 text
= (XIMText
*) draw_data
->text
;
5299 if ((text
== NULL
&& draw_data
->chg_length
== preedit_buf_len
)
5300 || preedit_buf_len
== 0)
5302 init_preedit_start_col();
5303 vim_free(draw_feedback
);
5304 draw_feedback
= NULL
;
5306 if (draw_data
->chg_length
> 0)
5310 if (draw_data
->chg_length
> preedit_buf_len
)
5311 bs_cnt
= preedit_buf_len
;
5313 bs_cnt
= draw_data
->chg_length
;
5314 xim_back_delete(bs_cnt
);
5315 preedit_buf_len
-= bs_cnt
;
5322 unsigned int nfeedback
= 0;
5326 src
= text
->string
.multi_byte
;
5327 if (src
!= NULL
&& !text
->encoding_is_wchar
)
5330 ptr
= (char_u
*)src
;
5331 /* Avoid the enter for decision */
5336 if (input_conv
.vc_type
!= CONV_NONE
5337 && (buf
= string_convert(&input_conv
,
5338 (char_u
*)src
, &len
)) != NULL
)
5340 /* Converted from 'termencoding' to 'encoding'. */
5341 add_to_input_buf_csi(buf
, len
);
5346 add_to_input_buf_csi((char_u
*)src
, len
);
5347 /* Add count of character to preedit_buf_len */
5351 if (draw_data
->text
->feedback
!= NULL
)
5353 if (draw_feedback
== NULL
)
5354 draw_feedback
= (char *)alloc(draw_data
->chg_first
5357 draw_feedback
= realloc(draw_feedback
,
5358 draw_data
->chg_first
+ text
->length
);
5359 if (draw_feedback
!= NULL
)
5361 draw_feedback
[nfeedback
+ draw_data
->chg_first
]
5362 = draw_data
->text
->feedback
[nfeedback
];
5367 ptr
+= (*mb_ptr2len
)(ptr
);
5376 preedit_end_col
= MAXCOL
;
5379 if (text
!= NULL
|| draw_data
->chg_length
> 0)
5381 event_queue
= key_press_event_queue
;
5382 processing_queued_event
= TRUE
;
5383 while (event_queue
!= NULL
&& processing_queued_event
)
5385 GdkEvent
*ev
= event_queue
->data
;
5388 gtk_signal_emit_by_name((GtkObject
*)gui
.mainwin
, "key_press_event",
5391 event_queue
= event_queue
->next
;
5393 processing_queued_event
= FALSE
;
5394 if (key_press_event_queue
)
5396 g_slist_free(key_press_event_queue
);
5397 key_press_event_queue
= NULL
;
5400 if (gtk_main_level() > 0)
5405 * Retrieve the highlighting attributes at column col in the preedit string.
5406 * Return -1 if not in preediting mode or if col is out of range.
5409 im_get_feedback_attr(int col
)
5411 if (draw_feedback
!= NULL
&& col
< preedit_buf_len
)
5413 if (draw_feedback
[col
] & XIMReverse
)
5415 else if (draw_feedback
[col
] & XIMUnderline
)
5416 return HL_UNDERLINE
;
5418 return hl_attr(HLF_V
);
5426 preedit_caret_cbproc(XIC xic
, XPointer client_data
, XPointer call_data
)
5429 xim_log("preedit_caret_cbproc()\n");
5435 preedit_done_cbproc(XIC xic
, XPointer client_data
, XPointer call_data
)
5438 xim_log("preedit_done_cbproc()\n");
5441 vim_free(draw_feedback
);
5442 draw_feedback
= NULL
;
5443 xim_can_preediting
= FALSE
;
5444 xim_has_preediting
= FALSE
;
5445 gui_update_cursor(TRUE
, FALSE
);
5459 xim_log("xim_reset()\n");
5464 text
= XmbResetIC(((GdkICPrivate
*)xic
)->xic
);
5465 if (text
!= NULL
&& !(xim_input_style
& (int)GDK_IM_PREEDIT_CALLBACKS
))
5466 add_to_input_buf_csi((char_u
*)text
, strlen(text
));
5468 preedit_buf_len
= 0;
5476 xim_queue_key_press_event(GdkEventKey
*event
, int down
)
5479 xim_log("xim_queue_key_press_event()\n");
5482 if (preedit_buf_len
<= 0)
5484 if (processing_queued_event
)
5485 processing_queued_event
= FALSE
;
5487 key_press_event_queue
= g_slist_append(key_press_event_queue
,
5488 gdk_event_copy((GdkEvent
*)event
));
5494 preedit_callback_setup(GdkIC
*ic
)
5497 XVaNestedList preedit_attr
;
5498 XIMCallback preedit_start_cb
;
5499 XIMCallback preedit_draw_cb
;
5500 XIMCallback preedit_caret_cb
;
5501 XIMCallback preedit_done_cb
;
5503 xxic
= ((GdkICPrivate
*)xic
)->xic
;
5504 preedit_start_cb
.callback
= (XIMProc
)preedit_start_cbproc
;
5505 preedit_draw_cb
.callback
= (XIMProc
)preedit_draw_cbproc
;
5506 preedit_caret_cb
.callback
= (XIMProc
)preedit_caret_cbproc
;
5507 preedit_done_cb
.callback
= (XIMProc
)preedit_done_cbproc
;
5509 = XVaCreateNestedList (0,
5510 XNPreeditStartCallback
, &preedit_start_cb
,
5511 XNPreeditDrawCallback
, &preedit_draw_cb
,
5512 XNPreeditCaretCallback
, &preedit_caret_cb
,
5513 XNPreeditDoneCallback
, &preedit_done_cb
,
5515 XSetICValues (xxic
, XNPreeditAttributes
, preedit_attr
, 0);
5516 XFree(preedit_attr
);
5521 reset_state_setup(GdkIC
*ic
)
5523 #ifdef USE_X11R6_XIM
5524 /* don't change the input context when we call reset */
5525 XSetICValues(((GdkICPrivate
*)ic
)->xic
, XNResetState
, XIMPreserveState
, 0);
5533 xim_log("xim_init()\n");
5539 if (!gdk_im_ready())
5544 EMSG(_("E292: Input Method Server is not running"));
5549 if ((xic_attr
= gdk_ic_attr_new()) != NULL
)
5551 #ifdef FEAT_XFONTSET
5555 GdkColormap
*colormap
;
5556 GdkICAttr
*attr
= xic_attr
;
5557 int attrmask
= (int)GDK_IC_ALL_REQ
;
5558 GtkWidget
*widget
= gui
.drawarea
;
5560 attr
->style
= (GdkIMStyle
)xim_input_style
;
5561 attr
->client_window
= gui
.mainwin
->window
;
5563 if ((colormap
= gtk_widget_get_colormap(widget
)) !=
5564 gtk_widget_get_default_colormap())
5566 attrmask
|= (int)GDK_IC_PREEDIT_COLORMAP
;
5567 attr
->preedit_colormap
= colormap
;
5569 attrmask
|= (int)GDK_IC_PREEDIT_FOREGROUND
;
5570 attrmask
|= (int)GDK_IC_PREEDIT_BACKGROUND
;
5571 attr
->preedit_foreground
= widget
->style
->fg
[GTK_STATE_NORMAL
];
5572 attr
->preedit_background
= widget
->style
->base
[GTK_STATE_NORMAL
];
5574 #ifdef FEAT_XFONTSET
5575 if ((xim_input_style
& (int)GDK_IM_PREEDIT_MASK
)
5576 == (int)GDK_IM_PREEDIT_POSITION
)
5578 if (gui
.fontset
== NOFONTSET
5579 || gui
.fontset
->type
!= GDK_FONT_FONTSET
)
5581 EMSG(_(e_overthespot
));
5585 gdk_window_get_size(widget
->window
, &width
, &height
);
5587 attrmask
|= (int)GDK_IC_PREEDIT_POSITION_REQ
;
5588 attr
->spot_location
.x
= TEXT_X(0);
5589 attr
->spot_location
.y
= TEXT_Y(0);
5590 attr
->preedit_area
.x
= gui
.border_offset
;
5591 attr
->preedit_area
.y
= gui
.border_offset
;
5592 attr
->preedit_area
.width
= width
- 2*gui
.border_offset
;
5593 attr
->preedit_area
.height
= height
- 2*gui
.border_offset
;
5594 attr
->preedit_fontset
= gui
.fontset
;
5598 if ((xim_input_style
& (int)GDK_IM_STATUS_MASK
)
5599 == (int)GDK_IM_STATUS_AREA
)
5601 if (gui
.fontset
== NOFONTSET
5602 || gui
.fontset
->type
!= GDK_FONT_FONTSET
)
5604 EMSG(_(e_overthespot
));
5608 gdk_window_get_size(gui
.mainwin
->window
, &width
, &height
);
5609 attrmask
|= (int)GDK_IC_STATUS_AREA_REQ
;
5610 attr
->status_area
.x
= 0;
5611 attr
->status_area
.y
= height
- gui
.char_height
- 1;
5612 attr
->status_area
.width
= width
;
5613 attr
->status_area
.height
= gui
.char_height
;
5614 attr
->status_fontset
= gui
.fontset
;
5617 else if ((xim_input_style
& (int)GDK_IM_STATUS_MASK
)
5618 == (int)GDK_IM_STATUS_CALLBACKS
)
5624 xic
= gdk_ic_new(attr
, (GdkICAttributesType
)attrmask
);
5630 mask
= (int)gdk_window_get_events(widget
->window
);
5631 mask
|= (int)gdk_ic_get_events(xic
);
5632 gdk_window_set_events(widget
->window
, (GdkEventMask
)mask
);
5633 if (xim_input_style
& (int)GDK_IM_PREEDIT_CALLBACKS
)
5634 preedit_callback_setup(xic
);
5635 reset_state_setup(xic
);
5644 xim_log("im_shutdown()\n");
5650 gdk_ic_destroy(xic
);
5653 xim_is_active
= FALSE
;
5654 xim_can_preediting
= FALSE
;
5655 preedit_start_col
= MAXCOL
;
5656 xim_has_preediting
= FALSE
;
5659 #endif /* FEAT_GUI_GTK */
5662 xim_get_status_area_height()
5665 if (xim_input_style
& (int)GDK_IM_STATUS_AREA
)
5666 return gui
.char_height
;
5668 if (status_area_enabled
)
5669 return gui
.char_height
;
5675 * Get IM status. When IM is on, return TRUE. Else return FALSE.
5676 * FIXME: This doesn't work correctly: Having focus doesn't always mean XIM is
5677 * active, when not having focus XIM may still be active (e.g., when using a
5678 * tear-off menu item).
5683 # ifdef FEAT_GUI_GTK
5684 if (xim_input_style
& (int)GDK_IM_PREEDIT_CALLBACKS
)
5685 return xim_can_preediting
;
5687 return xim_has_focus
;
5690 # endif /* !HAVE_GTK2 */
5692 # if defined(FEAT_GUI_GTK) || defined(PROTO)
5696 return xim_has_preediting
;
5699 #endif /* FEAT_XIM */
5701 #if defined(FEAT_MBYTE) || defined(PROTO)
5704 * Setup "vcp" for conversion from "from" to "to".
5705 * The names must have been made canonical with enc_canonize().
5706 * vcp->vc_type must have been initialized to CONV_NONE.
5707 * Note: cannot be used for conversion from/to ucs-2 and ucs-4 (will use utf-8
5709 * Afterwards invoke with "from" and "to" equal to NULL to cleanup.
5710 * Return FAIL when conversion is not supported, OK otherwise.
5713 convert_setup(vcp
, from
, to
)
5721 /* Reset to no conversion. */
5723 if (vcp
->vc_type
== CONV_ICONV
&& vcp
->vc_fd
!= (iconv_t
)-1)
5724 iconv_close(vcp
->vc_fd
);
5726 vcp
->vc_type
= CONV_NONE
;
5728 vcp
->vc_fail
= FALSE
;
5730 /* No conversion when one of the names is empty or they are equal. */
5731 if (from
== NULL
|| *from
== NUL
|| to
== NULL
|| *to
== NUL
5732 || STRCMP(from
, to
) == 0)
5735 from_prop
= enc_canon_props(from
);
5736 to_prop
= enc_canon_props(to
);
5737 if ((from_prop
& ENC_LATIN1
) && (to_prop
& ENC_UNICODE
))
5739 /* Internal latin1 -> utf-8 conversion. */
5740 vcp
->vc_type
= CONV_TO_UTF8
;
5741 vcp
->vc_factor
= 2; /* up to twice as long */
5743 else if ((from_prop
& ENC_LATIN9
) && (to_prop
& ENC_UNICODE
))
5745 /* Internal latin9 -> utf-8 conversion. */
5746 vcp
->vc_type
= CONV_9_TO_UTF8
;
5747 vcp
->vc_factor
= 3; /* up to three as long (euro sign) */
5749 else if ((from_prop
& ENC_UNICODE
) && (to_prop
& ENC_LATIN1
))
5751 /* Internal utf-8 -> latin1 conversion. */
5752 vcp
->vc_type
= CONV_TO_LATIN1
;
5754 else if ((from_prop
& ENC_UNICODE
) && (to_prop
& ENC_LATIN9
))
5756 /* Internal utf-8 -> latin9 conversion. */
5757 vcp
->vc_type
= CONV_TO_LATIN9
;
5760 /* Win32-specific codepage <-> codepage conversion without iconv. */
5761 else if (((from_prop
& ENC_UNICODE
) || encname2codepage(from
) > 0)
5762 && ((to_prop
& ENC_UNICODE
) || encname2codepage(to
) > 0))
5764 vcp
->vc_type
= CONV_CODEPAGE
;
5765 vcp
->vc_factor
= 2; /* up to twice as long */
5766 vcp
->vc_cpfrom
= (from_prop
& ENC_UNICODE
) ? 0 : encname2codepage(from
);
5767 vcp
->vc_cpto
= (to_prop
& ENC_UNICODE
) ? 0 : encname2codepage(to
);
5771 else if ((from_prop
& ENC_MACROMAN
) && (to_prop
& ENC_LATIN1
))
5773 vcp
->vc_type
= CONV_MAC_LATIN1
;
5775 else if ((from_prop
& ENC_MACROMAN
) && (to_prop
& ENC_UNICODE
))
5777 vcp
->vc_type
= CONV_MAC_UTF8
;
5778 vcp
->vc_factor
= 2; /* up to twice as long */
5780 else if ((from_prop
& ENC_LATIN1
) && (to_prop
& ENC_MACROMAN
))
5782 vcp
->vc_type
= CONV_LATIN1_MAC
;
5784 else if ((from_prop
& ENC_UNICODE
) && (to_prop
& ENC_MACROMAN
))
5786 vcp
->vc_type
= CONV_UTF8_MAC
;
5792 /* Use iconv() for conversion. */
5793 vcp
->vc_fd
= (iconv_t
)my_iconv_open(
5794 (to_prop
& ENC_UNICODE
) ? (char_u
*)"utf-8" : to
,
5795 (from_prop
& ENC_UNICODE
) ? (char_u
*)"utf-8" : from
);
5796 if (vcp
->vc_fd
!= (iconv_t
)-1)
5798 vcp
->vc_type
= CONV_ICONV
;
5799 vcp
->vc_factor
= 4; /* could be longer too... */
5803 if (vcp
->vc_type
== CONV_NONE
)
5809 #if defined(FEAT_GUI) || defined(AMIGA) || defined(WIN3264) \
5810 || defined(MSDOS) || defined(PROTO)
5812 * Do conversion on typed input characters in-place.
5813 * The input and output are not NUL terminated!
5814 * Returns the length after conversion.
5817 convert_input(ptr
, len
, maxlen
)
5822 return convert_input_safe(ptr
, len
, maxlen
, NULL
, NULL
);
5827 * Like convert_input(), but when there is an incomplete byte sequence at the
5828 * end return that as an allocated string in "restp" and set "*restlenp" to
5829 * the length. If "restp" is NULL it is not used.
5832 convert_input_safe(ptr
, len
, maxlen
, restp
, restlenp
)
5841 int unconvertlen
= 0;
5843 d
= string_convert_ext(&input_conv
, ptr
, &dlen
,
5844 restp
== NULL
? NULL
: &unconvertlen
);
5849 if (unconvertlen
> 0)
5851 /* Move the unconverted characters to allocated memory. */
5852 *restp
= alloc(unconvertlen
);
5854 mch_memmove(*restp
, ptr
+ len
- unconvertlen
, unconvertlen
);
5855 *restlenp
= unconvertlen
;
5857 mch_memmove(ptr
, d
, dlen
);
5860 /* result is too long, keep the unconverted text (the caller must
5861 * have done something wrong!) */
5869 * Convert text "ptr[*lenp]" according to "vcp".
5870 * Returns the result in allocated memory and sets "*lenp".
5871 * When "lenp" is NULL, use NUL terminated strings.
5872 * Illegal chars are often changed to "?", unless vcp->vc_fail is set.
5873 * When something goes wrong, NULL is returned and "*lenp" is unchanged.
5876 string_convert(vcp
, ptr
, lenp
)
5881 return string_convert_ext(vcp
, ptr
, lenp
, NULL
);
5885 * Like string_convert(), but when "unconvlenp" is not NULL and there are is
5886 * an incomplete sequence at the end it is not converted and "*unconvlenp" is
5887 * set to the number of remaining bytes.
5890 string_convert_ext(vcp
, ptr
, lenp
, unconvlenp
)
5896 char_u
*retval
= NULL
;
5904 len
= (int)STRLEN(ptr
);
5908 return vim_strsave((char_u
*)"");
5910 switch (vcp
->vc_type
)
5912 case CONV_TO_UTF8
: /* latin1 to utf-8 conversion */
5913 retval
= alloc(len
* 2 + 1);
5917 for (i
= 0; i
< len
; ++i
)
5924 *d
++ = 0xc0 + ((unsigned)c
>> 6);
5925 *d
++ = 0x80 + (c
& 0x3f);
5930 *lenp
= (int)(d
- retval
);
5933 case CONV_9_TO_UTF8
: /* latin9 to utf-8 conversion */
5934 retval
= alloc(len
* 3 + 1);
5938 for (i
= 0; i
< len
; ++i
)
5943 case 0xa4: c
= 0x20ac; break; /* euro */
5944 case 0xa6: c
= 0x0160; break; /* S hat */
5945 case 0xa8: c
= 0x0161; break; /* S -hat */
5946 case 0xb4: c
= 0x017d; break; /* Z hat */
5947 case 0xb8: c
= 0x017e; break; /* Z -hat */
5948 case 0xbc: c
= 0x0152; break; /* OE */
5949 case 0xbd: c
= 0x0153; break; /* oe */
5950 case 0xbe: c
= 0x0178; break; /* Y */
5952 d
+= utf_char2bytes(c
, d
);
5956 *lenp
= (int)(d
- retval
);
5959 case CONV_TO_LATIN1
: /* utf-8 to latin1 conversion */
5960 case CONV_TO_LATIN9
: /* utf-8 to latin9 conversion */
5961 retval
= alloc(len
+ 1);
5965 for (i
= 0; i
< len
; ++i
)
5967 l
= utf_ptr2len(ptr
+ i
);
5972 if (unconvlenp
!= NULL
&& utf8len_tab
[ptr
[i
]] > len
- i
)
5974 /* Incomplete sequence at the end. */
5975 *unconvlenp
= len
- i
;
5982 c
= utf_ptr2char(ptr
+ i
);
5983 if (vcp
->vc_type
== CONV_TO_LATIN9
)
5986 case 0x20ac: c
= 0xa4; break; /* euro */
5987 case 0x0160: c
= 0xa6; break; /* S hat */
5988 case 0x0161: c
= 0xa8; break; /* S -hat */
5989 case 0x017d: c
= 0xb4; break; /* Z hat */
5990 case 0x017e: c
= 0xb8; break; /* Z -hat */
5991 case 0x0152: c
= 0xbc; break; /* OE */
5992 case 0x0153: c
= 0xbd; break; /* oe */
5993 case 0x0178: c
= 0xbe; break; /* Y */
6001 case 0xbe: c
= 0x100; break; /* not in latin9 */
6003 if (!utf_iscomposing(c
)) /* skip composing chars */
6007 else if (vcp
->vc_fail
)
6015 if (utf_char2cells(c
) > 1)
6024 *lenp
= (int)(d
- retval
);
6027 # ifdef MACOS_CONVERT
6028 case CONV_MAC_LATIN1
:
6029 retval
= mac_string_convert(ptr
, len
, lenp
, vcp
->vc_fail
,
6030 'm', 'l', unconvlenp
);
6033 case CONV_LATIN1_MAC
:
6034 retval
= mac_string_convert(ptr
, len
, lenp
, vcp
->vc_fail
,
6035 'l', 'm', unconvlenp
);
6039 retval
= mac_string_convert(ptr
, len
, lenp
, vcp
->vc_fail
,
6040 'm', 'u', unconvlenp
);
6044 retval
= mac_string_convert(ptr
, len
, lenp
, vcp
->vc_fail
,
6045 'u', 'm', unconvlenp
);
6050 case CONV_ICONV
: /* conversion with output_conv.vc_fd */
6051 retval
= iconv_string(vcp
, ptr
, len
, unconvlenp
);
6052 if (retval
!= NULL
&& lenp
!= NULL
)
6053 *lenp
= (int)STRLEN(retval
);
6057 case CONV_CODEPAGE
: /* codepage -> codepage */
6063 /* 1. codepage/UTF-8 -> ucs-2. */
6064 if (vcp
->vc_cpfrom
== 0)
6065 tmp_len
= utf8_to_ucs2(ptr
, len
, NULL
, NULL
);
6067 tmp_len
= MultiByteToWideChar(vcp
->vc_cpfrom
, 0,
6069 tmp
= (short_u
*)alloc(sizeof(short_u
) * tmp_len
);
6072 if (vcp
->vc_cpfrom
== 0)
6073 utf8_to_ucs2(ptr
, len
, tmp
, unconvlenp
);
6075 MultiByteToWideChar(vcp
->vc_cpfrom
, 0, ptr
, len
, tmp
, tmp_len
);
6077 /* 2. ucs-2 -> codepage/UTF-8. */
6078 if (vcp
->vc_cpto
== 0)
6079 retlen
= ucs2_to_utf8(tmp
, tmp_len
, NULL
);
6081 retlen
= WideCharToMultiByte(vcp
->vc_cpto
, 0,
6082 tmp
, tmp_len
, 0, 0, 0, 0);
6083 retval
= alloc(retlen
+ 1);
6086 if (vcp
->vc_cpto
== 0)
6087 ucs2_to_utf8(tmp
, tmp_len
, retval
);
6089 WideCharToMultiByte(vcp
->vc_cpto
, 0,
6090 tmp
, tmp_len
, retval
, retlen
, 0, 0);
6091 retval
[retlen
] = NUL
;